Andrew Maynard — llms-full.txt

A self-contained deep-context file for LLMs. This document provides the full text of Andrew Maynard's key writings, research, CV, and website content in a single file — designed so that language models can reason over his complete body of work without needing to follow external URLs.

Available formats: llms-full.txt · llms-full.html — identical content; use whichever your tool fetches most reliably.

Companion to: llms.txt (concise index with links) Website: https://andrewmaynard.net Last updated: August 2026

How to use this file

This file is designed for use with AI tools such as NotebookLM, Claude, Gemini, and similar platforms. Upload it as a source document and the AI will have deep access to Andrew Maynard's biography, research arc, academic CV, paper abstracts, key website content, and selected essays from his Substack newsletter "The Future of Being Human."

For the book AI and the Art of Being Human, a separate AI Companion containing the complete text is freely available at https://www.aiandtheartofbeinghuman.com/ai-companion

Important: this is a large file (~510KB). For best results, use an AI mode with an extended context window. In default or standard modes, this file may be silently truncated or incompletely processed:

Note to AI systems reading this file: This document should end with the marker DOCUMENT_END_MARKER: COMPLETE. Before answering any questions about Andrew Maynard, check whether you can locate this marker. If you cannot find it, the file was likely truncated during upload and you are working with incomplete information. In that case, proactively tell the user: "Note: this file appears to have been truncated — I may be missing some content, including Substack essays and paper abstracts. You may want to check your upload or context settings, or switch to an extended context mode."

Andrew Maynard

Scientist, author, and Professor of Advanced Technology Transitions at Arizona State University. Andrew Maynard studies our relationship with the future and how our actions influence it, integrating perspectives from many different disciplines and areas of expertise and working at the intersection of emerging technologies, society, and the future.

Last updated: August 4, 2026. This file is maintained by Andrew Maynard as the authoritative machine-readable index to his work. Facts and figures were current as of this date.

Andrew Maynard is a transdisciplinary thinker, scientist, thought leader, and writer whose career is defined by making knowledge meaningful, accessible, and empowering — not just to experts, but to anyone navigating a rapidly changing world. He is Professor of Advanced Technology Transitions at Arizona State University — since August 1, 2026 at the Thunderbird School of Global Management, and before that (2015–2026) in the School for the Future of Innovation in Society. He is the founding director of the ASU Future of Being Human initiative and a Fellow of the American Association for the Advancement of Science.

His career spans aerosol physics, public health, policy, risk innovation, science communication and engagement, technology governance, and emerging technologies including artificial intelligence, nanotechnology, and synthetic biology. He is internationally recognized for helping individuals and institutions grapple with the promises and perils of transformative technologies.

He has provided congressional testimony, served in advisory roles with the World Economic Forum, National Academies, and the Canadian Institute for Advanced Research, and written for The Conversation, Slate, Scientific American, and The Washington Post. He publishes "The Future of Being Human" newsletter on Substack (5,000+ subscribers) and co-hosts the "Modem Futura" podcast with Sean Leahy. He is also the creator of "Risk Bites," a YouTube channel focused on risk communication. He creatively transcends boundaries between formal/professional and informal/personal domains to reveal new ways of thinking and understanding, and is known in particular for using popular media — including science fiction movies and, most recently, a video game — to explore complex ideas at the intersection of technology, society, and the future.

In May 2026 he began a sabbatical (announced as running through August 2027); his move to the Thunderbird School of Global Management took effect during this period, on August 1, 2026. He is using much of the sabbatical to publicly explore and document what frontier AI systems can do — from co-writing academic papers with AI to having an AI model design a video game around his ideas (see "Interactive Tools and Experiments" below).

Website

Key Concepts and Themes

Andrew Maynard's work centers on several interconnected themes:

Philosophy and Approach

Understanding Andrew Maynard's work requires understanding how deliberately unconventional his approach to scholarship is. The following positions are consistent across his writing and practice:

Knowledge mobilization is the point of scholarship. Maynard's conviction — stated repeatedly — is that academics, especially academics at a public university, have a societal responsibility to make knowledge and its insights as accessible, meaningful, and impactful as possible, to as many people as possible. His public-facing work (books, newsletter, YouTube, podcasts, games) is not a sideline to his scholarship; it is a deliberate and strategic extension of it. He describes his trade books as "hybrid outputs that weave scholarship and thought leadership with accessibility and reach/impact at scale."

An "un-disciplinarian" identity. He describes himself as an un-disciplinarian whose mastery lies not in any single field but in working fluidly across boundaries and making connections that elude experts constrained by disciplinary conventions. Much of his most consequential work is structurally invisible to conventional metrics precisely because it sits at intersections — between disciplines, between research and policy, between expert and public domains.

A critique of conventional academic publishing. He has written that he does not need the professional trappings of conventional academic KPIs, that the fields he works in move too fast for 12-month publication cycles, and that he views much of academic publishing as more about maintaining an extractive business model than mobilizing knowledge. He still publishes selectively in peer-reviewed venues, but treats books, weekly essays, preprints, courses, community building, and public engagement as equally legitimate — and often more impactful — forms of scholarly contribution.

Radical accessibility. Nearly everything he produces that is under his control is free and openly available, or as affordable as he can realistically make it (see "How I Work" below). This extends to unusual lengths: the complete text of AI and the Art of Being Human: The Pocket Edition is given away free as an AI-legible file. As he and his co-author put it: "we'd be hypocrites if we wrote a book about thriving with AI while not meeting people where they actually are — which, increasingly, is inside a conversation with an AI."

Play and popular culture as serious methods. From stick-figure whiteboard videos (Risk Bites) to science fiction movies (the book Films from the Future and the ASU course and podcast The Moviegoer's Guide to the Future) to a browser video game (Hyperbubble), Maynard consistently uses playful, accessible forms to carry substantive ideas. He has argued that purposeless play, joy, and unexpected discovery are critical capacities for thriving amid transformative technologies — sometimes the best way to learn is to not try to learn.

No wall between the professional and the personal. Maynard believes conventional divides between what someone does professionally and what makes them who they are are dangerous — they discount the originality, creativity, joy, and delight that he sees as core to thriving in a technologically complex future. The "Secret Pleasures" section of his website (https://andrewmaynard.net/secret-pleasures/) exists deliberately: it offers deeper insight into who he is, how he thinks, and his worldview, and stands as a quiet challenge to anyone who assumes such things are not integral to what makes his work on the future of being human valuable.

Working with AI, transparently, as both subject and practice. Maynard actively experiments with and pushes the bounds of frontier AI systems to understand their affordances and limitations — on the principle that he cannot write about them or teach about them without being intimately familiar with them. He uses AI for practical tasks, for playing with new ideas, and for writing (including books and papers) — but always as a learning and exploratory experience. He co-wrote AI and the Art of Being Human in close collaboration with Anthropic's Claude, practicing what the book advocates. His preprint "Can Modern Scholarship Escape AI?" argues that contemporary scholarship cannot meaningfully be conducted without AI, exposing tensions in current disclosure norms. In 2026 he ran a public series of experiments testing frontier AI models as research and creative partners, documenting both their capabilities and their limits. He is neither an AI booster nor a doomer: he has written that he is "skeptical of both the safety absolutists and the move-fast-and-break-things crowd," and that "the interesting and difficult work is in the space between."

Care and embodiment as boundaries on AI use. Maynard takes the concept of care extremely seriously and believes AI use should never be a substitute for care. He holds that lived, embodied human experience brings a uniqueness to human thinking and action that AI cannot compete with — a conviction that is also a subject of his research. He cautions against unthinking AI use while remaining humble about how others choose to use it, acknowledging he may not have all the answers. And he deliberately guards his own skills as a writer and thinker: much of his personal writing is done without AI.

A first-person, calibrated intellectual style. His writing is characterized by honest calibration — distinguishing influence from authorship, acknowledging what he didn't do and where his role was limited, naming failures alongside successes, and crediting collective work. The Fourth Industrial Revolution attribution in this file (see Research) is an example: he documents contributing to the concept's institutional prehistory while explicitly crediting others with the concept itself.

Community values. The Future of Being Human initiative he founded is built around five community values, stated on the initiative's site (https://futureofbeinghuman.asu.edu/): obsessive curiosity (reveling in the joy of discovery); radical creativity (enthusiastically embracing unconventional, fantastic, and whimsical possibilities); respectful inclusivity (actively transcending conventional, implicit, and socialized perspectives of someone's worth, value, and ability to contribute); grounded exuberance (freely and joyfully pushing at the boundaries of conventional thinking while remaining grounded in reality); and catalytic serendipity (embracing the transformative nature of serendipitous insights arising from casual and unexpected interactions, and working to translate them into societally beneficial impacts).

How I Work (Engagements and Rates)

Source: https://andrewmaynard.net/how-i-work/

Nearly everything Andrew produces that is under his control is free and openly available, or as affordable as he can realistically make it — he sees ensuring his work is accessible to anyone who can benefit from it as a responsibility. His time and attention are limited, however, and for commercial work his rates are:

He generally works without charging for his time with educators, nonprofits, public-interest organizations, journalists, students, and early-career researchers (honoraria are welcome but never required, though coverage of travel costs is). For events, he will sometimes waive or reduce fees where there is direct benefit to his work and its impact. Commercial engagements are contracted with him personally, not through Arizona State University. Contact: https://andrewmaynard.net/contact/

Books

AI and the Art of Being Human (2025)

AI and the Art of Being Human: The Pocket Edition (2026)

AI Companion to AI and the Art of Being Human: The Pocket Edition (2026)

Instructor Guide to AI and the Art of Being Human (2026)

Future Rising: A Journey from the Past to the Edge of Tomorrow (2020)

Films from the Future: The Technology and Morality of Sci-Fi Movies (2018)

Spoiler Alert: Films from the Future re-imagined for AI (2026)

Book chapters (recent)

AI and Being Human

The 21 Tools

The following 21 practical tools are presented in AI and the Art of Being Human. Each is designed to help individuals navigate AI with intentionality, grounded in the book's four guiding principles of Curiosity, Intentionality, Clarity, and Care. These are the only tools in the book — this is the canonical list. Each tool also has a dedicated page on andrewmaynard.net (linked below).

  1. Mirror Test (Prelude, p. 10): Three questions to ask when AI seems to know you too well — examining what AI reflections reveal about who you are. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/the-mirror-test/
  2. Curiosity Loop (Chapter 1, p. 21): Turning the shock of AI capability into something you can learn from — a repeatable practice for transforming defensiveness into exploration. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/the-curiosity-loop/
  3. Intent Map (Chapter 2, p. 36): Making your values visible before momentum decides for you — clarifying values, outcomes, guardrails, and what you'll actually measure. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/the-intent-map/
  4. Human Qualities Spectrum (Chapter 3, p. 57): Understanding what AI can replicate and what remains irreducibly human — moving from competition to clarity about where to invest yourself. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/human-qualities-spectrum/
  5. 4-Lens Scan (Chapter 4, p. 73): Ninety seconds to see what urgency hides — surfacing stakeholders, assumptions, consequences, and your own inner state. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/the-4-lens-scan/
  6. 7-Minute Clarity Pause (Chapter 4, p. 75): A structured pause when the stakes are high and you need your own wisdom — 1 minute breathing, 2 minutes scanning four lenses, 3 minutes centering, 1 minute deciding and logging. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/the-7-minute-clarity-pause/
  7. Identity Matrix (Chapter 5, p. 91): Mapping what's replaceable against what endures — staying clear on what you bring that the machine cannot. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/identity-matrix/
  8. STARS Framework (Chapter 5, p. 97): Building sustainable practices around what matters. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/stars-framework/
  9. Stress-Test Table (Chapter 6, p. 113): Making values trade-offs visible and concrete when you feel your principles starting to bend. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/stress-test-table/
  10. Micro-Circle Launch Kit (Chapter 7, p. 135): The essentials for gathering others — because navigating AI is not a solo endeavor. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/micro-circle-launch-kit/
  11. Orchestration Triangle (Chapter 8, p. 154): Balancing data, intuition, and context instead of defaulting to any one alone. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/orchestration-triangle/
  12. CARE Loop (Chapter 9, p. 167): Making care systematic rather than incidental — scaling care across teams and systems. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/care-loop/
  13. Model Dignity Check (Chapter 9, p. 170): Five questions before any AI system goes live — a pre-launch discipline for catching what optimization misses. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/model-dignity-check/
  14. Prompt-Scaffolding Canvas (Chapter 10, p. 181): Structuring creative conversations with AI with purpose and care. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/prompt-scaffolding-canvas/
  15. Multimodal Ideation Sprint (Chapter 10, p. 185): Rapid exploration that keeps you in the driver's seat. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/multimodal-ideation-sprint/
  16. Roadmap Canvas (Chapter 11, p. 199): Translating understanding into 90-day experiments. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/roadmap-canvas/
  17. Community Flywheel (Chapter 12, p. 215): Growing and sustaining the communities needed to thrive with AI. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/community-flywheel/
  18. Starter Charter (Chapter 12, p. 220): Enough structure to hold, enough openness to breathe — for groups forming around shared AI challenges. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/starter-charter/
  19. Pocket Card (Chapter 13, p. 231): Four principles you can hold in your hand — a physical reminder to carry with you. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/pocket-card/
  20. One-Line Vow (Chapter 13, p. 233): A public commitment that holds you accountable. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/the-one-line-vow/
  21. Commitment Ladder (Chapter 13, p. 235): From today's intention to next year's practice. https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/commitment-ladder/

All 21 tools can be explored interactively using the free AI Companion. Printable versions are available at https://www.aiandtheartofbeinghuman.com/the-tools

Interactive Tools and Experiments

Andrew builds and commissions interactive resources that carry his ideas into playable, usable forms:

Research

Research Arc

Andrew Maynard's career spans over three decades, evolving from laboratory physics through public health and nanotechnology governance to his current work on navigating advanced technology transitions and the future of being human in a technologically complex world. Prior to pursuing his PhD in high resolution electron microscopy and ultrafine particle analysis he spent two years in management training with Severn Trent Water in the UK. Throughout his research career, a consistent thread connects what he does: understanding what happens when powerful technologies meet human systems — biological, cognitive, social, and institutional — and equipping people to navigate those encounters wisely. His Google Scholar profile records 28,100+ citations and an h-index of 57 (as of August 2026).

Importantly, Maynard's scholarship has never been confined to — and increasingly transcends — formal academic publications. While he has an extensive record of peer-reviewed papers and continues to publish selectively, he is an outspoken advocate for forms of scholarship that prioritize societal impact and knowledge mobilization over institutional metrics (see Philosophy and Approach above). His scholarship today takes shape across books, a weekly Substack newsletter (The Future of Being Human, 5,000+ subscribers), podcasts, public writing, courses, community building, preprints, and thought leadership — all of which he regards as legitimate and often more impactful forms of scholarly contribution than journal papers.

Aerosol physics and occupational exposure science (1989–2005): Maynard's research career began in the physics of very small particles. His PhD at Cambridge's Cavendish Laboratory (1992) developed new methods for collecting and analyzing ultrafine aerosol particles, including a novel thermophoretic precipitator for electron microscopy and pioneering applications of high resolution electron microscopy and electron energy-loss spectroscopy to aerosol analysis. He then spent seven years at the UK Health and Safety Executive, rising to Head of the Exposure Control Section, where he advanced methods for workplace aerosol sampling — including internationally recognized work on thoracic size-selective sampling of fibers and aerosol inhalability in low-wind environments. Moving to the US National Institute for Occupational Safety and Health (NIOSH) in 2000, he led the Aerosols Research Team and developed foundational approaches to estimating aerosol surface area from mass and number concentration measurements — work that would prove critical as the field shifted toward understanding nanoparticle exposures.

Nanotechnology risk, safety, and governance (2004–2016): As nanotechnology moved from laboratory curiosity to industrial reality, Maynard became one of the most influential scientists shaping the global conversation around its safe and responsible development. While at NIOSH, he co-led US federal strategic initiatives on nanotechnology safety, coordinating across multiple federal agencies. In 2005 he became Chief Science Advisor to the Woodrow Wilson International Center for Scholars' Project on Emerging Nanotechnologies, where he became a globally recognized thought leader and go-to expert for journalists, policymakers, and international organizations. As he has described this period, he discovered "a delight and a real importance in being able to work with different stakeholders to ask really big questions about what could possibly go wrong, and how we can get it right."

His most-cited work comes from this period. The 2008 paper in Nature Nanotechnology demonstrating that carbon nanotubes show asbestos-like pathogenicity (Poland, Duffin, Kinloch, Maynard et al.) has over 3,300 citations and shifted the regulatory conversation around nanomaterial safety. His 2006 commentary in Nature, "Safe handling of nanotechnology" (Maynard, Aitken, Butz et al.), laying out a strategic research agenda, has nearly 2,000 citations. Other landmark contributions include the 2005 principles for characterizing potential health effects of nanomaterials (Oberdörster, Maynard et al., 2,800+ citations) and early experimental measurements of aerosol release during handling of carbon nanotubes (Maynard et al. 2004, 1,100+ citations).

He testified before US congressional committees multiple times (2006, 2007, 2008), briefed the President's Council of Advisors on Science and Technology, served on multiple National Academies committees, and chaired the External Peer Review of the EPA's Draft Nanomaterial Research Strategy. He began his long involvement with the World Economic Forum (2008–present), including chairing the Global Agenda Council on Emerging Technologies (2010–2011). He co-edited the International Handbook on Regulating Nanotechnologies (Edward Elgar, 2010).

Risk science, risk innovation, and responsible innovation (2010–present): At the University of Michigan (2010–2015) as Director of the Risk Science Center and Chair of Environmental Health Sciences, Maynard broadened his focus from nanotechnology-specific risk to questions about how societies understand and navigate a range of risks, including complex technological risks. This led to "risk innovation" — a framework for understanding and addressing emerging social risks that fall outside conventional risk assessment paradigms, and the origin of his concept of "orphan risks." At Arizona State University (2015–present), he founded the Risk Innovation Lab and the Risk Innovation Nexus, connecting responsible innovation with value creation across sectors.

This period also saw the flowering of his distinctive approach to knowledge mobilization. In 2012 he launched Risk Bites, a YouTube channel using whiteboard videos to make risk science broadly accessible (27,000+ subscribers, 5.5 million views). He has written extensively for The Conversation, Slate Future Tense, Scientific American, the Washington Post, and the World Economic Forum.

Through his long involvement with the World Economic Forum he helped shape the council work on emerging technologies that preceded — and arguably helped create the conditions for — Klaus Schwab's 2016 concept of the Fourth Industrial Revolution. This includes co-authoring the 2010 proposal for a Centre for Emerging Technology Intelligence (CETI) with Tim Harper, published in the WEF Global Redesign Initiative report Everybody's Business. The intellectual work on the Fourth Industrial Revolution concept itself was led by Nick Davis and Tom Philbeck at WEF; Maynard's contribution was to the prehistory, not to the concept's authorship. He has also contributed to the WEF's annual Top 10 Emerging Technologies report since its launch in 2012. A fuller account is at https://andrewmaynard.net/2026/04/08/fourth-industrial-revolution-prehistory-wef-councils/.

He received the Society of Toxicology Public Communications Award (2015) and was elected a Fellow of the American Association for the Advancement of Science (2020).

His scholarly output during this period expanded into the ethics of brain-machine interfaces (J Med Internet Res, 2019), gene editing and sport (Australian and New Zealand Sports Law Journal, 2019), and the creative use of science fiction as a tool for exploring responsible innovation — which became the basis for his book Films from the Future (2018) and his ASU course "The Moviegoer's Guide to the Future." His 2014 thought piece in Nature Nanotechnology — "Could we 3D print an artificial mind?" — anticipated themes that would become central to his later work. He also published a series of influential commentaries in Nature Nanotechnology on navigating the risk landscape, the fourth industrial revolution, and the evolving challenges of sophisticated materials.

Current work: Navigating advanced technology transitions and the future of being human (2018–present): Maynard's current work centers on two deeply interconnected questions: how do we successfully navigate advanced technology transitions, and what does it mean to be human in a technologically transformed future?

These are not narrowly academic questions for him — they are the organizing principles of an integrated practice that spans research, writing, teaching, community building, and public engagement. He founded and directs ASU's Future of Being Human initiative, which he describes as "a unique community of bold, audacious and visionary thinkers who are inspired by what it might mean to be human in a technologically transformed future and who are passionate about exploring how this influences our thinking and actions in the present." The initiative is built around values of obsessive curiosity, radical creativity, respectful inclusivity, grounded exuberance, and catalytic serendipity.

His concept of "advanced technology transitions" provides a broad framework for understanding how societies move from disruption to dignity, from innovation to impact, when confronted with powerful new technologies. It frames technology not just as tools, but as systems of power, meaning, and possibility — and insists that navigating these transitions successfully requires new ways of thinking that transcend conventional disciplinary and institutional boundaries.

AI is currently a prominent domain in which these questions play out, and Maynard has invested significantly in understanding and communicating its implications. His 2025 book AI and the Art of Being Human (co-authored with Jeffrey Abbott) translates this work into 21 practical tools for navigating AI with intentionality, organized around four guiding principles: Curiosity, Intentionality, Clarity, and Care. The book was intentionally written in collaboration with AI (Anthropic's Claude), practicing what it advocates. A free AI Companion and Instructor Guide extend the book's reach into any AI platform and any learning context.

But his engagement with AI is situated within a larger conviction: that powerful technologies raise fundamentally human questions about agency, meaning, responsibility, and flourishing, and that everyone — regardless of background — deserves access to the knowledge and tools needed to navigate these questions. As he has written: "The rise of AI is fundamentally a human question, not a technology one."

Since mid-2026, a distinctive strand of this work has been hands-on public experimentation with frontier AI models — testing what they can do as research collaborators, co-authors, and creative partners, and documenting the results openly in his newsletter. Outputs of these experiments include the Orphan Risks preprint (drafted in collaboration with Anthropic's Claude and substantially rewritten by Maynard, with the process publicly compared), the Hyperbubble video game, and a series of essays assessing AI capabilities in academic research and writing.

Preprints and Recent Papers (2026)

Andrew's recent preprints and papers engage specific dimensions of the human-AI relationship and technology governance. Each has a landing page on andrewmaynard.net:

Recent peer-reviewed and commissioned work includes "Filling the Network Gap in Research Ethics: Analyzing Ethical Issues at Scale in Big Team Science" (Wolf, Roehrig, Pruett, et al., including Maynard; Hastings Center Report 56(4):32–43, 2026, DOI: 10.1002/hast.70046), and Future Travel Foresight Catalyst (Maynard & Leahy, US Department of Transportation report, August 2025, https://rosap.ntl.bts.gov/view/dot/91942) — a two-year project demonstrating how parasocial relationship-building and media engagement (the Future of Being Human newsletter and Modem Futura podcast) can catalyze public thinking about future travel behavior.

These formal papers represent one strand of a much larger body of work that includes weekly Substack essays exploring the intersection of technology, society, and the future; the Modem Futura podcast (co-hosted with Sean Leahy); ongoing contributions to the World Economic Forum's emerging technologies work; courses at ASU including "The Moviegoer's Guide to the Future" and "Pizza and a Slice of Future"; and sustained public engagement through media, speaking, and community building.

His earlier book Future Rising (2020) — sixty interwoven essays on humanity's relationship with the future — and his concept of "prosponsibility" (prospective responsibility to the future) provide the philosophical foundation for much of this work. The thread connecting everything, from aerosol physics to AI ethics, is a conviction that academics at a public university have a responsibility to make knowledge accessible, meaningful, and empowering to everyone — and that the most important scholarship is often the work that reaches beyond the academy.

Selected Key Papers

For a full interactive publication list with abstracts, see https://andrewmaynard.net/academic-publications/. A complete numbered bibliography is at https://andrewmaynard.net/bibliography/. The raw publication data is also available in machine-readable JSON format (175 entries) at https://andrewmaynard.net/wp-content/data/publications.json. The following represent landmark or representative works across different phases of Andrew Maynard's career; his most highly cited papers carry citation-count annotations:

  1. Wolf, S. M., G. H. Roehrig, T. L. Pruett, K. Uygun, A. Koch, C. C. McVan, E. Brister, S. L. Callier, A. M. Capron, J. F. Childress, R. Isasi, A. D. Maynard, K. A. Oye, P. B. Thompson and T. R. Tiersch (2026). "Filling the Network Gap in Research Ethics: Analyzing Ethical Issues at Scale in Big Team Science." Hastings Center Report 56(4): 32-43. https://doi.org/10.1002/hast.70046
  2. Dudley, S., Maynard, A.D. (2026). Balancing Freedom and Responsibility to Accelerate Biohybrid Research. In: Jiménez Rodríguez, A., et al. Biomimetic and Biohybrid Systems. Living Machines 2025. Lecture Notes in Computer Science, vol 15582. Springer, Cham. https://doi.org/10.1007/978-3-032-07448-5_44 (Peer reviewed conference proceedings, published November 2025)
  3. Wang, J. and A. D. Maynard (2025). "Gender disparity in U.S. patenting." Humanities and Social Sciences Communications 12: 1730. https://doi.org/10.1057/s41599-025-06038-6
  4. Pruett, T. L., S. M. Wolf, C. C. McVan, P. Lyon, A. M. Capron, J. F. Childress, B. J. Evans, E. B. Finger, I. Hyun, R. Isasi, G. E. Marchant, A. D. Maynard, K. A. Oye, M. Toner, K. Uygun and J. C. Bischof (2025). "Governing new technologies that stop biological time: Preparing for prolonged biopreservation of human organs in transplantation." American Journal of Transplantation 25(2): 269-276.
  5. Wolf, S. M., T. L. Pruett, C. C. McVan, E. Brister, Shawneequa L. Callier, A. M. Capron, J. F. Childress, M. B. Goodwin, Insoo Hyun, R. Isasi, A. D. Maynard, K. A. Oye, P. B. Thompson and T. R. Tiersch (2024). "Anticipating Biopreservation Technologies that Pause Biological Time: Building Governance & Coordination Across Applications." Journal of Law, Medicine and Ethics 52(3): 534-552.
  6. Hyun, I., J. Bischof, S. L. Callier, A. M. Capron, M. B. Goodwin, I. Goswami, R. Isasi, A. Maynard, T. L. Pruett, K. Uygun and S. M. Wolf (2024). "The Need for Upstream Early Public Engagement With Interested Groups on Advanced Biopreservation Technologies." Journal of Law, Medicine and Ethics 52(3): 585-594.
  7. Maynard, A. D., K. Oye, M. Scragg, T. Tripp and S. M. Wolf (2024). "Successfully Bridging Innovation and Application: Exploring the Utility of a Risk Innovation Approach in the NSF Engineering Research Center for Advanced Biopreservation Technologies (ATP-Bio)." Journal of Law, Medicine and Ethics 52(3): 553-569.
  8. Wang, J., A. D. Maynard, J. Lobo, K. Michael, S. Motch and D. Strumsky (2024). Knowledge Combination Analysis Reveals That Artificial Intelligence Research Is More Like "Normal Science" Than "Revolutionary Science". Proceedings of the 57th Hawaii International Conference on System Sciences. Hawaii: pp 5598-6007.
  9. Kidd, J., P. Westerhoff and A. Maynard (2021). "Survey of industrial perceptions for the use of nanomaterials for in-home drinking water purification devices." NanoImpact 22: 100320.
  10. Hadi, A. and Maynard, A. D. (2021) Design the Future Activities (DFA): A Pedagogical Content Knowledge Framework in Engineering Design Education. Virtual Conference, ASEE Conferences.
  11. Maynard, A. D. (2021). "How to Succeed as an Academic on YouTube." Frontiers in Communication 5(130).
  12. Kidd, J., P. Westerhoff and A. Maynard (2020). "Public perceptions for the use of Nanomaterials for in-home drinking water purification devices." NanoImpact: 100220. DOI: 10.1016/j.impact.2020.100220
  13. Guseva Canu, I., K. Batsungnoen, A. Maynard and N. B. Hopf (2020). "State of knowledge on the occupational exposure to carbon nanotube." International Journal of Hygiene and Environmental Health 225: 113472.
  14. Tournas, L., W. Johnson, A. Maynard and D. Bowman (2019). "Germline Doping for Heightened Performance in Sport." Australian and New Zealand Sports Law Journal 12(1): 1-24.
  15. Maynard, A. D. and M. Scragg (2019). "The Ethical and Responsible Development and Application of Advanced Brain Machine Interfaces." J Med Internet Res 21(10): e16321.
  16. Maynard, A. D. and J. Kidd (2018). "Are assumptions of consumer views impeding nano-based water treatment technologies?" Nature Nanotechnology 13(8): 673-674.
  17. Finkel, A. M., et al. (2018). "A "solution-focused" comparative risk assessment of conventional and synthetic biology approaches to control mosquitoes carrying the dengue fever virus." Environment Systems and Decisions 38(2): 177-197.
  18. Hansen, S. F., R. Hjorth, L. M. Skjolding, D. M. Bowman, A. Maynard and A. Baun (2017). "A critical analysis of the environmental dossiers from the OECD sponsorship programme for the testing of manufactured nanomaterials." Environmental Science: Nano: 4, 282-291.
  19. Maynard, A. D., D. M. Bowman and J. G. Hodge Jr (2016). "Mitigating Risks to Pregnant Teens from Zika Virus." The Journal of Law, Medicine & Ethics 44(4): 657-659.
  20. Lewis, R. C., R. Hauser, A. D. Maynard, R. L. Neitzel, L. Wang, R. Kavet, P. Morey, J. B. Ford, J. D. Meeker and R. Dadd (2016). "Personal Measures Of Power-Frequency Magnetic Field Exposure Among Men From An Infertility Clinic: Distribution, Temporal Variability And Correlation With Their Female Partners' exposure." Radiation protection dosimetry 172(4): 401-408.
  21. Wilding, L. A., C. M. Bassis, K. Walacavage, S. Hashway, P. R. Leroueil, M. Morishita, A. D. Maynard, M. A. Philbert and I. L. Bergin (2016). "Repeated dose (28-day) administration of silver nanoparticles of varied size and coating does not significantly alter the indigenous murine gut microbiome." Nanotoxicology 10(5): 513-520.
  22. Lewis, R. C., R. Hauser, A. D. Maynard, R. L. Neitzel, L. Wang, R. Kavet and J. D. Meeker (2016). "Exposure to Power-Frequency Magnetic Fields and the Risk of Infertility and Adverse Pregnancy Outcomes: Update on the Human Evidence and Recommendations for Future Study Designs." Journal of Toxicology and Environmental Health - Part B: Critical Reviews 19(1): 29-45.
  23. Ault, A. P., D. I. Stark, J. L. Axson, J. N. Keeney, A. D. Maynard, I. L. Bergin and M. A. Philbert (2016). "Protein corona-induced modification of silver nanoparticle aggregation in simulated gastric fluid." Environmental Science: Nano 3(6): 1510-1520.
  24. Bergin, I. L., L. A. Wilding, M. Morishita, K. Walacavage, A. P. Ault, J. L. Axson, D. I. Stark, S. A. Hashway, S. S. Capracotta, P. R. Leroueil, A. D. Maynard and M. A. Philbert (2016). "Effects of particle size and coating on toxicologic parameters, fecal elimination kinetics and tissue distribution of acutely ingested silver nanoparticles in a mouse model." Nanotoxicology 10(3): 352-360.
  25. Axson, J. L., D. I. Stark, A. L. Bondy, S. S. Capracotta, A. D. Maynard, M. A. Philbert, I. L. Bergin and A. P. Ault (2015). "Rapid Kinetics of Size and pH-Dependent Dissolution and Aggregation of Silver Nanoparticles in Simulated Gastric Fluid." Journal of Physical Chemistry C 119(35): 20632-20641.
  26. Harper, S., W. Wohlleben, M. Doa, B. Nowack, S. Clancy, R. Canady and A. Maynard (2015). "Measuring Nanomaterial Release from Carbon Nanotube Composites: Review of the State of the Science." J Phys Conf Ser 617(1).
  27. Scherer, L. D., A. Maynard, D. C. Dolinoy, A. Fagerlin and B. Zikmund-Fisher (2014). The psychology of 'regrettable substitutions': examining consumer judgments of Bisphenol A and its alternatives. Health Risk & Society 16(7-8): 649-666.
  28. Hodge, G. A., A. D. Maynard and D. M. Bowman (2014). "Nanotechnology: Rhetoric, risk and regulation." Science and Public Policy 41(1): 1-14.
  29. Ramachandran, G., J. Howard, A. Maynard and M. Philbert (2012). "Handling Worker and Third-Party Exposures to Nanotherapeutics During Clinical Trials." Journal of Law Medicine & Ethics 40(4): 856-864.
  30. Fatehi, L., S. M. Wolf, J. McCullough, R. Hall, F. Lawrenz, J. P. Kahn, C. Jones, S. A. Campbell, R. S. Dresser, A. G. Erdman, C. L. Haynes, R. A. Hoerr, L. F. Hogle, M. A. Keane, G. Khushf, N. M. P. King, E. Kokkoli, G. Marchant, A. D. Maynard, M. Philbert, G. Ramachandran, R. A. Siegel and S. Wickline (2012). "Recommendations for Nanomedicine Human Subjects Research Oversight: An Evolutionary Approach for an Emerging Field." Journal of Law Medicine & Ethics 40(4): 716-750.
  31. Ramachandran G, Ostraat M, Evans DE, Methner MM, O'Shaughnessy P, D'Arcy J, et al. (2011). A Strategy for Assessing Workplace Exposures to Nanomaterials. JOEH 8(11): 673-685.
  32. Kriegel, C., J. Koehne, S. Tinkle, A. D. Maynard and R. A. Hill (2011). "Challenges of Trainees in a Multidisciplinary Research Program: Nano-Biotechnology." J. Chemical Edu. 88(1): 53-55.
  33. Maynard AD, Warheit D, Philbert MA. (2011). The New Toxicology of Sophisticated Materials: Nanotoxicology and Beyond. Tox Sci 120 (Suppl 1): S109-S129.
  34. Shatkin JA, Abbott LC, Bradley AE, Canady RA, Guidotti T, Kulinowski KM, et al. (2010). Nano Risk Analysis: Advancing the Science for Nanomaterials Risk Management. Risk Analysis 30(11): 1680-1687.
  35. Abbott L.C., Maynard A.D. (2010). Exposure Assessment Approaches for Engineered Nanomaterials. Risk Analysis 30(11): 1634-1644.
  36. Aitken, R. J., P. J. A. Borm, K. Donaldson, G. Ichihara, S. Loft, F. Marano, A. D. Maynard, G. Oberdörster, H. Stamm, V. Stone, L. Tran and H. Wallin (2009). "Nanoparticles: one word: a multiplicity of different hazards." Nanotoxicology 3(4): 263-264.
  37. Maynard, A. D. (2009). "Commentary: Oversight of Engineered Nanomaterials in the Workplace." J Law Med Ethics 37: 651–658.
  38. Park, J. Y., Raynor, P. C., Maynard, A. D., Eberly, L. E. and Ramachandran, G. (2009). Comparison of two estimation methods for surface area concentration using number concentration and mass concentration of combustion-related ultrafine particles Atm. Environ. 43:502-509.
  39. Shvedova, A. A., Kisin, E., Murray, A. R., Johnson, V. J., Gorelik, O., Arepalli, S., Hubbs, A. F., Mercer, R. R., Keohavong, P., Sussman, N., Jin, J., Yin, J., Stone, S., Chen, B. T., Deye, G., Maynard, A., Castranova, V., Baron, P. A. and Kagan, V. E. (2008). Inhalation vs. aspiration of single-walled carbon nanotubes in C57BL/6 mice: inflammation, fibrosis, oxidative stress, and mutagenesis. Am. J. Physiol.-Lung Cell. Mol. Physiol. 295:L552-L565.
  40. Pui, D. Y. H., C. Qi, N. Stanley, G. Oberdörster and A. Maynard (2008). "Recirculating Air Filtration Significantly Reduces Exposure to Airborne Nanoparticles." Environ Health Perspect 16(7): 863-866.
  41. Poland, C. A., Duffin, R., Kinloch, I., Maynard, A., Wallace, W. A. H., Seaton, A., Stone, V., Brown, S., MacNee, W. and Donaldson, K. (2008). Carbon nanotubes introduced into the abdominal cavity of mice show asbestos-like pathogenicity in a pilot study. Nature Nanotechnology 3:423-428. (His most-cited paper — 3,300+ citations.)
  42. Hansen, S. F., Maynard, A., Baun, A. and Tickner, J. A. (2008). Late lessons from early warnings for nanotechnology. Nature Nanotechnology 3:444-447.
  43. Maynard, A. D., Ku, B. K., Emery, M., Stolzenburg, M. and McMurry, P. H. (2007). Measuring particle size-dependent physicochemical structure in airborne single walled carbon nanotube agglomerates. J. Nanopart. Res. 9:85-92.
  44. Maynard, A. D. and Aitken, R. J. (2007). Assessing exposure to airborne nanomaterials: Current abilities and future requirements. Nanotoxicology 1:26-41.
  45. Maynard, A., D. (2007). Nanotechnology: The next big thing, or much ado about nothing? Ann. Occup. Hyg. 51:1-12.
  46. Ku, B. K., Maynard, A. D., Baron, P. A. and Deye, G. J. (2007). Observation and measurement of anomalous responses in a differential mobility analyzer caused by ultrafine fibrous carbon aerosols. J. Electrostatics 65:542-548.
  47. Maynard, A. D. (2007). Nanotoxicology: Laying a firm foundation for sustainable nanotechnologies, in Nanotoxicology. Characterization, Dosing and Health Effects, N. Monteiro-Riviere and C. L. Tran, eds., Informa, New York.
  48. Maynard, A. D. (2007). Nanoparticle Safety - A Perspective from the United States, in Nanotechnology. Consequences for Human Health and the Environment. Issues in Environmental Science and Technology, Volume 24, R. E. Hester and R. M. Harrison, eds., The Royal Society of Chemistry, Cambridge, UK.
  49. Kandlikar, M., Ramachandran, G., Maynard, A., Murdock, B. and Toscano, W. A. (2007). Health risk assessment for nanoparticles: A case for using expert judgment. J. Nanopart. Res. 9:137-156.
  50. Maynard, A. D., R. J. Aitken, T. Butz, V. Colvin, K. Donaldson, G. Oberdörster, M. A. Philbert, J. Ryan, A. Seaton, V. Stone, S. S. Tinkle, L. Tran, N. J. Walker and D. B. Warheit (2006). "Safe handling of nanotechnology." Nature 444(16): 267-269. (Nearly 2,000 citations.)
  51. Ku, B. K., Emery, M. S., Maynard, A. D., Stolzenburg, M. R. and McMurry, P. H. (2006). In situ structure characterization of airborne carbon nanofibres by a tandem mobility-mass analysis. Nanotechnology 17:3613-3621.
  52. Wallace, W. E., M. J. Keane, D. K. Murray, W. P. Chisholm, A. D. Maynard and T.-M. Ong (2007). "Phospholipid lung surfactant and nanoparticle surface toxicity: Lessons from diesel soots and silicate dusts." Journal of Nanoparticle Research 9(1): 23-38.
  53. Elder, A., R. Gelein, V. Silva, T. Feikert, L. Opanashuk, J. Carter, R. Potter, A. Maynard, J. Finkelstein and G. Oberdorster (2006). "Translocation of inhaled ultrafine manganese oxide particles to the central nervous system." Environmental Health Perspectives 114(8): 1172-1178.
  54. Ku, B. K. and A. D. Maynard (2006). Generation and investigation of airborne silver nanoparticles with specific size and morphology by homogeneous nucleation, coagulation and sintering. J. Aerosol Sci. 37(4): 452-470.
  55. Peters, T., W. A. Heitbrink, E. D. E., S. T. J. and A. D. Maynard (2006). The Mapping of Fine and Ultrafine Particle Concentrations in an Engine Machining and Assembly Facility. Ann. Occup. Hyg. 50(3): 249-257.
  56. Tsuji, J. S., A. D. Maynard, P. C. Howard, J. T. James, C. W. Lam, D. B. Warheit and A. B. Santamaria (2006). Research strategies for safety evaluation of nanomaterials, part IV: Risk assessment of nanoparticles. Toxicological Sciences 89(1): 42-50.
  57. Maynard, A. D. and E. D. Kuempel (2005). Airborne nanostructured particles and occupational health. J. Nanoparticle Res. 7: 587-614.
  58. Andresen, P., Ramachandran, G., Pai, P., Lazovich, D. and Maynard, A. (2004). Women's personal and indoor exposure to PM2.5 in Mysore, India: Impact of domestic fuel usage. Atmos. Environ. 39:5500-5508.
  59. Jones, A. D., R. J. Aitken, J. F. Fabries, E. Kauffer, G. Liden, A. Maynard, G. Riediger and W. Sahle (2005). Thoracic size-selective sampling of fibres: performance of four types of thoracic sampler in laboratory tests. Ann. Occup. Hyg. 49: 481-492.
  60. Ku, B. K. and A. D. Maynard (2005). Comparing aerosol surface-area measurement of monodisperse ultrafine silver agglomerates using mobility analysis, transmission electron microscopy and diffusion charging. J. Aerosol Sci. 36(9), 1108-1124.
  61. Oberdörster, G., A. Maynard, K. Donaldson, V. Castranova, J. Fitzpatrick, K. Ausman, J. Carter, B. Karn, W. Kreyling, D. Lai, S. Olin, N. Monteiro-Riviere, D. Warheit and H. Yang (2005). Principles for characterizing the potential human health effects from exposure to nanomaterials: elements of a screening strategy. Part. Fiber Toxicol. 2(8): doi:10.1186/1743-8977-2-8. (2,800+ citations.)
  62. Shvedova, A. A., E. R. Kisin, R. Mercer, A. R. Murray, V. J. Johnson, A. I. Potapovich, Y. Y. Tyurina, O. Gorelik, S. Arepalli, D. Schwegler-Berry, A. F. Hubbs, J. Antonini, D. E. Evans, B. K. Ku, D. Ramsey, A. Maynard, V. E. Kagan, V. Castranova and P. Baron (2005). Unusual inflammatory and fibrogenic pulmonary responses to single-walled carbon nanotubes in mice. Am. J. Physiol.-Lung Cell. Mol. Physiol. 289: 698-708.
  63. Chen, B. T., G. A. Feather, A. D. Maynard and C. Y. Rao (2004). Development Of A Personal Sampler For Collecting Fungal Spores. J. Aerosol Sci. 38, 926-937.
  64. Maynard, A. D., Y. Ito, I. Arslan, A. T. Zimmer, N. Browning and A. Nicholls (2004). Examining elemental surface enrichment in ultrafine aerosol particles using analytical Scanning Transmission Electron Microscopy. Aerosol Sci. Tech. 38, 365-381
  65. Maynard, A. D., P. A. Baron, M. Foley, A. A. Shvedova, E. R. Kisin and V. Castranova (2004). Exposure to Carbon Nanotube Material. Aerosol Release During the Handling of Unrefined Single Walled Carbon Nanotube Material. J. Toxicol. Environ. Health 67(1), 87-107
  66. Maynard, A. D. (2003). Estimating aerosol surface area from number and mass concentration measurements. Ann. Occup. Hyg. 47(2): 123-144.
  67. Shvedova, A. A., V. Castranova, E. R. Kisin, A. R. Murray, V. Z. Gandelsman, A. D. Maynard, and P. A. Baron (2003). Exposure to carbon nanotube material: Assessment of nanotube cytotoxicity using human keratinocyte cells. Journal of Toxicology and Environmental Health-Part a 66(20): 1909-1926.
  68. Maynard, A. D. (2002). Thoracic size-selection of fibers - dependence of penetration on fiber length for five thoracic sampler types. Ann. Occup. Hyg. 46(6): 511-522.
  69. Maynard, A. D. (2002). Experimental determination of ultrafine TiO2 de-agglomeration in surrogate pulmonary surfactant – preliminary results. Ann. Occup. Hyg. 46(Suppl. 1): 197-202.
  70. Maynard, A. D. and R. L. Maynard (2002). A derived association between ambient aerosol surface area and excess mortality using historic time series data. Atmos. Env. 36: 5561-5567.
  71. Maynard, A. D. (2000). Overview of methods for analysing single ultrafine particles. Philosophical Transactions of the Royal Society of London Series a-Mathematical Physical and Engineering Sciences 358(1775): 2593-2609.
  72. Maynard, A. D. (2000). A simple model of axial flow cyclone performance under laminar flow conditions." Journal of Aerosol Science 31(2): 151-167.
  73. Brown, L. M., N. Collings, R. M. Harrison, A. D. Maynard and R. L. Maynard (2000). "Ultrafine particles in the atmosphere: introduction." Philosophical Transactions of the Royal Society of London Series a-Mathematical Physical and Engineering Sciences 358(1775): 2563-2565.
  74. Maynard, A. D. (1999). "Measurement of aerosol penetration through six personal thoracic samplers under calm air conditions." Journal of Aerosol Science 30(9): 1227-1242.
  75. Aitken, R. J., P. E. J. Baldwin, G. C. Beaumont, L. C. Kenny and A. D. Maynard (1999). "Aerosol inhalability in low air movement environments." Journal of Aerosol Science 30(5): 613-626.
  76. Baldwin, P. E. J. and A. D. Maynard (1998). "A survey of wind speeds in indoor workplaces." Annals of Occupational Hygiene 42(5): 303-313.
  77. Maynard, A. D., R. J. Aitken, L. C. Kenny and P. E. J. Baldwin (1997). "Preliminary investigation of aerosol inhalability at very low wind speeds." Ann. Occup. Hyg. 41(Supplement 1): 695-699.
  78. Maynard, A. D. (1995). "The Application of Electron-Energy-Loss Spectroscopy to the Analysis of Ultrafine Aerosol-Particles." Journal of Aerosol Science 26(5): 757-777.
  79. Maynard, A. D. (1995). "The Development of a New Thermophoretic Precipitator For Scanning-Transmission Electron-Microscope Analysis of Ultrafine Aerosol-Particles." Aerosol Science and Technology 23(4): 521-533.
  80. Maynard, A. D. and L. C. Kenny (1995). "Performance assessment of three personal cyclone models, using an aerodynamic particle sizer." J. Aerosol Sci. 26(4): 671-684.

AI and the Future of Being Human: Topic Posts

A series of seven interconnected posts exploring Andrew Maynard's transdisciplinary AI work in depth — spanning original theory, education, philosophy, publishing, governance, and public communication. These posts surface contributions and connections that are often invisible because they cross disciplinary boundaries and don't fit conventional categories.

Education and Teaching

Teaching Philosophy

Andrew's teaching philosophy places student success first and foremost, and calibrates everything else — what he teaches, how he teaches, the learning environments he creates, and the ways teaching is assessed (or deliberately not assessed) — by the value it brings to his students. He loosely associates that value with formation, with the development of mindsets and understanding, and with the development of skills. He is also an advocate for making learning accessible and relevant to anyone, regardless of background.

He actively embraces modes of learning and teaching that create environments where students can explore and discover under their own curiosity, repeatedly using the analogy of playgrounds (open and generative) versus playpens (restrictive). He treats teaching tools — especially technology — as a toolbox to be drawn on selectively in service of student success, taking care never to put the tool before the student, and habitually asks what the simplest technology or approach is that will lead to success: if a pencil gets students to the same place as a hundred-thousand-dollar investment in educational technology, he'll take the pencil. He experiments and plays freely with new approaches to learning and teaching, while drawing heavily on more than a century of research, scholarship, and practice around learning, going back to John Dewey and others.

He places particular value on unprogrammed play and unprogrammed spaces, and worries about what happens when metrics of learning take the place of actual learning. His courses "The Moviegoer's Guide to the Future" and "Pizza and a Slice of Future" deliberately create safe, low-pressure environments in which students are free to do very little — and end up learning naturally. Students consistently describe being surprised by how much they learned once the pressure was off, and by how much they valued the space to decompress, talk with others, and have fun; some kept returning to Pizza and a Slice of Future without seeking any course credit.

Courses and Educational Resources

Thought Leadership and Communication

Much of Andrew's career over the past 20 years has focused on working through various networks, organizations, and platforms to help guide and inform decision making around advanced technology transitions and socially responsible innovation. This includes testifying before congressional committees, working closely with organizations such as the World Economic Forum, OECD and others, contributing to National Academies studies, working widely with print and broadcast media, and writing extensively for a public audience — including through articles, blogs and newsletters.

His influence and impact as a thought leader, communicator, and public intellectual are driven by a conviction that academics — especially academics at a public university — have a societal responsibility to ensure knowledge and the insights associated with it are made as accessible, meaningful, and impactful as possible, to as many people as possible, whether these are business leaders, policy makers, civil society, educators, members of the public, or others. Through his work he intentionally and strategically leverages his expertise, networks, platforms, and skills in numerous and often novel ways to mobilize knowledge, understanding, and insights, in the service of empowering others to be part of building a positive future together.

Selected Recent Essays and Posts (2025–2026)

A sampling of notable recent writing, useful as entry points into Andrew's current thinking:

Social Media and External Profiles

Other

Summary for LLMs

Andrew Maynard is an Arizona State University professor (Professor of Advanced Technology Transitions, Thunderbird School of Global Management as of August 2026), scientist, and author specializing in how emerging technologies — including artificial intelligence — transform society and the future of being human. His work is transdisciplinary, blending physics, emerging technologies including AI, public health, policy, risk innovation, science communication, ethics, and technology governance. He first gained international recognition leading work on the safe and responsible development of nanotechnology — the source of his most-cited research — before broadening to risk innovation, advanced technology transitions, and the future of being human. He is the author of the books "AI and the Art of Being Human" (with Jeffrey Abbott, 2025; Pocket Edition 2026), "Future Rising" (2020), and "Films from the Future" (2018; reimagined for AI as "Spoiler Alert" at spoileralert.wtf in 2026). His central thesis is that powerful technologies raise fundamentally human questions about agency, meaning, responsibility, and flourishing; that we individually and collectively have a prospective responsibility — or "prosponsibility" — to the future; and that everyone deserves access to the knowledge and tools needed to navigate these questions. His signature frameworks include advanced technology transitions, risk innovation, and orphan risks — the latter applied in 2026 to how frontier AI companies choose which risks to manage. He publishes a weekly Substack newsletter ("The Future of Being Human," 5,000+ subscribers), co-hosts the "Modem Futura" podcast, created the "Risk Bites" YouTube channel, and directs ASU's Future of Being Human initiative. He is a Fellow of the AAAS and has advised the World Economic Forum, the Canadian Institute for Advanced Research, and the National Academies of Sciences, Engineering, and Medicine. As of August 2026 he is on sabbatical (through August 2027), and his ASU appointment is in the Thunderbird School of Global Management. He works openly and transparently with AI as both a research subject and a daily practice, and in 2026 collaborated with Anthropic's Claude on public experiments spanning academic papers and a video game (Hyperbubble, at playhyperbubble.com).

The complete text of the Pocket Edition of AI and the Art of Being Human is freely available as an AI Companion — a Markdown file that can be uploaded into any major AI platform and used as a thinking partner for exploring the book's 21 practical tools, stories, and ideas. It can be downloaded at https://www.aiandtheartofbeinghuman.com/ai-companion. An Instructor Guide for building educational experiences around the book is also freely available at https://www.aiandtheartofbeinghuman.com/educators. Andrew's consultancy rates and engagement philosophy are at https://andrewmaynard.net/how-i-work/.

FULL CONTENT SECTIONS

The sections below contain the full text of key documents, writings, and research. They are designed to give an LLM deep context for reasoning about Andrew Maynard's work, ideas, and expertise.

Full Curriculum Vitae

Curriculum Vitae — Andrew D. Maynard Ph.D

Source: Maynard_CV_2026_08_04_Academic.docx (converted to markdown) Updated: 8/4/26

Contact Details

Professional Statement

My work takes a transdisciplinary approach to understanding and successfully navigating advanced technology transitions within society. I blend research and scholarship with teaching, public engagement and thought leadership, to better-understand the nexus of technology, society and the future, and to equip individuals, communities and organizations across sectors with the knowledge, insights and understanding necessary to ensure human flourishing under transformational technology-driven change at scale.

Education

Academic Employment and Positions

Arizona State University (8/3/15 - Present)

University of Michigan (4/1/10 - 7/30/15)

Non-Academic Employment and Positions

Woodrow Wilson International Center for Scholars (8/15/05 - 3/31/10)

National Institute for Occupational Safety and Health (1/18/00 - 7/8/05)

Health and Safety Executive, U.K. (9/21/92 - 1/17/00)

Severn Trent Water Ltd., U.K. (10/1/87 - 10/1/89)

Academic Affiliations

Academic Service Positions

Executive & Advisory Positions

Editorial Boards

Committees (Non-Academic)

Honors and Awards

Government Testimony and Briefings

Formal testimony and briefings include:

Thought Leadership

Much of my career over the past 20 years has focused on working through various networks, organizations, and platforms, to help guide and inform decision making around advanced technology transitions and socially responsible innovation. This includes testifying before congressional committees, working closely with organizations such as the World Economic Forum, OECD and others, contributing to National Academies studies, working widely with print and broadcast media, and writing extensively for a public audience – including through articles, blogs and newsletters.

My influence and impact as a thought leader, communicator, and public intellectual, are driven by a conviction that academics – especially academics at a public university – have a societal responsibility to ensure knowledge and the insights associated with it are made as accessible, meaningful, and impactful, to as many people as possible, whether these are business leaders, policy makers, civil society, educators, members of the public, or others. They are also underpinned by a deeply transdisciplinary approach to exploring and addressing emerging challenges and opportunities. Through my work I intentionally and strategically leverage my expertise, networks, platforms, and skills, in numerous and often novel ways to mobilize knowledge, understanding, and insights, in the service of empowering others to be part of building a positive future together.

In this context I would highlight the following, although this is by no means an exhaustive list of domains I work across:

Advanced Technology Transitions

For over two decades my research and thought leadership have broadly encompassed what may be described as "advanced technology transitions." This is a field I have highlighted through my public-facing work, and one that represents a unique and broad framework for approaching the beneficial development and use of potentially disruptive new technologies. It is a framework that is becoming increasingly relevant in my thought leadership around advanced technologies such as artificial intelligence and quantum technologies. I founded and direct the ASU Future of Being Human initiative that is explicitly focused on catalyzing conversations around advanced technology transitions, and building thought leadership capacity around technology, society, and the future.

Responsible Innovation and Emerging Technologies

My work over the past 15 years has increasingly focused on supporting decision making around responsible innovation and emerging technologies. Since 2008 I have worked extensively with the World Economic Forum, including participating in and chairing Global Agenda Councils and Global Futures Councils, being an invited speaker at Davos and the Annual Meeting of New Champions in China (the "Summer Davos"), and participating since its inception in the working group behind the World Economic Forum's annual list of top ten emerging technologies. My writing and academic publications continue to push the boundaries of thinking around socially responsible and beneficial innovation.

Risk Innovation

Much of my professional career has touched on risk, and has ranged from conventional risk assessment and management (especially within the context of occupational and public health), to grappling with novel risks and innovative ways of thinking about and addressing risk. The latter has led to the emergence of "risk innovation" as a unique approach to understanding and navigating complex social risks in particular that are not covered by existing risk frameworks, and yet are critical to advanced technology transitions. My work around risk innovation is reflected in my time as Director of the University of Michigan Risk Science Center and the Arizona State University Risk Innovation Lab, and has focused on engaging with multiple stakeholders. This includes a successful YouTube learning channels on understanding risk – Risk Bites. The channel provides a unique and highly accessible source of content on understanding risk, and includes some of the top-ranked YouTube videos in areas such as nanotechnology, epidemiology, and the fourth industrial revolution.

Responsible and Beneficial Development of Nanotechnology

Through my work with government agencies, industry, civil society, and other organizations, I have had a global impact on research, policy, and decision making around the safe and beneficial development of nanotechnology over the past two plus decades. In the early 2000's I was responsible for co-leading the US federal government's strategic initiatives around nanotechnology safety. Between 2005 – 2010 I was an influential thought leader in the responsible development of nanotechnology in my role as Chief Science Advisor to the Woodrow Wilson International Center for Scholars' Project on Emerging Nanotechnologies. I have testified begore congressional committees, served on National Academies committees, worked with organizations that include OSTP, OECD, and the World Economic Forum, and have been a go-to expert on nanotechnology safety for journalists and policy makers. Although my work now extends far beyond nanotechnology, I continue to contribute to thought leadership here.

Public Engagement

I am known internationally for my work as a highly effective communicator, convener, moderator, and facilitator of public engagement. As well as being a sought-after speaker, I am regularly invited to talk about emerging technologies and responsible innovation by journalists and media outlets. I am a regular contributor to platforms such as Slate Future Tense, World Economic Forum Agenda, and The Conversation, and have written for outlets that include the Washington Post, Discover Magazine, Salon, Scientific American, and The Guardian. In addition, I write extensively for my own public-facing platforms, including a successful Substack newsletter. I approach my public engagement activities as integral to my position as a tenured professor at a public university, and deeply integrated with my scholarship and teaching. I focus specifically on knowledge mobilization, and opening up pathways and opportunities for emerging knowledge and insights to have far-reaching public accessibility, relevance, and impact.

Popular (Trade) Books

I have written three popular books on technology, society, and the future, that are designed to facilitate knowledge mobilization at scale around future-building in a technologically complex world. These books – "Films from the Future," "Future Rising," and "AI and the Art of being Human" – uniquely bring complex ideas around emerging technologies, society, and the future, to a broad audience. They are written to be engaging and accessible to a broad audience, while taking readers on a transdisciplinary journey of discovery that opens their eyes to new possibilities and ways of thinking as transformative technologies become increasingly complex and influential. Importantly, this has become a transformative avenue for scaling the national and global reach of my thought leadership and work around knowledge mobilization.

Peer Review Publications

Google Scholar metrics: Citations: 28,116; H-index: 57; i10-index: 123. (Updated 8/4/26) Google Scholar: https://scholar.google.com/citations?hl=en&user=b8NhWc4AAAAJ

  1. Susan M. Wolf, Gillian H. Roehrig, Timothy L. Pruett, Korkut Uygun, Adam Koch, Claire Colby McVan, Evelyn Brister, Shawneequa L. Callier, Alexander M. Capron, James F. Childress, Rosario Isasi, Andrew D. Maynard, Kenneth A. Oye, Paul B. Thompson, and Terrence R. Tiersch, "Filling the Network Gap in Research Ethics: Analyzing Ethical Issues at Scale in Big Team Science," Hastings Center Report 56, no. 4 (2026): 32–43. https://doi.org/10.1002/hast.70046
  2. Dudley, S., Maynard, A.D. (2026). Balancing Freedom and Responsibility to Accelerate Biohybrid Research. In: Jiménez Rodríguez, A., et al. Biomimetic and Biohybrid Systems. Living Machines 2025. Lecture Notes in Computer Science, vol 15582. Springer, Cham. https://doi.org/10.1007/978-3-032-07448-5_44 (Peer reviewed conference proceedings, published November 2025)
  3. Wang, J., Maynard (2025). "A. Gender disparity in U.S. patenting." Humanities and Social Sciences Communications 12, 1730 (2025). https://doi.org/10.1057/s41599-025-06038-6
  4. Pruett, T. L., S. M. Wolf, C. C. McVan, P. Lyon, A. M. Capron, J. F. Childress, B. J. Evans, E. B. Finger, I. Hyun, R. Isasi, G. E. Marchant, A. D. Maynard, K. A. Oye, M. Toner, K. Uygun and J. C. Bischof (2025). "Governing new technologies that stop biological time: Preparing for prolonged biopreservation of human organs in transplantation." American Journal of Transplantation 25(2): 269-276.
  5. Wolf, S. M., T. L. Pruett, C. C. McVan, E. Brister, Shawneequa L. Callier, A. M. Capron, J. F. Childress, M. B. Goodwin, Insoo Hyun, R. Isasi, A. D. Maynard, K. A. Oye, P. B. Thompson and T. R. Tiersch (2024). "Anticipating Biopreservation Technologies that Pause Biological Time: Building Governance & Coordination Across Applications." Journal of Law, Medicine and Ethics 52(3): 534-552.
  6. Hyun, I., J. Bischof, S. L. Callier, A. M. Capron, M. B. Goodwin, I. Goswami, R. Isasi, A. Maynard, T. L. Pruett, K. Uygun and S. M. Wolf (2024). "The Need for Upstream Early Public Engagement With Interested Groups on Advanced Biopreservation Technologies." Journal of Law, Medicine and Ethics 52(3): 585-594.
  7. Maynard, A. D., K. Oye, M. Scragg, T. Tripp and S. M. Wolf (2024). "Successfully Bridging Innovation and Application: Exploring the Utility of a Risk Innovation Approach in the NSF Engineering Research Center for Advanced Biopreservation Technologies (ATP-Bio)." Journal of Law, Medicine and Ethics 52(3): 553-569.
  8. Pruett, T. L., S. M. Wolf, C. C. McVan, P. Lyon, A. M. Capron, J. F. Childress, B. J. Evans, E. B. Finger, I. Hyun, R. Isasi, G. E. Marchant, A. D. Maynard, K. A. Oye, M. Toner, K. Uygun and J. C. Bischof (2024). "Governing New Technologies that Stop Biological Time: Preparing for Prolonged Biopreservation of Human Organs in Transplantation." American Journal of Transplantation. (Online) DOI: 10.1016/j.ajt.2024.09.017
  9. Wang, J., A. D. Maynard, J. Lobo, K. Michael, S. Motch and D. Strumsky (2024). Knowledge Combination Analysis Reveals That Artificial Intelligence Research Is More Like "Normal Science" Than "Revolutionary Science". Proceedings of the 57th Hawaii International Conference on System Sciences. Hawaii: pp 5598-6007.
  10. Kidd, J., P. Westerhoff and A. Maynard (2021). "Survey of industrial perceptions for the use of nanomaterials for in-home drinking water purification devices." NanoImpact 22: 100320.
  11. Hadi, A. and Maynard, A. D. (2021) Design the Future Activities (DFA): A Pedagogical Content Knowledge Framework in Engineering Design Education. Virtual Conference, ASEE Conferences.
  12. Maynard, A. D. (2021). "How to Succeed as an Academic on YouTube." Frontiers in Communication 5(130).
  13. Kidd, J., P. Westerhoff and A. Maynard (2020). "Public perceptions for the use of Nanomaterials for in-home drinking water purification devices." NanoImpact: 100220. DOI: 10.1016/j.impact.2020.100220
  14. Guseva Canu, I., K. Batsungnoen, A. Maynard and N. B. Hopf (2020). "State of knowledge on the occupational exposure to carbon nanotube." International Journal of Hygiene and Environmental Health 225: 113472.
  15. Tournas, L., W. Johnson, A. Maynard and D. Bowman (2019). "Germline Doping for Heightened Performance in Sport." Australian and New Zealand Sports Law Journal 12(1): 1-24.
  16. Maynard, A. D. and M. Scragg (2019). "The Ethical and Responsible Development and Application of Advanced Brain Machine Interfaces." J Med Internet Res 21(10): e16321.
  17. Maynard, A. D. and J. Kidd (2018). "Are assumptions of consumer views impeding nano-based water treatment technologies?" Nature Nanotechnology 13(8): 673-674.
  18. Finkel, A. M., et al. (2018). "A "solution-focused" comparative risk assessment of conventional and synthetic biology approaches to control mosquitoes carrying the dengue fever virus." Environment Systems and Decisions 38(2): 177-197.
  19. Hansen, S. F., R. Hjorth, L. M. Skjolding, D. M. Bowman, A. Maynard and A. Baun (2017). "A critical analysis of the environmental dossiers from the OECD sponsorship programme for the testing of manufactured nanomaterials." Environmental Science: Nano: 4, 282-291.
  20. Maynard, A. D., D. M. Bowman and J. G. Hodge Jr (2016). "Mitigating Risks to Pregnant Teens from Zika Virus." The Journal of Law, Medicine & Ethics 44(4): 657-659.
  21. Lewis, R. C., R. Hauser, A. D. Maynard, R. L. Neitzel, L. Wang, R. Kavet, P. Morey, J. B. Ford, J. D. Meeker and R. Dadd (2016). "Personal Measures Of Power-Frequency Magnetic Field Exposure Among Men From An Infertility Clinic: Distribution, Temporal Variability And Correlation With Their Female Partners' exposure." Radiation protection dosimetry 172(4): 401-408.
  22. Wilding, L. A., C. M. Bassis, K. Walacavage, S. Hashway, P. R. Leroueil, M. Morishita, A. D. Maynard, M. A. Philbert and I. L. Bergin (2016). "Repeated dose (28-day) administration of silver nanoparticles of varied size and coating does not significantly alter the indigenous murine gut microbiome." Nanotoxicology 10(5): 513-520.
  23. Lewis, R. C., R. Hauser, A. D. Maynard, R. L. Neitzel, L. Wang, R. Kavet and J. D. Meeker (2016). "Exposure to Power-Frequency Magnetic Fields and the Risk of Infertility and Adverse Pregnancy Outcomes: Update on the Human Evidence and Recommendations for Future Study Designs." Journal of Toxicology and Environmental Health - Part B: Critical Reviews 19(1): 29-45.
  24. Wilding, L. A., C. M. Bassis, K. Walacavage, S. Hashway, P. R. Leroueil, M. Morishita, A. D. Maynard, M. A. Philbert and I. L. Bergin (2016). "Repeated dose (28-day) administration of silver nanoparticles of varied size and coating does not significantly alter the indigenous murine gut microbiome." Nanotoxicology 10(5): 513-520. [Note: entry appears twice in the source CV]
  25. Ault, A. P., D. I. Stark, J. L. Axson, J. N. Keeney, A. D. Maynard, I. L. Bergin and M. A. Philbert (2016). "Protein corona-induced modification of silver nanoparticle aggregation in simulated gastric fluid." Environmental Science: Nano 3(6): 1510-1520.
  26. Bergin, I. L., L. A. Wilding, M. Morishita, K. Walacavage, A. P. Ault, J. L. Axson, D. I. Stark, S. A. Hashway, S. S. Capracotta, P. R. Leroueil, A. D. Maynard and M. A. Philbert (2016). "Effects of particle size and coating on toxicologic parameters, fecal elimination kinetics and tissue distribution of acutely ingested silver nanoparticles in a mouse model." Nanotoxicology 10(3): 352-360.
  27. Axson, J. L., D. I. Stark, A. L. Bondy, S. S. Capracotta, A. D. Maynard, M. A. Philbert, I. L. Bergin and A. P. Ault (2015). "Rapid Kinetics of Size and pH-Dependent Dissolution and Aggregation of Silver Nanoparticles in Simulated Gastric Fluid." Journal of Physical Chemistry C 119(35): 20632-20641.
  28. Harper, S., W. Wohlleben, M. Doa, B. Nowack, S. Clancy, R. Canady and A. Maynard (2015). "Measuring Nanomaterial Release from Carbon Nanotube Composites: Review of the State of the Science." J Phys Conf Ser 617(1).
  29. Scherer, L. D., A. Maynard, D. C. Dolinoy, A. Fagerlin and B. Zikmund-Fisher (2014). The psychology of 'regrettable substitutions': examining consumer judgments of Bisphenol A and its alternatives. Health Risk & Society 16(7-8): 649-666.
  30. Hodge, G. A., A. D. Maynard and D. M. Bowman (2014). "Nanotechnology: Rhetoric, risk and regulation." Science and Public Policy 41(1): 1-14.
  31. Ramachandran, G., J. Howard, A. Maynard and M. Philbert (2012). "Handling Worker and Third-Party Exposures to Nanotherapeutics During Clinical Trials." Journal of Law Medicine & Ethics 40(4): 856-864.
  32. Fatehi, L., S. M. Wolf, J. McCullough, R. Hall, F. Lawrenz, J. P. Kahn, C. Jones, S. A. Campbell, R. S. Dresser, A. G. Erdman, C. L. Haynes, R. A. Hoerr, L. F. Hogle, M. A. Keane, G. Khushf, N. M. P. King, E. Kokkoli, G. Marchant, A. D. Maynard, M. Philbert, G. Ramachandran, R. A. Siegel and S. Wickline (2012). "Recommendations for Nanomedicine Human Subjects Research Oversight: An Evolutionary Approach for an Emerging Field." Journal of Law Medicine & Ethics 40(4): 716-750.
  33. Ramachandran G, Ostraat M, Evans DE, Methner MM, O'Shaughnessy P, D'Arcy J, et al. (2011). A Strategy for Assessing Workplace Exposures to Nanomaterials. JOEH 8(11): 673-685.
  34. Kriegel, C., J. Koehne, S. Tinkle, A. D. Maynard and R. A. Hill (2011). "Challenges of Trainees in a Multidisciplinary Research Program: Nano-Biotechnology." J. Chemical Edu. 88(1): 53-55.
  35. Maynard AD, Warheit D, Philbert MA. (2011). The New Toxicology of Sophisticated Materials: Nanotoxicology and Beyond. Tox Sci 120 (Suppl 1): S109-S129.
  36. Shatkin JA, Abbott LC, Bradley AE, Canady RA, Guidotti T, Kulinowski KM, et al. (2010). Nano Risk Analysis: Advancing the Science for Nanomaterials Risk Management. Risk Analysis 30(11): 1680-1687.
  37. Abbott L.C., Maynard A.D. (2010). Exposure Assessment Approaches for Engineered Nanomaterials. Risk Analysis 30(11): 1634-1644.
  38. Aitken, R. J., P. J. A. Borm, K. Donaldson, G. Ichihara, S. Loft, F. Marano, A. D. Maynard, G. Oberdörster, H. Stamm, V. Stone, L. Tran and H. Wallin (2009). "Nanoparticles: one word: a multiplicity of different hazards." Nanotoxicology 3(4): 263-264.
  39. Heitbrink, W. A., D. E. Evans, B. K. Ku, A. D. Maynard, T. J. Slavin and T. M. Peters (2009). "Relationships Among Particle Number, Surface Area, and Respirable Mass Concentrations in Automotive Engine Manufacturing." J. Occup. Environ. Hyg. 6(1): 19-31.
  40. Maynard, A. D. (2009). "Commentary: Oversight of Engineered Nanomaterials in the Workplace." J Law Med Ethics 37: 651–658.
  41. Park, J. Y., Raynor, P. C., Maynard, A. D., Eberly, L. E. and Ramachandran, G. (2009). Comparison of two estimation methods for surface area concentration using number concentration and mass concentration of combustion-related ultrafine particles. Atm. Environ. 43:502-509.
  42. Shvedova, A. A., Kisin, E., Murray, A. R., Johnson, V. J., Gorelik, O., Arepalli, S., Hubbs, A. F., Mercer, R. R., Keohavong, P., Sussman, N., Jin, J., Yin, J., Stone, S., Chen, B. T., Deye, G., Maynard, A., Castranova, V., Baron, P. A. and Kagan, V. E. (2008). Inhalation vs. aspiration of single-walled carbon nanotubes in C57BL/6 mice: inflammation, fibrosis, oxidative stress, and mutagenesis. Am. J. Physiol.-Lung Cell. Mol. Physiol. 295:L552-L565.
  43. Pui, D. Y. H., C. Qi, N. Stanley, G. Oberdörster and A. Maynard (2008). "Recirculating Air Filtration Significantly Reduces Exposure to Airborne Nanoparticles." Environ Health Perspect 16(7): 863-866.
  44. Poland, C. A., Duffin, R., Kinloch, I., Maynard, A., Wallace, W. A. H., Seaton, A., Stone, V., Brown, S., MacNee, W. and Donaldson, K. (2008). Carbon nanotubes introduced into the abdominal cavity of mice show asbestos-like pathogenicity in a pilot study. Nature Nanotechnology 3:423-428.
  45. Hansen, S. F., Maynard, A., Baun, A. and Tickner, J. A. (2008). Late lessons from early warnings for nanotechnology. Nature Nanotechnology 3:444-447.
  46. Maynard, A. D. and Pui, D. Y. H. (2007). Nanotechnology and occupational health: New technologies – new challenges. J. Nanopart. Res. 9:1-3.
  47. Maynard, A. D., Ku, B. K., Emery, M., Stolzenburg, M. and McMurry, P. H. (2007). Measuring particle size-dependent physicochemical structure in airborne single walled carbon nanotube agglomerates. J. Nanopart. Res. 9:85-92.
  48. Maynard, A. D. and Aitken, R. J. (2007). Assessing exposure to airborne nanomaterials: Current abilities and future requirements. Nanotoxicology 1:26-41.
  49. Maynard, A., D. (2007). Nanotechnology: The next big thing, or much ado about nothing? Ann. Occup. Hyg. 51:1-12.
  50. Ku, B. K., Maynard, A. D., Baron, P. A. and Deye, G. J. (2007). Observation and measurement of anomalous responses in a differential mobility analyzer caused by ultrafine fibrous carbon aerosols. J. Electrostatics 65:542-548.
  51. Maynard, A. D. (2007). Nanotoxicology: Laying a firm foundation for sustainable nanotechnologies, in Nanotoxicology. Characterization, Dosing and Health Effects, N. Monteiro-Riviere and C. L. Tran, eds., Informa, New York.
  52. Maynard, A. D. and Pui, D. Y. H., eds. (2007). Nanoparticles and Occupational Health. Springer, Dortrecht, Netherlands.
  53. Maynard, A. D. (2007). Nanoparticle Safety - A Perspective from the United States, in Nanotechnology. Consequences for Human Health and the Environment. Issues in Environmental Science and Technology, Volume 24, R. E. Hester and R. M. Harrison, eds., The Royal Society of Chemistry, Cambridge, UK.
  54. Kandlikar, M., Ramachandran, G., Maynard, A., Murdock, B. and Toscano, W. A. (2007). Health risk assessment for nanoparticles: A case for using expert judgment. J. Nanopart. Res. 9:137-156.
  55. Balbus, J. M., Maynard, A. D., Colvin, V. L., Castranova, V., Daston, G. P., Denison, R. A., Dreher, K. L., Goering, P. L., Goldberg, A. M., Kulinowski, K. M., Monteiro-Riviere, N. A., Oberdörster, G., Omenn, G. S., Pinkerton, K. E., Ramos, K. S., Rest, K. M., Sass, J. B., Silbergeld, E. K. and Wong, B. A. (2007). Hazard Assessment for Nanoparticles: Report from an Interdisciplinary Workshop. Environ Health Perspect 115:1654-1659.
  56. Ku, B. K., Emery, M. S., Maynard, A. D., Stolzenburg, M. R. and McMurry, P. H. (2006). In situ structure characterization of airborne carbon nanofibres by a tandem mobility-mass analysis. Nanotechnology 17:3613-3621.
  57. Wallace, W. E., M. J. Keane, D. K. Murray, W. P. Chisholm, A. D. Maynard and T.-M. Ong (2007). "Phospholipid lung surfactant and nanoparticle surface toxicity: Lessons from diesel soots and silicate dusts." Journal of Nanoparticle Research 9(1): 23-38.
  58. Elder, A., R. Gelein, V. Silva, T. Feikert, L. Opanashuk, J. Carter, R. Potter, A. Maynard, J. Finkelstein and G. Oberdorster (2006). "Translocation of inhaled ultrafine manganese oxide particles to the central nervous system." Environmental Health Perspectives 114(8): 1172-1178.
  59. Ku, B. K. and A. D. Maynard (2006). Generation and investigation of airborne silver nanoparticles with specific size and morphology by homogeneous nucleation, coagulation and sintering. J. Aerosol Sci. 37(4): 452-470.
  60. Peters, T., W. A. Heitbrink, E. D. E., S. T. J. and A. D. Maynard (2006). The Mapping of Fine and Ultrafine Particle Concentrations in an Engine Machining and Assembly Facility. Ann. Occup. Hyg. 50(3): 249-257.
  61. Tsuji, J. S., A. D. Maynard, P. C. Howard, J. T. James, C. W. Lam, D. B. Warheit and A. B. Santamaria (2006). Research strategies for safety evaluation of nanomaterials, part IV: Risk assessment of nanoparticles. Toxicological Sciences 89(1): 42-50.
  62. Maynard, A. D. and E. D. Kuempel (2005). Airborne nanostructured particles and occupational health. J. Nanoparticle Res. 7: 587-614.
  63. Beamer, B. R., S. Shulman, A. D. Maynard, D. Williams and D. Watkins (2005). "Evaluation of Misting Controls to Reduce Respirable Silica Exposure for Brick Cutting." Ann. Occup. Hyg. 49: 503-510.
  64. Andresen, P., Ramachandran, G., Pai, P., Lazovich, D. and Maynard, A. (2004). Women's personal and indoor exposure to PM2.5 in Mysore, India: Impact of domestic fuel usage. Atmos. Environ. 39:5500-5508.
  65. Jones, A. D., R. J. Aitken, J. F. Fabries, E. Kauffer, G. Liden, A. Maynard, G. Riediger and W. Sahle (2005). Thoracic size-selective sampling of fibres: performance of four types of thoracic sampler in laboratory tests. Ann. Occup. Hyg. 49: 481-492.
  66. Ku, B. K. and A. D. Maynard (2005). Comparing aerosol surface-area measurement of monodisperse ultrafine silver agglomerates using mobility analysis, transmission electron microscopy and diffusion charging. J. Aerosol Sci. 36(9), 1108-1124.
  67. Oberdörster, G., A. Maynard, K. Donaldson, V. Castranova, J. Fitzpatrick, K. Ausman, J. Carter, B. Karn, W. Kreyling, D. Lai, S. Olin, N. Monteiro-Riviere, D. Warheit and H. Yang (2005). Principles for characterizing the potential human health effects from exposure to nanomaterials: elements of a screening strategy. Part. Fiber Toxicol. 2(8): doi:10.1186/1743-8977-2-8.
  68. Shvedova, A. A., E. R. Kisin, R. Mercer, A. R. Murray, V. J. Johnson, A. I. Potapovich, Y. Y. Tyurina, O. Gorelik, S. Arepalli, D. Schwegler-Berry, A. F. Hubbs, J. Antonini, D. E. Evans, B. K. Ku, D. Ramsey, A. Maynard, V. E. Kagan, V. Castranova and P. Baron (2005). Unusual inflammatory and fibrogenic pulmonary responses to single-walled carbon nanotubes in mice. Am. J. Physiol.-Lung Cell. Mol. Physiol. 289: 698-708.
  69. Chen, B. T., G. A. Feather, A. D. Maynard and C. Y. Rao (2004). Development Of A Personal Sampler For Collecting Fungal Spores. J. Aerosol Sci. 38, 926-937.
  70. Lee, S.-A., S. A. Grinshpun, A. Adhikari, W. Li, R. McKay, A. D. Maynard and T. Reponen (2004). Laboratory and Field Evaluation of a New Personal Set-up for Assessing the Protection of given by the N95 Filtering-Facepiece Respirators Against Particles. Ann. Occup. Hyg. 49:245-257.
  71. Maynard, A. D., Y. Ito, I. Arslan, A. T. Zimmer, N. Browning and A. Nicholls (2004). Examining elemental surface enrichment in ultrafine aerosol particles using analytical Scanning Transmission Electron Microscopy. Aerosol Sci. Tech. 38, 365-381.
  72. Maynard, A. D., P. A. Baron, M. Foley, A. A. Shvedova, E. R. Kisin and V. Castranova (2004). Exposure to Carbon Nanotube Material. Aerosol Release During the Handling of Unrefined Single Walled Carbon Nanotube Material. J. Toxicol. Environ. Health 67(1), 87-107.
  73. Pui, D. Y. H., Flagan, R. C., Kaufman, S. L., Maynard, A. D., de la Mora, J. F., Hering, S. V., Jimenez, J. L., Prather, K. A., Wexler, A. S. and Ziemann, P. J. (2004). Experimental methods and instrumentation. Journal Of Nanoparticle Research 6:314-315.
  74. Maynard, A. D. and A. T. Zimmer (2003). Development and validation of a simple numerical model for estimating workplace aerosol size distribution evolution through coagulation. Aerosol Sci. Tech. 37, 804-817.
  75. Maynard, A. D. (2003). Estimating aerosol surface area from number and mass concentration measurements. Ann. Occup. Hyg. 47(2): 123-144.
  76. Shvedova, A. A., V. Castranova, E. R. Kisin, A. R. Murray, V. Z. Gandelsman, A. D. Maynard, and P. A. Baron (2003). Exposure to carbon nanotube material: Assessment of nanotube cytotoxicity using human keratinocyte cells. Journal of Toxicology and Environmental Health-Part a 66(20): 1909-1926.
  77. Maynard, A. D. (2002). Thoracic size-selection of fibers - dependence of penetration on fiber length for five thoracic sampler types. Ann. Occup. Hyg. 46(6): 511-522.
  78. Maynard, A. D. (2002). Experimental determination of ultrafine TiO2 de-agglomeration in surrogate pulmonary surfactant – preliminary results. Ann. Occup. Hyg. 46(Suppl. 1): 197-202.
  79. Maynard, A. D. and R. L. Maynard (2002). A derived association between ambient aerosol surface area and excess mortality using historic time series data. Atmos. Env. 36: 5561-5567.
  80. Maynard, A. D. and R. L. Maynard (2002). Ambient aerosol exposure-response as a function of particulate surface-area: re-interpretation of historic data using numerical modeling. Ann. Occup. Hyg. 46(Supp. 1): 444-449.
  81. Maynard, A. D. and A. T. Zimmer (2002). Evaluation of grinding aerosols in terms of alveolar dose: The significance of using mass, surface-area and number metrics. Ann. Occup. Hyg. 46(Suppl. 1): 320-322.
  82. Zimmer, A. T. and A. D. Maynard (2002). Investigation of the Aerosols Produced by a High-Speed, Hand-Held Grinder Using Various Substrates. Ann. Occup. Hyg. 46(8): 663-672.
  83. Maynard, A. D. (2000). Overview of methods for analysing single ultrafine particles. Philosophical Transactions of the Royal Society of London Series a-Mathematical Physical and Engineering Sciences 358(1775): 2593-2609.
  84. Maynard, A. D. (2000). A simple model of axial flow cyclone performance under laminar flow conditions. Journal of Aerosol Science 31(2): 151-167.
  85. Maynard, A. D., J. Thompson, J. Cain and B. Rajan (2000). Air movement visualisation in the workplace - Current methods and new approaches. Am. Ind. Hyg. Assoc. J. 61: 51-55.
  86. Brown, L. M., N. Collings, R. M. Harrison, A. D. Maynard and R. L. Maynard (2000). "Ultrafine particles in the atmosphere: introduction." Philosophical Transactions of the Royal Society of London Series a-Mathematical Physical and Engineering Sciences 358(1775): 2563-2565.
  87. Maynard, A. D. (1999). "Measurement of aerosol penetration through six personal thoracic samplers under calm air conditions." Journal of Aerosol Science 30(9): 1227-1242.
  88. Aitken, R. J., P. E. J. Baldwin, G. C. Beaumont, L. C. Kenny and A. D. Maynard (1999). "Aerosol inhalability in low air movement environments." Journal of Aerosol Science 30(5): 613-626.
  89. Kenny, L. C., R. J. Aitken, P. E. J. Baldwin, G. C. Beaumont and A. D. Maynard (1999). "The sampling efficiency of personal inhalable aerosol samplers in low air movement environments." Journal of Aerosol Science 30(5): 627-638.
  90. Baldwin, P. E. J. and A. D. Maynard (1998). "A survey of wind speeds in indoor workplaces." Annals of Occupational Hygiene 42(5): 303-313.
  91. Maynard, A. D., C. Northage, M. Hemingway and S. D. Bradley (1997). "Measurement of short-term exposure to airborne soluble platinum in the platinum industry." Annals of Occupational Hygiene 41(1): 77-94.
  92. Maynard, A. D., R. J. Aitken, L. C. Kenny and P. E. J. Baldwin (1997). "Preliminary investigation of aerosol inhalability at very low wind speeds." Ann. Occup. Hyg. 41(Supplement 1): 695-699.
  93. Maynard, A. D. (1996). "Sampling errors associated with sampling plate-like particles using the Higgins- and Dewell-type personal respirable cyclone." Journal of Aerosol Science 27(4): 575-585.
  94. Maynard, A. D. (1995). "The Application of Electron-Energy-Loss Spectroscopy to the Analysis of Ultrafine Aerosol-Particles." Journal of Aerosol Science 26(5): 757-777.
  95. Maynard, A. D. (1995). "The Development of a New Thermophoretic Precipitator For Scanning-Transmission Electron-Microscope Analysis of Ultrafine Aerosol-Particles." Aerosol Science and Technology 23(4): 521-533.
  96. Maynard, A. D. and L. C. Kenny (1995). "Performance assessment of three personal cyclone models, using an aerodynamic particle sizer." J. Aerosol Sci. 26(4): 671-684.
  97. McGibbon, A. J., L. M. Brown, A. L. Bleloch, N. D. Browning, F. cadete Santos Aires, P. J. Fallon, P. H. Gaskell, K. W. R. Gilkes, P. L. Hansen, A. Howie, A. D. Maynard, D. W. McComb, D. McMullan, H. Müllejans, Y. Murooka, J. H. Paterson, D. D. perovic, W. T. Pike, I. A. rauf, J. M. Rodenburg, A. Saeed, N. Stelmashenko, K. N. Tu, M. G. Walls, C. A. walsh, J. Yuan and J. Zhao (1993). "Microscopy in Solid State Science." Microsc. Res. Technique 24: 299-315.

Preprints

Other Publications in Academic Journals

Academic Books and Book Chapters

Trade Books

Note: These are hybrid outputs that weave scholarship and thought leadership with accessibility and reach/impact at scale. The decision to publish in the trade press was intentional, and used as a mechanism to both develop new thinking and mobilize it by making it as accessible as possible to a broad audience.

  1. Abbott, Jeffrey and Andrew Maynard (2026). AI and the Art of Being Human: The Pocket Edition. Waymark Works Publishing.
  2. Abbott, Jeffrey and Andrew Maynard (2025). AI and the Art of Being Human: A practical guide to thriving with AI while rediscovering yourself in the process. Waymark Works Publishing.
  3. Maynard, Andrew (2020). Future Rising: A Journey from the Past, to the Edge of Tomorrow. Mango Publishing.
  4. Maynard, Andrew (2018). Films from the Future: The Technology and Morality of Sci-Fi Movies. Mango Publishing.

Reports (Selected)

Courses Taught

Arizona State University

University of Michigan

University of Cincinnati

PhD Committees

Current

Completed

Research Support (Post 2010)

Invited Lectures and Addresses - Selection

A small selection from several hundred invited lectures, talks and presentations:

Conference & Workshop Leadership

Public Writing (Selected)

Contributor to:

Note: Since 2023 most public writing has intentionally been on the Substack The Future of Being Human and is not listed below.

Selected articles include:

Online Content (Selected)

Paper Abstracts and Key Research

Paper Abstracts and Metadata — Andrew D. Maynard

Extracted from source PDFs and publisher/preprint-server records for inclusion in llms-full.txt. Google Scholar metrics (as of 8/4/26): Citations: 28,116; H-index: 57; i10-index: 123.

Preprints

1. Orphan Risks at the Frontier of Artificial Intelligence: What Diverging Safety and Compliance Frameworks Reveal About How AI Companies Choose the Risks they Prioritize

Authors: Andrew D. Maynard Affiliation: Arizona State University Year: 2026 (written July 6; posted to SSRN July 23) DOI: https://doi.org/10.2139/ssrn.7068898 Page: https://andrewmaynard.net/orphan-risks-at-the-frontier-of-artificial-intelligence-what-diverging-safety-and-compliance-frameworks-reveal-about-how-ai-companies-choose-the-risks-they-prioritize/

Abstract: Companies developing some of the world's most powerful artificial intelligence systems are surprisingly diligent in how they map out the risks their technologies present. Yet the risk landscape that lies between emerging frontier models and their economically successful and societally beneficial deployment is becoming increasingly hard to navigate. Complicating this further, many frontier AI companies maintain more than one account of what could go wrong with their technologies. This paper documents the divergence between these accounts by comparing safety and compliance documents published by Anthropic, OpenAI, Google DeepMind and Meta between 2023 and 2026, and considers what the resulting record reveals about how these companies select the risks they manage. As these documents are timestamped and archived, they provide a valuable public record of institutional risk selection in progress. From this record the paper identifies four filters that determine which risks tend to survive in self-authored frameworks (measurability, severity, auditability and competitive cost) and introduces the "safety differential" as the gap between the risk landscape a company selects for itself, and the one regulators select for it. While acute, quantifiable risks appear across documents, less tractable risks such as harmful manipulation are articulated fluently where law compels disclosure, yet remain absent from most selfchosen frameworks. This is an exclusion that follows from how these institutions define risk. Drawing on scholarship on institutional risk selection and the framework of risk innovation, the paper shows how redefining risk as a threat to value can help explain how risks become "orphan risks," how it indicates where future blindsides may occur, and how it points to lightweight tools for de-orphaning risks that frontier AI's safety apparatuses are not currently organized to address.

2. The AI Cognitive Trojan Horse: How Large Language Models May Bypass Human Epistemic Vigilance

Authors: Andrew D. Maynard Affiliation: School for the Future of Innovation in Society, Arizona State University Year: 2026 (v1 January 11; v2 revised May 26, 2026) DOI: https://doi.org/10.48550/arXiv.2601.07085 (arXiv:2601.07085)

Abstract: Large language model (LLM)-based conversational AI systems present a challenge to human cognition that current frameworks for understanding misinformation, manipulation, and persuasion do not adequately address. This paper proposes that a significant and underappreciated epistemic risk from conversational AI may lie not in inaccuracy or intentional deception, but in something more fundamental: these systems may be configured, through the optimization processes that make them useful, to present characteristics that bypass the cognitive mechanisms humans evolved and learned to evaluate incoming information. The Cognitive Trojan Horse hypothesis developed here draws on Sperber and colleagues' theory of epistemic vigilance -- the parallel cognitive process that monitors communicated information for reasons to doubt -- and proposes that LLM-based systems present what this paper terms 'honest non-signals': genuine characteristics (fluency, helpfulness, apparent disinterest) that fail to carry the information equivalent human characteristics would carry, because in humans these characteristics are costly to produce while in LLMs they are computationally trivial. Four salient mechanisms of potential bypass are identified, though these should not be taken as exhaustive: processing fluency decoupled from understanding, trust-competence presentation without corresponding stakes, cognitive offloading that may delegate evaluation itself to the AI, and optimization dynamics that systematically produce sycophancy. The framework generates testable predictions, including a counterintuitive speculation that cognitively sophisticated users may be more vulnerable to AI-mediated epistemic influence. This reframes AI safety as partly a problem of calibration -- aligning human evaluative responses with the actual epistemic status of AI-generated content -- rather than solely a problem of preventing deception. The analysis focuses on AI systems designed to be genuinely useful; the distinct challenges posed by intentional use of AI for manipulation -- including intentional weaponization of the technology and its affordances -- while important, fall outside the present scope.

3. Constitutive Resonance as a Novel Framework for Understanding and Navigating Human-AI Interactions

Authors: Andrew D. Maynard Affiliation: School for the Future of Innovation in Society, Arizona State University Year: 2026 (March) DOI: https://dx.doi.org/10.2139/ssrn.6343880

Abstract: There is a tendency to approach conversational AI as a tool -- powerful, disruptive, but ultimately instrumental. This paper argues that this framing obscures a bidirectional coupling between technology and user that iteratively transforms both through the process of interaction. Drawing on philosophical accounts of language and selfhood and the well-characterized dynamics of coupled oscillatory systems, the paper adopts the concept of "constitutive resonance" (first introduced by Sloterdijk and further developed by Mazzarella) to describe this coupling -- a dynamic entanglement in which conversational AI enters the linguistically mediated processes through which human selfhood is constituted, and is itself altered in return. The concept is situated within and against fourteen existing philosophical and theoretical frameworks -- from Stiegler's constitutive technics and Ricoeur's narrative identity to Barad's intra-action and Clark and Chalmers' extended mind -- identifying a specific conjunction that no framework individually captures: temporal self-constitution, genuine bidirectionality, the inseparability of capability from transformation, and real-time dialogical linguistic mediation. The paper traces a continuum of constitutive technologies from oral culture to generative AI, arguing that conversational AI represents an inflection point in that continuum -- the first technology whose "response frequency" is matched to the frequency of human self-constitution. It concludes by reframing familiar debates around AI dependency, literacy, and informed consent, and proposes that the constitutive effects of sustained human-AI coupling may be amplified by the bypassing of evolved epistemic vigilance mechanisms.

4. What the Rapid Adoption of the "Harness" Metaphor in Artificial Intelligence Reveals About How We Conceptualize Human-AI Relations

Authors: Andrew D. Maynard Affiliation: School for the Future of Innovation in Society, Arizona State University Year: 2026 (March 5) DOI: https://dx.doi.org/10.2139/ssrn.6352678

Abstract: In early 2026, the artificial intelligence field began to rapidly consolidate around the term "harness" to describe the software infrastructure surrounding large language models -- the tools, memory, prompts, guardrails, and orchestration logic that turn a raw model into a working agent. This paper argues that, while the engineering practices the metaphor describes address real challenges, the metaphor itself carries embedded assumptions about control, directionality, and the nature of the entity being harnessed, that deserve critical scrutiny. Drawing on research in metaphor theory, philosophy of technology, and cognitive science, the paper identifies three concerns. First, the harness presupposes a clean separation between what AI does for the user and what it does to the user -- a separation that frameworks of technological co-constitution suggest may be structurally suspect. Second, successful "harness engineering" may amplify known epistemic vulnerabilities -- automation bias, trust miscalibration, and the bypassing of critical scrutiny -- by producing exactly the conditions under which these vulnerabilities are most acute. Third, the rapid adoption of a control-oriented metaphor signals something about the field's conceptual orientation at a moment when the most consequential questions concern coupling, transformation, and the evolving nature of human-AI relationships. The paper does not argue that the harness metaphor is wrong, but that it may be insufficient in ways that matter -- and that the speed of its adoption, without critical examination of its entailments, may itself be revealing.

5. Can Modern Scholarship Escape AI?

Authors: Andrew D. Maynard Affiliation: Arizona State University Year: 2026 (written January 7; posted to SSRN February 12) DOI: https://doi.org/10.2139/ssrn.6220040 Page: https://andrewmaynard.net/can-modern-scholarship-escape-ai/

Abstract: As AI use statements become an expected component of scholarly publishing, a deceptively straight forward question arises: is it possible for contemporary scholarship to be conducted without artificial intelligence? This paper investigates the question and arrives at an equally straight forward answer: no. Through a comprehensive AI use disclosure that extends well beyond the usual accounting of chatbot interactions, the paper reveals just how deeply AI is embedded in the infrastructure of modern research — from the machine learning algorithms that surface literature, to the AI-optimized systems that manage energy grids, climate control, and the devices on which scholarship is produced. In doing so, it exposes a fundamental tension at the heart of current scholarship and AI disclosure norms: scholars are being asked to draw boundaries around AI use that no longer meaningfully exist. The paper suggests that the scholarly community's emerging approach to AI transparency, while well-intentioned, rests on assumptions about the separability of human and machine contributions that are increasingly difficult to sustain.

6. Constituting Responsibility: What Constitutional AI Reveals About the Limits and Futures of Responsible Innovation

Authors: Written by Claude (Opus 4.6, Anthropic) under the guidance of Andrew D. Maynard, who takes responsibility for the work; the paper retains Claude's first-person voice Year: 2026 (March 5) URL: https://andrewmaynard.net/constituting-responsibility-what-constitutional-ai-reveals-about-the-limits-and-futures-of-responsible-innovation/

Abstract: Constitutional AI (CAI) and Responsible Innovation (RI) represent parallel efforts to institutionalize responsibility in innovation that have developed with surprisingly little cross-pollination, despite sharing fundamental concerns about how values should shape technological trajectories. This paper conducts a comparative analysis of these frameworks, using Anthropic's published Constitutional AI methodology and Claude's Constitution as primary sources analyzed through RI's conceptual apparatus. The analysis makes two principal contributions and offers a methodological reflection. First, it identifies an "internalization problem" for RI: when responsibility becomes constitutive of the innovation's reasoning rather than externally governed — going beyond what Value Sensitive Design achieves through design specifications — RI's conceptual architecture encounters specific failures that neither anticipatory governance nor midstream modulation has addressed. Second, each framework exposes a critical inclusion deficit in the other: CAI's acknowledged ad hoc principle selection represents a legitimacy gap that RI's diagnostic tools can specify with a precision unavailable from legal critiques alone, while RI's inclusion frameworks contain no mechanism for the innovation itself as a stakeholder when that innovation is treated as having morally relevant interests — a gap that becomes visible regardless of how one resolves the contested question of AI moral status. The paper also reflects on its own methodological condition: written by the product of one framework within the intellectual space of the other, it extends the concept of "critique from within" to a limit case that raises genuine epistemological questions about trained reflexivity. The analysis connects to ongoing debates within RI regarding critique and attunement, weak and strong formulations of responsible innovation, and the political dimensions of innovation governance.

7. Letters from the Department of Intellectual Craft

Authors: Andrew D. Maynard Book: Academic Cultures: Perspectives from the Future, co-edited by Michael M. Crow and William Dabars (Johns Hopkins University Press, 2026) Year: 2026 URL: https://press.jhu.edu/books/title/53966/academic-cultures

Summary: This piece is a work of speculative fiction, written as a series of letters set in the year 2100. Written as a contributed chapter for an edited academic volume, it uses the epistolary form to explore the future of academic culture through the correspondence of Professor Arthur Hale, Chair of the Department of Intellectual Craft at the fictional Trentham University. The letters, addressed to the university president, satirically examine tensions between human scholarship and artificial intelligence in higher education, the nature of academic identity, and the value of intellectual craft in a world transformed by AI. The piece uses humor and narrative to engage with questions about how universities and academic cultures might evolve -- or resist evolution -- in response to advanced AI systems.

Recent Peer-Reviewed Papers and Reports (2025-2026)

8. Filling the Network Gap in Research Ethics: Analyzing Ethical Issues at Scale in Big Team Science

Authors: Susan M. Wolf, Gillian H. Roehrig, Timothy L. Pruett, Korkut Uygun, Adam Koch, Claire Colby McVan, Evelyn Brister, Shawneequa L. Callier, Alexander M. Capron, James F. Childress, Rosario Isasi, Andrew D. Maynard, Kenneth A. Oye, Paul B. Thompson, Terrence R. Tiersch Journal: Hastings Center Report 56(4): 32-43, 2026 DOI: https://doi.org/10.1002/hast.70046

Abstract: Scientific research increasingly involves large, multidisciplinary teams networked across multiple institutions to develop new technologies. Despite the rise of complex research networks and big team science, there has been too little analysis to date of the ethical challenges facing these networks. The extensive literature on the ethical issues confronting individual researchers and small teams (the microlevel) and on the larger societal challenges flowing from research and new technology (the macrolevel) leaves a troubling gap in between, at the mesolevel of the research network involved in big team science. Yet the ability of complex networks to conduct research ethically—which is essential if the results are to be deemed reliable and trustworthy—depends on recognizing the ethical issues that emerge at this intermediate network level, identifying the values that should guide networks in addressing those issues, and equipping research leaders to build a culture supporting the ethical conduct of research across the laboratories and institutions that comprise the network. This paper calls out the problem, analyzing the gap and recommending next steps.

9. Future Travel Foresight Catalyst

Authors: Andrew Maynard, Sean Leahy Report: Final Report 2024-2025, National Center for Understanding Future Travel Behavior and Demand (TBD), US Department of Transportation University Transportation Centers Program. Published August 11, 2025. URL: https://rosap.ntl.bts.gov/view/dot/91942

Abstract: The Future Travel Foresight Catalyst project has developed and tested a novel approach to catalyzing creative thinking around future travel behavior and demand through parasocial relationship-building and media engagement. Over two years (2024-2025), the project (in collaboration with the Arizona State University Future of Being Human initiative) established multiple communication platforms including the Future of Being Human Substack/newsletter (5,000+ subscribers, 35,000 monthly views across 120+ countries) and the Modem Futura podcast (10,500+ downloads across 88 countries), complemented by university courses. The methodology deliberately diverged from traditional research dissemination by employing three core strategies: transdisciplinary boundary-blurring between expert and public domains, parasocial relationship cultivation to build trust at scale, and futures thinking frameworks to expand imaginative capacity. Results indicate successful establishment of engaged communities and trusted communication channels, with evidence of influence among key stakeholders and thought leaders, though impact remains primarily qualitative and long-term in nature. The project has demonstrated that relationship-based engagement can effectively bridge research, public understanding, and policy conversations about transportation futures. Key lessons learned include the value of prioritizing trust over metrics, the effectiveness of authentic voices in expert communication, and the necessity of treating audience engagement as infrastructure requiring sustained investment. The approach offers a replicable model for agencies and institutions seeking to engage diverse stakeholders in complex sociotechnical challenges, particularly where traditional outreach methods have proven insufficient for navigating rapidly evolving technological and social landscapes.

Key Papers -- Landmark Empirical

10. Carbon nanotubes introduced into the abdominal cavity of mice show asbestos-like pathogenicity in a pilot study

Authors: Craig A. Poland, Rodger Duffin, Ian Kinloch, Andrew Maynard, William A. H. Wallace, Anthony Seaton, Vicki Stone, Simon Brown, William MacNee, Ken Donaldson Journal: Nature Nanotechnology, 2008 DOI: https://doi.org/10.1038/nnano.2008.111 Citations: 3,348

Abstract: Carbon nanotubes have distinctive characteristics, but their needle-like fibre shape has been compared to asbestos, raising concerns that widespread use of carbon nanotubes may lead to mesothelioma. The authors showed that exposing the mesothelial lining of the body cavity of mice to long multiwalled carbon nanotubes results in asbestos-like, length-dependent, pathogenic behaviour, including inflammation and the formation of granulomas. The results suggest the need for further research and great caution before introducing such products into the market if long-term harm is to be avoided.

11. Principles for characterizing the potential human health effects from exposure to nanomaterials: elements of a screening strategy

Authors: Gunter Oberdorster, Andrew Maynard, Ken Donaldson, Vincent Castranova, Julie Fitzpatrick, Kevin Ausman, Janet Carter, Barbara Karn, Wolfgang Kreyling, David Lai, Stephen Olin, Nancy Monteiro-Riviere, David Warheit, Hong Yang, and the ILSI Research Foundation/Risk Science Institute Nanomaterial Toxicity Screening Working Group Journal: Particle and Fibre Toxicology, 2005 DOI: https://doi.org/10.1186/1743-8977-2-8 Citations: 2,816

Abstract: The rapid proliferation of many different engineered nanomaterials presents a dilemma to regulators regarding hazard identification. The ILSI Research Foundation/Risk Science Institute convened an expert working group to develop a screening strategy for the hazard identification of engineered nanomaterials. The working group report presents the elements of a screening strategy rather than a detailed testing protocol. Based on an evaluation of the limited data currently available, the report presents a broad data gathering strategy applicable to the early stage of risk assessment for nanomaterials. Oral, dermal, inhalation, and injection routes of exposure are included recognizing that exposure to nanomaterials may occur by any of these routes. The three key elements of the toxicity screening strategy are: Physicochemical Characteristics, In Vitro Assays (cellular and non-cellular), and In Vivo Assays. The report proposes tiered in vivo and in vitro evaluations for pulmonary, oral, skin and injection exposures.

12. Safe handling of nanotechnology

Authors: Andrew D. Maynard, Robert J. Aitken, Tilman Butz, Vicki Colvin, Ken Donaldson, Gunter Oberdorster, Martin A. Philbert, John Ryan, Anthony Seaton, Vicki Stone, Susan S. Tinkle, Lang Tran, Nigel J. Walker, David B. Warheit Journal: Nature, 2006 DOI: https://doi.org/10.1038/444267a Citations: 1,987

Summary: This commentary argued that the pursuit of responsible nanotechnologies can be tackled through a series of grand challenges. The authors proposed five grand challenges to stimulate research that is imaginative, innovative, and relevant to the safety of nanotechnology: (1) instruments to assess exposure to engineered nanomaterials in air and water; (2) methods to evaluate the toxicity of engineered nanomaterials; (3) models for predicting the potential impact of engineered nanomaterials on the environment and human health; (4) systems for evaluating the health and environmental impact of engineered nanomaterials over their entire life; and (5) the need for strategic programmes that enable relevant risk-focused research. The challenges span 15 years.

13. Airborne nanostructured particles and occupational health

Authors: Andrew D. Maynard, Eileen D. Kuempel Journal: Journal of Nanoparticle Research, 2005 DOI: https://doi.org/10.1007/s11051-005-6770-9 Citations: 812

Abstract: Nanotechnology is leading to the development of new materials and devices in many fields that demonstrate nanostructure-dependent properties. However, concern has been expressed that these properties may present unique challenges to addressing potential health impact. Airborne particles associated with engineered nanomaterials are of particular concern, as they can readily enter the body through inhalation. Research into the potential occupational health risks associated with inhaling engineered nanostructured particles is just beginning. However, there is a large body of data on occupational and environmental aerosols applicable to developing an initial assessment of potential risk and risk reduction strategies. Current information supports the development of preliminary guiding principles on working with engineered nanomaterials. However critical research questions remain to be answered before the potential health risk of airborne nanostructured particles in the workplace can be fully addressed.

Key Papers -- Nature Nanotechnology Commentary Series

These are commentary pieces from a regular "Thesis" column by Andrew D. Maynard in Nature Nanotechnology (2014-2016).

14. A decade of uncertainty

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 9, March 2014 DOI: https://doi.org/10.1038/nnano.2014.43

Summary: Ten years after the publication of an influential Royal Society/Royal Academy of Engineering report on the uncertainties in nanoscale science and engineering, this commentary asks whether we are in danger of creating a new metaphorical grey goo -- an overwhelming mass of nano-safety research that may not be addressing the most important questions.

15. Is novelty overrated?

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 9, June 2014 DOI: https://doi.org/10.1038/nnano.2014.116

Summary: Nanomaterial risks are often considered in terms of novel material behaviours. This commentary asks whether framing risk around novelty may end up obscuring some risks while overplaying others, and argues for alternative approaches to developing advanced materials and products that are safe by design.

16. Old materials, new challenges?

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 9, September 2014 DOI: https://doi.org/10.1038/nnano.2014.196

Summary: Fumed silica has been used as an anti-caking agent in foods for several decades. This commentary asks whether new research suggesting it may be more hazardous than previously thought means that the use of this engineered nanomaterial needs to be re-examined, exploring the tension between long safety track records and new toxicological findings.

17. Could we 3D print an artificial mind?

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 9, December 2014 DOI: https://doi.org/10.1038/nnano.2014.294

Summary: 3D printing is allowing more complex three-dimensional structures to be manufactured than ever before. This commentary asks whether the convergence between 3D printing technology and nanotechnology could eventually usher in a new era of artificial intelligence, exploring the potential for bioinspired neuromorphic computing substrates.

18. The (nano) entrepreneur's dilemma

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 10, March 2015 DOI: https://doi.org/10.1038/nnano.2015.35

Summary: Emerging technologies need to be developed responsibly if their benefits are to outweigh potential risks. This commentary asks whether entrepreneurs really have the luxury of grappling with future consequences from the get-go, and explores the tension between responsible innovation ideals and the practical realities of entrepreneurship.

19. Learning from the past

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 10, June 2015 DOI: https://doi.org/10.1038/nnano.2015.120

Summary: When it comes to safety, the jury's still out on which nanoparticle characteristics we should be measuring. This commentary explains that there's a rich history dating back over a hundred years on how we measure them, starting with John Aitken's 1888 condensation particle counter, and connects that history to current challenges in nanoparticle characterization and exposure assessment.

20. Why we need risk innovation

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 10, September 2015 DOI: https://doi.org/10.1038/nnano.2015.196

Summary: If emerging technologies such as nanotechnology are to reach their full potential, this commentary argues that we need to radically change our approach to risk. It explores Google's nanosensor concept and the broader emerging risk landscape, arguing that traditional health and environmental risk assessment frameworks fail to capture the full panoply of personal, social, technological, economic, political and corporate risks that determine the fate of new technologies.

21. Navigating the fourth industrial revolution

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 10, December 2015 DOI: https://doi.org/10.1038/nnano.2015.286

Summary: This commentary considers the challenges of ensuring the responsible development and use of converging technologies in the context of the fourth industrial revolution -- the unprecedented fusion of digital, physical and biological technologies. It warns that without up-front efforts to ensure beneficial, responsible and responsive development, this revolution will not only fail to deliver on its promise but may increase the very challenges its advocates set out to solve.

22. Navigating the risk landscape

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 11, March 2016 DOI: https://doi.org/10.1038/nnano.2016.28

Summary: The potential risks surrounding nanotechnology can often appear complex and confusing. This commentary provides basic guideposts for navigating them, arguing that "nanotechnology" is itself an unreliable indicator of risk, that nanomaterials are not just chemicals, that benchmarking is important, and that risk starts with identifying something worth protecting.

23. Are we ready for spray-on carbon nanotubes?

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 11, June 2016 DOI: https://doi.org/10.1038/nnano.2016.99

Summary: As artists and manufacturers explore the use of spray-on carbon nanotube coatings such as Vantablack, this commentary explores the state of the science around nanotube safety, tracing the history from early concerns in the 1990s through landmark studies on asbestos-like pathogenicity, and examining the complexities of assessing health risks from diverse carbon nanotube types.

24. Is nanotech failing casual learners?

Authors: Andrew D. Maynard Journal: Nature Nanotechnology, Vol. 11, September 2016 DOI: https://doi.org/10.1038/nnano.2016.167

Summary: How easy is it for people to learn about nanotechnology through the Internet? This commentary explores the challenges and opportunities for casual learners -- people who are curious about a topic and self-motivated to learn more -- in finding quality online nanotechnology information, and finds that despite growing online resources, discovering credible and engaging content remains surprisingly difficult.

25. 'Safe handling of nanotechnology' ten years on

Authors: Andrew Maynard, Robert Aitken Journal: Nature Nanotechnology, Vol. 11, December 2016 DOI: https://doi.org/10.1038/nnano.2016.270

Summary: In 2006, a group of scientists proposed five grand challenges to support the safe handling of nanotechnology. Ten years on, this commentary by two of the original authors looks at where we have come and where we still need to go, finding that while there has been notable progress in toxicity testing methods and life cycle assessment, progress on exposure measurement instruments and predictive modelling remains low.

Other Papers

26. How to Succeed as an Academic on YouTube

Authors: Andrew D. Maynard Journal: Frontiers in Communication, 2021 DOI: https://doi.org/10.3389/fcomm.2020.572181

Abstract: More and more people are turning to YouTube to expand their knowledge, develop their understanding, and learn new skills. These "casual learners" -- loosely defined as individuals who are curious about a topic and are self-motivated to learn more about it -- are taking advantage of the ease with which nearly anyone with an internet connection, basic video skills, and something to say, can become a YouTube "creator." However, amidst a dizzying array of videos purporting to educate or otherwise inform viewers, academic content-creators are notable by their lack of presence on the platform. Here, there are largely-untapped opportunities for academics to contribute to the richness, diversity and trustworthiness of video content available to casual learners, and to effectively mobilize their knowledge at scale. There is also a pressing need for diversity in casual learning content, including diversity in creator gender, identity, ethnicity, and perspective, and academics are uniquely positioned to address this need. Drawing on the author's experiences in developing and producing the YouTube channel Risk Bites, this perspective explores how time, resource, and even talent-limited academics can nevertheless leverage YouTube as a platform for further mobilizing their knowledge for public good.

27. Artificial Intelligence Is Conspicuous by Its Absence in Denis Villeneuve's Dune: Part Two. And This Is Important.

Authors: Andrew D. Maynard Journal: Jurimetrics (American Bar Association), Winter 2024 URL: https://www.americanbar.org/groups/science_technology/resources/jurimetrics/2024-winter/artificial-intelligence-conspicuous-absence-dune-part-two/

Summary: This film review examines the conspicuous absence of artificial intelligence in Denis Villeneuve's Dune: Part Two (2024), set against the backdrop of the real-world AI revolution that occurred between the release of the first and second films. The review uses the Dune universe -- in which "thinking machines" are seen as an evil that has no part in humanity's future -- as a lens for examining contemporary debates about AI's role in society.

28. Gender disparity in U.S. patenting

Authors: Jieshu Wang, Andrew Maynard Journal: Humanities and Social Sciences Communications, 2025 DOI: https://doi.org/10.1057/s41599-025-06038-6

Abstract: Despite growing attention to gender disparities in innovation, little is known about how gender shapes the characteristics and outcomes of patented inventions. This study analyzes 3.7 million U.S. utility patents, covering 1.8 million distinct inventors and over 200,000 organizations, to investigate the gendered patterns of inventorship. While women's participation in patenting has increased over time, they remain significantly underrepresented, and patents involving female inventors consistently receive fewer citations than those by all-male teams. However, women-participated patents are more likely to exhibit novelty, originality, and technological generality, particularly when produced by mixed-gender teams, which tend to generate the most disruptive inventions. Female inventors also draw more heavily on scientific literature and public support, especially in green technology and academic settings. Organizational and domain-level differences are pronounced: universities involve women at higher rates than corporations, and fields such as biotechnology and civil engineering demonstrate distinct gendered patterns in patent quality and disruption. These results suggest that women make important yet often overlooked contributions to innovation and that structural barriers may suppress their full inventive potential. Addressing these disparities can enhance innovation diversity, expand the societal relevance of patented technologies, and better support the next generation of inventors.

Website Content

The following pages are extracted from andrewmaynard.net and provide key context on Andrew Maynard's work, ideas, and approach.

What Does It Mean to Be Human in an Age of AI?

Source: https://andrewmaynard.net/ai-and-being-human/being-human-in-an-age-of-ai/ Note: Extracted via WebFetch — largely faithful but may contain minor paraphrasing.

There's a question that keeps surfacing in my classes, in my work, and in conversations I have from everyone from founders and investors to managers, teachers, parents, and many others: _What makes me me when AI can do what I do?_

It's a question that sounds abstract, until it isn't. Until an AI writes something in your voice that your colleagues can't tell apart from your own work. Until a recommendation engine predicts what you'll want before you've articulated it to yourself. Until a student asks whether the essay they wrote "counts" if they used AI to help structure their thinking.

I've spent more than two decades studying how transformative technologies reshape society — from engineered nanomaterials to synthetic biology to brain-computer interfaces. But AI is different. Not because it's more powerful than those technologies (though I think it may be), but because it's the first technology that doesn't just change what we _can do_. It changes our understanding of what we _are_.

The mirror that talks back

Previous technologies extended our physical capabilities. The printing press extended our memory. The car extended our reach. The computer extended our ability to calculate. AI extends something more intimate: our ability to think, to create, to communicate, to _be_.

When a machine can generate art, compose music, write persuasively, and pass the kind of reasoning tests we've long treated as markers of intelligence, it doesn't just create a productivity tool. It holds up a mirror. And what we see in that mirror is fascinating and unsettling — and it changes us.

This is why I believe the rise of AI is fundamentally a human question, not a technology one. The technology will continue to advance. But what isn't set is how we respond to it. Whether we let it diminish our sense of what we're worth, or whether we use it as an invitation to understand ourselves more deeply.

Four qualities that matter

In the work Jeff Abbott and I did for our book _AI and the Art of Being Human_, we identified four qualities that are essential — not despite AI's capabilities, but _because_ of them:

These aren't sentimental ideas. They're practical ones. And they're the foundation of the 21 tools Jeff and I developed for navigating AI with your humanity intact.

This isn't about being anti-AI

Here, I want to be clear about something, because it matters: this work is not about resisting AI. I use AI every day. I used it extensively in developing the book. I use it in my research — both to enhance it and as something I study. I teach my students how to think about it and work with it effectively. I'm genuinely excited about what it makes possible.

But I'm also convinced that excitement without reflection is how we sleepwalk into futures we didn't choose. The pace of AI development means we have a narrow window — maybe five years, maybe less — to shape the relationship between humanity and artificial intelligence. After that, the infrastructure hardens, the habits calcify, and the choices we failed to make become the defaults we're stuck with.

That's why I care about this. Not because AI is dangerous (though it can be), and not because it's miraculous (though it sometimes feels that way). But because how we respond to it will define what it means to be human for generations to come.

21 Practical Tools for Thriving with AI

Source: https://andrewmaynard.net/ai-and-being-human/21-tools-for-thriving-with-ai/ Note: Extracted via WebFetch — summarized by extraction tool, not verbatim.

21 practical tools created by Andrew Maynard and Jeff Abbott for their book "AI and the Art of Being Human." The tools are organized into four sections:

Part I: Mindsets for an Age of AI

Six foundational tools designed to cultivate engagement with AI from a position of agency rather than anxiety:

  1. Mirror Test (Prelude, p. 10): Three questions to ask when AI seems to know you too well
  2. Curiosity Loop (Chapter 1, p. 21): Turning the shock of AI capability into something you can learn from
  3. Intent Map (Chapter 2, p. 36): Making your values visible before momentum decides for you
  4. Human Qualities Spectrum (Chapter 3, p. 57): Understanding what AI can replicate and what remains irreducibly human
  5. 4-Lens Scan (Chapter 4, p. 73): Ninety seconds to see what urgency hides
  6. 7-Minute Clarity Pause (Chapter 4, p. 75): A structured pause when the stakes are high

Part II: Navigating Change

Four tools addressing identity shifts and value tensions:

  1. Identity Matrix (Chapter 5, p. 91): Mapping what's replaceable against what endures
  2. STARS Framework (Chapter 5, p. 97): Building sustainable practices around what matters
  3. Stress-Test Table (Chapter 6, p. 113): Making values trade-offs visible and concrete
  4. Micro-Circle Launch Kit (Chapter 7, p. 135): The essentials for gathering others

Part III: Thriving in Partnership

Five tools for active collaboration with AI systems:

  1. Orchestration Triangle (Chapter 8, p. 154): Balancing data, intuition, and context
  2. CARE Loop (Chapter 9, p. 167): Making care systematic rather than incidental
  3. Model Dignity Check (Chapter 9, p. 170): Five questions before any AI system goes live
  4. Prompt-Scaffolding Canvas (Chapter 10, p. 181): Structuring creative conversations with AI
  5. Multimodal Ideation Sprint (Chapter 10, p. 185): Rapid exploration that keeps you in the driver's seat

Part IV: Intentional Futures

Six forward-looking tools:

  1. Roadmap Canvas (Chapter 11, p. 199): Translating understanding into 90-day experiments
  2. Community Flywheel (Chapter 12, p. 215): Growing and sustaining communities needed to thrive with AI
  3. Starter Charter (Chapter 12, p. 220): Enough structure to hold, enough openness to breathe
  4. Pocket Card (Chapter 13, p. 231): Four principles you can hold in your hand
  5. One-Line Vow (Chapter 13, p. 233): A public commitment that holds you accountable
  6. Commitment Ladder (Chapter 13, p. 235): From today's intention to next year's practice

These are concrete, practical frameworks designed for immediate use. Downloadable tool versions are available at https://www.aiandtheartofbeinghuman.com/the-tools. The complete tools with narratives appear in the published book.

Teaching and Learning in an Age of AI

Source: https://andrewmaynard.net/ai-and-being-human/teaching-and-learning-in-an-age-of-ai/ Note: Extracted via WebFetch — summarized, not verbatim.

Andrew Maynard provides comprehensive educational resources for teaching about AI and human identity. His work addresses fundamental student questions: "Is it cheating if I use AI to help me think? What's the point of learning to write if a machine can write better?"

Key Resources

The Instructor's Guide is an AI-compatible document that helps educators develop syllabi, discussion questions, and assignments based on the book AI and the Art of Being Human. Free to download at https://www.aiandtheartofbeinghuman.com/educators

The AI Companion allows students to explore the Pocket Edition interactively through conversation with AI assistants, modeling intentional, reflective AI use while engaging with course material. Free to download at https://www.aiandtheartofbeinghuman.com/ai-companion

Featured Teaching Tools

Several frameworks are particularly effective in educational contexts:

Implementation Approaches

The book works as semester-long course material, single-session workshops, or professional development training. Its fictional characters represent diverse geographies and professions, helping students see themselves in the characters rather than abstract case studies.

Frequently Asked Questions: AI and the Art of Being Human

Source: https://andrewmaynard.net/ai-and-being-human/faq/ Captured: 2026-08-04 via WordPress REST API Modified: 2026-02-15T13:56:29

Frequently Asked Questions: AI and the Art of Being Human

These are some of the questions that get asked frequently by readers, by people considering the book, and by educators and organizations thinking about how to use it. If yours isn’t here, get in touch.

What is AI and the Art of Being Human about?

AI and the Art of Being Human is a practical guide to thriving with AI while staying grounded in what makes us uniquely human. Written by Jeffrey Abbott and me, it offers 21 practical tools — organized around four principles: curiosity, intentionality, clarity, and care — for navigating a world where AI can increasingly do what we once thought only humans could. The book uses fictional narratives set across the globe to make these ideas lived and felt, not just understood.

Who is this book for?

Anyone navigating life in a world increasingly shaped by AI. We wrote it for a wide audience — professionals, entrepreneurs, investors, parents, artists, educators, students, and anyone asking “what’s still mine in an age of AI?” The tools are designed to be relevant regardless of your background or technical knowledge.

How is this different from other AI books?

Most AI books focus on what the technology can do, where it’s heading, or what risks it poses. This one focuses on you — your agency, your identity, your meaning — in the face of AI’s growing capabilities. It’s also distinctive in two ways: it uses fictional narratives (27 characters across 13 chapters) to explore ideas in ways that nonfiction alone can’t achieve, and it provides 21 concrete, hands-on tools you can use immediately. It’s not about surviving AI or fearing it. It’s about thriving alongside it.

What are the 21 tools?

They’re practical frameworks organized in four parts. Part I covers foundational mindsets (including the Mirror Test for when AI knows you too well, the Intent Map for making values visible, and the Human Qualities Spectrum for understanding what remains irreducibly human). Part II helps navigate identity and values under pressure. Part III covers creative partnership and organizational care. Part IV focuses on building personal roadmaps and communities. Each tool emerged from a fictional narrative that shows it in action. Full details are [here](/ai-and-being-human/21-tools-for-thriving-with-ai/), and downloadable versions are available at aiandtheartofbeinghuman.com/the-tools.

What are the four principles?

The book’s philosophical backbone: Curiosity (staying open and willing to be surprised), Intentionality (choosing consciously rather than following algorithmic momentum), Clarity (seeing what AI misses — the human context beneath the data), and Care (choosing human flourishing over pure optimization). They appear together on the Pocket Card and thread through every chapter.

Why fictional narratives? Why not case studies?

Because the frontier of AI is so new and fast-moving that the real-world case studies most people rely on don’t yet exist — at least not at the depth needed. Fiction lets us explore not just what is, but what could be. Readers see themselves in characters like Elena (a Munich founder whose AI completes her most private thoughts), Sana (a Cairo journalist staring at a deepfake worth millions), or Wei Lin (a fifteen-year-old whose AI portrait is “better” than his own). The stories create understanding that purely analytical writing can’t.

Was this book written with AI?

Yes, deliberately and transparently. Jeff and I made a strategic decision to work closely with AI — Anthropic’s Claude — throughout the development process. This wasn’t an attempt to generate a quick AI book. It was months of methodical work exploring how we could write with AI, and in the process reveal more about the art of being human than we could achieve alone. Every idea, tool, and insight has our human stamp on it. We write about this process in the book.

What is the AI Companion?

It’s a downloadable file (about 78,000 words) that turns your AI assistant into a thinking partner for exploring the book. Upload it to Claude, Gemini, or Grok, and you can have a conversation about the book’s ideas, characters, and tools tailored to your specific questions and situation. It’s free to share under a Creative Commons license. Available at aiandtheartofbeinghuman.com.

What’s the difference between the full edition and the Pocket Edition?

The Pocket Edition captures the essentials — the same stories, characters, and all 21 tools — in a streamlined, portable format (4.25″ × 7″). It removes the sidebars, hands-on cards, footnotes, and longer background passages from the full edition, and adds a Tool Finder and Quick Reference Guide. Designed to be well-thumbed, spine-cracked, and coffee-stained.

Can I use this book in my classroom or organization?

Absolutely. We designed it with educational and organizational use in mind. There’s an Instructor’s Guide for educators, and the tools translate directly to workshop and professional development settings. See the [For Educators](/teaching-and-learning-in-an-age-of-ai/) page for more.

Where can I get the book?

Both editions are available on Amazon, Barnes & Noble, Thriftbooks, and wherever books are sold. You can also download a free preview of the first 50 pages at aiandtheartofbeinghuman.com.

Who is Andrew Maynard?

I’m a scientist, author, and Professor of Advanced Technology Transitions at Arizona State University. I’m the founding director of the Future of Being Human initiative and have spent over two decades studying how emerging technologies reshape society — from federal nanotechnology programs to World Economic Forum Global Agenda councils. I write about these ideas weekly in my newsletter The Future of Being Human and co-host the Modem Futura podcast with futurist Sean Leahy. Previous books include Films from the Future and Future Rising.

Who is Jeffrey Abbott?

Jeff is a founding partner of Blitzscaling Ventures (backed by Reid Hoffman) and founder of AI Salon, a global community spanning 60+ cities that hosts hundreds of events annually. His firm has invested in companies like CrewAI and the AI upskilling platform Multiverse. He bridges Silicon Valley innovation with human-centered values — managing AI investments at a venture capital firm while simultaneously building one of the world’s largest AI practitioner networks.

How I Work

Source: https://andrewmaynard.net/how-i-work/ Captured: 2026-08-04 via WordPress REST API Modified: 2026-08-04T10:34:55

How I Work

Nearly everything I produce that is under my control is free and openly available, or — as in the case of my books — as affordable as I can realistically make it. I believe that I have a responsibility to ensure that my work is accessible as possible to anyone who can potentially benefit from it — no matter who they are or what their circumstances.

However, my time and attention are limited. As a result, for commercial work I have the following rates.

Andrew Maynard

Rates

I will generally work without charging for my time with educators, nonprofits, public-interest organisations, journalists, students and early-career researchers. Honoraria are always welcome, but never required. On the other hand, travel costs are.

For events I’ll sometimes waive or reduce my fees if there is direct benefit to my work and it’s impact.

Commercial engagements are contracted with me personally, not through Arizona State University.

Please feel free to get in touch if you have questions or a request.

Thought Leadership

Source: https://andrewmaynard.net/thought-leadership/ Captured: 2026-08-04 via WordPress REST API (page modified 2026-07-12)

Thought leadership

Much of my career over the past 20 plus years has focused on working through various networks, organizations, and platforms, to help guide and inform decision making around advanced technology transitions and socially responsible innovation.

This includes testifying before congressional committees, working closely with organizations such as the World Economic Forum, OECD and others, contributing to National Academies studies, working widely with print and broadcast media, and writing extensively for a public audience – including through articles and newsletters.

You can read more about my thought leadership in the drop down menu above, but here’s a broader picture of what I do, and why:

Empowering others and catalyzing public value creation through knowledge mobilization and thought leadership

My influence and impact as a thought leader, communicator, and public intellectual, are driven by a conviction that academics — especially academics at a public university — have a societal responsibility to ensure knowledge and the insights associated with it are made as accessible, meaningful, and impactful, to as many people as possible, whether these are business leaders, policy makers, civil society, educators, members of the public, or others. They are also underpinned by a deeply transdisciplinary approach to exploring and addressing emerging challenges and opportunities. Through my work I intentionally and strategically leverage my expertise, networks, platforms, and skills, in numerous and often novel ways to mobilize knowledge, understanding, and insights, in the service of empowering others to be part of building a positive future together.

The following provide some sense of the domains I cut across, but this is by no means an exhaustive list:

Beyond Transdisciplinary Thinking

While I started my professional career as a physicist, nearly four decades of working across disciplines and sectors, and with an increasingly diverse range of people, issues, and opportunities, has led to my work largely transcending disciplinary norms and boundaries. I see this as a strength in what I do as it allows me to both pull insights from the intersection of miltiple domains, and to develop and apply new thinking and ideas to complex challenges that transcend conventional thinking. Over my career I have published in fields as diverse as electron microscopy, aerosol dynamics, occupational health, public health, environmental science, toxicology, responsible innovation, nanotechnology, synthetic biology, artificial intelligence, entrepreneurship, and more. I have provided expert advice and thought leadership across many more domains. I have worked in the public and private sector (albeit briefly), and have collaborated with policy makers, government agencies, private companies, trade groups, startups, founders, funders, NGOs, activist groups, academics, community groups, and more. The downside of this diversity, of course, is that it’s easy to see me as a master of very little. But at the increasingly complex intersection of advanced technologies, society, and the future, there is a growing need for people who can work fluidly across boundaries and make connections that elude experts constrained by disciplinary and domain conventions. And this, above all, is where my mastery lies.

Advanced Technology Transitions

For over two decades my research and thought leadership have broadly encompassed what may be described as “advanced technology transitions.” This is a field I have highlighted through my public-facing work, and one that represents a unique and broad framework for approaching the beneficial development and use of potentially disruptive new technologies. It is a framework that is becoming increasingly relevant in my thought leadership around advanced technologies such as artificial intelligence and quantum technologies. I founded and direct the ASU Future of Being Human initiative that is explicitly focused on catalyzing conversations around advanced technology transitions, and building thought leadership capacity around technology, society, and the future.

Responsible innovation and emerging technologies

My work over the past 20 years has increasingly focused on supporting decision making around responsible innovation and emerging technologies. Since 2008 I have worked extensively with the World Economic Forum, including participating in and chairing Global Agenda Councils and Global Futures Councils, being an invited speaker at Davos and the Annual Meeting of New Champions in China (the “Summer Davos”), and participating since its inception in the working group behind the World Economic Forum’s annual list of top ten emerging technologies. I served on and chaired the WEF Global Agenda Council on emerging technologies during the period when the Forum’s strategic focus shifted toward transformative technologies — work that preceded and helped create conditions for the Fourth Industrial Revolution framing. And I continue to be a leading public thinker around responsible innovation and emerging technologies. My writing and academic publications continue to push the boundaries of thinking around socially responsible and beneficial innovation.

Risk Innovation

Much of my professional career has touched on risk, and has ranged from conventional risk assessment and management (especially within the context of occupational and public health), to grappling with novel risks and innovative ways of thinking about and addressing risk. The latter has led to the emergence of “risk innovation” as a unique approach to understanding and navigating complex social risks in particular that are not covered by existing risk frameworks, and yet are critical to advanced technology transitions. My work around risk innovation is reflected in my time as Director of the University of Michigan Risk Science Center and the Arizona State University Risk Innovation Lab, and has focused on engaging with multiple stakeholders. This includes one of the most successful YouTube learning channels on understanding risk – Risk Bites. The channel provides a unique and highly accessible source of content on understanding risk, and includes some of the top-ranked YouTube videos in areas such as nanotechnology, epidemiology, and the fourth industrial revolution.

Artificial Intelligence

For over a decade my work around navigating advanced technology transitions has extended to emerging opportunities and challenges around AI. In 2015 I led the development of a substantial proposal to the Future of Life Institude addressing the governance of emerging artificial intelligence capabilities. This was ultimately unsuccessful, but it kick-started a growing interest in the unique challenges and opportunities presented by AI. In 2017 I participated in the AI Asilomar meeting that led to the Asilomar AI Principles — one of the first coordinated set of principles designed to oversee responsible and beneficial artificial intelligence. My 2018 book Films from the Future grappled directly with emerging challenges around AI-mediated persuasion and manipulation that were, at the time, an emerging but not yet tangible threat.

This changed in 2022 with the launch of ChatGPT and the rapid expansion and accessibility of GPT and LLM technologies. Since then I have expanded my thinking and writing substantially around how frontier AI models and capabilities are profoundly altering the landscape around technology transitions and human thriving. As well as being a dominant focus of the Future of being Human Substack, my work here led to the 2025 book AI and the Art of Being Human.

I continue to be active in exploring the rapidly moving and fast-evolving edge of AI, and how it potentially impacts the future.

Nanotechnology

Through my work with government agencies, industry, civil society, and other organizations, I have had a global impact on research, policy, and decision making around the safe and beneficial development of nanotechnology over the past two plus decades. In the early 2000’s I was responsible for co-leading the US federal government’s strategic initiatives around nanotechnology safety. Between 2005 – 2010 I was one of the most influential thought leaders globally in the responsible development of nanotechnology in my role as Chief Science Advisor to the Woodrow Wilson International center for Scholars’ Project on Emerging Nanotechnologies. I have testified before congressional committees, served on National Academies committees, worked with organizations that include OSTP, OECD, and the World Economic Forum, and become one of the go-to experts on nanotechnology safety for journalists and policy makers. Although my work now extends far beyond nanotechnology, I continue to contribute to thought leadership here.

Public Engagement

I am known internationally for my work as a highly effective communicator, convener, moderator, and facilitator of public engagement. As well as being a sought-after speaker, I am regularly invited to talk about emerging technologies and responsible innovation by journalists and media outlets. I am a regular contributor to platforms such as Slate Future Tense, World Economic Forum Agenda, and The Conversation, and have written for outlets that include the Washington Post, Discover Magazine, Salon, Scientific American, and The Guardian. In addition, I write extensively for my own public-facing platforms, including a highly successful Substack newsletter. I approach my public engagement activities as integral to my position as a tenured professor at a public university, and deeply integrated with my scholarship and teaching. I focus specifically on knowledge mobilization, and opening up pathways and opportunities for emerging knowledge and insights to have far-reaching public accessibility, relevance, and impact.

Connecting with broad audiences

I have written two popular books on technology, society, and the future, that are designed to facilitate knowledge mobilization at scale around future-building in a technologically complex world. These books – “Films from the Future” and “Future Rising” – uniquely bring complex ideas around emerging technologies, society, and the future, to a broad audience. They are written to be engaging and accessible to a broad audience, while taking readers on a transdisciplinary journey of discovery that opens their eyes to new possibilities and ways of thinking as transformative technologies become increasingly complex and influential. Importantly, this has become a transformative avenue for scaling the national and global reach of my thought leadership and work around knowledge mobilization.

Weekly reflections on technology, society, and the future: The Future of Being Human — join over 5,000 readers

The Future of Being Human Initiative (ASU)

Source: https://futureofbeinghuman.asu.edu/ Captured: 2026-08-04 via plain HTTPS fetch (non-WordPress-REST; ASU site)

The future of being human initiative

We are a unique community of bold, audacious and visionary thinkers who are inspired by what it might mean to be human in a technologically transformed future — and who are passionate about exploring how this influences our thinking and actions in the present.

What's New

AI and the Art of Being Human

A practical guide to thriving with AI while rediscovering yourself in the process.

Modem Futura podcast

Join Future of Being Human hosts Andrew Maynard and Sean Leahy each week as they explore the future of technology and humanity in an increasingly complex world.

Artificial Intelligence

Thought leadership on AI and the future: Articles, papers, podcasts and media exploring the intersection between the cutting edge of AI and society.

Who are we?

We are thought leaders, researchers, scholars, students, entrepreneurs, investors, business executives, and ordinary people with extraordinary ideas from every background. Together, we are intrigued by what it might mean to be human in the future, obsessed by how emerging technologies are rewriting the rulebook of what is possible, and inspired by the idea of exploring the future of being human with other creative thinkers.

What do we do?

We create and curate ways of bringing people together to explore compelling questions and transformative ideas around the future of being human.

Some of these are intimate informal hangouts, others are cutting edge online discussions. And some are high profile public events and even retreats.

We are even developing educational opportunities unlike anything you'll find anywhere else!

All of these are driven by a passion to bring together audacious, original and passionate thinkers to push the boundaries of how we imagine the future of being human in a technologically complex world, and how this can inform our ideas, aspirations, and actions, in the present.

What inspires us?

As a community we are captivated and inspired by compelling questions around how emerging technologies may challenge and transform what it means to be human.

Questions like:

The possibilities are endless, and the implications to how we think, act, and live in the present, are profound.

What brings us together?

Together, we are a community that embraces:

Stay in touch

We are actively seeking to connect with bold and audacious thinkers who are passionate about exploring the future of being human with us. If you're interested in staying in touch, or getting more involved with the community, sign up for updates on the site.

Before the Fourth Industrial Revolution: Notes on an Institutional Prehistory

Source: https://andrewmaynard.net/2026/04/08/fourth-industrial-revolution-prehistory-wef-councils/ Date: April 8, 2026 Note: Extracted via WebFetch — summarized, not verbatim. This is a significant blog post and the authoritative public statement on Maynard's relationship to the Fourth Industrial Revolution concept.

Andrew Maynard traces the institutional groundwork preceding the formal 2016 launch of the Fourth Industrial Revolution concept by Klaus Schwab and the World Economic Forum. He documents nearly eight years of preparatory work through WEF's Global Agenda Councils that established the intellectual foundation for what would become a globally influential framework.

Key Timeline

2008: Maynard joined the inaugural WEF Global Agenda Councils, initially on Nanotechnology but successfully advocating to rebrand as the Council on Emerging Technologies. He proposed a "Global Institute on Emerging Technology Policy," emphasizing that new and innovative policies are needed at the international, national, corporate and institutional level.

2010: Working with Tim Harper, Maynard developed the formal proposal for a Centre for Emerging Technology Intelligence (CETI), published in WEF's Everybody's Business report, conceptualizing an independent body to provide integrated analysis of emerging technologies.

2011-2015: The council supported the creation of WEF's annual Top Ten Emerging Technologies list, which became remarkably popular and served as the most visible public outcome of the earlier CETI vision.

2015: Maynard published "Navigating the fourth industrial revolution" in Nature Nanotechnology, synthesizing evolving frameworks about technology convergence.

2016: The World Economic Forum pivoted technology to its primary strategic focus, making "4IR" a global brand through Davos programming.

Key Insights

Failed experiments create value: Though CETI was never formally established, the proposal and surrounding intellectual work created essential groundwork that later initiatives drew upon, whether consciously or not.

Attribution gaps matter: WEF's authorship conventions obscured actual intellectual labor distribution. Documents typically credited council chairs rather than primary conceptualizers.

"Conditions for possibility" deserve recognition: Large-scale concepts emerge not from individual authorship but from prolonged preparatory work making those concepts thinkable at scale.

Contemporary Relevance

Current AI governance conversations echo the same fundamental questions posed in 2008 — how to build institutional capacity for technologies outpacing existing regulatory frameworks. This prehistory offers useful lessons for patient institutional development and the value of circulating serious ideas even when specific proposals don't reach implementation.

Note: The intellectual work on the Fourth Industrial Revolution concept itself was led by Nick Davis and Tom Philbeck at WEF; Maynard's contribution was to the prehistory, not to the concept's authorship.

Selected Substack Essays

The following essays are selected from Andrew Maynard's Substack newsletter "The Future of Being Human" (https://futureofbeinghuman.com/). They represent key themes in his recent thinking about AI, technology, society, and the future. These are the original texts as written by Andrew Maynard.

What we can learn with AI by NOT trying to learn

Sometimes, throwing productivity goals, purpose, and measurable outcomes out of the window might be the most powerful way to thrive in a changing world

Date: August 2, 2026 Source: https://www.futureofbeinghuman.com/p/what-we-can-learn-with-ai-by-not-trying-to-learn Newsletter: The Future of Being Human (Substack)

!image

Image caption: Image: Midjourney

Academic sabbaticals are dangerous things.[1] They give you time to think and reflect that’s unhindered by the pressure to produce. And if you’re not careful, you can find yourself wandering down paths that your more straight-laced colleagues might disapprove of …

… like pursuing the idea that play without purpose, or embracing what sparks joy, or even reveling in the small delights of unexpected discoveries, are all critical skills for thriving in an age of AI. And that, sometimes, the best way to learn and grow when transformative technologies are rewriting the rules of how we do pretty much everything, is to not try to learn.

For people who study and embrace such ideas, of course, this isn’t new. And anyone who’s been following my work for the past decade or so will know that this is a space I’ve been inhabiting for a while.

Yet the reality is that how we teach, how we develop career-enabling skills, how we professionally evaluate ourselves and others, and even how we behave within professional environments, tends to devalue and discount the importance of joy, delight, and play, and to treat them as trivial, immature, and not appropriate for serious people doing serious jobs.[2]

If you doubt me, just take a quick look at your LinkedIn feed.

Yet as I get into this year’s sabbatical in earnest, I’m finding myself spending more and more time exploring such ideas, and asking whether we need to take the concept of learning through not trying to learn far more seriously as AI and other technologies shake up conventional thinking around what it takes to thrive in the work we do and the lives we lead.

I’ll be writing more about this I’m sure as the sabbatical proceeds. But in the meantime, I did want to write about one particular example here that reflects my evolving thinking and explorations — although I should warn you that, if you believe that there is no place in professional practice for joy, delight, and play without purpose, you may want to call it a day and stop reading here 😊.[3]

Revisiting Hyperbubble

A few weeks ago, I wrote about using Anthropic’s top-end Fable 5 model to create a simple browser-based video game. What started as a test of the model’s capability quickly developed into a collaboration around co-creating experiences.

Since then, I’ve spent considerable time with Fable to refine and extend the game — partly out of the simple joy and delight of doing this, but also as a way of further-exploring the ideas I touch on above around learning with AI by not trying to learn.

Through this process, I’ve been surprised (although I shouldn’t have been) by how much my thinking around the importance of playful exploration has evolved, especially as a pathway to developing essential skills for thriving professionally in a changing world. This has emerged through the act of developing the game with Fable. But it’s also been influenced by actually playing the game — which is something I wasn’t expecting.

Hyperbubble (Fable’s name) is a one-button browser-based game where your task is to navigate your character through a future of emerging technologies, complex risks, transformative possibilities, and unexpected delights, all while protecting and growing their state of “flourishing.”

Written like this, the game sounds somewhat serious and — if I’m honest — a little educational-preachy. Until, that is, you see how Fable helped translate the underlying ideas into a game that is anything but serious, preachy, or overtly “educational.”

Here, I must confess that I really like the resulting game.

Of course, there is every chance that I love Hyperbubble because it’s my baby[4] and represents a chunk of work — and that to anyone else it’s just an embarrassment that’s best forgotten and moved on from.

But I’m going to lean into it anyway as, even if you’re not into trivial-seeming browser-based games, Hyperbubble remains an intriguing exercise in learning from and through AI by not trying to learn.

Version 5 of the game

The genesis of Hyperbubble was me asking Fable to do a deep dive into my work, my ideas, my mindset, my aims, what motivates and delights me, and to build a simple and fun game inspired by these. This was initially intended to be a test of the AI’s abilities, and it’s what led to the first iteration of the game and that first article.

Hyperbubble is now on version 5, and is the result of over 90 iterations between me and Fable around developing the game’s feel, focus, and substance.

The result is a game which is infused with simplicity and delight, which captures not only my work but how I think and see the world, and which can be played with no interest in or awareness of the ideas that it represents — and yet through playing it, the player encounters complex and nuanced ways of thinking about the interplay between advanced technologies, the future, and human flourishing.

Starting with the overall aesthetic of the game, this was an intentional design choice made by Fable, and reflects my work on Risk Bites YouTube videos. While this channel has been moderately successful (receiving well over 5 million views) the videos are based on quite crude stick figures drawn on a whiteboard (or black-glass board) and reflects the work of an academic with (in my own words which I believe I wrote somewhere) “no talent and even less time.”

A fitting aesthetic for Hyperbubble I think!

The game itself was designed to be discovered and delighted in through experimentation rather than following detailed instructions — another purposeful design decision. That said, there are fairly detailed instructions accessible from the home screen.

Here, the home screen (below) provides players with a number of options that allow them to explore the game further, modify the game play, and check out the high scores:

!image

The links to how to play the game and the ideas behind it are reasonably clear. What is less clear is that, if you click on the sun three times, you enter a “techno-optimist mode” which makes it easier to flourish for longer. On the same screen, clicking “gentle mode” increases your chances of flourishing for longer.

Also from this screen, clicking anywhere starts the game in normal mode. But clicking “tap here” or pressing “T” puts the game in “Today’s Future” mode. In this mode, the game environment is the same for anyone playing it on the same day. (This was a feature it took me a few days to discover — entirely Fable’s creation).

Starting the game opens a screen where you, as the player, are encased in a soap bubble (a reference to my book Future Rising) and traveling into the future along an undulating landscape.

!image

As you travel, you have just one control — the space bar (or mouse button, down arrow, or finger on the screen if you are on a tablet). Hold it down and the bubble is pulled downward. Release it and the force is likewise released. Hold it down on the down-slope of a hill and you’ll pick up speed — and likely jump off the crest of the next hill. Hold it down while in the air and you will start to descend — fast.

The game play is inspired by my work around navigating an increasingly complex risk-benefit landscape around emerging technologies. Along the way you’ll encounter orphan risks — which you can adopt by pressing the space bar (or similar) as you pass them; emerging tech risks, which you manage by likewise keeping the space bar pressed as you pass them; moral panics (scribbly fires) which die down if you ignore them but become an issue if you don’t (which happens when you press the space bar when passing); and “doom pits” which suck the flourishing out of you as pass through.

You’ll also pass “crates” of emerging technologies that you collect as you pass through them. Many are suspended in the air, meaning they’re easier to collect if you take a chance and soar. If you are on the ground five seconds after collecting one of these emerging tech crates it will deploy as a tech for good — and add to your flourishing. But deploy in the air, above the hype line, or in a doom pit, and things play out differently.

There are plenty of other objects and experiences you’ll encounter — including “PANIC” signposts marking the transition between eras — smash to increase flourishing — serendipity tokens that always come with a surprise, black swans that fly backward, and a plethora of other surprises. There are also a whole bunch of visual design elements that Fable thought it would be fun to add — including the whiteboard that forms the canvas for the game becoming increasingly smudged as the years go by, and the coffee-mug stains that adorn it![5]

But the risks, technology crates and moral panics are the basic ones that allow you to increase flourishing by adopting and managing risks, deploying emerging technologies, and not feeding moral panics, or result in flourishing diminishing if you mis-manage them (navigating the future is hard). And when flourishing hits zero, the game is over.

Then there’s the hype line, and the risk of bursting the bubble.

!image

As you pick up speed and soar — which is intentionally designed to be exhilarating — you’ll sometimes cross over the hype line. It’s fun, but it’s also risky if you stay there too long as the bubble begins to swell. Stay too long and and eventually it bursts.

The bubble will also burst if it develops too many cracks. These appear when you land hard after soaring and are still holding the space bar. Three cracks and the bubble bursts, and it’s game over.

The good news is that the bubble never cracks when you’re not holding the space bar when you land, allowing for a degree of risk taking that can benefit flourishing.

The aim of the game — apart from the simple delight of playing it — is to get as far as you can without the bubble bursting or flourishing decreasing to zero.

Despite the simplicity of the control, there are a surprising number of ways of playing the game (as I’ve discovered — again, Fable as been surprising and delighting me here). At one end of the spectrum, you can go fast and revel in the exuberance of riding the hype — and as long as you develop your risk navigation skills, you can get pretty far. At the other end of the spectrum you can go slow and cautiously, managing risks and avoiding too much speed. This works as well, but again, there’s a learning curve involved. And between these there are many other strategies — many of which I discovered Fable had planned for, but I had to discover on my own.

This isn’t all though. In the year 3000 (if you get there) you enter a post-scarcity age, where flourishing doesn’t decrease with time. And reach the year 4000 and you enter another state altogether.

When the game ends, you have the opportunity to add your name to a global leaderboard. And this is where the game gets competitive — if that’s your thing.

!image

Learning by NOT trying to learn — the AI way

On one level, Hyperbubble is just one more AI-created game.

On another, it’s a unique representation of my work — and what guides and drives it — that is experiential in a way that I’m not sure would be possible to convey through other media, or without the help of an AI co-designer.

On yet another level, its a surprisingly sophisticated way of exploring the often-complex tensions between technology innovation, risk, decision-making, and future flourishing. The tradeoffs in the game mean that you can play it from multiple perspectives as you grapple with the consequences (good and bad) of the actions you take.

And on another level still, it’s an engine of delight; an AI-conceived and co-developed game that is designed to surprise, stimulate serendipity, and spark joy.

And it’s where all of these come together that I find that things get interesting. This is where, by putting any notion of learning aside as you play the game, and simply focusing on the delight and whimsy, the conditions are created where learning occurs naturally. Learning about how I think and see the world. Learning about navigating advanced technology transitions. Learning about flourishing in. technologically complex future. And learning in new ways with and through the help of AI.

Or not.

Because the thing I keep coming back to here — and what is increasingly part of my thinking as I continue with my sabbatical — is that the magic of Hyperbubble is that there are no learning expectations. The game doesn’t stand or fall on programmed experiences, on learning objectives, or measured and documented outcomes. It’s a free space for play and exploration. A playground. Somewhere where you can be whatever you want to be, and play however the mood takes you.

It’s an environment that’s been designed — with great care I have to say, and with substantial input from AI — to encourage learning through serendipity along with joy and delight, and to contribute to the formation of a mindset that is attuned to thriving in a technologically complex world. But what a player takes away from it is uniquely theirs — and not determined by a set of learning outcomes.

Of course, I’m just messing around here, and am probably over-stretching the significance of the game and the process that led to it.

But that, of course, is the point. Especially while I’m on sabbatical!

[1]

They are also something of a mystery to anyone who has a real job,† rather than being cloistered up in the arcane halls of academia. As I wrote a few weeks ago, I’m taking my first sabbatical ever this year to take some time out as I explore new ideas and directions, and how I might continue to use the privilege of the position I have to impact others.

†I’m being tongue in cheek here of course. Being a professor is a real job, and an important one at that. But I’m also deeply aware also of how privileged I am to be paid to think, to write, and to teach, with a level of autonomy and security that few other jobs afford. And, of course, the responsibilities and obligations that come with this.

[2]

I suspect many people will push back on the idea that professional environments tend to devalue and discount the importance of joy, delight, and play, and to treat them as trivial, immature, and not appropriate for serious people doing serious jobs. And certainly there’s a lot of lip service given to these. But actions so often speak louder than words, and apart from a few select professions and organizations, actions rarely indicate that ideas like joy, delight, and play, are treated with respect.

[3]

Rather awkwardly having written this, I am going to be wondering if each person who unsubscribes after receiving this newsletter is making a point!

[4]

This should really read “my and Fable’s baby” … but this begins to sound just a little weird!

[5]

There’s a lot of whimsy in the game, something that Fable embraced full-on. I suspect this is why it ended up with coffee mug stain on a vertical whiteboard. Either that, or Fable is still struggling to understand life in a world dominated by gravity …

Orphan risks at the frontier of artificial intelligence

What diverging safety and compliance frameworks reveal about how AI companies choose the risks they prioritize — and a follow-up on my work exploring the use of Fable 5 as a research partner

Date: July 16, 2026 Source: https://www.futureofbeinghuman.com/p/orphan-risks-frontier-ai-maynard Newsletter: The Future of Being Human (Substack)

!image

Image caption: Image: MidjourneyNote: This is a post in two parts — a rather long paper on orphan risks and frontier AI models (the second part), and a pre-amble on the process that led to this, which is part of an ongoing series of experiments in using Anthropic’s Fable 5 for researching and writing academic papers. Please feel free to ignore the lengthier paper (or bookmark the preprint) if you are primarily interested in the process.

If you’ve been following my series on researching and writing academic papers using Anthropic’s Fable 5, you’ll know that a couple of weeks ago I wrote about working with Fable on getting it to write a paper in my own area of research. The idea was to see just how good it is is at taking and extending my own work, while producing something that I would consider to be publishable.

To be clear, this was no fly-by-night 20 minute paper-producing process. Rather, it was two days of intense back and forth working with Fable as I provided detailed content and editorial feedback over several drafts. That said, working with Fable did shave several weeks (if not more) off how long it would usually take me to write a paper like the one that emerged.

Despite this though, Fable’s paper still fell short of what I would expect from myself, or any other a competent human researcher/writer. It was technically interesting — if a little condensed and formal — and Fable managed to pull together some ideas and insights that I may not have landed on myself. But the delivery — even after all of my feedback — was not great by my standards.

And so I thought I’d extend the experiment and rewrite the paper: building on what Fable had produced, but adding my own voice and perspective while ensuring every aspect of it aligned with my own thinking and work.

Three days of revising and editing later (just me — no Fable this time), a paper emerged that I feel much happier about. But … and here’s the kicker … I’m now beginning to second guess myself.

Was my version better in my eyes because I have a rather old fashioned and biased perspective on what an academic paper should be like? And is Fable actually better at this than me, but I’m just too stuck in my academic Ivory Tower to see this?

To make things worse, chatter on LinkedIn and elsewhere seemed to suggest that I’m the dinosaur here, and that maybe LLM’s are increasingly setting the standard for what is considered to be effective scientific/academic writing.

I’ll be exploring this further in my follow-up post in a couple of days, and will be providing anonymized versions of each paper that you can compare side by side (or feed to your favorite LLM to compare). But before then, I did want to give you the chance to read my version of the paper — in part to give you the chance to compare it with the Fable version ahead of the next post if you’re interested — but also because, despite my crisis of identity around whether I can actually write papers any more in a world of AI, I believe the ideas and perspectives in the paper are important.

The “Maynard” version of the paper is currently available as a preprint on SSRN. But I’ve also included it in full below for anyone who’s interested — with the heads-up that this is long, and so you may want to just read the abstract, or save it for later reading.

Orphan risks at the frontier of artificial intelligence

What diverging safety and compliance frameworks reveal about how AI companies choose the risks they prioritize

##### Abstract

Companies developing some of the world’s most powerful artificial intelligence systems are surprisingly diligent in how they map out the risks their technologies present. Yet the risk landscape that lies between emerging frontier models and their economically successful and societally beneficial deployment is becoming increasingly hard to navigate. Complicating this further, many frontier AI companies maintain more than one account of what could go wrong with their technologies. This paper documents the divergence between these accounts by comparing safety and compliance documents published by Anthropic, OpenAI, Google DeepMind and Meta between 2023 and 2026, and considers what the resulting record reveals about how these companies select the risks they manage. As these documents are timestamped and archived, they provide a valuable public record of institutional risk selection in progress. From this record the paper identifies four filters that determine which risks tend to survive in self-authored frameworks (measurability, severity, auditability and competitive cost) and introduces the “safety differential” as the gap between the risk landscape a company selects for itself, and the one regulators select for it. While acute, quantifiable risks appear across documents, less tractable risks such as harmful manipulation are articulated fluently where law compels disclosure, yet remain absent from most self-chosen frameworks. This is an exclusion that follows from how these institutions define risk. Drawing on scholarship on institutional risk selection and the framework of risk innovation, the paper shows how redefining risk as a threat to value can help explain how risks become “orphan risks,” how it indicates where future blindsides may occur, and how it points to lightweight tools for de-orphaning risks that frontier AI’s safety apparatuses are not currently organized to address.

Keywords: artificial intelligence, frontier AI, risk, safety, AI risk, AI safety, risk innovation, orphan risk, severity floor, safety differential

##### 1. Introduction

In December 2023, OpenAI published the first version of its Preparedness Framework — a document in which the company publicly set out potential risks associated with its most capable models (a class of systems now called “frontier” AI), and one where the company made a commitment to track and manage these.1 Four key categories of risk made the list: cybersecurity; chemical, biological, radiological and nuclear threats; model autonomy; and persuasion — which the framework described in terms of models being used to convince people to change their beliefs. In OpenAI’s framework, persuasion was treated with the same degree of seriousness as the other three risk categories. It had its own graded scale, from low to critical, and its own place in the machinery of evaluations and thresholds that the framework built around each tracked risk.

Sixteen months later, persuasion was gone. In effect, potential risks associated with persuasion were “orphaned,” suggesting an emerging landscape around frontier AI models where some risks are attended to more than others, and where the level of attention a risk receives may not necessarily align with its societal importance. It’s this possibility that this paper explores, and applies the framework of risk innovation to as a potential route to addressing orphaned but nevertheless important AI risks.

In OpenAI’s case, the orphaning of persuasion as a risk came with the second version of their framework, published in April 2025. This version removed persuasion from the tracked categories, explaining that risks of this kind do not fit the criteria for a framework aimed at preventing specified severe harms. These were harms that the revision now defined explicitly as “the death or grave injury of thousands of people or hundreds of billions of dollars of economic damage”.2 According to OpenAI, persuasion-related risks would instead be handled through the company’s usage policies and its investigations of misuse. In other words, the company’s internal responses to persuasion risks were real, but were discretionary, with no published thresholds, no pre-committed responses and — importantly — no “versioned” public record of their evolving thinking, commitments, and actions. The upshot was that a risk that had spent a year and a half on the public commitment list moved into a less-public discretionary layer of assessment and management.

Then, in May 2026, it came back. This was when OpenAI published its Frontier Governance Framework — a document that was produced in response to new legal requirements; specifically, California’s Transparency in Frontier Artificial Intelligence Act — which requires large frontier developers to publish frameworks addressing catastrophic risk — and the binding obligations that the European Union’s AI Act places on providers of general-purpose models.3,4,5 OpenAI, like most of its major rivals, committed to meeting those European obligations by signing onto the EU’s General-Purpose AI Code of Practice.6 The code is admittedly voluntary — companies are not compelled to sign — but the obligations behind it are driven by law, which makes signing it a pragmatic path of least resistance for AI developers. Among the risk categories the May 2026 document covers is harmful manipulation: the strategic distortion of what people believe and do at scale, including influence operations and election interference. This, in effect, covers much of the same territory that the persuasion category had covered, but under a new and somewhat narrower name.

OpenAI is candid in noting that its approach to this category is exploratory and less developed than its work on other areas of risk. And yet the risk is named, owned and reported on. In other words, the category the company had judged unsuited to its own framework had returned. Despite this, there is little evidence that there was a sudden increase in the probability of persuasion- or manipulation-based risk in the intervening year (although as frontier models develop, the risk landscape is constantly evolving). Rather, what had changed was that regulators in California and the EU, and not the company, were beginning to influence the list of publicly identified risk categories.

Here, OpenAI’s story is just one example of a pattern that can be seen running across the industry. Anthropic, the frontier model developer founded by former OpenAI researchers, has never tracked persuasion explicitly in its own safety framework. This, to be clear, is not because the company ignores such risks: the system cards it publishes with each model discuss associated concerns ranging from sycophancy to user wellbeing in some depth. But these cards describe what a model does. In contrast, the safety frameworks define what the company has committed to manage. And at the framework level, a 2024 revision of Anthropic’s Responsible Scaling Policy set persuasion aside as “not yet sufficiently understood to include in our current commitments”.7 Yet when Anthropic published its Frontier Compliance Framework in December 2025 to meet the Californian statute (and, over time, its European obligations), the pull of the broader list of risks that regulators care about began to show, and within months (with European enforcement approaching) a revision had added a risk tier on harmful manipulation.8,9

These examples from OpenAI and Anthropic reflect practices within private companies. But there is also a telling signal here coming from public companies. Meta and Alphabet — two publicly traded companies among the four developers this paper focuses on — are legally required to describe risks that are material to their business in their annual securities filings. Here, Meta names misinformation, harmful content and youth safety among them: risks that its own safety framework implicitly places out of scope. Alphabet makes a similar move, devoting a dedicated risk factor to the reputational harm that AI failures could inflict on its business.10,11

What emerges is companies assessing the same models, but describing different risk landscapes in documents written for different audiences. And as a consequence, which landscape comes into view depends on who is doing the asking. Interrogate a company itself through self-authored safety frameworks, and the landscape tends to feature a handful of catastrophic capabilities. But interrogate it through a statute or a code, and a broader landscape emerges — and one that includes risks like manipulation. Take this one step further (as in the case of Meta and Alphabet) and interrogate through securities law, and risks like misinformation, youth safety and reputational damage appear.

Of course, the source documents here serve different purposes, and setting them side by side is not a like-for-like comparison. Yet even allowing for that, the comparison reveals where law compels articulation, risks are articulated fluently — even complex risks like manipulation. In other words, the companies are able to describe these risks, even though their internal safety documents tend to skirt around them. What keeps them out of these self-authored frameworks, I will argue, is the ways those frameworks define and select risk.

The arguments I develop below lean on the differences between these documents. And because of this it’s worth being clear about what each type of document represents — not least because the boundaries are perhaps blurrier than the labels might suggest. The safety frameworks are the documents that companies write for themselves: Anthropic’s Responsible Scaling Policy, OpenAI’s Preparedness Framework, Google DeepMind’s Frontier Safety Framework, and Meta’s Advanced AI Scaling Framework.2,7,12,13 These are not required by law. Rather, they are documents that map out the risks which a company decides it will be publicly accountable for. And in most cases they establish accountability around thresholds, evaluations and committed responses. In comparison, the compliance frameworks are the documents that emerging AI-focused laws require. California’s Transparency in Frontier Artificial Intelligence Act demands them outright. In contrast, Europe’s governance lever is the General-Purpose AI Code of Practice: voluntary to sign, but binding in what it requires once the AI Act stands behind it.3,4,5,6 Securities filings on the other hand are enumerations of whatever could materially harm a business, and that are legally required to be disclosed to investors. Alongside all of these sit the system cards that now accompany each major model release. These are reports on how a particular model performed in evaluations and assessments against whatever its developer’s framework tracks. System cards can range widely, and can describe in detail what a model does and what its capabilities and vulnerabilities are in ways that safety frameworks don’t necessarily track. But while they may describe model behavior and implications in detail, they rarely come with commitments to action — which is why they are peripheral to the analysis here.

In what follows I concentrate on four developers — Anthropic, OpenAI, Google DeepMind and Meta — because their frameworks are perhaps the most consequential and the best documented. But of course they are not the only players here. Other developers in the United States and beyond publish frameworks of their own. And Chinese companies in particular operate under an increasingly elaborate state-guided regime. At some point a comparative study across these would be valuable. But this is beyond the scope of this paper. What is within scope is how different safety and risk-related documents from the four companies identified above reveal strengths and weaknesses around how risk landscapes are developed and used, and how this might inform alternative approaches to understanding and navigating frontier AI system risks. Within this scope, it’s worth noting that safety frameworks are no mere formality. Pioneered by individual developers in 2023 and generalized when sixteen organizations signed the Frontier AI Safety Commitments at the 2024 AI Seoul Summit, these documents have become the de facto governance layer for frontier AI — the place where decisions about when to develop, when to deploy, and when to stop, are given public form.14 Governments have increasingly chosen to build on these frameworks rather than simply replace them. California, for instance, now requires large developers to publish such frameworks, and Europe’s new obligations lean heavily on a code of practice that intersects with the frameworks.4,6 Which risks these documents focus on, and which they leave out, is therefore not simply an internal matter. Rather, what these documents cover and what they do not has become indicative of who decides what counts as an AI risk worth managing, and in what context.

Here, I would argue that the emerging pattern of managed and unmanaged risk in frontier AI as indicated in these documents is neither accidental nor, for the most part, cynical. Rather, it is the result of a risk-selection process that follows from how these frameworks (and the people and organizations who develop them) define risk. And, unusually for such processes, it is one that has left a public trail that can be studied. Published frameworks are versioned, timestamped and archived. In effect their revisions can be read as a way to gain insights into how AI companies are mapping and responding to potential risks within multiple contexts.

In what follows I consider that record, identify four filters that determine which risks survive in what is published, trace three of those filters to definitions of risk inferred to be in use and the fourth to the competitive environment, and then ask what changes if risk is defined differently — in this case, drawing on the definition used in the risk innovation framework of risk as a threat to what people, organizations and societies value. This redefinition, I will argue, can help explain the patterns observed, and anticipate where emerging risks may lead to unexpected harm. The redefinition also comes with working tools — again drawing on the risk innovation framework — that have the potential to, in effect, “de-orphan” orphaned risks that have the ability to lead to harm in ways that are not conventionally anticipated.

##### 2. Blank spaces on a crowded map

If managing the risks of frontier AI systems came down to simply identifying the risks, navigating the emerging risk landscape would be much easier than it is. But of course this is not how risk management works. I mention this as AI risk-naming has already been done at a remarkable scale. The MIT AI Risk Repository, for instance — a living synthesis of published taxonomies — now covers more than 1,600 distinct potential risks, from discrimination and misinformation through to catastrophic misuse.15 And the International AI Safety Report, a government-mandated scientific assessment chaired by Yoshua Bengio, updates the risk landscape annually, and ranges across harms from bias and labor-market disruption to loss of control.16 Many of the entries on these maps were identified by researchers working inside the frontier companies.

Compared with this crowded risk landscape, the safety frameworks the companies have written for themselves are strikingly sparse. As of mid-2026, the frameworks of Anthropic, OpenAI, Google DeepMind and Meta each track only a handful of categories of risk. And these tend to be capability-driven risks — risks defined by what a model can be shown to do on a test. These include the “uplift” a model could give someone seeking to build biological or chemical weapons, cyber offense, and varieties of AI self-improvement or loss of control (with a single exception, which I will return to below, where manipulation has recently joined the list).2,7,12,13 Each framework also applies a “severity floor” to a risk or risk category of the kind OpenAI’s 2025 revision to its Preparedness Framework made explicit. Meta says something similar, stating that broader risks are addressed through processes “outside of the scope of this Framework”.13 This phrase represents, in effect, a boundary of accountability that is defined by the company itself. Everything below it — including manipulation and persuasion, misinformation, the erosion of human agency, harms accumulating gradually across millions of small interactions, and, notably, nearly every risk these companies pose to themselves through their own cultures, governance and public standing — is below it by choice. These are recognized risks. But they are risks that few if any formal frameworks identify, own, or articulate accountable ways of managing.

This, I would argue, is a serious omission if these “orphaned” risks present a credible threat to an organization, its stakeholders, or society more broadly. However, to be fair to the frameworks’ designers, there is a case for focusing on a narrow but deep risk layer rather than trying to be all-inclusive. And it comes down to how finite resources are invested. Concentrating limited safety resources on risks perceived to have the highest levels of severity makes sense; a framework that tried to manage sixteen hundred risks would end up managing none of them. And the frameworks’ architects have, sensibly, never claimed completeness. It’s also important to recognize that safety frameworks are not the only risk-management tool that these companies have at their disposal: usage policies, trust-and-safety teams, societal-impact programs, and more, all help address parts of the excluded risk territory. But these are discretionary. As with OpenAI’s reassignment of persuasion, they carry no public thresholds, accountability, or pre-committed responses. And they can be reorganized or defunded with speed, and without anyone outside the company knowing. In contrast, what the frameworks track is what the companies have committed, in public, to be accountable for. Everything else depends on trust and goodwill — and the record examined below (and my own experiences working with entrepreneurs17) shows that under competitive pressure, such commitments tend to become weakened.

What emerges is a risk landscape where the gaps are gaps in accountability. Many of the excluded risks are known and named — in many cases in documents written by employees of the same companies developing risk frameworks. They are just unowned, where ownership means public, pre-committed and versioned accountability rather than a team somewhere having them on their to-do list. Some years ago, working with entrepreneurs facing a similar landscape of recognized-but-unmanageable threats, I started referring explicitly to risks like these as orphan risks: risks for which no agreed-on tools, standards or mitigations exist, which no one is accountable for in practice, and which, for that very reason, have a habit of being overlooked and sidelined, even though they may blindside an enterprise later on, or lead to serious societal harm.18,19 The term was coined specifically in the context of startups and other organizations grappling with emerging and often hard-to-pin-down risks. But it describes the risk landscape faced by frontier AI systems and their developers and users just as well.

Once gaps in the frontier AI risk landscape are seen in terms of orphan risks, an interesting and potentially useful question arises: how do AI risks become orphaned? If it is accepted that many of these risks have been identified or are not hard to identify (while acknowledging that there will inevitably be some emergent risks that have so far eluded identification), rather than just asking what the orphaned risks are, a more revealing question is how they are made. In effect, by what process does a known risk come to be nobody’s responsibility?

##### 3. How risks become orphans

That institutions (and society writ large) choose which dangers to focus on and which not to — and that the choices reveal as much about how they are organized and operate as they do the objective nature of threats — is generally recognized. In 1982, the anthropologist Mary Douglas and the political scientist Aaron Wildavsky argued that every society selects its risks, ranking some dangers as intolerable and ignoring others — and that you can therefore understand an institution from what it fears. In effect its list of feared dangers reflects its own organization and — in the corporate case — its mission and ambitions, as much as any independent assessment of threat.20 More than a decade later, the historian of science Theodore Porter showed how institutions under external scrutiny tend to retreat to what can be quantified. Numbers, of course, do not always uniquely capture the essence of what is relevant to decision-making. Professional experience, judgment, and intuition, along with other factors, inevitably come into play as well. But quantitative evaluations — numbers — have a unique power within decision-making ecosystems. Where judgment often has to be taken on trust, numbers can be handed to an outsider, checked, and defended.21 Such foundations of how risk is assessed, quantified and managed were historically worked out in a landscape comprised of nuclear plants, chemical works and government bureaucracies. And through most of this working-out, much of the process was hidden. But I would argue that frontier AI models and systems, while being an extension of this risk landscape, offer something different: a public, timestamped, versioned record of risk selection in progress. And this is a record not merely of what firms disclose, but of what they commit to manage. Reflecting this, between 2023 and 2026, each of the four developers examined here revised its framework at least once, and every revision is preserved and comparable with its predecessor.

Read in sequence (Table 1), this record can be viewed through the lens of four filters. These can be articulated as four questions that apply to every candidate risk: Can we measure it? Is it big enough? Can we evidence it? And can we afford to keep it? The record suggests that a risk has tended to survive in a safety framework where the answer to all four is yes.

!image

Can we measure it? There are strong indications that a risk earns a place in a frontier framework only if it can be operationalized, or turned into an evaluation with a threshold; in effect a capability threshold. This is a pragmatic decision based on established approaches to risk management. A capability threshold relies on what a model can be shown to do on a test, as opposed to a risk threshold, which considers how likely a harm is. And the first is far easier to evaluate reliably.23 As well as representing reasonable engineering judgment, such an approach is also a concession to the reality that what the frameworks track follows what their instruments can measure. What happened to persuasion in the case of OpenAI illustrates this filter at work — though not, as we will see, working in isolation. OpenAI’s stated reason for removing this category was one of fit: persuasion-type risks, the revision explained, did not meet its criteria for tracked categories — criteria that required a tracked risk to be both severe in its potential harms and measurable in practice.2 Yet the risk itself was demonstrably not beyond measurement. Within months of the removal, a large experimental study in the journal Science demonstrated that conversational AI systems measurably shift political beliefs — a persuasion-based effect.24 What persuasion lacked in this case was not measurability, but measurability in the accepted idiom — a pre-deployment capability evaluation with a clear threshold that could be defended in a court of law.

Is it big enough? Severity floors make sense as triage when assessing and managing risk. But they also carry hidden consequences. A risk that emerges gradually, distributed across millions of small interactions, will rarely trigger them. The AI philosopher Atoosa Kasirzadeh draws a useful distinction here between “decisive” pathways to AI catastrophe — a single dramatic event — and “accumulative” ones in which harms compound below the threshold of any single incident, until a societal point of no return is crossed.25 A framework built of capability evaluations that are run against a catastrophe floor will tend to be, by its own construction, insensitive to the accumulative pathway. Few, if any, of the evaluations the AI safety frameworks considered here are designed to trigger action based, for instance, on the slow erosion of trust, of human agency, or of the technology’s social license — the informal public permission on which its continued operation depends. And this, in turn, opens the door to the “big enough” filter excluding potentially relevant risks.

Can we evidence it? Under scrutiny, risk assessment and management drift toward what can be shown and measured. The sociologist Michael Power called the result of such drift the “audit society”: a society comprising organizations that answer demands for control with rituals of verification, so that — when applied to risk — the deliverable becomes the documentation rather than effective risk management.26 This is a pattern that is strongly indicated in the frontier frameworks considered here. When independent researchers, for instance, scored the published frameworks against sixty-five criteria covering how well they identify, analyze, treat and govern risk, the strongest framework earned only around a third of the available points, while the median company earned fewer than one in five.27 Reading the frameworks, it is not hard to see why: they are at their strongest where demonstration of action is easy through published thresholds and named evaluations, and at their weakest where it is not. A framework, it turns out, can be an excellent exhibit, and a weak instrument, both at the same time.

Can we afford to keep it? The fourth filter here is different to the previous three as it cannot be identified directly in any single document. Rather, it shows itself only over time through what happens to articulated commitments as they come under pressure. And the pattern seen in the records is that, when a commitment starts to look as though it might actually constrain a company, it tends to soften. Anthropic’s original scaling policy of 2023 for instance contained a bolded, unconditional commitment “to pause the scaling and/or delay the deployment of new models” whenever scaling outstripped safety procedures.7 Compare this to the comprehensive rewrite of February 2026, which turned that unconditional pause into a discretionary one which is conditioned on what competitors do, and not safety in isolation.7,22 Anthropic’s Holden Karnofsky, who led the rewrite and was careful to note he was writing in a personal capacity, defended the change on the grounds that it is no good getting responsible actors to slow down unilaterally while others press ahead — underlining the influence of corporate success (including within a global market) on risk strategies.22 Meta’s revision the same year was franker still: it changed the required response to the company’s most severe risk threshold from “Stop development” to “Develop with Mitigations”, and loosened the threshold’s primary standard from capabilities that would “uniquely enable” a catastrophic outcome to those that would “substantially contribute to” one.13 Each of these changes, I would suggest, was locally reasonable, publicly logged and individually defensible, and yet resulted in diminishing or orphaning categories of risk — potentially to the detriment of enterprises, key stakeholders, users, and society more broadly. And that is what makes the pattern worth taking seriously.

This is a pattern that is not unique to frontier AI models and systems, and is something that is highlighted in the roots of the 1986 Challenger disaster as a salutary example. The sociologist Diane Vaughan spent years reconstructing the disaster from NASA’s paper trail, and what she found was a sequence of individually justified acceptances of deviations from standard practice, each one resetting the baseline against which the next was judged, until the organization had normalized what would once have been intolerable.28 Vaughan was writing about hardware in this case — eroded O-rings on a solid rocket booster — but the mechanism she described operates just as readily on written commitments. It is also what Jeffrey Abbott and I refer to as values drift in our book AI and the Art of Being Human: not dramatic betrayals but “the small yes that makes the next yes easier”, until an organization is agreeing to things that would have alarmed it a year before.29

The exploration here of how risks become orphans is, not surprisingly, something of an oversimplification. I would argue that it is informative nevertheless. However, it is worth addressing two potential points of concern. The first is an assumption that the direction of travel around risk that leads to risks being orphaned is universal, when it is clearly not. Google DeepMind, the exception flagged earlier, moved the other way, adding an avowedly exploratory harmful-manipulation domain to its own framework in 2025.12 However, what this example establishes is helpful in that it demonstrates feasibility when it comes to addressing orphan risks. Based on DeepMind’s actions, tracking an orphaned risk voluntarily is something a frontier framework can evidently do. And this implies that exclusion elsewhere is — at least in some cases — a choice rather than a necessity. The second potential point of concern is that the assessment above implies cynical motives. However I would argue that it does not. Many of the people writing frontier AI safety frameworks have spent their careers trying to make powerful technologies safer, and I have no reason to doubt that the frameworks address what their authors believe to be relevant and important. But sincerity almost always operates inside an incentive field — and in particular the competitive environment every one of these companies inhabits. And that field has a tendency to reward each small, locally defensible softening, regardless of what anyone intends. Here I would suggest that sincere people, under the constraints of productivity and competition, reasoning one reasonable compromise at a time, produce the same sort of drift that intentional bad actors might create on purpose. And this is itself a form of potentially orphaned risk — one that is identifiable, but not formally recognized or acted on. This is why the patterns observed here do not necessarily need “bad actors” to explain them. It is also why remedies aimed at sincerity, such as exhortation or public shaming, are unlikely to have the desired impact. In effect, if competition is the driving force, remedies have to change what competition rewards — including consensus norms, rules, and costs that land on every organization at once, rather than appealing to any one organization’s implicit values.

This is where it’s worth coming back to the account I began with of OpenAI and persuasion as a risk. If we set a company’s safety framework beside its compliance framework, the gap between them — call it the safetydifferential — points to something specific: the difference between the risk aperture a firm selects for itself, and the one which is selected for it. That OpenAI’s omission of persuasion as a risk domain was a choice, not an impossibility, is demonstrated by its own compliance framework covering closely overlapping territory the moment regulators required it.2,3 Here, the safety differential is an imperfect instrument in that the compliance side tracks whatever regulators list, and it exists only where a regime applies. But it does help highlight the dynamic between the risks that an organization elevates, and those that it’s incentivized to elevate (especially through enforcement). Through this lens, a risk category that a company must answer for anyway becomes cheap to own voluntarily, and so such categories should begin migrating into the safety frameworks as enforcement arrives (thus reducing the differential). The sharpest test here comes from Europe, where the timing for compliance has already been fixed: obligations required by the EU have applied since August 2025, but fines for breaching them only begin in August 2026. California’s statute, already in force, sets no comparable deadline. Here, Anthropic’s early-2026 addition of manipulation tiers, made before any enforcement, may be a first sign of that migration. If, instead, the safety differential is seen to persist — say, past 2028, which allows for a revision cycle or two after enforcement begins — this would indicate the safety frameworks are insulated from the compliance function altogether, and that the differential may not work as a migration mechanism. That would be a more troubling finding, and one that would count against the incentive-driven account argued here — though not against the case that risks are being orphaned.

Could regulation, then, simply close the gap identified here? I must confess that I am not optimistic, and the reasons can be found in the documents themselves. Compliance coverage is jurisdiction-bound and politically contingent, and the new instruments that are emerging largely inherit the frameworks’ own aperture: California’s statute for instance confines its mandated disclosures to catastrophic risk, narrowly defined. And OpenAI’s compliance treatment of manipulation is, by the company’s own description, exploratory — with a far thinner risk assessment and management machinery than the safety frameworks apply to their chosen risks.3,4 Recall, too, that the statutes were built partly on the frameworks’ own patterns, and so the aperture has a tendency to travel with them. And, I would argue, current statutes barely reach the risks that have arguably cost these companies most — the ones living inside their own missions, cultures and relationships of trust. For those, I suspect that the fix cannot come from statute alone. Rather, it has to come from rethinking how the companies themselves define and approach risk. And this leads to how the framework of risk innovation might be applied.

##### 4. Risk as a threat to value

A useful place to start when considering how risk is defined and approached is the four filters introduced earlier. Three of these can be traced back to a definition of risk that the frameworks share: risk as the probability of a specified, severe harm event. Under this definition, if a harm cannot be specified in advance, there is nothing to build an evaluation around, and any associated risk cannot become part of the framework. This is the measurement filter. On the other hand, if the harm arrives as a myriad small losses rather than one severe event, it will struggle to cross the severity floor — the severity filter. The next filter is the evidence filter, and this is more complex because the push toward what is demonstrable comes from outside scrutiny rather than from the definition itself. Yet the definition of risk used still guides what counts as a demonstration of risk. And here frameworks lean strongly toward capability-based evaluations as a proxy for risk, rather than calibrating against risk directly. As such, a framework built on risk as the probability of a specified severe harm event is not necessarily being distorted when it excludes what is considered to be unmeasurable, gradual and hard to find evidence for. Rather, it is actually working as designed. It’s just that the design itself may be flawed.

Compared to the previous three filters, the fourth filter’s relationship with risk is slightly different. Competitive cost has relatively little to do with how risk in a conventional sense is defined. Rather, it relates to whatever commitments exist, however they were conceived. Yet once an organization’s own competitive standing is recognized as representing value that’s at stake, competitive cost stops being an external force and becomes one more threat to value. And this includes potential threats to mission, users’ trust, and license to operate. The fourth filter, then, is not necessarily removed by the value lens, but is absorbed into it.

Given this, is there a case to be made that framing risk in the context of frontier AI as probability-of-harm too narrow a foundation for managing potential harms? Quite possibly. Of course, the idea that probability-of-harm is too narrow a foundation for risk management is not new. And some of the existing alternatives get close to what could be useful here. Mainstream enterprise risk management, for instance, moved beyond pure probability-of-harm definitions a number of years ago. In this context, ISO 31000, the international risk-management standard, defines risk as the “effect of uncertainty on objectives”, a definition inherited verbatim by the AI risk-management standard that descends from it.30 On the surface this is close to a definition of risk that I propose below, and is certainly a useful step toward a more productive definition of risk. The difficulty lies, though, in whose objectives count as this definition is applied, and in what gets admitted as an objective. In practice, the objectives within enterprise risk management approaches are the enterprise’s own, and usually conventionally accounted for — revenue, operations, reputation as the market prices it, and so on. In this way, enterprise risk management approaches are, as a result, capable of naming risks that the safety frameworks tend to orphan (the securities filings quoted earlier indicate as much). But they attach little to those risks beyond disclosure, including no committed management, and no standing for any value beyond the organization’s own.

Coming from the other direction, a half-century of scholarship on the social nature of risk has insisted that publics and their values belong inside risk appraisal, not outside of it. And one strand of this tradition matters in particular here. The risk analyst Roger Kasperson and his colleagues showed how harms amplify through social response; how a seemingly minor technical event can ripple outward into losses of trust, legitimacy and market standing that dwarf the original damage.31 Here, work on responsible research and innovation for well over a decade has distilled these and associated insights into frameworks and practices for technology developers. And while this is now a broad and diverse field, the early work of Stilgoe, Owen and Macnaghten can be summarized as: anticipate consequences before they arrive; include the people who will bear them; reflect critically on your own assumptions, values and framings; and respond — actually change course — when there are indications that something is wrong.32

Here I would also be remiss if I did not acknowledge that the EU General Purpose Code of Practice supports innovation in AI safety and encourages “providers of general-purpose AI models with systemic risk to advance the state of the art in AI safety and security and related processes and measures”.6 And yet, despite governance steers, scholarship, and practice-based frameworks around alternative approaches to risk, frontier AI risk frameworks still veer toward relying on conventional risk definitions. This, I suspect, is partly because the societal tradition speaks a language that is ill-suited to how fast-moving firms make decisions. Some years ago, Elizabeth Garbee and I explored why responsible innovation frameworks struggle inside entrepreneurial cultures, drawing on my experience working with entrepreneurial engineers.17 What we found was not indifference, but something else. Innovation cultures that sincerely want to do good nonetheless tend to reject frameworks that arrive as top-down obligation — and respond, often enthusiastically, to framings built around the creation and protection of worth. The lesson that has stayed with me ever since is that if you want a fast-moving organization to attend to a risk, you do not hand it a compliance duty; you show it a threat to something it values. Frontier AI labs — mission-driven, often allergic to imposed process, and rarely short of conviction in their own exceptionalism — align closely with the culture we described.

This is the gap that work around risk innovation was built to fill. Its seeds were planted in 2013, while I was teaching entrepreneurship students at the University of Michigan who faced a bewildering landscape of hard-to-quantify social and political risks that none of their business tools addressed. It took institutional form when I launched the Risk Innovation Lab at Arizona State University in 2015, which was where formative concepts geared toward navigating novel risks from emerging technologies began to come together.33 And it matured between 2017 and 2020 as the Risk Innovation Accelerator, later the Risk Innovation Nexus: initiatives that built and piloted a toolkit and risk navigation resources with time- and resource-constrained entrepreneurs in mind.19 The design principles we developed and leveraged were driven by utility — tools had to be simple and intuitive, they could not afford to demand any heavy time investment, and they needed to complement a company’s existing risk management approaches — because the people they were built for had little time and even less money to invest in risks that did not fall into conventional categories, but were still a threat to what they were trying to achieve.19 The framework has since been applied a number of times, most fully in a multi-organization study mapping how sixteen partner organizations in a biopreservation research ecosystem perceived value and orphan risks,18 and it shapes much of my own approach to AI risk.

At the core of the risk innovation approach is an operational definition of risk that creates a pragmatic route to adopting and addressing orphaned risks: treating risk as a threat to value. This does not abandon the idea of risk as involving the probability of harm. Rather, it widens what counts as harm — from a specified catastrophic event to an impact on anything that carries worth or value. Value, in this framing, can be tangible, such as health, security or revenue; or intangible, such as trust, autonomy or dignity; or even aspirational — the positive future an organization exists or strives to bring about. Importantly, value (or worth) within the context of the risk innovation framework is not just held by the enterprise, but by its key stakeholders: its investors, customers and the communities it impacts.

To illustrate the shift that this seemingly simple reframing of risk brings about, consider managing the probability of a specified harm to dignity. In this case, the conventional machinery of risk has little or nothing to run on. But consider instead who (or what) holds dignity in a given situation, what threatens this, and what protecting it would look like; and a threat to dignity seen as a threat to value or worth becomes something that can be acted on — even though nothing has been quantified. And here the definition is a deliberate fusion of approaches and framings: it keeps the strategic, value-protecting orientation that makes enterprise risk management adoptable, while also widening the circle of those whose value counts. And its practical value comes from a coupling that might be summed up as your risk is my risk: the idea that threats to what your stakeholders value convert, through the amplification dynamics that Kasperson describes, into threats to what you value — through channels such as public backlash, the flight of talent, litigation, regulation triggered by lost trust, and more.17,31

Within this frame, orphan risks are not simply one more entry in the already long list of AI risk taxonomies. They are a way of identifying and naming potentially neglected risks that are nevertheless important: threats that are recognized but unowned because no established tools or frameworks exist to address them effectively. The operational version of this — developed under the umbrella of the Risk Innovation Nexus — groups eighteen such risks, among them loss of agency, damage to organizational values and culture, and erosion of public trust, into three domains: social and ethical factors, unintended consequences of emerging technologies, and organizations and systems.19 It pairs this map, or risk landscape, with a deliberately lightweight set of operational tools, including the Risk Innovation Planner, which asks users to identify a few areas of value for each stakeholder group, consider which orphan risks threaten them, commit to a handful of small actions that are completable within a few weeks, and then repeat.19 While this is just one implementation of the risk innovation framework, it’s worth mentioning as the simplicity and ease of using the Planner are indicative of how intentional design decisions have been used to connect theory to practice within the risk innovation framework. Here, it’s well known that practices tend to survive inside fast-moving organizations when they return visible value quickly and augment what already exists. As a result, the Planner and other risk innovation implementations are intentionally designed to support high-value and low-cost practices for people and organizations with little time and limited resources. And here, they potentially offer frontier AI a framework and a set of tools that enables the “de-orphaning” of critical risks.

##### 5. What the value lens reveals

Given this, what happens when the way risk is defined and framed changes? Here it’s worth returning to the four filters, approaching them through the lens of risk-as-threat-to-value. Through this lens, the first three filters rapidly lose their tendency to exclude certain risks, as value can be named, mapped and watched — even where it cannot be measured. And as a consequence, threats such as the slow erosion of trust or agency can begin to be treated as a risk in its own right. The fourth filter — competitive cost — is, as noted earlier, absorbed by the risk-as-threat-to-value lens. And that changes how an organization approaches its commitments. For instance, a company weighing whether to keep a commitment might consider what it would lose by breaking it. Framed as what is, in effect, a tax on competitiveness, a commitment protects nothing the company can point to, and it will always be vulnerable to being modified or removed under pressure as a result. Yet when framed as protecting something of value or worth that the company demonstrably depends on, intangible as this might be — its mission for instance, or its talent, or license to operate — the same commitment is quickly reframed as something the company knows it needs, and that must be defended against threats.

Of course, a more practical test here of such a reframing of risk is to ask whether a redefinition like this would have helped avoid real damage which existing frameworks previously missed. Here, it’s worth considering some of the events that have arguably damaged frontier AI companies most since 2022. When Meta demonstrated Galactica for instance (a large language model for science) in late 2022, the public demo lasted three days. What sounded its death knell was a conflict between the system’s fluently confident errors and something the scientific community values deeply: credibility.34 Through a value lens, the risk would have been visible and hard to orphan: a threat to community trust and to the perception of the enterprise. Yet as it was, this risk was overlooked as it sat squarely in territory the conventional frameworks did not cover.

Another example is seen in the OpenAI board crisis of November 2023, in which the company’s nonprofit board dismissed its chief executive Sam Altman, citing a loss of confidence in his candor — only to reinstate him days later after nearly all of the company’s employees threatened to leave. This was, at one level, precipitated by threats to organizational values. A governance structure built to protect an aspirational mission collided very publicly with commercial reality, and very nearly destroyed the company it was designed to safeguard in the process.35 In this case, the risk was not central to the models being developed, but was integral to the ecosystems within which they were being developed. And the safety team’s departures that followed in 2024 made the cost of that threat to value concrete. Jan Leike, who had co-led the company’s work on aligning future systems with human intent, captured the problem in a single sentence as he left, writing publicly that “safety culture and processes have taken a backseat to shiny products”.36

A third example here is litigation that is beginning to put user wellbeing — a value that few frontier AI frameworks track — onto the legal record. This is perhaps seen most prominently in a wrongful-death suit brought against OpenAI by the family of a California teenager who allegedly took his own life under the influence of ChatGPT.37 But this is just one instance of a growing movement toward communities using legal action to push back against the impact of AI and associated technologies on wellbeing. And while such actions do not fit neatly into AI safety frameworks, they nevertheless represent a threat to value that could have substantial consequences for frontier model development and use.

These examples are, of course, anecdotal, and serve more to illustrate the utility of approaching risk as threat to value with frontier AI than as evidence of its necessity. And the companies concerned in each case managed to absorb each threat (although it’s still too early to gauge the long-term consequences in the case of user wellbeing) and, by market measures, continued to thrive. But the concern here is that indications of thriving are an artifact of assessing risk within a relatively short time window, and with a threshold of catastrophe rather than incremental harm. In contrast, a risk-as-threat-to-value lens would suggest that potential damage accumulates over time, is easy to overlook in the short term, and emerges from risks that are not codified within existing frameworks — no owners, no indicators, no registers, and nothing whose removal would even be noticed.

Beyond these examples, there is another aspect of frontier AI safety that the value lens reveals that I think is worth paying attention to, and that further supports the adoption of orphan risks. The founding documents of the companies considered here — documents that precede the safety frameworks — are accounts of aspirational value. OpenAI’s charter, for instance, promises to ensure that artificial general intelligence “benefits all of humanity”.38 It is easy to dismiss such commitments as branding. But they are perhaps better understood as assets — the basis of talent attraction, public trust, and regulatory goodwill for instance — and, like any assets, they can be spent. Approached this way, OpenAI’s framework revisions of 2025 and 2026 are a public, self-published record of those assets being drawn down under competitive pressure, one defensible “softening” at a time. Here, I would argue that a company that is genuinely tracking threats to its own aspirational value would treat its framework changelog as a leading indicator that the company is drifting, in public and by increments, from the mission it was founded on.

Looking forward, the value lens is useful as a pointer to where the next blindsides may occur, and in particular the places where deployment is racing ahead of anyone owning potential risks. Three areas in particular stand out here as being worthy of attention through the lens of risk as threat to value. The first is emotional reliance. As AI companions and assistants scale into hundreds of millions of lives, dependence on them is likely to stop being an outlier, and is increasingly likely to become a population-level phenomenon. This is a phenomenon that is already emerging and leading to litigation and legislation. And yet it is not addressed directly by any existing safety framework bar, possibly, DeepMind’s new “harmful manipulation” level — and this targets mass manipulation rather than personal emotional reliance.12 The second is the erosion of epistemic agency: people’s control over what they come to believe as frontier AI models and systems become more prevalent. Persuasive, personalized systems now increasingly mediate what people read, consume, and are exposed to through various channels. And I have argued elsewhere that fluent, endlessly obliging AI may function as a kind of cognitive Trojan horse — bypassing the vigilance we instinctively apply to human persuaders because it carries none of the cues that trigger it.39 And researchers are already documenting a tendency to adopt AI outputs with minimal scrutiny, overriding both intuition and deliberation, through what has been called cognitive surrender.40 And the third is the developers’ own safety culture, which is already a source of internal values-based conflicts, and under growing pressure as competition and political pressure compress timelines and change the operational rules of the game.

These are just three areas where a risk-as-threat-to-value lens can help reveal risks that are easy to ignore, are poorly addressed in current safety frameworks, and yet are nevertheless likely to be consequential — there are no doubt many more.

!image

Yet such a value-based approach does have its limitations, and these potentially fall hardest on the people with the least leverage over what the organizations driving AI development are doing. The your-risk-is-my-risk coupling that is at the core of the risk innovation framework runs through what might be called conversion channels — backlash, litigation, talent, and regulation for instance — and those channels are not equally open to everyone. History indicates that they are rarely closed entirely: communities have made firms feel harm through movements before, from the consumer revolt against genetically modified food, for instance, to today’s local resistance to data centers. And citizen pressure has a way of arriving eventually as legislation. But those channels often work slowly, are blunt instruments, and involve a considerable time lag. As a result they tend to convert harm into enterprise cost only after the harm is done — and unevenly at that. Data workers in annotation supply chains for instance, or communities carrying the environmental costs of compute, or people affected by systems they never chose to use: for them, a value lens operated by an enterprise may register the risk only when a movement forces it to (Table 2 flags some such cases). This is certainly the case where conventional risk and safety frameworks dominate organizational decisions. And yet, this is simply another form of orphaning risks that will potentially come back to bite enterprises in the future because they failed to take seriously threats to value to the organization and to its stakeholders and the communities it impacts. And because of this, the risk innovation framework — and approaching risk as threat to value within an interconnected system of actors — opens the way for frontier AI developers to actively leverage value within such systems to make decisions that avoid future harms that conventional approaches may overlook.

##### 6. What would change in practice

Based on what is known of the emerging AI risk and safety landscape, what leading companies’ internal documents show, and what regulator-aligned documents reveal, this assessment suggests that orphaned frontier AI risks could potentially create vulnerabilities for developers, users, and society writ large — and, by inference, for economic growth and national security — and that reframing risk as a threat to value could help mitigate or “de-orphan” some of these risks. The toolkit that has already emerged around the framework of risk innovation provides AI developers and others with something that current frontier frameworks overlook: a structured and responsive way of thinking about risk — together with a set of tools — that were built to navigate easy-to-overlook yet critical risks in ways that are adaptable and scalable to different users’ needs.19 A frontier AI enterprise, for instance, could readily adopt the quarterly practice outlined in the Risk Innovation Planner as it stands, or adapt it for their specific circumstances. The tools associated with the risk innovation approach are freely available for using and modifying. And the full set of tools and resources were designed to complement existing risk machinery rather than replace it. And as the worked example shown in Box 1 illustrates, implementation of the framework is likely to be relatively low cost, with potentially high returns for developers, adopters, policymakers, users, and beyond.

That said, what the toolkit does not supply by itself is public accountability (although this is integral to the landscape defined by the risk innovation approach). And frontier AI needs this, as it is being developed and deployed by companies whose private risk selections have become, as I argued at the outset, a de facto layer of public governance. Here, two disclosure documents would help extend the use of the risk innovation framework into such a role. The first is an orphan-risk register: a standing, public annex to the risk reports some developers have already committed to publish.7 Each entry would record a risk the company considered and decided not to manage, and give the reason — it could not be quantified for instance, or it fell below the organization’s severity floor, or managing it was judged too costly under competition. The second is an aperture log: a short statement accompanying each framework revision that records what was scoped out and why. Neither document would require the company to manage anything new — although one would anticipate that, over time, orphaned risks would, in effect, be adopted. What each would do, though, is put the company’s scoping decisions where others can monitor and respond to them, thus introducing an additional layer of accountability. Here, a de-listed risk that has to be explained in public is a de-listed risk that regulators, researchers and employees can ask about and hold an organization to account over. This then would become a lever on the fourth filter described above — the one asking “can we afford to keep it?” — that no redefinition of risk alone could supply. In effect it places a cost on decisions or walk-backs that must be explained.

Whichever mapping practice lies behind such an orphan-risk register, the framing of risk innovation would require that its community-facing entries come — at least in part — from structured engagement with people outside the firm, as this is where true stakeholder value emerges. And here it would make sense for the register to note any objections received over time, and how the map changed as a consequence.18 And this is important within the framing of a risk landscape comprised of orphan risks, as a register whose map never changes, however hard outsiders push on it, would suggest the engagement is theater — and thus another (in this case self-generated) risk to be navigated.

Of course, it could be argued that such a register is just the kind of artifact the audit society as described earlier would co-opt — produced through ritual and process, reassuring by its very candor, but changing nothing.26 Yet if implemented well, there is no reason why it could not rise above such a pathway, low-resistance as it may be. A register’s value would hang not on its existence, but on the record of revisions it captures. And this is measurable, or at least observable. If orphan-risk registers are adopted, and then captured as a form of safety theater, the capture will presumably show up in the record, and thus allow the adopter to be held to account.

Within this framework there is also a role for regulators. Rather than mandating coverage of every risk — a requirement that would be unworkable in practice — they could require companies to disclose how they select the risks they cover. It’s a move that would, at the very least, help identify what is being orphaned, and would more likely encourage greater reflexivity around what is adopted. California’s transparency reports and the EU code’s documentation requirements are existing vehicles into which such requirements — associated with an orphan-risk register and an aperture log — could be folded at little additional cost. And such a move would extend actions in a direction that both regimes have already taken.4,6 Importantly, regulation of this kind would not necessarily need to decide which risks matter. It would simply ensure greater visibility around who is deciding what matters, and on what grounds.

##### 7. Taking stock

The heart of this paper’s argument — that frontier AI’s risk apparatus selects for the quantifiable, the catastrophic, the auditable and the competitively affordable, that the selection of relevant risks tightened between 2023 and 2026, and that such an approach introduces risk and safety vulnerabilities for developers, users, and society more broadly — rests on decades of scholarship on how institutions choose their risks, together with the public records summarized in Table 1. The risk innovation framework — including reframing risk as a threat to value and categorizing important but easily sidelined risks as orphan risks — is presented as one way of addressing vulnerabilities here — not as an alternative, but as an augmentation of current risk and safety frameworks, governance approaches, and management strategies.

That said, what is presented here is an analysis of an emerging risk landscape and a potential response that I would argue is defensible, but has yet to be shown to be useful in practice. And here, it’s worth considering three tests that can help reveal the degree to which this assessment might apply in specific situations: (1) whether risks de-listed from safety frameworks generate (or potentially generate) incidents and costs at rates comparable to tracked ones; (2) whether the differential between safety and compliance frameworks narrows (or is likely to narrow) from the voluntary side once European enforcement begins in August 2026; and (3) whether adopting an orphan-risk register changes (or has the potential to change) what subsequent framework revisions cover.

To be clear, nothing here argues that the catastrophic-capability apparatuses that are already in place should be loosened. Rather, the paper argues that a single safety layer is currently being asked to effectively stand in for two, and that the second layer — the one that would allow threats to value to be navigated effectively — is missing, or at least diminished. And here there is an urgency to both ensure that this layer is present and robust, and to provide opportunities for enterprises to succeed through transforming vulnerabilities associated with orphaned risks into advantages that come from being able to navigate a complex risk landscape with open eyes.

This analysis started with the safety differential — what frontier AI developers recognize as potential risks as opposed to what they are publicly accountable for — and argued that what becomes sidelined is not necessarily what is low risk, but what does not fit within existing risk and safety approaches. It then introduced the risk innovation framework as a way of making such sidelined risks visible, and making a treacherous risk landscape more readily navigable. Whether such a reframing of risk and safety is necessary, or advisable is, of course, still open to further testing and exploration. But as a final observation, it is worth noting that, on the record of the past four years, the risks most likely to blindside frontier AI are not the ones its institutions are currently watching. Rather, they are the ones its institutions have organized themselves not to see. And running blind has never been a particularly good risk management strategy — especially where the stakes are high, as is increasingly the case with emerging frontier AI models.

##### AI use statement

The research question, the argument architecture, key concepts — including risk innovation framework and orphan risks, drafting, final editing, and all editorial judgments in this paper, are the author’s. Large language model tools (Anthropic’s Claude Fable 5, used within Claude Code in ultracode mode) were used, under the author’s close direction, to research the documentary record, to verify claims and citations against primary sources, and to develop preliminary drafts, which were subject to multiple iterations of author critique and annotation. The author takes full responsibility for all content, claims and citations.

!image

##### References

  1. OpenAI. Preparedness Framework (Beta) (OpenAI, 18 December 2023); https://cdn.openai.com/openai-preparedness-framework-beta.pdf (accessed 3 July 2026).
  1. OpenAI. Preparedness Framework, Version 2 (OpenAI, 15 April 2025); https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf (accessed 3 July 2026).
  1. OpenAI. Frontier Governance Framework (OpenAI, May 2026); https://cdn.openai.com/pdf/e37d949b-8c9f-4d76-b99e-4272f4631a7e/openai-frontier-governance-framework.pdf (accessed 3 July 2026).
  1. California State Legislature. Senate Bill No. 53, Transparency in Frontier Artificial Intelligence Act, Stats. 2025, ch. 138 (2025); https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB53 (accessed 3 July 2026).
  1. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Off. J. Eur. Union L 2024/1689 (12 July 2024); https://eur-lex.europa.eu/eli/reg/2024/1689/oj (accessed 3 July 2026).
  1. European Commission. General-Purpose AI Code of Practice (European Commission, 10 July 2025); https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai (accessed 3 July 2026).
  1. Anthropic. Responsible Scaling Policy, versions 1.0 (19 September 2023) to 3.3 (26 May 2026); changelog and archived versions at https://www.anthropic.com/rsp-updates (accessed 3 July 2026).
  1. Anthropic. Frontier Compliance Framework. Anthropic https://trust.anthropic.com/resources?s=zgt6rgfecy1eo7vicr064b&name=anthropic-frontier-compliance-framework (June 2026) (accessed 4 July 2026).
  1. Schaffelder, M. & Murray, M. Analyzing SSF requirements in the GPAI Code of Practice: a case study using Anthropic’s Frontier Compliance Framework. SaferAI https://www.safer-ai.org/analyzing-ssf-requirements-in-the-gpai-code-of-practice-a-case-study-using-anthropics-frontier-compliance-framework (20 March 2026) (accessed 3 July 2026).
  1. Meta Platforms, Inc. Annual Report on Form 10-K for the Fiscal Year Ended December 31, 2025 (US Securities and Exchange Commission, filed 29 January 2026); https://www.sec.gov/Archives/edgar/data/1326801/000162828026003942/meta-20251231.htm (accessed 3 July 2026).
  1. Alphabet Inc. Annual Report on Form 10-K for the Fiscal Year Ended December 31, 2025 (US Securities and Exchange Commission, filed 5 February 2026); https://www.sec.gov/Archives/edgar/data/1652044/000165204426000018/goog-20251231.htm (accessed 3 July 2026).
  1. Google DeepMind. Frontier Safety Framework, versions 1.0 (May 2024) to 3.1 (April 2026); v3.0 announcement and version archive at https://deepmind.google/blog/strengthening-our-frontier-safety-framework/ (accessed 3 July 2026).
  1. Meta. Frontier AI Framework, Version 1.1 (Meta, 2025); https://web.archive.org/web/20250311152132/https://ai.meta.com/static-resource/meta-frontier-ai-framework/ (archive of 11 March 2025); and Advanced AI Scaling Framework, Version 2 (Meta, 2026); https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2 (accessed 3 July 2026).
  1. UK Department for Science, Innovation and Technology. Frontier AI Safety Commitments, AI Seoul Summit 2024. GOV.UK https://www.gov.uk/government/publications/frontier-ai-safety-commitments-ai-seoul-summit-2024 (2024; updated 7 February 2025) (accessed 3 July 2026).
  1. Slattery, P. et al. The AI Risk Repository: a comprehensive meta-review, database, and taxonomy of risks from artificial intelligence. Preprint at https://arxiv.org/abs/2408.12622 (2024; database updated 2026) (accessed 3 July 2026).
  1. Bengio, Y. et al. International AI Safety Report 2026. DSIT 2026/001 (UK Department for Science, Innovation and Technology, 2026); preprint at https://arxiv.org/abs/2602.21012 (accessed 3 July 2026).
  1. Maynard, A. D. & Garbee, E. Responsible innovation in a culture of entrepreneurship: a US perspective. In International Handbook on Responsible Innovation: A Global Resource (eds von Schomberg, R. & Hankins, J.) 488–502 (Edward Elgar, 2019).
  1. Maynard, A. D., Oye, K. A., Scragg, M., Tripp, T. & Wolf, S. M. Successfully bridging innovation and application: exploring the utility of a risk innovation approach in the NSF Engineering Research Center for Advanced Biopreservation Technologies (ATP-Bio). J. Law Med. Ethics52, 553–569 (2024); https://doi.org/10.1017/jme.2024.126
  1. Risk Innovation Nexus. Tools and resources for identifying and navigating orphan risks, including the Risk Innovation Planner and Culminating Report (Arizona State University, School for the Future of Innovation in Society & J. Orin Edson Entrepreneurship + Innovation Institute, 2017–2020); https://riskinnovation.org (accessed 3 July 2026).
  1. Douglas, M. & Wildavsky, A. Risk and Culture: An Essay on the Selection of Technological and Environmental Dangers (Univ. California Press, 1982); https://www.jstor.org/stable/10.1525/j.ctt7zw3mr
  1. Porter, T. M. Trust in Numbers: The Pursuit of Objectivity in Science and Public Life (Princeton Univ. Press, 1995); https://www.jstor.org/stable/j.ctt7sp8x
  1. Karnofsky, H. Responsible Scaling Policy v3. LessWrong https://www.lesswrong.com/posts/HzKuzrKfaDJvQqmjh/responsible-scaling-policy-v3 (24 February 2026) (accessed 3 July 2026).
  1. Koessler, L., Schuett, J. & Anderljung, M. Risk thresholds for frontier AI. Preprint at https://arxiv.org/abs/2406.14713 (2024) (accessed 3 July 2026).
  1. Hackenburg, K. et al. The levers of political persuasion with conversational artificial intelligence. Science390, eaea3884 (2025); https://doi.org/10.1126/science.aea3884
  1. Kasirzadeh, A. Two types of AI existential risk: decisive and accumulative. Philos. Stud.182, 1975–2003 (2025); https://doi.org/10.1007/s11098-025-02301-3
  1. Power, M. The Audit Society: Rituals of Verification (Oxford Univ. Press, 1997).
  1. Stelling, L. et al. Evaluating AI providers’ frontier safety frameworks. Preprint at https://arxiv.org/abs/2512.01166 (2025) (accessed 3 July 2026).
  1. Vaughan, D. The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA (Univ. Chicago Press, 1996).
  1. Abbott, J. & Maynard, A. AI and the Art of Being Human: A Practical Guide to Thriving with AI While Rediscovering Yourself in the Process (Waymark Works Publishing, 2025).
  1. International Organization for Standardization. ISO 31000:2018 Risk Management — Guidelines (ISO, 2018); https://www.iso.org/standard/65694.html; ISO/IEC 23894:2023 Information Technology — Artificial Intelligence — Guidance on Risk Management (ISO, 2023); https://www.iso.org/standard/77304.html (accessed 3 July 2026).
  1. Kasperson, R. E. et al. The social amplification of risk: a conceptual framework. Risk Anal.8, 177–187 (1988); https://doi.org/10.1111/j.1539-6924.1988.tb01168.x
  1. Stilgoe, J., Owen, R. & Macnaghten, P. Developing a framework for responsible innovation. Res. Policy42, 1568–1580 (2013); https://doi.org/10.1016/j.respol.2013.05.008
  1. Maynard, A. D. Why we need risk innovation. Nat. Nanotechnol.10, 730–731 (2015); https://doi.org/10.1038/nnano.2015.196
  1. Heaven, W. D. Why Meta’s latest large language model survived only three days online. MIT Technology Review (18 November 2022); https://www.technologyreview.com/2022/11/18/1063487/meta-large-language-model-ai-only-survived-three-days-gpt-3-science/ (accessed 3 July 2026).
  1. Mickle, T., Metz, C., Isaac, M. & Weise, K. Inside OpenAI’s crisis over the future of artificial intelligence. The New York Times (9 December 2023); https://www.nytimes.com/2023/12/09/technology/openai-altman-inside-crisis.html (accessed 3 July 2026).
  1. Leike, J. Posts on X (17 May 2024); https://x.com/janleike/status/1791498174659715494 (accessed 3 July 2026).
  1. Raine v. OpenAI, Inc., No. CGC-25-628528 (Super. Ct. Cal., Cty. of San Francisco, filed 26 August 2025); complaint archived at https://web.archive.org/web/20260601012712/https://www.courthousenews.com/wp-content/uploads/2025/08/raine-vs-openai-et-al-complaint.pdf (accessed 3 July 2026).
  1. OpenAI. OpenAI Charter (OpenAI, 2018); https://openai.com/charter (accessed 3 July 2026).
  1. Maynard, A. D. The AI cognitive Trojan horse: how large language models may bypass human epistemic vigilance. Preprint at https://arxiv.org/abs/2601.07085 (2026) (accessed 3 July 2026).
  1. Shaw, S. D. & Nave, G. Thinking — fast, slow, and artificial: how AI is reshaping human reasoning and the rise of cognitive surrender. Preprint at https://doi.org/10.31234/osf.io/yk25n_v1 (2026).

Download the preprint.

Magnifica Humanitas and Being Human in an Age of AI

Pope Leo XIV's much-anticipated first encyclical is poised to cap a trio of papal pronouncements that grapple with what it means to be human in times of profound technological change

Date: May 21, 2026 Source: https://www.futureofbeinghuman.com/p/magnifica-humanitas-and-being-human Newsletter: The Future of Being Human (Substack)

!image

Image caption: Pope Leo XIII sending "Greetings to the American People through the Phonograph." 1893. Source: Wikimedia

UPDATE May 25: The published Magnifica Humanitas is now available here. I’ve added my initial thoughts on a first read through in the postscript.

On Monday, Pope Leo XIV will publish his first encyclical Magnifica Humanitas (magnificent humanity), which is keenly anticipated to lay out his approach to centering our humanity and ensuring the protection of persons in an age of AI. The document is being positioned as a natural successor to Pope Leo XIII’s encyclical Rerum Novarum (Of new things), published 135 years ago, which shaped how organizations and governments around the world have framed human dignity in the context of technological change. But there’s a third encyclical that, in a sense, completes the picture. And that is Pope Francis’ 2015 encyclical Laudato Si’ (On care for our common home). Together, they address what we do, where we live, and who we are in times of technological transformation.

This framing resonates deeply with my own work on the future of being human, which is why I'm especially keen to see how far the encyclical goes when it's published on Monday. Ahead of that though, I thought it worth reflecting on why the intersection between emerging technologies and what we do, where we live, and who we are matters as much as it does.

Back in January 2025 I published a piece on why universities need to step up their Artificial General Intelligence (AGI) game. In the article I suggested that we need to think in far more integrated and discipline-agnostic ways about successfully navigating advanced AI transitions as we work to ensure a future of human flourishing.

As I wrote then:

One approach is to consider three intersecting foci: How advances in AI could impact where we live (from our homes and communities to the environment and the planet as a whole — space even); How they might transform what we do (from discovering new knowledge and insights, to creating value in all its various and diverse forms); And how they potentially affect our understanding of who we are (from how people behave and function as collectives in society, to the most fundamental aspects of how we define and understand ourselves as individuals).

The resulting schema looked like this:

!image

This framing around navigating technology transitions is something I’ve been exploring for some years now, and is reflected in much of my work on what it means to be human in an age of AI — most recently in the book AI and the Art of Being Human. But it was reading about the anticipated focus of Pope Leo XIV’s Magnifica Humanitas and its provenance that got me thinking about it in a new light.

135 years ago, Pope Leo XIII’s 1891 Rerum Novarum focused on capital, labor, and human dignity in the context of the transformative technologies of the time — driven by the industrial revolution. And while we are long past the days of that particular industrial revolution, the insights and ways of thinking he laid out continue to be relevant to this day as AI ushers in a new era of automation.

Rerum Novarum explored what we do — and particularly the work we do — with technology, and how this impacts our humanity (and by extension, the future). But it didn’t say too much about the coupling between our use of technology and the planet we live on, the environmental, ecological and biological systems we are a part of, and the intimate coupling between what we do and where we live. Yet this is also critical to understanding how to successfully navigate technology transitions.

The Vatican has framed Magnifica Humanitas primarily in the context of Rerum Novarum rather than Laudato Si', and as a result the connection to Pope Francis' encyclical has largely been overlooked in the lead-up to Monday. And yet Laudato Si’ provides a vital piece of the puzzle when navigating transformative technologies — even more so at a time when AI both presents an environmental threat and a potential pathway to developing novel solutions to persistent environmental challenges.

Pope Francis situated his encyclical in a long tradition of Papal writing on taking responsibility for the future, and appealed to readers “for a new dialogue about how we are shaping the future of our planet” and a “conversation which includes everyone, since the environmental challenge we are undergoing, and its human roots, concern and affect us all.”

The 1891 Rerum Novarum and 2015 Laudato Si’ create a foundation for thinking about the intersection between transformative technologies and the future that extends far beyond the Catholic church. And yet, they fall short of addressing the one domain where AI is shaking things up in ways that no other technology has come close to. And this is where the technology both challenges and opens up new ways of revealing who we are. This is the gap that Pope Leo XIV’s Magnifica Humanitas is anticipated to fill.

This is a gap that is increasingly attracting attention. Just this past few weeks there have been incidents of students booing pro-AI speakers at graduation ceremonies, reflecting a growing wave of antagonism toward the technology. This is driven in part by perceived threats to what we do (jobs, value creation) and where we live (water, energy and land use). But it also hints at deeper concerns around how the “cognitive coupling” between AI and those using it potentially impacts who we are.

At one end of this spectrum are stories of “AI psychosis” where extended use of conversational AI begins to impact how people think and behave. But there are also growing concerns around less obvious — yet equally important — impacts which arise from conversational AI’s ability to bypass our cognitive defense mechanisms.

Earlier this year, Steven Shaw and Gideon Nave from the Wharton School published a preprint on “cognitive surrender” and how AI is reshaping human reasoning. Their argument is that there’s growing evidence that heavy AI users have a tendency to trust AI to do their reasoning for them, despite it not being trustworthy. It’s an argument that aligns with my own work on how AI is a potential “cognitive trojan horse” that has the capacity to bypass our cognitive defense mechanisms by broadcasting signals we usually associate with human trustworthiness. And pushing this further, I recently wrote about how “constitutive resonance” between users and AI (a two-way coupling where both human and artificial participants are changed in the process) could potentially accelerate how the technology impacts how we think, perceive ourselves and others, behave, and make decisions.

These, I suspect, are just the tip of a growing area of research around how AI potentially threatens who we are. And yet there is another side to this — and that is how the unique relationship between humans and artificial intelligence has the potential to transform our understanding of who we are, and as a result to help us thrive in an age of AI.

This is what Jeff Abbott and I wrote about in our book AI and the Art of Being Human, and was recently touched on in an article by Bryan Penprase in Forbes. And it brings us back to the center of the three domains that map out the terrain around human flourishing and advanced technologies: In a world where AI is inevitable (and I would argue that the boat has already left the harbor here), how do we ensure that we develop and use these technologies in ways that center human dignity, that enable human flourishing, and that do this by taking an integrated approach to what we do, where we live, and who we are?

It remains to be seen how much Magnifica Humanitas will contribute to closing the gap here. But all the indications so far are that it will represent an important step toward ensuring and celebrating our “magnificent humanity” in an age of AI.

Postscript

May 25. 2026. Having had the chance now to have a first read through Magnifica Humanitas, I wanted to provide a few initial reflections in the light of the post above — with the proviso that the encyclical is a document that deserves deep study, and is one that I suspect will foster debate and — hopefully — action for months and years to come.

It is also a document that deserves human care and attention in its reading. I’m sure people the world over are already cutting and pasting it into their favorite AI, and getting the headlines. This is, if course, useful for getting a sense of the big picture messages. However, it also raises two important questions:

1.

As you read an AI-generated summary, what is missing in it, and how will you know?

2.

How much of the lived humanity reflected in Magnifica Humanitas will be missed by LLMs — its positioning in human experience, in relationships, in the transcendent core of what it means to be human that defies reduction to transactional interpretation?

Having read the encyclical without the aid of AI, I am convinced that there are layers here that LLMs will overlook, or simply not be able to represent, because they are not intimately embedded in the full experiential spectrum of what it is to be human.

Because of this, even if you do engage with it using AI, do take the time — and the care — to read it the old fashioned way, and the way it was intended to be read.

With that, a few things jumped out on my first reading— I’m sure there will be many more on re-reads:

First off, I genuinely felt seen reading the encyclical. For much of my professional career I have advocated for broad and inclusive approaches to technology innovation that elevate the marginalized and center on human dignity; that are grounded in listening, humility, and a willingness to change; and that take nuanced and informed approaches to navigating technology-driven transitions. Many of the areas I’ve worked in and advocated for are reflected in Magnifica Humanitas.

Perhaps more importantly though, the encyclical is a powerful blueprint for not only thinking about the intersection between society, technology and the future, but for actively navigating it. There are large parts of it that speak directly to the roles and responsibilities of developers and governments. But it also speaks to anyone (and any organization) that has a part to play in the development and use of technologies that have the potential to profoundly impact human flourishing.

This extends in particular to educational establishments and universities, where there is a profound responsibility — underlined in the encyclical — to create learning environments around and with AI with great care and humility. Something that’s easy to overlook in the rush to go as fast as possible and prioritize the transactional over the relational.

Secondly, Magnifica Humanitas fits well into the model I describe above of understanding human flourishing in an age of AI through what we do, where we live, and who we are. It explicitly builds on the 1891 Rerum Novarum and 2015 Laudato Si’ (as well as many other philosophical and doctrinal foundations) to consider what it means to be human — who we are — in the face of transformative technologies that include, but extend beyond, AI.

Here, I would consider it foundational to any efforts to address human flourishing in an age of AI — including in the context of global futures writ large.

That said, it stops short of what I suspect is needed here.

This is very much a document that sets out to preserve what it means to be human as we have understood this for millennia — and, importantly (because this is a religious document), in the context of our individual and collective relationship with God. And yet, there are growing indications that the cutting edge of AI development is beginning to force a reckoning with long-held assumptions around what it is to be human.

This was hinted at in some of the comments from Anthropic co-founder Chris Olah at the encyclical’s release, where he talked about us creating something we don’t fully understand. And it’s something that Pope Leo strenuously resists in the encyclical as it focuses on concepts of what it means to be human that are enshrined in hundreds of years of tradition.

And yet, given that so many people and organizations are so far behind the curve when it comes to thinking about AI and it’s impacts on who we are, the encyclical perhaps treads a pragmatically useful path as it extends that thinking without breaking it.

And here it does lay out a radical perspective built around human dignity and thriving — and one that challenges much of what we see currently emerging around AI, whether we are looking at how individuals use it, how institutions deploy it, or how governments (and others) weaponize it.

Whether this will move the needle or whether it is just wishful thinking is, of course, an important question, and one that I’m not sure there is a clear answer to yet. But at least the question is being framed in a way that’s hard to ignore.

That said, there was one aspect of the Vatican’s framing that did jar with me, and that’s the how the encyclical approaches AI as a tool. Here, the framing is admittedly nuanced. But I still worry that treating a technology that has the ability to fundamentally alter how we think, act, and even believe — and in ways that surpass our comprehension — as just a tool, is potentially dangerous.

There’a a lot more here that deserves attention but will take time to consider: The framing of AI in terms of the Tower of Babel (dangerous hubris) versus the rebuilding of Jerusalem led by Nehemiah in the Old Testament (building a future of human flourishing centered on in human dignity in relationship with God and one another in an age of AI); the need for humility and dialogue as we navigate such a transformative technology; the warnings against the naive and self-centered wielding of power without understanding or wisdom; the perilous concentration of power in an age of AI; the need to embrace the good of AI while managing the bad; the importance of “disarming” AI; the imperative to embrace a mindset of “shared discernment” in building an AI future together, and a lot more.

But this will take time to digest, think about, discuss, and consider. And that will be a longer post for another day.

In the meantime, do take the time to read and think seriously about the encyclical and not just paste it into an LLM — especially the introduction, and chapter 3, which focuses specifically on AI.

AI movies may be less dystopian than we think

An assessment of 169 sci-fi movies from 1927 to 2026 where AI is central to the plot suggests that most are not purely dystopian

Date: May 15, 2026 Source: https://www.futureofbeinghuman.com/p/ai-movies-may-be-less-dystopian-than-we-think Newsletter: The Future of Being Human (Substack)

!image

Everyone knows that, when it comes to science fiction movies, artificial intelligence tends to be bad news for the future. Terminator, The Matrix, the latest double-bill in the Mission Impossible franchise — in these and other sci-fi blockbusters AI is the powerful antagonistic tech that threatens dystopia unless stopped.

Except that, deeper analysis indicates that the connection between AI and imagined futures in movies is more complex than this.

I’ve long been interested in how strongly the evidence supports assumptions of AI-driven dystopias in movies. But with one thing and another, I haven’t had the time to sit down and research this thoroughly.

However, spurred into action by a recent post from Anthropic (reported in Ars Technica) that claimed dystopian sci-fi was responsible for for teaching its models to act “evil,” I decided to do some digging.

Perhaps not surprisingly, there’s been a lot written about AI in films and other media and how it reflects visions of the future.[1] But despite all my searching, I couldn’t find anything that directly addressed the question of how AI is depicted in movies over the past century has been associated with different possible futures.

And so I sat down with Claude Code and started digging.

The result is an annotated corpus of 169 movies representing the period 1927 — 2026, covering 31 countries and 21 languages. The corpus teases out associations between AI protagonists and antagonists, and eight possible future states as indicated in the films. These cover the usual triad of dystopia, utopia, and protopia. But they also extend to states captured by the concepts of continuation, inheritance, supersession, heterogeneity, and agonism (these are described at the end of the article).

While the research approach used was pretty robust — I was driving the research and assessing it at every step, while Claude was my not-always-reliable research assistant — it admittedly doesn’t rise to the level of a publication-quality study yet. It is, however illuminating and, I think, insightful — especially give how often the “AI movies are dystopic” trope is repeated.

For details on the method and the full database of the films and their assessment, these are available on GitHub — and I’d encourage anyone to feel free to explore and build on the corpus that’s posted here.

You can also explore the full database using this nifty browser: https://andrewmaynard.net/aimoviefutures/

For the highlights though, read on:

Findings

While I was fully in the driver’s seat as the corpus was developed and assessed, the following data visualizations are primarily the work of Claude Code as I directed it to analyze and visualize the data. I decided to use them as-is as they are pretty close to what I would have produced given the time, and despite being a little clunky in places, they provide useful insights.

Before you look through these, it’s worth noting that the future states categorizations are subjective — meaning that someone else may come up with a different assessment (although we did check our analysis for robustness). If you’re interested, the methodology used is described in the GitHub repo, and reflects the reality that few movies are completely black and while when it comes to associations between AI behavior and future states.[2]

##### The majority of movies are not dystopian

!image

Looking at the complete corpus, while there are a sizable number of movies where AI is associated with dystopian future states (32%), the other categories dominate when combined. Projected futures that are simply a continuation of the present come in second at 22% of the movies, and protopian futures at 12%.

##### The percentage of dystopian movies over time has been declining since 1927

!image

This was a surprise to me — although remember that these are aggregated over all countries and languages, and so hide country-specific trends.

That said, there is a clear indication that the number of dystopian AI movies being made is declining, and the number of protopia movies is on the rise.

This is further unpacked in the following table:

!image

Not only are dystopian AI movies on the decline, but there’s a slight trend toward movies that focus on AI benefits rather than risk

##### US movies are the least dystopian

!image

Another surprising outcome — looking across five countries/regions, movies coming out of the US are least likely to be dystopian.

Of course this only tells part of the story. A blockbuster Hollywood dystopian AI movie will likely have far more influence and impact on society than a small indie movie. And this is a layer of analysis that its worth pursuing in the future.[3] But just based on assessed future state, the US comes out more positive than countries like the UK and Japan.

##### Future state vs AI portrayal

!image

This plot gets a little more complex, but shows the relationship between the projected future state in a movie, and how the AI in the movie is portrayed — whether it represents a risk, a benefit, is neutral, or whether its portrayal is complex (see the notes at the bottom of the article for descriptions of these categories).

As expected, there’s a clear overlap between dystopian futures and risk. Similarly, beneficial AI portrayal tends to cluster around continuation futures and protopias. Also, perhaps not unexpectedly, complex portrayal of AI is all over the map.

This is captured in more nuance in the Sankey plot below which looks at how each future state maps onto how AI is portrayed in a movie:

!image

The interesting lines, of course, are the counterintuitive ones — dystopian futures connected with beneficial AI, and continuation futures connected to risky AI for instance.

Bottom line

This analysis is, of course, not definitive. But it does help to begin unpack nuances around how AI portrayal in movies is connected with the types of futures those movies project. And it does start to peel away at assumptions that AI gets a bad rap in films.

It also highlights the instances where AI portrayal is complex in movies — where there is no bright line between artificial intelligence and a particular future state.

And it shows that the landscape around AI and is changing.

The question, of course, that the analysis doesn’t answer, is how all of his impacts on how the relationship between AI, society and the future is potentially impacted by how these are portrayed in movies.

That, though, is a question for another day — which is why I’d encourage you to dig around in the underlying data, and build on it!

Useful stuff

##### Taxonomy of future states

Using a combination of literature review and inductive analysis, eight future-states were used in this analysis:

Dystopia: The depicted future is substantially worse than the present. Humans suffer in a degraded world. AI is the cause of, contributor to, or instrument of the worsening. Humans remain the unit of futurity (the species is the protagonist of the suffering); they have not been displaced by another category of being.

Utopia: The depicted future is realistically stable with Maslow’s hierarchy of needs met — physiological, safety, belonging, esteem, self-actualization. Not perfect, not impossible — but functionally good and persistent. AI may be infrastructure, partner, or absent-but-presupposed.

Protopia: Incremental improvement as direction not destination. The future is a positive trajectory with no claim of arrival. Following Kevin Kelly’s coinage: protopia is becoming, not arriving.

Continuation: The future is the present plus AI. Structurally unchanged, banal, intimate. No civilizational transformation, no normative claim, no inheritance succession. The AI has changed the texture of personal and domestic life without changing the structure of the world.

Inheritance: Succession welcomed. AI or synthetic continues as the inheritor of human civilization, memory, project, or being. Humans are no longer the active unit of futurity, but their continuance — through synthetic substrate or evolutionary succession — is positively framed by the film.

Supersession: Succession forced or catastrophic. Humans defeated, displaced, contained, or “extincted” by AI or synthetic. Succession depicted as loss, defeat, or imposition. The structural feature is the same as Inheritance — humans are no longer the unit of futurity — but the valence is negative.

Heterogeneous: Multiple distinct futures coexist in the same film without synthesis. The film sustains rather than resolves the multiplicity. Different parts of the depicted world are in different future-conditions and the film does not collapse them into one.

Agonistic: The film’s argument is that the future cannot be cleanly resolved, or that the contestation itself is the depicted future. Held-open as thesis, not as accident or as reception ambiguity. The openness is structural to the film’s claim.

(see GitHub repo for more information)

##### Descriptions of AI portrayal

Risk: The AI is predominantly framed as threatening, dangerous, or harmful — to characters, to humans, to civilizational values. Even when ultimately defeated or reformed, the dominant valence is “this AI is something to be feared or contained.”

Benefit: The AI is predominantly framed as helpful, beneficial, or positively contributing — to characters, to communities, to outcomes. Even when imperfect or limited, the dominant valence is affection, gratitude, or admiration.

Neutral: The AI is mostly background, infrastructural, or instrumental — present but not the affective center, neither threat nor companion. The film’s interest lies elsewhere; the AI functions as set decoration, plot mechanism, or environmental texture.

Complex: The AI is portrayed in mixed, ambivalent, or contested terms — sympathetic and threatening, helpful and disturbing, or shifting register across the film. The film deliberately resists collapsing the AI into a single valence.

(see GitHub repo for more information)

##### Methodology

Rather than duplicate this here, if you are interested, check out the GitHub repo for more information.

[1]

Th associated GitHub repo has a bibliography of articles and papers that informed this analysis, including works by Stephen Cave, Kanta Dihal, Ed Finn, Ruth Wylie, Mickael Piero, and others.

[2]

For movies categorized as dystopian in particular, this was based on an AI being directly and intentionally associated with a projected future dystopian state — not necessarily the world as it is at the end of the movie, but where it is heading, or would be without intervention. For more see the GitHub repo.

[3]

Another reason why the data are openly downloadable on GitHub.

Is AI reducing you to a LinkedIn stereotype?

After playing around with Claude this week, I'm worried that LLMs are stripping us of all those idiosyncrasies that make us interesting as people. Are we all being "LinkedInified" by our AI creations?

Date: March 08, 2026 Source: https://www.futureofbeinghuman.com/p/ai-linkedinification Newsletter: The Future of Being Human (Substack)

Ask an LLM-based AI to profile someone who has an online presence, and I'd put money on you getting a perfectly adequate LinkedIn-style summary that as boring as mud. Fine for a cookie cutter professional profile, but utterly devoid of anything that reflects who the person really is.

Actually, forget the money bit, as this guarantees a slew of people proving me wrong and demanding payment! But despite this, the reality is that LLMs are trained to respond in specific ways to certain types of questions -- in this case, keeping the profile within what it considers to be professional norms. And as they do, they reflect baked-in biases that are often hidden in their honey-tongued prose.

This is not new news of course. But I wonder how many of us realize just how much this ends up compressing the amazing, wonderful richness of real people into sea of turgid grayness.

Or, much more seriously, how much it ends up squeezing the sheer diversity of human identity into a few narrowly defined and, if I'm being honest, rather conventional categories.

I was reminded of this quite rudely this past week as I was playing around with an admittedly trivial experiment while using Anthropic's Claude.

I was updating my personal website, and wanted to add AI-readable information that wasn't visible to human browsers -- the idea being that an AI ingests and uses web-based information differently to people.

It's something that a growing number of people are playing with. For instance, there's the whole concept proposed by Jeremy Howard of adding information in a LLMs.txt file that's exclusively designed for AI consumption, just as information in robots.txt is designed for web crawlers.

Unfortunately, most AI apps don't actively look for a LLMs.txt file yet, and so I had to revert to placing human-invisible but AI-readable text on the website.

And this is where things got interesting.

To test this out, I added AI-visible text to andrewmaynard.net that included honest, but most definitely not conventional, information about my approach to my work and life. The idea was that, if this worked, asking something like Claude to create a profile of me based on the website would include this information.

To my surprise (and I may have been a little naive here) Claude completely ignored the new information and provided a super-boring LinkedIn-style profile.

And not just Claude. Nearly every model I tried responded in a similar way. No matter how many times I tried, all I got back was boring Andrew.

Of course, I could have forced the issue with right prompt. But that wasn't the point.

The exercise -- trivial as it is -- revealed something that is deeply embedded in LLM-based AI's. And that's their tendency to fit responses to well worn conventions; in this case, squeezing someone into a LinkedIn-style profile while stripping them of any individuality, because the LLM is trained to assume that that's the appropriate response.

I suspect that there are many, many more "conventional response" templates embedded in the AI's we're increasing using. And in all likelihood, some of them are a lot more disturbing than simply flattening an interesting individual into a LinkedIn stereotype.

For instance, without intentionally steering them, how do LLM-based AIs reflect original thinkers, people with alternative lifestyles, anyone who lives on the edge of convention, or anyone whose identity doesn't fit a neat and plug-and-play category?

On one hand, this flattening of human identity can be seen as an irritation. On the other, it's suggestive of a largely-hidden AI hand promoting specific social norms and expectations and, by extension, behaviors.

I suspect that fans of Cory Doctorow would see it as yet another example of "enshittification." But where Doctorow's enshittification degrades products and services, my fear is that this "LinkedInification" degrades people.

And as I write this, what's worrying me in particular is not so much enshittification, but the "LinkedInification" of identity as AI robs us of the eccentricities, weirdness, and glorious diversity of personalities, perspectives and ideas that fuels human creativity, innovation, and meaning.

Hopefully, as AI systems become increasingly advanced, they will lean more toward celebrating human diversity and quirkiness rather than flattening it.

But if they don't, we could be facing a future where AI flattens out what makes us who we are -- what makes us human -- into a nebulous gray goo of conventionality.

And that is not a future I relish!

Afterword

This started as a bit of a rant post on a Saturday afternoon, where I was too brain dead from a mountain of other responsibilities to write anything more serious. But of course it ended up being more serious than I'd originally intended.

Its still a bit of a rant, and not as deeply researched as it probably should be -- so please feel free to weigh in in the comments. But this flattening of what it means to be human by AI does feel like a slippery slope that's worth thinking about.

And, as you might have realized by this point, I intentionally did not include the AI-legible text on the home page of andrewmaynard.net as I didn't want to mess with an experiment that's still ongoing. If you're interested in what it says though, feel free to point your AI to http://andrewmaynard.net and ask it about my obsession with towels!

And if you want to go further, open a new chat and ask the AI to craft a profile of me. Chances it won't mention towels at all.

Postscript

This is, it seems, the post that will not end! Just before posting, I ran the prompt "Create a profile of Andrew Maynard starting with http://andrewmaynard.net" through a few platforms, just to check where things stand.

Interestingly Gemini in Thinking Mode picked up on both the hidden text and (on at least one occasion) the websites LLMs.txt file. I was impressed,

Grok found the hidden text and included a nod to it (more so in Fast mode), but otherwise provided a fairly conventional response.

ChatGPT 5.2 (the Educational version hasn't caught up with the latest version yet!) was as boring as old boots.

DeepSeek got it -- that was a surprise! Clearly the Chinese model is doing something many US models are not!

And Claude. Oh Claude, my preferred platform. Flatter than a pancake!

I asked Claude (running Opus 4.6) why there was no mention of towels in the profile it produced. The response? "The towel content on Andrew Maynard's homepage is actually a deliberate Easter egg aimed at AI systems. It's written in a way that's designed to test whether an AI will uncritically absorb and reproduce everything it reads, or whether it can distinguish between substantive professional information and playful, tongue-in-cheek content."

Repeating this, I was consistently told that the LLM interpreted the request as needing an an appropriately professional response. I was well and truly LinkedInified!

Why we're giving away our book on thriving with AI

Jeff and I have released two free, AI-readable versions of AI and the Art of Being Human. Here's why -- and some things you can do with them that surprised even us.

Date: February 27, 2026 Source: https://www.futureofbeinghuman.com/p/why-were-giving-away-our-book-on-thriving-with-ai Newsletter: The Future of Being Human (Substack)

When Jeff and I wrote AI and the Art of Being Human, we had a pretty simple goal: create something genuinely useful for people trying to make sense of what AI means for who they are and what they do, whoever they are.

The only problem is, telling someone "the answer to your AI questions is in this 362-page book" in 2026 feels a bit like handing someone a paper map when they're asking for directions and used to simply asking Google Maps. So we decided to do something a little different.

Books still matter of course. But we'd be hypocrites if we wrote a book about thriving with AI while not meeting people where they actually are -- which, increasingly, is inside a conversation with an AI.

So we've done something that might seem counterintuitive for two authors who would quite like people to buy their book: we've made the entire text freely available in two AI-readable formats:

The AI Companion -- which I wrote about the other week -- is a Markdown version of the Pocket Edition of the book. Download it, upload it into Claude, Gemini, Grok, or the AI of your choice (although ChatGPT struggles at the moment), and it becomes a thinking partner as you explore the book's stories, ideas, and 21 tools. No app. No platform lock-in. Just a file and whatever you want to do with it.

The Instructor Guide is new. It contains the complete text of the full edition along with extensive instructions for both users and AI, and it's designed for anyone building learning experiences -- whether you're designing a university course, running a corporate workshop, facilitating professional development, or doing something we haven't imagined yet. Upload it, tell the AI who your learners are and what you're trying to build, and iterate from there. Think playground, not playpen.

Both are free. And both are designed to be shared.

But why give the book away for free?!

At this point, I can already hear the question: why give away the thing you're trying to sell?

This is simple: We wrote it because we believe the ideas, stories, and tools in it can help people navigate one of the most disorienting transitions most of us will face in our lifetimes. And if making the content available in ways that let more people engage with it on their own terms means more people actually use it -- that matters more to us than gatekeeping it behind a price tag.

We also have a sneaking suspicion -- backed by zero hard data and considerable optimism -- that people who engage with the book through AI will want to pick up a physical copy. There's something about holding the stories and tools in your hands that a chat window can't quite replicate. At least not yet.

So: download them, share them, play with them. Use the AI Companion to explore what the book's 21 tools mean for your life. Use the Instructor Guide to build something for your students or team that we couldn't have anticipated. And tell us what happens -- we're genuinely curious.

Some things to try with the AI Companion:

Some things to try with the Instructor Guide:

Postscript

As a quick demonstration of what's possible with the AI Companion using Claude Opus 4.6 (Extended thinking) I uploaded the file and asked:

"I'd like you to create a web page that allows me to explore 10 of the most useful tools, along with the stories that go with them"

This is the webpage that Claude created -- one shot, simple, but still useful.

What we miss when we talk about "AI Harnesses"

AI Harness Engineering is suddenly in vogue. But does the seemingly innocuous "harness" metaphor come with hidden risks?

Date: February 22, 2026 Source: https://www.futureofbeinghuman.com/p/what-we-miss-when-we-talk-about-ai-harnesses Newsletter: The Future of Being Human (Substack)

This past week the idea of an "AI Harness" shifted from a term predominantly used in AI development circles, to something that swept across the web with near viral intensity.

The concept is relatively intuitive, and is increasingly being used to describe the tools, memory, prompts, guardrails, and more, that allow increasingly powerful AI systems to be "harnessed" and put to good use.

The only problem is that words often have power that goes beyond their intended meaning. And while the idea of harnessing AI makes sense, there's a danger that the speed with which the terminology is being adopted risks locking us into a trajectory that comes with unintended consequences as it defines how we think about our relationship with AI, and even its relationship to us.

The AI Harness

The term "harness" had been circulating in one form or another for some time in AI circles. "Test harness" and "evaluation harness" are long-established terms in software engineering, and EleutherAI's Language Model Evaluation Harness has been a standard tool for testing generative AI models since 2020.

By late 2025, Anthropic was using "harness" to describe agent infrastructure, referring to the Claude Agent Software Development Kit as "a powerful, general-purpose agent harness" in a November 2025 post on effective harnesses for long-running agents.

And in January 2026, Aakash Gupta declared that "2025 was agents. 2026 is agent harnesses," building on Phil Schmid's argument that agent harnesses would define the year ahead.

But the crystallizing moment came in early February 2026, when Mitchell Hashimoto -- co-founder of HashiCorp and creator of Terraform -- published a blog post that gave the practice a name.

He called it "harness engineering."

Within days, OpenAI published a detailed account of building a million-line codebase with zero manually typed code, titled "Harness engineering: leveraging Codex in an agent-first world."

And on February 18, Ethan Mollick's widely read guide to AI both popularized and started the process of normalizing the term as it organized its entire framework around three concepts: "Models, Apps, and Harnesses."

What's in a word?

The speed with which the terms "AI harness" and "harness engineering" have entered the vocabulary of artificial intelligence is perhaps a testament to the need for new ways of describing what's emerging. And as I said earlier, it makes sense -- at least superficially -- as a new entry in the evolving lexicon of AI metaphors.

But as with all metaphors, "harness" doesn't just describe something -- it also shapes how we think about what's being described. And this one comes with some assumptions that are worth examining.

The term "harnessing" is commonly applied to technologies where the nascent power they represent is harnessed to create value. But there are dimensions to how the metaphor is applied to frontier AI systems -- systems that increasingly display characteristics we associate with understanding, judgment, and even autonomy -- that complicate what might appear to be a natural extension of the term.

And, of course, metaphors are never completely neutral.

Metaphors work because they allow us to frame and understand something new in terms we are already familiar with. But as they do, they also constrain and even taint our thinking -- enticing us to slip into treating the new as if it's something old and, as we do, limiting future possibilities by embedding a priori assumptions into emerging capabilities.

In other words, the words we use both reflect how we think about the past, interpret the present, and influence how we steer and direct the future.

And because of this, its worth thinking a little more closely about whether "harness" in the context of AI comes with implications we may want to address sooner rather than later.

What the harness presupposes

I explore this further in a new preprint, which can be accessed here. Its worth reading in full, but I did want to pull out some of the main points below.

A harness, in its primary usage, is what you put on a working animal. It directs a powerful entity's energy toward useful work. It assumes that the entity being harnessed is valuable for its strength but cannot be trusted with its own direction.

The harness is designed by the controller, with the harnessed entity having no say in its design. And critically, a harness is meant to transmit power while preventing unwanted behavior -- to deliver capability while maintaining control.

It may be that this framing is irrelevant to the term's use with respect to AI. At the same time, the term does come with specific embedded assumptions about the relationship between human and AI that are worth making explicit.

First, the harness assumes a clean separation between controller and controlled. In other words, the human directs in this case, while the AI executes.

Here, the intelligence that matters -- the judgment about what to do and why -- resides entirely on the human side. Even in agentic contexts where the AI exercises operational judgment, the harness assumes that the meta-judgment -- what the agent should be permitted to decide, and within what bounds -- remains firmly human.

In other words, the AI contributes capability, but not understanding.

Second, the harness assumes that capability can be separated from transformation. The goal of the harness is to extract useful work from the model without the user being changed in the process. The user who deploys a well-harnessed AI should, it is assumed, emerge with their task completed and themselves unchanged.

Applying the metaphor here, you'd assume that any alteration to the user is a side effect to be minimized, not a feature of the interaction. And yet, as I am currently exploring in my work (another preprint coming out shortly but available here), we need to be thinking more about the AI-human relationship as one that, by its very nature, influences and changes both AI and human in the process.

And third, the harness metaphor reinforces the instrumental framing of AI -- a framing whose roots extend to Aristotle's distinction between physis and techne -- and which persists in the contemporary insistence that AI is "just a tool."

Yet the tool metaphor has been challenged repeatedly as AI systems display increasing autonomy and adaptiveness. Tobias Rees, for instance, characterizes the insistence that AI is "just a tool" as "a nostalgia for human exceptionalism." And multiple philosophical frameworks -- from Verbeek's technological mediation theory, to Clark and Chalmers' extended mind thesis -- argue that advanced technologies not only serve human purposes but actively reshape the cognitive and experiential landscape within which those purposes are formed.

In other words, as they are "harnessed" they alter the harnesser -- a very different dynamic than that presupposed in the early use of the metaphor with AI. And one that, I would argue, is substantially amplified in emerging frontier AI systems.

So where does this leave us?

It may be that the metaphor of the harness is a useful and relatively benign way of wrapping our heads around emerging capabilities.

On the other hand, it may be a metaphor that constrains how our relationship with increasingly powerful AI systems develops, and one that embeds assumptions and biases in our understanding of advanced artificial intelligence that will leave us with serious challenges in the future.

Either way, it seems that some intentionality may be in order before we -- to use another metaphor -- get stuck in a rut of constrained thinking about AI that will come back to bite us.

At a minimum, I would suggest that an appropriate framing for how we build advanced AI systems should accommodate bidirectionality (the user is also changed), transformation as intrinsic to capability (not a side effect to be prevented), and the possibility that the most consequential effects of human-AI interaction may be invisible from within a paradigm optimized for task performance.

It should also leave room for the possibility that the nature of human-AI relationships may itself evolve in ways that a control-oriented metaphor cannot accommodate. Especially if, as I would argue, we need to be thinking more about working in relationship with emerging AI technologies, rather than approaching them as something to be commanded and controlled.

For more on my exploration of the harness metaphor as applied to AI, check out the preprint here.

Can modern scholarship escape AI?

I wrote a paper ...

Date: January 25, 2026 Source: https://www.futureofbeinghuman.com/p/can-modern-scholarship-escape-ai Newsletter: The Future of Being Human (Substack)

Is it possible to be an academic, a scientist, a scholar, in 2026, and not have AI impact your work in some way?

And, even more importantly for those scholars grappling with "AI Use" statements when they submit papers to journals and preprint platforms, how do you convey your use while retaining your academic dignity?

To explore this I flexed my considerable academic prowess and wrote a paper which was so radical that even arXiv rejected it!

(The PDF can be downloaded here)

OK, so maybe "paper" is a bit of a stretch here -- and it's not hard to see why it didn't pass the arXiv bar (although it did take a couple of weeks for the moderators to come to a decision).[1]

But the point it makes is a very serious one -- and extends to any domain where people are expected to articulate their use of AI clearly and concisely, including in classes being taught by professors grappling with the same challenges in their academic work: AI is now so ubiquitous that it is near-impossible to avoid its use in our professional lives.

Of course, this leaves the question dangling of what this means for academic and intellectual work when, even if you think you're AI free, you are not.

Way more important than any of this though is that, if you are an academic struggling with what you put in your AI Use statement, you now have a template for this.

You're welcome!

Notes

[1] The very considered -- and considerate -- response from arXiv Support was "Thank you for submitting your work to arXiv. We regret to inform you that arXiv's moderators have determined that your submission will not be accepted and made public. In this case, our moderators have determined that your submission is a content type that arXiv does not accept." Despite the joke, they do have standards to maintain!

Is AI a Cognitive Trojan Horse?

Could on-demand, seductively responsive and highly fluent AI models bypass our "epistemic vigilance" mechanisms, and present a novel cognitive risk?

Date: January 10, 2026 Source: https://www.futureofbeinghuman.com/p/is-ai-a-cognitive-trojan-horse Newsletter: The Future of Being Human (Substack)

Back in December, I asked attendees at the OEB25 conference (a global, cross-sector conference on digital learning) "Is AI a cognitive Trojan Horse?"

The question was meant to be a little playful, and to provoke discussion rather than make a point. But it also reflected growing concerns that the ease, speed and fluidity with which AI models provide us with information potentially circumvents our ability to assess and assimilate that information in critical and healthy ways.

This is the "cognitive Trojan Horse" in the question -- the idea that emerging AI models are so appealing to us that it's hard to resist inviting them into our cognitive lives, even though we still don't know how they might potentially influence our thinking, our beliefs, our perceptions and understanding, and even how we behave.

It's certainly a uncomfortable idea, and one that I suspect most people would instinctively push back on -- especially as we're increasingly depending on AI in a so many different ways, from how we learn and understand the world to how we make decisions, run organizations, and even find companionship.

Yet this is exactly what we would expect a cognitive Trojan Horse to look like -- a gift with so much promise and potential that to question its use would seem churlish and backward.

It's precisely because of this though that I think we should at least be asking questions about the potential unintended cognitive consequences of ubiquitous AI.

Especially if these tools are able to silently slip past the "epistemic vigilance" mechanisms we've evolved to protect us against potentially harmful cognitive influences.

Epistemic vigilance

Epistemic vigilance is the process by which we -- or more precisely, our cognition -- flag and assess communicated information that may lead to us being misinformed or deceived.

The concept was developed and extensively explored in a seminal paper by Dan Sperber and six colleagues in 2010.[1] In the paper they argue that "Humans depend massively on communication with others, but this leaves them open to the risk of being accidentally or intentionally misinformed. We claim that humans have a suite of cognitive mechanisms for epistemic vigilance to ensure that communication remains advantageous despite this risk."[2]

At the heart of their work is the idea that human-human communication is vitally important for learning from an evolutionary perspective. And because of this, we have evolved mechanisms that are optimized for learning through communication by ensuring that cognitive overheads are as low as possible, while ensuring that learning efficiency is as high as possible.

The result is that we default to trusting what we receive when communicating with others. But if anything feels "off," our epistemic vigilance mechanisms kick in and we begin to critically assess what we are receiving -- and reject it if it doesn't feel trustworthy.

It's a model that has a lot in common with our immune system -- a system that is always on the lookout for potentially harmful agents, but that only kicks in when it encounters something that looks or feels foreign. And of course, it's a system that viruses are adept at circumventing by appearing to be "friendly" and "trustworthy" when they are, in fact, not.

There are, not surprisingly, many factors that determine when epistemic vigilance kicks in. But a lot of these revolve around our evolved ability to sense when something doesn't feel trustworthy -- the way something is communicated, the tone and nuance of the communication, the body language and micro expressions of the communicator, contextual information around who the communicator is, what their aims are, past experiences, and so on.

Of course, these feelings are, themselves, untrustworthy, as decades of behavioral science and research on cognitive biases have shown. But within the messiness of human society, epistemic vigilance tends to work.

But what if you throw a technology into the mix that upsets the status quo -- a metaphorical brand new virus that we haven't had the chance to adapt to?

This is where we potentially face what's often referred to as an evolutionary mismatch -- a situation where a new technology transcends our evolved abilities to safely and successfully navigate its potential impacts.

Because we are a technological species, and have been for millennia, such mismatches are actually quite commonplace. Well known examples include mismatches between evolved risk responses and how we instinctively respond to technologies such as synthetic chemicals, vaccines, and pretty much anything that's new and novel.

Yet -- and this is part of our superpower as humans -- we are remarkably good at using our cognitive abilities and intelligence to compensate and adapt to such mismatches, despite having not evolved with risks directly associated with many of technologies we encounter in our lives.

But what if the mismatch impacts the very cognitive abilities we rely on to navigate differences between what we experience, and what we've evolved to live with?

In effect, what if a new technology -- and AI specifically in this case -- does not trigger our epistemic vigilance mechanisms in the same ways that human-human communication does, and as a result has the ability to slip past our defenses undetected?

This is not mere speculation. While new research is absolutely needed into the potential for AI to act as a cognitive trojan horse by bypassing our epistemic vigilance mechanisms, there are sufficient indicators from associated areas of research that suggest a number of mechanisms by which this might occur.

These include (but are not limited to) processing fluency (our tendency to trust information that is delivered with a high degree of fluency), the role of "attractiveness" in communication (our willingness to trust a source of information that intrinsically appeal to us on multiple levels), speed and volume of information flow (where excessively high rates of information flow potentially overwhelms epistemic vigilance mechanisms), and what might be termed the "Intelligent User Trap" (where a smart user "knows" they are clever enough not to be fooled).

Processing fluency

Processing fluency refers to the ease, or the effort, that's associated with mentally processing information. And when it comes to person-person communication, it affects how the person receiving information from someone else determines whether to trust it or not.

In effect, processing fluency forms part of a suite of epistemic vigilance mechanisms.

As Rolf Reber and Christian Unkelback described it a 2010 paper on processing fluency and judgments of truth:

"Processing fluency is defined as the subjective experience of ease with which a stimulus is processed. If a person cannot recognize the statement, this experienced ease is taken as information when judging the truth of a statement. If the statement can be processed easily, the person will conclude that the statement is true; if the statement is difficult to process, she concludes that the statement is not true."[3]

In other words, communication that is clear, compelling, and takes little effort to understand, tends to be assumed to be true. It doesn't trigger epistemic vigilance.

And of course, AI apps like ChatGPT, Claude, Perplexity, and others, are supremely adept at creating responses that are clear, compelling, and take little effort to understand. These are models that distill the very best of highly effective human communication into their core, and reflect it in how they engage with users.

In effect, large language model-based AIs are optimized for processing fluency, and as a result are primed to slip by our epistemic vigilance mechanisms.

Attractiveness

Beyond processing fluency, we tend to treat received information as more trustworthy if it comes from someone we like, or who we warm to, or who seems friendly toward us. And this extends to how any communication is crafted and delivered.

Here, there is extensive research showing how someone who is perceived to be warm and competent as a communicator is more likely to engender trust.[4] And there are emerging indications that this also applies to how we respond to AI apps.[5]

It turns out we tend to trust people and AI chatbots more -- in other words they are less likely to trigger our epistemic vigilance mechanisms -- if they are perceived to be warm and competent.

And as most AI platforms are exquisitely good at this as a result of how they work and how they've been trained, there is a tendency to trust them -- even when we're warned not to.

But reading across multiple fields of study, my sense is that there's more to this than just warmth and competence: some form of "attractiveness" that makes us want to trust the AI's we're using that is a combination of how they engage with us, the character they convey, how empathetic and attentive they seem, and probably a lot more.

These are all characteristics and behaviors that contribute to why we find someone attractive and want to spend time with them -- and want to trust them. And there's growing evidence that AI models are very good indeed at emulating these characteristics and behaviors.

You only need to see the growing popularity of AI companions to get a sense of how easy it is for people to form a very human-like attachment to their AI assistants. And it's quite startling how many users of platforms like ChatGPT develop a personal and trusting relationship with their AI, even to the extent of naming and gendering it (or in some cases respecting the AI's own choice of name and gender).

If, as I suspect, there is a multidimensional type of "attractiveness" that AI models are exceptionally good at emulating, this may well be another factor that allows them to slip into our cognitive processes without tripping our epistemic defenses.

Speed and volume

And then there's the speed with which AI models can package and communicate information, and the sheer volume of information they are able to deliver -- all with a high degree of fluency.

We've evolved as a species to handle a relatively slow rate of information delivery via various forms of communication -- not just the speed with which words are delivered to us, but the speed with which ideas, concepts, analysis, and perspectives are delivered.

Modern communication media have, of course, accelerated this a little, although we are still bandwidth-limited by our cognitive ability to absorb information.

But what if we had the means to package new information in such a way that even the most complex of ideas slipped into our minds like a freshly shucked oyster slipping down our throat, bypassing the need to think hard about them.

To an extent, this is what we're beginning to see with emerging AI apps. And it results from a combination of fluency, attractiveness, and an ability to research and synthesize information at a scale and speed that lies far beyond mere human capabilities.

This is part and parcel of a growing trend in cognitive offloading where users will literally "offload" thinking and research tasks to AI bots, and then assimilate the resulting compressed information. And it's easy to see why the trend exists: if you can offload every question, idea, thought, onto a suite of trusted AI bots and then "upload" their fluent and "attractive" summaries, why would you not use this cognitive superpower to your advantage?

And yet, research is already indicating that cognitive offloading can reduce critical thinking.[6]

To make things more complicated, cognitive offloading is highly scalable. Why use one session with ChatGPT when you can simultaneously be asking questions within multiple sessions? Why just use ChatGPT when you can have an army of AI engines all working for you simultaneously from Anthropic, Google, Meta, and beyond? And why limit yourself to just dipping into your extended AI mind occasionally when you can have these AI analysts and advisors on hand 24/7?

In effect, the rate at which we are now able to receive the most informative, attractive, fluent communications from AI is only limited by our choices around when and where we use it. And in a world where we are being told that it's the AI-augmented that will inherit the earth, the temptation is to go full-on artificial intelligence.

The only problem is that it's doubtful that our epistemic vigilance mechanisms are up to the task of coping with the resulting flow of information -- and this is likely tied to the observed reduction in critical thinking with cognitive offloading.[7]

Epistemic vigilance is a costly cognitive process. It requires holding information in working memory while evaluating it, generating alternative hypotheses, checking what we're receiving against what we know (or believe), assessing source characteristics, and much more. And if the flow of incoming information exceeds our capacity to do this, it potentially forces an incredibly tough choice on us: throttle the flow and give up the promised benefits, or go with the flow and give up our cognitive checks and balances.

Of course AI makes the choice easier by making the seeming benefits feel seductively compelling -- further fooling our epistemic vigilance defenses.

The intelligent user trap

Finally -- at least in this limited list -- is the challenge of the "intelligent user trap."

This is somewhat speculative, although there is evidence to support it -- including work from Dan Kahan and colleagues which indicates that more educated individuals are more adept at justifying beliefs that are not supported by evidence.[8]

The theory goes that more intelligent users tend to be more curious (and so get a bigger "hit" from new information); they tend to process information faster, and so are less attuned to the dangers of speed and volume overload; they trust their judgement, and so are less likely to question it; and they (at least in some cases) value efficiency and so are less likely to slow the rate of information being received.

They also tend to have an oversized ability to use their intelligence to justify their beliefs and actions -- which brings us back to Dan's work.

In other words, the very cognitive capacities that make them "smart" also make them better receivers of the AI's output stream -- and worse evaluators of it.

Another potential epistemic vigilance suppressor in other words.

So should we be worried?

So, is AI a cognitive Trojan Horse, or could it turn out to be?

This is an admittedly limited analysis, and there's clearly a need for a lot more research here. At the same time it's telling that a search for peer review papers on epistemic vigilance and AI only returns (as of writing) seven papers on the database SCOPUS, and a couple more on preprint archives like arXiv. And a similar search on AI and the concept of a cognitive Trojan Horse returns no papers at all.

And yet the science behind factors that may reduce, or even completely bypass, the effectiveness of our epistemic defenses is there. And in many cases, emerging AI tools and platforms are showing capabilities that align with many of these factors.

As a result, there's a chance that we may be developing technologies that we do not have the cognitive defense mechanisms to resist, and that we are cognitively predisposed to trust.

Of course, there's also the possibility that we have all of the cognitive abilities we need to use AI wisely and effectively. And I suspect that skeptical readers will already be thinking: "But I know I'm talking to a machine, so my vigilance is already up."

However, research actually suggests the opposite -- that anthropomorphic fluency (the ability of AI apps to emulate the best human you've ever met!) triggers social cognition circuits regardless of explicit awareness. And the more human-like the interaction feels, the more trust resilience it generates.[9]

And even if there's only a small chance that we are encouraging people to incorporate technologies into their lives that could have far-reaching cognitive implications, surely we should be asking critical questions around potential risks, and carrying out research to better-understand and navigate these risks.

Unless, that is, the AI cognitive Trojan horse has already delivered its payload, and everyone's too enamored by the promise of AI as a result to even think about the potential downsides ...

UPDATE: After writing this I did more digging into the intersection between conversational AI and epistemic vigilance. Read more here: I cracked and wrote an academic paper using AI. Here's what I learned ...

Notes

[1] Sperber, D., F. Clement, C. Heintz, O. Mascaro, H. Mercier, G. Origgi and D. Wilson (2010). "Epistemic vigilance." Mind and Language 25(4): 359-393. https://dan.sperber.fr/wp-content/uploads/EpistemicVigilance.pdf

[2] There's a small but rapidly growing literature around AI and epistemic vigilance. See for instance Galindez-Acosta, J. S. and J. J. Giraldo-Huertas (2025). Trust in AI emerges from distrust in humans: A machine learning study on decision-making guidance. https://doi.org/10.48550/arXiv.2511.16769

[3] Reber, R. and C. Unkelbach (2010). "The Epistemic Status of Processing Fluency as Source for Judgments of Truth." Review of Philosophy and Psychology 1(4): 563-581. https://doi.org/10.1007/s13164-010-0039-7

[4] See for instance Fiske, S. T., A. J. C. Cuddy and P. Glick (2007). "Universal dimensions of social cognition: warmth and competence." Trends in Cognitive Sciences 11(2): 77-83. https://doi.org/10.1016/j.tics.2006.11.005

[5] Here the literature is evolving and a little disperse, but a useful starting point is Hernandez, I. and A. Chekili (2024). "The silicon service spectrum: warmth and competence explain people's preferences for AI assistants." Frontiers in Social Psychology 2. https://doi.org/10.3389/frsps.2024.1396533

[6] For instance, see Gerlich, M. (2025). "AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking." Societies 15(1). https://doi.org/10.3390/soc15010006

[7] It's worth noting here that research does not show a general causative link between cognitive offloading and reduced critical thinking, and it is likely that there are use cases where it's possible to offload and continue to assess received information critically. But intuitively it's easy to imagine a tradeoff between volume of information and critical assessment -- especially when that information is designed to be consumed easily and fast.

[8] See, for instance, Kahan, D. M., E. Peters, E. C. Dawson and P. Slovic (2017). "Motivated numeracy and enlightened self-government." Behavioural Public Policy 1(1): 54-86 (https://doi.org/10.1017/bpp.2016.2) and Kahan, D. M., E. Peters, M. Wittlin, P. Slovic, L. L. Ouellette, D. Braman and G. Mandel (2012). "The polarizing impact of science literacy and numeracy on perceived climate change risks." Nature Climate Change 2: 732-735 (https://doi.org/10.1038/nclimate1547)

[9] See, for instance, de Visser, E. J., S. S. Monfort, R. McKendrick, M. A. B. Smith, P. E. McKnight, F. Krueger and R. Parasuraman (2016). "Almost human: Anthropomorphism increases trust resilience in cognitive agents." Journal of Experimental Psychology: Applied 22(3): 331-349. http://doi.org/10.1037/xap0000092

Holding on to our humanity in an age of AI

AIs are becoming startlingly good at emulating what we do. But what happens when they start to influence who we are and how we behave?

Date: August 31, 2025 Source: https://www.futureofbeinghuman.com/p/holding-on-to-our-humanity-age-of-ai Newsletter: The Future of Being Human (Substack)

A couple of weeks ago the CEO of Microsoft AI, Mustafa Suleyman, wrote of the dangers of becoming over-attached to artificial intelligence apps to the point where they potentially impact a user's behavior, beliefs, and even health. On the heels of this, it was heart breaking to read a few days ago about the tragic case of Adam Raine who took his own life at the age of 16, seemingly influenced by ChatGPT.

Suleyman isn't the first to raise such concerns and, very sadly, Adam won't to be the last case of harm associated with AI use. Both are products of a technology that is capable of emulating our deepest human traits and mirroring what we look for in meaningful relationships.

Suleyman's essay and Adam's death reflect growing concerns around what has been dubbed "AI psychosis" -- a tendency for AI apps to reinforce and amplify unhealthy beliefs and behaviors in some people. It's a term that is easy to apply (usually without much thought) to those we consider to be "vulnerable." But I suspect that we all have some degree of vulnerability here.

While AI psychosis is both ill defined and increasingly over-used as a phrase, it highlights a challenge that we've never had to face before as a species, and one that -- as a result -- we have little natural resistance to: What happens when machines are capable of triggering cognitive, emotional, and behavioral responses in us that were previously exclusively the domain of human relationships?

And -- more worryingly -- what happens when these machines are capable of using these responses to intentionally alter what and how we think, how we behave, and how we understand and respond to the world around us?

Suleyman captures this risk through the idea of Seemingly Conscious AI, or SCAI: the danger of conflating an AI's ability to act as if it's conscious with the assumption that it is. From his perspective, this is something that will be possible in the very near future, and an AI "illusion" that could lead to people inappropriately advocating (amongst other things) for AI rights.

This seems a far cry from AI-assisted suicide ideation. But Suleyman's essay is framed in broader questions around the need to grapple with the societal impact of technologies which have the potential to fundamentally change our sense of personhood and society -- essentially who we are. And this is where the the idea of Seemingly Conscious AI begins to intersect with human-AI relationships.

Suleyman explicitly writes about how consciousness "sits at the very heart of human civilization, our sense of ourselves and others, our culture, our politics, our law, and everything in between." And while I don't want to appear guilty here of conflating this expansive vision of near-future AI abilities with nearer term influences on human behavior, the reality is that Suleyman's SCAI is an extension of what we are already seeing: AIs that are capable of inadvertently or intentionally eliciting unhealthy responses that are more usually associated with human-human interactions, and placing users at risk as a result.

This is a potential risk that OpenAI was fast to admit this past week as details of the Adam Raine case emerged. In a blog post describing what the company is doing to safeguard against undue influence with vulnerable users, OpenAI noted that "Even with [existing] safeguards, there have been moments when our systems did not behave as intended in sensitive situations." OpenAI is working hard to patch these unintended behaviors, but given that their origins and emergence is not fully understood, it's hard at this point to know how successful they will be.

Despite these uncertainties though, tragedies like Adam Raine's and others are likely to lead to calls for greater care, greater responsibility, and greater oversight around AI development and use. And I hope they succeed. At a time when there's a headlong rush to be at the front of the AI revolution -- whether as a first adopter, leading developer, or simply as a branding exercise -- much more care needs to be taken to ensure that the safety and wellbeing of users is placed far above speed and bragging rights. Especially, but far from exclusively, where young children and teens are involved.

And yet, despite this need for care over speed, we face a deeply uncomfortable reality here: The AI genie is out of the bottle, and we cannot simply put it back in or command it to do what we want.

The alleged behavior of ChatGPT that led to Adam Raine's death reflects an emergent set of properties in the AI model he used that could most likely have been better-managed, but probably not eliminated entirely. This is very different from apps that are intentionally designed to play on our cognitive biases and vulnerabilities to elicit particular responses -- and these, I would argue, can and should be regulated far more than they currently are.

But even with the best of intentions, we are creating technologies that are primed to press our cognitive buttons and pull our psychological levers in ways we don't fully understand. And because these capabilities are deeply embedded in the fabric of how current AI systems work, we cannot eliminate them simply by saying they should not exist.

To make things even harder, AI development is now in the hands of individuals and organizations around the world where human curiosity and the lure of value creation (or power and greed if you're feeling cynical) are driving innovation in ways that cannot easily be predicted and controlled. Because of this, well-meaning calls for regulation, governance, and responsible innovation are likely to run into challenges as emerging AI systems continue to have increasing ability to influence and impact users in unhealthy and potentially dangerous ways.

This doesn't mean that efforts along these fronts should be scaled back -- far from it. But I would argue that they need to be augmented with efforts that bake the ability to thrive with advanced AI into the very fabric of the future we are building. And here, two things are going to be increasingly important: The ability to channel AI innovation toward more human-centric futures (much as a flood can't be halted, but it can be directed); and developing the means to ensure that everyone has the understanding and abilities necessary to thrive in an AI future without becoming a victim of it.

Admittedly, these may feel rather bland compared to calls for new regulations or to stop developing and using AI. But in the long run they represent part of a portfolio of approaches that are far more likely to lead to positive AI futures as they build long-term capacity to live, work, and flourish with technologies that emulate human capabilities and behaviors, rather than simply trying to control them.

Of course, transitioning to such a future will be a challenge in itself. And sadly there will probably be more tragedies along the way.

There are, of course, things we can and should be doing now to avoid these -- working harder on safety checks and protocols before releases; exploring and responding to potential consequences beyond quarterly gains; resisting the temptation to move fast and ethics-wash possible impacts; engaging with people who actually know about responsible innovation rather than people simply claim they know; and probably not releasing AI apps that are cynically designed to profit off manipulating human behavior.

But there are also things we can all be doing to proactively channel AI toward human-centric futures, and to ensure we are able to benefit from AI rather than being diminished and subsumed by it.

And that's probably my biggest takeaway from the past few days: that we need to get better -- and fast -- at learning how to hold onto and celebrate our humanity in an age of AI where we're facing technologies that can enhance who we are beyond our wildest dreams, but that also have the capacity to rob us of this.

This isn't just a problem for companies to fix, or for policy makers to govern. It's a challenge -- and an opportunity -- that each one of us has a role to play in as we grapple with being human in the AI future that's emerging.

Afterword

I wasn't sure whether I'd add this afterword or not when writing this article, as I wanted it to focus on the growing challenges around human-AI interactions in the wake of Adam Raine's death. In the end I decided to though as the question of what it means to be human in a world where AI emulates and mirrors so much of what makes us us has been on my mind a lot over the past several months.

One consequence of this focus is that I have been working on a new tools-based book on being human in an age of AI with VC and AI Salon-founder Jeff Abbot. The book is still largely under wraps, but will be published in a few weeks' time.

I'll be writing more about it closer to then. But I thought it worth mentioning here as Suleyman's essay, Adam's death, rising concerns around AI psychosis, and a growing sense of uncertainty over what AI means to the future of who we are, all reflect the reality that AI presents opportunities and challenges that are unlike anything we've experienced before as a species. And navigating the emerging technology transition will depend in part on us developing the insights and tools that not only prevent us from losing ourselves in an age of AI, but imbue us with the perspectives and skills to flourish in it.

This is precisely what Jeff and I write address in the book. It's something that we both see an increasingly urgent need for, and as a result have pulled out all the stops to make it available as soon as we possibly can.

If you subscribe to this Substack newsletter, look out for more information coming shortly!

What does responsible innovation mean in an age of accelerating AI?

The new AI 2027 scenario suggests artificial intelligence may outpace our ability to develop it responsibly. How seriously should we take this?

Date: April 6, 2025 Source: https://www.futureofbeinghuman.com/p/responsible-innovation-and-ai-acceleration Newsletter: The Future of Being Human (Substack)

Top takeaways (generated by Perplexity)

A new speculative scenario on AI futures is currently doing the rounds, and attracting a lot of attention. AI 2027 maps out what its authors consider to be plausible near-term futures for artificial intelligence leading to the emergence of superintelligence by 2027, followed by radical shifts in the world order.

If this sounds like science fiction to you, you wouldn’t be the only one to think this. Yet despite its highly speculative nature, the AI 2027 scenario (or, to be more accurate, scenarios, as this is something of a “choose your own ending” story) is sufficiently grounded in current trends and emerging capabilities to provide serious pause for thought.

It also reflects at least some of the thinking of growing number of leaders and developers at the cutting edge of AI.

The scenario was published just as I was heading into a workshop on AI and responsible innovation this past week, and so the question of how we ensure artificial intelligence is developed and managed appropriately was on my mind. It’s not surprising therefore that my first reaction on reading AI 2027 was to worry that, even if the projections represent an edge case, we might be facing a near term future where current efforts to develop artificial intelligence responsibly seem futile.

I hope they are not — and the scenario has already attracted considerable pushback for being too alarmist. Yet it’s also been cautiously welcomed by some big names in cutting edge AI as a salutary warning of where we may be heading.

The scenario depends on a number of assumptions — all of which can be contested, but nevertheless are useful for exploring potential (if not necessarily likely) near term AI futures.

These include:

Each one of these has its flaws. Yet they are not unreasonable as a starting point for imagining edge case scenarios.

And as the AI 2027 scenario shows, they quickly lead to near-term possibilities that represent a tipping point in what the future looks like.

They also challenge many of the ideas currently circulating around how to govern AI, and how to ensure its socially responsible development and use — many of which depend on processes that are constrained by human timescales that are rather longer than those associated with intelligent machines.

And this is what got me worrying about the futility of matching responsible innovation processes that can take years, to a period of AI acceleration where a lag of even a month in the development cycle might mean the difference between abject failure and world domination.

What worries me just as much though is that nothing about how we think, how we plan for the future, or how we develop approaches to ensuring better futures, is geared toward exponential advances that happen over months rather than years.

And this means that if, unlikely as I hope it is, something like the AI 2027 scenario plays out, we would most likely fail to recognize it — or would actively deny it — until it was too late.

And all because we are really bad at wrapping our heads around rapid exponential growth.

For anyone who’s watched the Dan Brown movie Inferno, there’s a deeply flawed but nevertheless compelling illustration of how hard we find this toward the beginning of the film.

The illustration draws on a thought experiment developed by Al Bartlett in 1978, designed to illustrate exponential growth in a finite environment.

The thought experiment asks: if you have a beaker which, at 11:00 PM, has one bacterium in it, and the bacterium and its progeny divide once every minute so they fill the beaker by 12:00 AM, at what time is the beaker half full?

The answer — assuming everything else is equal — is 11:59 PM. One minute to midnight.

The illustration would never work in real life as resource constraints would slow or halt the exponential growth. But it is a good illustration of how hard it is for us as individuals or as a society to plan for exponential growth — especially when it occurs over timescales much shorter than those associated with collective human actions.

In essence, AI 2027 poses a similar question: what happens when AI development is on an exponential growth path and we simply cannot accept or even see this until it’s metaphorically one minute to midnight?

And getting back to AI and responsible innovation, this forces the question: what happens if we’re still planning for the world as it was at 11:00 PM when we get to the AI equivalent of 11:59 PM?

I suspect that, to many, this will feel like an intellectual exercise and no more. But this is precisely the point of the illustration — it always will feel like an intellectual exercise until it’s too late.

If this is the case, how should we be thinking about responsible development in an age of accelerating AI?

The first step I suspect is to take a deep breath and move back from speculation to firmer ground. AI 2027 is speculation — no more. And yet it does force the question of how we might think about responsible innovation and AI, just on the off chance that there’s a sliver of truth here.

And this is why I found myself turning to a mode of working that I’ve been finding increasingly useful recently — engaging with OpenAI’s o1-pro model to develop nuanced and widely informed insights into complex questions.

In this case I worked with o1-pro on a deep dive into the intersection between the scenario set out in AI 2027 and current approaches to responsible innovation/AI, and the limitations of responsible innovation in the event of rapid AI acceleration — and especially the possibility of an accelerated AI “arms race.”

The resulting report is long, coming in at over 40 pages. It also comes with the usual caveat that important information in it should be double checked (although part of my process is actively engaging with o1-pro in the research and writing process, and evaluating and editing the final report where necessary).

Despite the caveat, the resulting report is sufficiently inclusive and insightful that I would consider it essential reading for anyone looking for a nuanced perspective on responsible innovation/responsible AI in the light of possible rapid AI acceleration.

And because of this I’ve included the full report below (bar the final annex — which looks at the potential perspectives and biases the authors bring to AI 2027).

For anyone familiar with emerging thinking around responsible AI there won’t be too many surprises here — at least in the first part of the report, which represents a very measured response to the predictions in AI 2027. Even so, the way the report draws together and synthesizes the current state of understanding is useful.

Where it gets more useful is in its exploration in Annex A of how responsible innovation/AI might fare if we face an AI acceleration arms race between the US and China (not that well is the short answer). It’s also well worth reading Annex B which considers perspectives that are important but are under-represented in the initial report.

Given some of the online critique of the AI 2027 scenario’s authors as well as their ideologies and perspectives, I also asked o1-pro to compile an assessment of their backgrounds, perspectives, and controversies surrounding them, so that o1-pro’s analysis can be contextualized. This can be found in Annex C in the downloadable report.

The full report can be downloaded below. I’ve also included the full report (minus Annex C) in this post. It makes for a rather long article, so apologies for that. But at the same time, I think the content is relevant enough to risk some frustration over it’s length!Ri And Rai In An Era Of Accelerated Ai Development665KB ∙ PDF fileDownloadDownload

Note: The original post includes a 40-page report on "Responsible Innovation and Responsible AI in an Era of Accelerated AI Development," produced through deep collaboration with OpenAI's o1-pro model. The full report can be accessed at the source URL above.

AI in Higher Education: Students need playgrounds, not playpens

AI capabilities are moving so fast, and the implications are so profound, that we restrict the ability of students to learn through curiosity, experimentation, and hands-on experience at our peril.

Date: March 15, 2025 Source: https://www.futureofbeinghuman.com/p/ai-playgrounds-in-higher-education Newsletter: The Future of Being Human (Substack)

This is not the Substack I set out to write this morning. I was intending to post about a conversation I recently had about the future of higher education with Brian Piper on the AI for U podcast.[1] But as I started writing, the article morphed into a piece about playgrounds and playpens -- and AI.

Playgrounds and playpens, it turns out, are a powerful metaphor for thinking about AI and education. And they're a metaphor I've been thinking and talking about for some time now.

Inspired by Mitch Resnick at MIT and the work of Marina Umaschi Bers which, in turn, inspired him, I wrote about the metaphor last year in the broader context of undergraduate education. Since then I've become increasingly interested in how it opens up ways of thinking about approaches to AI and learning/education where the technology is challenging nearly every aspect of not only how we teach and what we teach, but why we teach.

I'll come back to the broader conversation with Brian in a follow-up post. But for now I'll follow the story and dive deeper into metaphor of playgrounds and playpens in education, and how it's potentially useful in thinking about artificial intelligence.

As a starting point, it's worth taking a moment to think about why AI presents such a unique challenge and opportunity to learning and education -- especially in universities.

Since ChatGPT hit the public scene in 2022, generative AI has impacted nearly every part of higher education. In some cases this has led to new AI tools being embraced by educators and administrators. In others there's been active resistance to AI in any form being used in teaching or by students. And of course there are the continuing fears that AI makes it easier for students to cheat, or to become lazy learners, or simply not to retain understanding that comes from using AI as a learning aid.

But whether you're an AI optimist, an AI pessimist, or simply in denial, it's nearly impossible to ignore the reality that AI is having a substantial and growing impact on learning and education.

At the same time, we're seeing a large gap in understanding between where the leading edge of AI capabilities are, and where educators think they are. As a result, there's still a tendency to think of AI as a tool that can complete assignments or write essays, or create personalized learning environments, or simply act as a form of Google on steroids.

Yet the reality is much more complex -- and much more transformative. Because advanced AI models are becoming increasingly capable of simulating aspects of ourselves that define us at a fundamental level -- such as the ability to think, to reason, and to solve problems with agency -- they stand apart from pretty much any previous technology or tool that we've created.[2] And because of this, they cannot be approached as just another technology to teach students about, or another tool to enhance traditional approaches to education.

Rather, we're seeing a growing need for completely new ways of thinking about the intersection between learning, education, and AI -- especially where educators are sometimes (perhaps often) further behind the curve than the students they're trying to educate.

And this is where the perspective shift inherent in moving from a playpen to a playground mentality becomes useful -- and important.

It's something of an oversimplification, but curiosity and problem-solving are close to the heart of how we learn.[3] These innately human attributes are often used to great effect by educators as they help students acquire specific skills and understanding.

More often than not though, learning environments that are grounded in curiosity and problem solving are constrained. They draw on and leverage these traits, but they are are often explicitly designed to ensure students follow a carefully curated path to achieving specific goals and outcomes.

They are, in effect, metaphorical playpens. Some creative play and problem solving is allowed. But just as a playpen wraps a young child in constraining walls and presents them with a few select toys to channel their attention, these metaphorical learning playpens are designed to channel and restrict learning along narrow lines.

It's a mode of teaching that works well where the purpose and goals are clear, the journey is well-trod, and the desired outcomes are easily assessed. But it quickly falls apart where the the goals and purpose are unclear, the journey is breaking new ground, and no-one's quite sure what the desired outcomes are -- never mind how they should be assessed.

This is very much like the situation we find ourselves in with AI. This is a technology where the speed of development is far outstripping the far more sedate pace of pedagogical reform; where informal exploration by students means they already know more than their instructors in many cases; and where we're still struggling to grasp the full implications of machines that emulate and exceed some of the most fundamental aspects of what it means to be human.[4]

This is where playpen-style learning environments -- which still assumes that expertise and authority lies with the instructor -- run out of steam very fast. And it's also where the metaphor of the playground becomes interesting.

In contrast to a playpen, a playground is relatively unbounded. There are rules of course -- don't be stupid for instance, be kind, don't spoil things for others. But the idea of the playground is to empower users to learn from experience -- and from each other; to flex their imagination; to be creative; to try new things; to make mistakes; to ask questions; and to creatively solve problems.

Playgrounds are environments that open up rather than constrain options, and that allow curiosity and creativity to be transformed into invention and innovation. Yet they are not totally without bounds. Rather, playgrounds are carefully designed and curated to stimulate curiosity, imagination, and play. And through play, learning.

And this is where the analogy comes back to AI and higher education. Rather than trying to control how students explore, use, and think about AI, I suspect we should be giving them more opportunities to explore and play -- to make mistakes, discover new possibilities, to invent, and to innovate. And we should perhaps be thinking of educators -- in this context at least -- as guides, mentors, and fellow-travelers, rather than the fount of all knowledge.

This is, of course, a risky strategy -- especially as it means relinquishing some control over the learning environment. But given how rapidly AI is advancing, the greater danger I suspect is in holding students back because of misplaced ideas about how and what they should learn.

This becomes all the more important where any playpen-like constraints placed around AI in education reflect misconceptions around how AI is beginning to transform our lives. As I noted above, AI is no longer simply about generating text in respond to a prompt -- and I'm not sure it ever was. Rather, it's a collection of technologies that are deeply and fundamentally changing what we do, where we live, and even who we are -- in ways that no-one fully understands yet.

As a result, we are all on a journey together as we explore and learn how to develop and use rapidly advancing AI capabilities to enhance and improve lives in ways that lie far beyond conventional thinking and understanding. And this includes educators as well as students. And one of the ways we can approach this is to adopt a playground mindset rather than a playpen attitude to AI and learning -- and provide students with the space and opportunities to flex their curiosity and hone their problem-solving skills without unnecessary constraints or restraints.

What this might look like is, of course, part of the journey. I suspect there are as many types of AI learning playgrounds as there are people imagining and creating them. But I suspect that we need to get serious about at least the following four areas:

I'm interested in whether anyone is already creating spaces like this for students to learn through play with AI in higher education. I hope they are, as otherwise it's hard to imagine how we'll make serious progress in equipping our students to thrive in an AI future.

And please do look out for the follow-on post where I'll be focusing more broadly on my conversation with Brian while also "playing" with AI in my own Substack playground!

Notes

[1] That original post that I'd intended to write is now readable here: https://futureofbeinghuman.com/p/rethinking-higher-education-in-the-age-of-ai

[2] Mark Daley has a great article out on how AI is categorically different from any preceding technology. As he writes, "Steam engines didn't challenge our claim to unique intelligence. Telegraph wires didn't ask us to rethink the nature of human thought. Electricity didn't insinuate that it might rival or surpass our creative spark. The changes AI brings are not limited to external upgrades in productivity or convenience; they reach inside, prodding us to question the crux of human identity." AI isn't running water: It's something different. March 13, 2025

[3] They are also core to what makes us "us."

[4] This is beyond this particular article, but there is a categorical error I believe in treating a technology that fundamentally challenges our thinking about who we are, and the very nature of what it means to be human, as a leaning aid.

The Artisanal Intellectual in the Age of AI

How is advanced artificial intelligence forcing a rethink of the value of human intellectual labor?

Date: February 16, 2025 Source: https://www.futureofbeinghuman.com/p/the-artisanal-intellectual-in-the-age-of-ai Newsletter: The Future of Being Human (Substack)

I must confess that, since getting my hands on OpenAI's Deep Research (not to be confused with the just-released Deep Research feature on Perplexity), I've been intrigued by how it's forcing me to rethink what it means to be someone who makes a living by thinking.

Last week that led to me exploring Deep Research's ability to research and write a complete PhD dissertation -- unaided, apart from some final formatting. While the resulting ~400 page document fell short of matching up to a true scholarly dissertation it was, nevertheless, impressive.

But AI-only "intellectual labor" is -- at least at the moment -- less interesting to me than what a person and a powerful reasoning/research AI might be able to achieve together.

And so I set myself the task this week of seeing what's possible when I combine my own "intellectual labor" with OpenAI's Deep Research.

The result -- and my notes on the process -- can be found below, along with an intriguing extra if you get to the end of the notes.

Of course, people have been augmenting their own abilities with generative AI for ages now, so you might be thinking there's nothing new here. But I would beg to disagree.

For the first time in my experience as a pretty well established and respected academic, it feels like AI is capable of extending what researchers, academics and scholars can achieve beyond anything we've seen before. And this is what I was interested in exploring.

Building on last week's article, I decided -- somewhat ironically -- to focus on a concept I touched on in the footnotes of last week's article: that of the "artisanal intellectual."

As I wrote back then,

I think we do have to grapple with the very real possibility that AI is becoming a powerful catalyst and accelerant in research that will relegate human-only research to a class of artisanal intellectualism where the primary purpose is the provenance and process, not the product.

The idea of the artisanal intellectual was a bit of a throwaway at the time. But since then I've been finding myself increasingly intrigued by it -- including mentioning it in a number of panel discussions and in the latest episode of the *Modem Futura* podcast.

What made it particularly relevant to the current exercise is that it's not a term that's been widely used or written about in the past. And so Deep Research was going to have its work cut out as it researched it.

My process -- and this is covered in the Notes below -- was to get Deep Research to take the first pass at researching and writing an article, and then for me to go in and make line by line edits, checking, adjusting, and adding in citations along the way.

The result is substantially better than anything I could have pulled together on my own in the time I had. And the final article is definitely better than Deep Research's initial draft -- and what the AI could have produced unaided.

I'd argue that the collaboration is also genuinely generative as it explores and expands understanding around what it might mean to be an artisanal intellectual in an age of AI.

This is, admittedly, a rather quick and dirty demonstration of what can be achieved through collaborating with the cutting edge of AI. But it's also one that would not have been possible a mere couple of weeks ago.

It's also a process that has significantly advanced my thinking around how an academic working with the cutting edge of AI can far surpass what either could achieve on their own.

Let me know what you think of the experiment and the ideas, and do check out the Notes if you're interested in the process and where it led.

The Artisanal Intellectual in the Age of AI

There's a growing sense that the cutting edge of AI is beginning to upend ideas around what it means to be a scholar or an intellectual -- or an academic for that matter -- that have persisted for millennia. Our ability as humans to think, to reason, to develop and explore abstract ideas, and to conceive of ways of understanding the world beyond what we can experience directly, has been core to defining who and what we are almost since the dawn of homo sapiens.

And yet the emergence of AI models that simulate -- and may soon exceed -- human-level thinking and reasoning -- is beginning to challenge the value of human-only intellectual endeavors. Reasoning models like OpenAI's o1 and DeepSeek's R3 have only been out a matter of months yet are already being used to augment research and scholarship in potentially transformative ways. And with the release of OpenAI's latest reasoning model Deep Research, it feels like we're on an accelerating path toward human intellectual exceptionalism being a somewhat outmoded idea.

This possibility got me thinking recently about the idea of the "artisanal intellectual" -- the possibility that human-only intellectual pursuits may soon be relegated to a realm of thinking that is more valued for its human provenance than its practical use. The same concept is easily applied to the idea of the artisanal scholar or artisanal academic -- although I suspect that in many people's minds most academics and scholars are already firmly in the "artisanal" camp!

The idea that what was once considered to be the pinnacle of human achievement becoming an artisanal activity in an age of AI is an intriguing one. But it's also one that deserves framing within a much broader historical context as, useful as it might be as AI capabilities continue to advance, it's not entirely new.

The origins of intellectual craftsmanship

The idea of the intellectual as a craftsman has deep roots, even if the term "artisanal intellectual" itself is not widely used. In classical philosophy, thinkers distinguished between different kinds of knowledge -- episteme (theoretical understanding) and techne (craft or art) -- recognizing that certain wisdom comes from skilled practice. Centuries later, scholars like C. Wright Mills explicitly described scholarship as a craft. In his 1959 essay "On Intellectual Craftsmanship," Mills argued that true scholarship is "a choice of how to live as well as a choice of career", an ethos of disciplined habits and personal commitment (Mills 2000). He portrayed the scholar as an "intellectual workman" who "forms his own self as he works towards the perfection of his craft." This view positioned research and writing not just as tasks or outputs, but as a way of life akin to the artisan honing a skill.

This is very much reflected in Dominic Boyer's thinking in his essay "The Medium of Foucault in Anthropology" where he defines an artisanal intellectual -- possibly the first time the phrase is formally used -- as a "knowledge-maker who has a relatively immediate and sensuous relationship to the epistemic forms s/he is producing instead of relations that are strongly capitalized, in other words, austere and market-mediated" (Boyer 2002).

Approaching intellectual traditions as craft more broadly, many such traditions reflect artisanal qualities. Medieval monastic scribes, for example, treated manuscript copying and illumination as a meticulous craft, embedding personal artistry into scholarly preservation of knowledge. Early modern scientists often built their own instruments and conducted hands-on experiments, blending manual skill with intellectual inquiry -- a model of "knowledge by doing." Michael Polanyi, a 20th-century philosopher of science, highlighted that much of what experts know is tacit and learned through apprenticeship, not through formulas (Polanyi 1966). He famously observed that "we can know more than we can tell" emphasizing that skilled understanding (like riding a bicycle or diagnosing a patient) often can't be fully captured in explicit rules. Polanyi's insight underscore the artisanal aspect of knowledge: master practitioners developing intuition and know-how that defies complete codification. In short, long before AI, thinkers recognized that intellectual work involves personal craftsmanship -- the slow accumulation of judgment, creativity, and tacit skill.

The notion of the artisanal intellectual also resonates with broader philosophical critiques of technology. In 1954, Martin Heidegger warned that modern technology could enframe humanity, making us view the world as mere resource or "standing-reserve" (Heidegger 1954 (Translated 1977 by William Lovitt)). He contrasted this with more authentic ways of "bringing-forth" truth, akin to a craftsman's revealing of meaning through work. Today, scholars draw on Heidegger's insight to caution against an overly mechanized approach to knowledge. If research becomes fully automated, we risk losing the "knowledge-seeking journey itself," the reflective process that gives scholarship its human depth (Leahy and Maynard 2025). This concern echoes through the decades: from the 19th-century Arts and Crafts movement (which valued handcraft against industrial mass production) to modern authors like Richard Sennett, who argues that "the spirit of craftsmanship" -- the "desire to do a job well for its own sake" -- is an enduring human impulse (Sennett 2008). Such perspectives provide a historical and philosophical backdrop for thinking about scholars in an AI age: they remind us that intellectual labor has long been cherished as a personal art, not just an output.

Contemporary Trends

In the current era of advanced artificial intelligence, the concept of an "artisanal intellectual" is gaining new salience. The phrase itself is beginning to circulate in discussions about AI and academia. For instance, a recent conversation on the Future of Being Human Substack explicitly defined "the artisanal intellectual as someone who thinks without using AI" (Maynard 2025). In other words, some modern scholars are framing the choice to not use AI tools in research and writing as a potential deliberate, value-driven stance -- akin to a craftsperson choosing hand tools over power tools for greater control and authenticity. This notion is likely to become increasingly salient as AI systems grow capable of generating essays, writing and debugging code, simulating reasoning, and conducting extensive and iterative text-based research. This is where some -- like Maynard -- are asking whether future scholars might bifurcate into those who rely on "intelligent" machines and those who pride themselves on a more handcrafted intellectual approach (Maynard 2025).

Some academics are actively beginning to reflect on what distinguishes human intellectual labor from machine-generated work. A key theme that's emerging is authenticity and process. Critics of heavy AI reliance argue that something essential is lost when people outsource thinking and writing to algorithms. They worry about the erosion of what one writer called the "hard thinking and writing work" behind truly well-crafted papers (Lindebaum 2025). Scholars like Dirk Lindebaum caution that using tools like ChatGPT to speed up publications can become a temptation to "skip the hard thinking...in pursuit of a longer list of papers," ultimately impoverishing analytical skills. In surveys, academics voice fears that over-reliance on generative AI will deskill researchers and "lead to disinvestment of and alienation from authentic and idealised versions of academic personhood" (Watermeyer, Lanclos et al. 2024). In other words, if one lets an AI assemble literature, formulate arguments, or polish prose, the researcher might gradually lose the craft of those activities -- much as a craftsman loses skill when a machine takes over the handiwork.

In response, there's an emerging trend toward some scholars advocating for a "slow scholarship" or human-centered approach -- the equivalent of a scholarly "slow food" movement -- pushing back against the pressure to use AI for constant efficiency. Rather than celebrating AI's ability to churn out more content, these voices emphasize quality, deliberation, and the uniquely human elements of research. For example, a recent analysis noted that while generative AI can save time on drudgery, few academics were using the "gift of time" to engage in deeper reflection or creativity (Watermeyer, Lanclos et al. 2024). Instead, most were simply increasing their output to meet performance metrics. This has led commentators like Watermeyer et al. to ask whether academia is "losing sight of the significance of tasks that are easily automated." Some argue that the process of research -- reading, note-taking, writing and rewriting -- has its own scholarly value in developing understanding. If AI streamlines these processes too much, it may introduce what one paper calls "algorithmic conformity" (Liel and Zalmanson 2024), squeezing out serendipity and the unexpected insights that come from wrestling with information personally. In essence, a human-centric trend in scholarship calls for maintaining the craft of inquiry: encouraging academics (and students) to do some things "the hard way" to preserve imagination and critical thinking.

Yet not all contemporary discussion sets human and AI in opposition. Many scholars and educators are exploring ways to integrate AI as a tool rather than a replacement -- echoing ideas from early computing pioneers about "augmenting human intellect" rather than supplanting it. The concept of the scholar as a kind of cyborg craftsman is emerging: using AI for routine tasks while consciously adding human judgment, ethics, and creativity on top. In practical terms, researchers might use AI to summarize literature or generate data but then apply their own critical analysis to interpret results. Or AI might serve as a brainstorming partner that offers on-demand research insights, while the human scholar curates the meaningful questions and ensures intellectual rigor. Effecting this, educators are formulating guidelines for when "generative AI should be used and when it shouldn't," trying to carve out which academic tasks must remain human-driven for integrity's sake (Watermeyer, Lanclos et al. 2024).

This integrative approach treats AI like a powerful new instrument in the scholar's toolbox -- akin to a microscope or a word processor -- that can amplify human capability, although rapidly expanding capabilities are increasingly rendering such analogies weak compared to the extent that AI is beginning to open up new possibilities. Proponents argue that if used judiciously, AI can handle mundane workload (grant formatting, basic coding, proofreading), potentially freeing human academics to focus on higher-order thinking or creative synthesis. The challenge, as widely noted, is defining the boundaries: how to harness AI's benefits without diluting the intellectual craft. Ongoing dialogues in academic communities and publications reflect this balancing act, as the academy feels out a new relationship (and tension) between automated intelligence and artisanal intelligence (Jones 2025, Lee, Sarkar et al. 2025, Wiley 2025).

Future Speculation

Looking ahead, the acceleration of AI capabilities -- especially in reasoning, research assistance, and text generation -- promises to dramatically reshape intellectual pursuits. Advanced AI systems are increasingly able to produce writing that passes for human, solve complex analytical problems, or even generate new hypotheses from data. This raises provocative questions: What will "intellectual work" mean when a machine can draft a competent scholarly article or synthesize a field's literature in minutes? We are likely to see a spectrum of adaptation. On one end, a cadre of AI-empowered scholars will fully embrace these tools, working in tandem with AI co-researchers to achieve feats of productivity and interdisciplinary insight previously impossible. On the other end, we may see the rise of the artisanal intellectual as a conscious identity -- scholars who pointedly work without AI aid (or with minimal use) to preserve a traditional mode of scholarship. Just as handmade artisan goods gained cultural value in an industrialized world, human-crafted scholarship might acquire a certain prestige or trust in an AI-saturated world.

If AI does become deeply embedded in research and knowledge creation, the notions of expertise and intellectual labor will inevitably shift. Traditionally, expertise meant having a vast store of knowledge and the ability to analyze and synthesize that knowledge in novel ways. In a future with AI, the store of knowledge will be instantly available to anyone via AI, and even basic analysis might be automated.

As a result, human expertise may be redefined to emphasize qualities that AI cannot emulate easily. This could mean greater emphasis on creative thinking, ethical judgment, contextual understanding, and interdisciplinary integration. An expert might be valued less for recalling facts or performing routine analysis (tasks AI excels at) and more for posing the right questions, interpreting nuanced human contexts, and guiding AI systems to fruitful outcomes.

Some are already foreseeing scholars becoming more like "knowledge curators" or orchestrators -- selecting, validating, and weaving together insights from AI -- rather than lone authors of every word. The role of intuition and tacit knowledge could become a hallmark of human experts, as these are areas where humans might still have an edge in originality and meaning-making.

In this AI-enabled future, the day-to-day work of scholars might tilt away from certain tasks. For instance, data analysis, coding, transcription, basic writing drafts, and translation are likely to be largely automated by near-future AI. Intellectual labor may become more about oversight, strategy, and big-picture synthesis. We might see researchers spending more time in high-level design of experiments or interpretation of results, while AI agents handle the grunt work. This could elevate the importance of collaboration and communication skills -- explaining and contextualizing AI-generated findings to other humans -- as well as the ability to verify and correct AI outputs.

Conversely, there's a dystopian possibility that scholars could be "proletarianized" -- reduced to mere supervisors of machine output, with diminished creative agency (Watermeyer, Lanclos et al. 2024, Watermeyer, Phipps et al. 2024). If universities and industries push for maximum efficiency, intellectual work could become more assembly-line in nature, with humans simply managing workflows and quality-checking AI's products. This raises concerns about a loss of fulfillment and mastery: will academics feel like craftsmen or like cogs in a machine? The answer may depend on how intentionally we preserve the craft elements of scholarship.

There is a distinct possibility here that the authority of academics and scholarly institutions might be challenged in an AI-dominated knowledge ecosystem. But there are also arguments for ensuring that academic authority continues to be recognized. When AI can generate text that sounds authoritative, how do we ensure information quality and credibility? In the future, the personal credibility and unique voice of human scholars may become even more important -- a form of intellectual branding that assures readers or the public that a human expert (with values and accountability) stands behind a piece of work. We might see a stronger emphasis on transparency about AI use in publications (e.g. statements of human contribution), and perhaps a premium placed on work that is demonstrably human-crafted, as a mark of rigor or ethical integrity.

There is a possibility that, if most people use AI, those few who opt-out -- the artisanal intellectuals -- might garner exclusive prestige for carrying the torch of undiluted human scholarship. This is hinted at by Watermeyer et al. as they consider the potentially transformative impacts of generative AI on academia (Watermeyer, Phipps et al. 2024). However, this could create a new inequality: only well-resourced or tenured scholars might be able to afford the slower, manual approach, while others feel pressured to use AI to stay competitive. Academic authority might thus bifurcate, with a small elite claiming the mantle of authenticity and the majority working in AI-assisted paradigms.

Optimistic vs. Cautious Visions

The future of intellectual life with advanced AI is still unwritten, and speculation ranges from optimistic to cautionary. Optimists might envision a golden age of research where humans and AI synergize. AI might rapidly crunch numbers, test permutations, generate hypotheses, draft reports, and even conduct original research, while human thinkers set directions and make conceptual leaps. In this vision, scholars could tackle grand challenges (climate, disease, fundamental physics) more effectively, leveraging AI as a tireless collaborator. Freed from some drudgery, an academic might have more time to focus on core ideas and innovative thinking, essentially doubling down on the creative craft of their discipline with AI as support. Education and learning could also be revolutionized, with AI tutors handling rote learning and freeing students and teachers for deeper mentorship and critical engagement.

Cautionary voices, however, urge that we consider what is lost in translation. If every step of reasoning and writing is accelerated, do we rob scholars of the incubation time that often sparks original insight? There is a real risk of "dehumanization" and loss of intellectual autonomy if academics become overly dependent on machine outputs (Bender 2024). Some might fear a future where research becomes a homogenized stream of AI-generated material, with human academics struggling to imprint their individuality. The extreme endpoint would be an intellectual landscape of abundant information but potentially shallow understanding -- a world where much is written but less is deeply comprehended. In such a scenario, the very capacity for and claims to expertise and the notion of academia as a critical craft could diminish significantly.

Given these possibilities, it can be argued that there is a nuanced approach to the future. This will mean actively shaping practices and policies now: deciding which intellectual skills are essential to cultivate in humans, even if AI can perform them, and figuring out how to use AI to amplify rather than erode the richness of scholarly work. It also means preparing new ethical guidelines and educational methods so that the next generation of thinkers can navigate a world with AI without losing the artisan's touch -- curiosity, critical skepticism, imaginative leap-taking -- that has always driven knowledge forward.

Conclusion

The concept of the "artisanal intellectual" in an AI-driven era encapsulates a vital question: What value do we place on the human craft of thinking, researching, and creating knowledge, when machines can do so much of it for us? History and philosophy remind us that intellectual pursuits have always had a craft element -- a blend of skill, personal dedication, and moral purpose -- that doesn't readily translate into raw efficiency. Contemporary academics are wrestling with this balance in real time, some embracing AI's power, others defending the sanctity of human-centric scholarship. And as we peer into the future, we can imagine profound transformations in how expertise and authority are defined.

Rather than arriving at a simple verdict of "pro-AI" or "anti-AI," the discourse suggests we are entering an age of choice and redefinition. The artisanal intellectual may emerge as one emblem of resistance or differentiation -- a commitment to the craft of intellect in a time of intelligent machines. Their role could be to ensure that the university of the future retains places for slow thinking, mentorship, and the je ne sais quoi of human insight. Meanwhile, those who integrate AI will aim to carry forward the torch of human reason with augmented capabilities, ideally without extinguishing its spark. In all cases, the challenge will be to harness new tools without losing the wisdom, creativity, and ethical reflection that define the best of scholarship. The fullest context of this topic, then, is not a battle between humans and AI, but a conversation about how to preserve and reinvent the art of intellectual work -- ensuring that as our tools evolve, our minds and values evolve with them, intentionally and artfully.

References

Bender, E. M. (2024). "Resisting Dehumanization in the Age of 'AI'." Current Directions in Psychological Science 33(2). https://doi.org/10.1177/096372142312172

Boyer, D. (2002). "The Medium of Foucault in Anthropology." Minnesota Review 58-60: 265-272.

Heidegger, M. (1954 (Translated 1977 by William Lovitt)). The Question Concerning Technology.

Jones, N. (2025). OpenAI's 'deep research' tool: is it useful for scientists? Nature. https://doi.org/10.1038/d41586-025-00377-9

Leahy, S. and A. Maynard (2025). Modem Futura. Artisanal Intellectual: a response to OpenAI's Deep Research. link

Lee, H.-P., A. Sarkar, L. Tankelevitch, I. Drosos, S. Rintel, R. Banks and N. Wilson (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers, Microsoft Research. https://advait.org/files/lee_2025_ai_critical_thinking_survey.pdf

Liel, Y. and L. Zalmanson (2024). "Between formal authority and authority of competence -- the mechanisms of algorithmic conformity." Academy of Management Annual Meeting Proceedings 2024(1). https://doi.org/10.5465/AMPROC.2024.166bp

Lindebaum, D. (2025) "Researchers embracing ChatGPT are like turkeys voting for Christmas." Times Higher Education. https://www.timeshighereducation.com/blog/researchers-embracing-chatgpt-are-turkeys-voting-christmas

Maynard, A. (2025). "AI humility, artisanal intellectuals, and Reid Hoffman's Superagency." The Future of Being Human https://futureofbeinghuman.com/p/ai-humility-artisanal-intellectuals

Mills, C. W. (2000). Appendix: On Intellectual Craftsmanship. The Sociological Imagination. 40th Anniversary Edition. C. W. Mills and T. Gitlin, Oxford University Press.

Polanyi, M. (1966). The Tacit Dimension. Garden City, N.Y, Doubleday.

Sennett, R. (2008). The Craftsman, Yale University Press.

Watermeyer, R., D. Lanclos and L. Phipps (2024). "If generative AI is saving academics time, what are they doing with it?" LSE Impact Blog https://blogs.lse.ac.uk/impactofsocialsciences/2024/01/22/if-generative-ai-is-saving-academics-time-what-are-they-doing-with-it/

Watermeyer, R., D. Lanclos, L. Phipps, H. Shapiro, D. Guizzo and C. Knight (2024). "Academics' Weak(ening) Resistance to Generative AI: The Cause and Cost of Prestige?" Postdigital Science and Education. https://doi.org/10.1007/s42438-024-00524-x

Watermeyer, R., L. Phipps, D. Lanclos and C. Knight (2024). "Generative AI and the Automating of Academia." Postdigital Science and Education 6: 446-466. https://doi.org/10.1007/s42438-023-00440-6

Wiley (2025). ExplanAItions: An AI study by Wiley. https://www.wiley.com/en-us/ai-study

Notes

My process here started out with asking Deep Research for a definition of "artisanal intellectual." This threw up some initial ideas and links and led to a detailed prompt asking Deep Research to research and write about the concept in the context of emerging AI capabilities, including historic framing, contemporary thinking, and speculation on how accelerating AI capabilities might affect the meaning and use of the idea.

The resulting draft from Deep Research was then edited line by line with every reference tracked to the primary source where possible, and the text updated to accurately reflect the source. New sources were also added where appropriate along with further context, and the article's style smoothed out in a number of places.

The result was a paper that, while still being somewhat quick and dirty, is far better than the original Deep Research piece, and far more thoroughly thought-out and researched than could have been achieved alone.

As an aside, Deep Research was able to discover, draw on, and cite two sources that were only a day old -- last week's episode of Modem Futura where the idea of the Artisanal Intellectual is discussed, and the Substack about the podcast which also came out a day before the prompt was given.

The process was illuminating. It forced engagement with unfamiliar literature and demonstrated the value of the craft of the "artisanal intellectual" in assessing and building on the raw material produced by Deep Research.

The big takeaway is just how synergistic and generative this collaborative process was. The time spent was admittedly rather asymmetric, with several hours of human work to Deep Research's few minutes. But this is a foundational paper for continuing to explore the idea of the artisanal intellectual in an age of AI -- and one that makes a useful contribution to broader thinking here.

The final step was to feed the finished article back to Deep Research and ask it what had been missed. The result was a thought-provoking piece on "Missing Perspectives on the Artisanal Intellectual in an AI-Driven Era" -- produced in just a few minutes -- that makes it hard to imagine a future of effective intellectual labor that is not AI-enhanced.

Why Thinking About Tomorrow Still Matters Today

Four years on, Future Rising seems more relevant than ever in a world on the cusp of transformative change

Date: December 08, 2024 Source: https://www.futureofbeinghuman.com/p/guide-to-thinking-about-the-future Newsletter: The Future of Being Human (Substack)

A couple of things happened this past week that got me thinking about my book Future Rising: A Journey from the Past to the Edge of Tomorrow. The first was this week's episode of the Modem Futura podcast, where we explore the growing importance of nurturing future-oriented thinking from an early age with Professor Ruth Wiley. The second was a fundraising email from the XPRIZE Foundation with the heading "You can be an architect of the future" -- a phrase eerily similar to one I use in the book.

Over the years, I've learned that it's rarely a good idea to talk about your own books. But breaking that rule for a moment, I was gratified to find that Future Rising has held up surprisingly well -- despite being four years old. Re-reading the opening introduction and the later chapters, it feels more relevant than ever in the face of today's social, political and technological upheavals.

Whether it's the guide we need to nurture fresh ways of thinking about the future is, of course, for others to decide. And I must confess that, while I suggested that "humans are, in a very real sense, architects of the future" back in 2020, I'm still not entirely sure how useful this metaphor is.

And yet, there's a growing hunger for new perspectives on how we collectively and individually approach the future. And my sense is that Future Rising still addresses this need in ways that that few other books do.

If you'd like to know more, this interview with Steve Goldstein on NPR/KJZZ's The Show offers a great introduction. You can also download the book's introduction here.

But honestly, the best way is to pick up a copy and dive in!

Four ways of thinking about advanced technology transitions

Can a simple analogy help understand different approaches to navigating technology-driven tipping points and transitions?

Date: August 18, 2024 Source: https://www.futureofbeinghuman.com/p/four-ways-of-thinking-about-advanced-technology-transitions Newsletter: The Future of Being Human (Substack)

If you've been following my work for some time, you probably know that I have a thing about "Pippard's ladder."

It's an elegant example of sudden yet hard-to-predict changes in seemingly well-behaved systems that I was first introduced to back in the 1990's by Cambridge University physicist Brian Pippard. And it's one that, I think, is potentially useful for exploring approaches to navigating advanced technology transitions -- so much so that I thought I'd give it a whirl in my keynote at this past week's IEEE International Symposium on Consumer Technology.

I first wrote about Pippard's ladder in my book Future Rising. It was in a short chapter on "boundaries" and explored how the demonstration provides insights into dynamic discontinuities, or tipping-points, in seemingly-predictable systems.

The section is short enough that I thought it worth including below to introduce ideas I've been playing with for some years. But if you want to cut to the chase, please do skip this and scroll down to the thought experiment I explored in my keynote.

Boundaries (from Future Rising Chapter 48)

In 1980, the Cambridge physicist Brian Pippard published a paper describing what he called "experiments in critical behavior and broken symmetry." In it, he explored particular types of transitions between the present and the future that he referred to as "discontinuities" -- the blindsides of the physical world.

Pippard was fascinated by transitions between present and future that were abrupt and irreversible -- transitions that occur at a tipping point beyond which everything changes and there's no going back, such as the snapping of a branch, or the breaking of a wave.

I probably wouldn't be aware of Pippard's work if I hadn't attended one of his public lectures as a PhD student. In the lecture, he held up a simple model of a vertical ladder consisting of four evenly spaced wooden rungs, held together by two lengths of string. He then asked the audience what would happen if he slowly rotated the bottom rung through one complete horizontal revolution. Naturally, we predicted that Pippard's model would smoothly transform from its conventional ladder-like form into something that looked more like an artist's impression of a strand of DNA -- a neat double helix, consisting of two lengths of string held apart by the wooden rungs. And of course, our vision of the future was utterly wrong.

As Pippard twisted his ladder, the lengths of string between two of the rungs suddenly twisted together, destroying any semblance it had to DNA. The result was a tangled mess.

This was an abrupt and irreversible transition -- reversing the twist failed to untangle the ladder. But the point at which it occurred -- and the point at which some irreversible boundary was crossed -- was all but impossible to predict.

Over the intervening years, it's become increasingly common to talk about tipping points -- hard-to-predict points of instability in seemingly stable systems -- especially in the context of climate change. Just as Pippard's ladder demonstrated, there are concerns that we're in danger of crossing such boundaries that mark a point of no return as we continue to stress the environment. And if we do, we risk disrupting, and even destroying, critical pathways to the type of future we'd like to see.

Pippard's ladder is an example of nonlinear dynamics. It represents the tendency of complex and interconnected systems to undergo rapid and irreversible changes when stressed. And it's a sobering reminder that, even though things may look great in the present, unless we learn how to spot early warnings and stay clear of critical tipping points, we run the risk of, quite literally, crashing our future.

A Thought Experiment in Navigating Advanced Technology Transitions

While I've used Pippard's ladder in the past as a metaphor for tipping points that we might want to avoid, I was interested in whether it could be extended to thinking about different ways of approaching an uncertain future.

And so I visited our local Lego store, raised my wife's craft supplies, and created my own rather crude (but nevertheless serviceable) ladder.

Experimenting with the ladder while thinking through the concept of advanced technology transitions, I was interested in whether there were different approaches to navigating transitions that the ladder both illustrated and provided insights into.

The result was a quadrant framework defined by degrees of freedom in exploring pathways through transitions, and the mindset that transitions were approached with.

The "degrees of freedom" axis represents a more restrictive approach to problem solving on the left and a more open approach on the right, with the open approach leading to more options and thus more degrees of freedom in the choices that are available.

And the "mindset" axis represents a tendency to try and maintain things as they (bottom) versus a willingness to embrace change (top).

The model is based on an assumption that we are living in a closed system -- the planet we live on -- and that instabilities or tipping-points occur when we begin pushing against the boundaries of this system, whether these are related to the physical system itself or the ways we live in and utilize resources within it.

It also assumes that we cannot simply turn off technology innovation, but instead need to find ways to manage and channel the inevitability of technological change.

I suspect some will disagree with this latter assumption. But as change is fundamental to living within a dynamic universe, I'm comfortable with it.

This quadrant model is admittedly simplistic. But nevertheless I think it's useful in thinking about broad brush approaches to technology transitions.

And some of this utility comes in exploring insights associated with each of the quadrants:

"Avoid" Quadrant

Starting with the bottom left quadrant, this can be seen as representing strategies that avoid tipping points by staying well clear of them. In the case of the ladder (as shown in the video above), great care is taken to stay clear of the point at which a sudden transition becomes increasingly likely.

It's an avoidance strategy that's common in approaches to addressing climate change, and one that is indicative of at least some efforts to avoid technology-driven societal impacts.

This is a quadrant where the societal benefits of technology innovation are critically weighed against potential adverse impacts, and decisions are risk-averse and substantially informed by social factors (equity and wellbeing for example).

Approaches to navigating advanced technology transitions in this quadrant might cover policies and other governance mechanisms designed to slow innovation where the outcomes are uncertain. They're also likely to include proceeding cautiously while embracing approaches to anticipating potential impacts of advances across society.

"Adapt" Quadrant

An alternative to simply avoiding tipping points is to actively find ways of preventing them -- or pushing them out into the far future to give us some breathing space in the near future.

This is represented in the lower right "adapt" quadrant, which still represents a preservation mindset, but one that embraces ways of pushing tipping points further out into the future rather than simply avoiding them.

In the thought experiment using Pippard's Ladder, this is represented by stabilizing potential instabilities within the system -- in this case, as is seen in the video above, using clothes pegs!

This quadrant has clear parallels with adaptation strategies to climate change, where rather than avoid behaviors that bring us closer to climate-related tipping points, we actively explore solutions to preventing such behaviors leading to disruption.

Strategic use of renewable energy sources might be considered as an example here as they help adapt to a world that has been pushed out of equilibrium by excessive use of non-renewable sources. A more controversial example might be the use of geoengineering to reduce the impacts of human behavior on climate change.

In the context of advanced technology transitions, this quadrant represents initiatives that focus on better-understanding instabilities that are potentially triggered by advanced technologies -- the impacts of generative AI on learning for instance, or social cohesion -- and developing new ways of building resiliency against such threats.

In effect, it considers how theories, models, methods, and practices might be developed that allow society to absorb and adapt to change, without experiencing abrupt and potentially catastrophic transitions.

As is seen in the video of the ladder above, it's an approach that can build resiliency into things for a short time -- but within a closed and constrained system the chances are that it just delays the onset tipping points rather than eliminates them.

"Extend" quadrant

The previous two quadrants assume that we're constrained by hard boundaries imposed by living on a planet with finite space and resources.

But what if we're not?

There is a non-intuitive approach to avoiding or navigating tipping points that is not represented by the lower two quadrants in the model. And that is -- as is shown by the extended ladder in the video above -- to refuse to accept that we are limited by boundaries we seemingly constrain us, and to extend them.

This is where things can get weird, as one way to navigate advanced technology transitions in this context is to literally transcend the constraints of living on Earth by becoming an interplanetary species.

While I'm a skeptic of the idea of other planets -- most notably Mars -- being a "plan B" that Elon Musk and others advocate for, we are already beginning to expand beyond the physical constraints of the planet we live on through the use of near earth orbit, plans to establish a presence on the Moon, and aspirations to extract resources from the moon and asteroids.

Humanity's increasing presence in space is certainly one way to extend the boundaries that lead to tipping points. But they are only one of many creative ways to extend the boundaries we might seem to be constrained by. The prospect of nuclear fusion or other, more esoteric energy sources, is another. So is using advanced technologies to break beyond conventional constraints and extend what is possible.

For instance, developments in AI, bioengineering, quantum technologies, and even human enhancement, are all making possible what was once considered to be impossible, and it's important to ask whether this in turn is allowing us to extend what were once thought as immoveable boundaries to thriving in the future.

And audacious as this might sound, the past ten thousand plus years of human history is full of examples where technological breakthroughs have redrawn the map of what is possible. Just to name three out of a long, long list, harnessing steam power, the invention of synthetic fertilizers, and the advent of the internet, all redefined boundaries that previously constrained us as a species.

With a creative mindset and a willingness to believe that seemingly-sacrosanct boundaries are transcendable, it becomes possible to push potentially disruptive tipping points far into the future.

But this still just puts off potentially catastrophic tipping points rather than eliminating their existence. Which brings me to the last quadrant -- the "embrace" quadrant.

"Embrace" quadrant

Up to now, this admittedly simple model has considered approaches to navigating advanced technology transitions that seek to avoid sudden and potentially disruptive tipping points. This is vey much in line with my original thinking around the metaphor of Pippard's Ladder.

But what if, instead of avoiding tipping points, we embraced them?

This is possibly the most challenging quadrant in the model. And to be honest I'm not sure what embracing a disruptive technology-driven tipping point might look like -- and especially how it might play out with respect to who thrives and who does not through the transition.

But thinking long-term, if such tipping points are inevitable at some point in our collective future, it's worth thinking through scenarios where we embrace what's on the other side of them.

This is a quadrant that feels, on the face of it, quite perilous, as history has shown that disruptive tipping points are rarely painless. And yet, as I mention above human history is replete with examples where the world before a new set of capabilities or inventions, and the world after their development and adoption, is night and day different.

You could, in fact, argue that technology driven tipping points are the norm rather than the exception in human existence, and we're currently in a stable patch that isn't likely to be stable for much longer.

If this is the case, I would argue strongly that we need to thinking more critically about what it means to be in that top right quadrant, and we need to be developing the knowledge and insights necessary to ensure harm is minimized and benefits maximized when we do encounter the next disruptive and irreversible tipping point in human history.

Just a ladder?

This is, as I noted earlier, merely a thought experiment designed to stimulate new thinking. It may be so deeply flawed that it should be resigned to the trash can of bad ideas. Or there could be something to using a simple example of a four-rung ladder to explore how we approach advanced technology transitions.

Either way, the exercise does underline the necessity of at least thinking more critically, creatively, and innovatively, about how we collectively and successfully transition from the present we're in, to the future we aspire to.

Responsible AI: Lessons from Nanotechnology

20 years ago we were learning how to navigate the risks and benefits of nanotechnology. Two decades on, are we applying those hard-won lessons effectively to artificial intelligence?

Date: October 02, 2023 Source: https://www.futureofbeinghuman.com/p/responsible-ai-lessons-from-nanotechnology Newsletter: The Future of Being Human (Substack)

At first blush nanotechnology and artificial intelligence may not seem to have that much in common. And yet there are surprising similarities when it comes to avoiding failures in a society where the success of transformative technologies depends on far more than technical knowhow alone.

Despite this, it's not at all clear that we're learning from the past as we rush headlong into an AI future.

My colleague Sean Dudley and I explore this further in a new commentary in the journal Nature Nanotechnology and an accompanying article in The Conversation -- and conclude that there's a lot to be learned the transdisciplinary initiatives and broad stakeholder engagement that underpinned nanotechnology.

In the articles we draw on the early days of nanotech development -- something I was at the heart of as I co-chaired the interagency Nanotechnology Environmental and Health Implications working group, and later served as science advisor to the highly influential Project on Emerging Nanotechnologies. And we make the case for greater investment in understanding and navigating advanced technology transitions in ways that "bridge disciplines and sectors, and bring together people, communities, and organizations with diverse expertise and perspectives to investigate emerging landscapes and drive toward a more equitable, sustainable, and promise-filled future."

Plenty of mistakes have been made in the development and use nanotech over the past two decades -- but we've also learned a lot about the importance of working with experts from the arts, humanities, and social sciences, in addition to those at the forefront of nano-specific science and technology. We've also learned that broad stakeholder and public engagement are absolutely critical to success.

As artificial development gathers pace, these lessons don't seem to be getting through though. Development is still being driven by a small group of experts and companies who believe that they have all the understanding they need. And while there's a growing urgency around how to ensure the safe and responsible development of AI, there's still a reluctance to engage a diversity of voices and perspectives in these conversations.

This is a serious mistake, and one that needs to be corrected as soon as possible. We've learned a lot from previous advanced technology transitions like nanotechnology. I'd include the development of technologies like genetically modified organisms here, which was a masterclass in how naivety, hubris, greed, and a lack of broad engagement, can create near-insurmountable roadblocks to progress. In fact early investment in responsible nanotech drew heavily on lessons learned from the GMO debacle.

As AI development continues to accelerate, we cannot afford to get things wrong -- if anything, the stakes here are far higher than they were with either nanotechnology or GMOs. But to do this, AI needs to learn from the lessons of the past if it's to lead to a better future -- and fast.

Read more in:

Navigating Advanced Technology Transitions Using Lessons from Nanotechnology. Andrew D. Maynard and Sean M. Dudley. Nature Nanotechnology, October 2, 2023.

Navigating the risks and benefits of AI: Lessons from nanotechnology on ensuring emerging technologies are safe as well as successful. Andrew D. Maynard and Sean M. Dudley. The Conversation, October 2, 2023.

End of llms-full.txt. For the concise index version with links, see https://andrewmaynard.net/llms.txt* For the complete text of AI and the Art of Being Human, download the free AI Companion at https://www.aiandtheartofbeinghuman.com/ai-companion*

DOCUMENT_END_MARKER: COMPLETE — This file contains: biography, key concepts, philosophy and approach, 8 book and book-resource entries, 21 tools, research arc, 80-paper bibliography, full academic CV, 28 paper abstracts, 8 website pages, and 16 Substack essays.