S7: Books and other items — what they add to the map#

Supplement to 05-maynard-risk-and-ai-map.md (the map is read, not edited). Share S7: the Future Rising files, the two AI and the Art of Being Human files, the 2025 USDOT foresight report, and four co-authored biopreservation and research-ethics papers. Prepared 26 September 2026. It deals only with Andrew Maynard’s own thinking. It makes no comparison with any other material.

How to read this file#

• Coverage. Future Rising (all five files) and both AI and the Art of Being Human files were read in full. The USDOT report was read in full, although the brief allowed a skim. The four biopreservation and research-ethics papers were skimmed for risk governance and anticipatory engagement. Their governance and engagement sections were read closely.
• Citations. Items are cited by filename and page.
• Future Rising: printed book page (“FR p.150”), plus the file that holds it.
• AI and the Art of Being Human: the Pocket Edition’s [Page X] markers (“Pocket p.109”). For the 50-page preview, the printed book page (“Preview p.48”).
• USDOT report: printed report page.
• Journal articles: journal page.
• Labels. “My reading” marks interpretation. The map’s two Andrew-supplied notes are called Note 1 (quantitative risk assessment stays a foundation that has been built on, not abandoned) and Note 2 (the lack of quantitative methods for AI reflects humility about the hubris of risk assessment, joined to the need to keep grappling with emerging issues).

Provenance at a glance#

A note on the index. README.md describes 2020_Future-Rising_Subscriber-Excerpts.pdf as “a two-page AI-use disclosure”. That is wrong. It is an 88-page PDF: a preface letter to andrewmaynard.net subscribers, followed by 21 chapters (1–3, 17, 19–23, 39–43, 46, 51, 53, 55, 59, 60) and the Afterword. 2020_Future-Rising_Excerpts.pdf (3 pp.) is that preface letter alone. The letter is undated. It postdates publication and probably comes from 2021–24.

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

Provenance#

Sole-authored book, published in 2020. Taken together, the five files give the Introduction, Part I in full (chs 1–17), chs 19–23, 39–43, 46, 51, 53, 55, 59 and 60, the Afterword, the full table of contents and the subscriber preface. That is 33 of 60 chapters. Missing are ch.18 and chs 24–38, 44–45, 47–50, 52, 54 and 56–58, which include Fear, Loss, Despair, Stories, Acceleration, Singularity, Threat, Blindside, Restraint, Boundaries, Cataclysm and Responsibility. Until now the map knew the book only through his posts, above all the one-quotation-per-chapter post of 2024-09-08. The full text here adds the arguments around those quotations.

The argument#

The book is a sequence of 60 short reflections. They move from cosmology (Earthrise, Origins, Entropy, Emergence) to evolved “future-senses” (Anticipation, Instinct, Causality, Memory, Learning, Intentionality, Intelligence, Knowledge, Reason). Part 2, “Uniquely Human”, covers Faith, Imagination, Curiosity, Creativity, Art and the rest. Part 3, “Building the Future”, covers Complexity, Hubris, Delusion, Perception, Deception, Threat, Blindside and Change. Part 4, “The Edge of Tomorrow”, covers Humanity, Morality, Stewardship and Futurerise.

The thesis runs throughout. Humans are “profoundly talented architects of our own future” (FR p.18). But our power to shape the future now exceeds our understanding of how to do so responsibly. So we need a “reset” in how we think about the future “and our responsibility to it” (Excerpts.pdf p.3). The “we” is “everyone, not just experts and policy makers (who are often stymied by their own blinkered views)” (Excerpts.pdf p.2).

Key concepts (his terms)#

• Future-senses: instinct, anticipation and intuition as evolved ways of foreseeing the future (FR p.55).
• Architects of the future; stewards of the future (FR pp.18, 190, 210).
• Mastery of cause and effect: the power of science and engineering, never “by default, benign” (FR p.58).
• Complexity: unpredictable in detail, predictable in broad terms (FR pp.146–148).
• Hubris: productive but dangerous, the source of “false hope” (FR pp.149–151).
• Delusion: intelligence makes us better at rationalising (FR pp.152–153).
• Perception: it matters, but it is “strictly speaking” not everything (FR p.154).
• Deception and “fake future” artists, including deepfakes (FR pp.156–158).
• The “language of change”: mathematics as a tool of foresight, “tempered with humility” (FR pp.165–167).
• The inclusive “we” (FR pp.191–193).
• The gap between what we can do and what we should be doing (FR p.210).
• Futurerise (FR p.212).

What this adds to the map#

Bearing on Note 1 (quantitative risk assessment as a foundation built on). This is the clearest direct evidence in the supplement, and it is his own pre-2023 prose. - Change (ch.46) is a short defence of mathematical modelling as foresight. “Using the mathematics and science of change, it becomes possible to get a sense of where observed trends and trajectories are heading”, and “The same science and math help identify impending dangers, as change threatens to take away what we value” (FR pp.166–167). Quantitative analysis is tied here directly to threats to value. The two definitions of risk that the map calls unintegrated (§8, tension 3) sit together in a single sentence. - The same chapter sets the limit and the ethic in one move. The language of change is “far-from-perfect” because “our lives are defined by more than mathematics alone” (FR p.166). “When tempered with humility and guided by our humanity, our technical mastery of change can help set boundaries around what we don’t know or cannot predict” (FR p.167). This is Note 1 and Note 2 in one passage: quantitative tools are kept and valued, and they are bounded by humility. - Knowledge (ch.16) praises the Apollo engineers. Their knowledge let mission leaders “predict the future of a journey into the unknown with impressive accuracy”. The chapter contrasts them with the attitude “We don’t know that it won’t work, so let’s just try it and see” (FR pp.72–73). Rigorous prediction, where it is possible, is the standard, not the enemy. - Complexity (ch.39): “Of course, there remains a lot that we can predict”. His examples are the consequences of not vaccinating and, “in broad terms at least”, human effects on the environment (FR p.148). His scepticism is aimed at precision in complex systems, not at prediction as such. - Perception (ch.42) opens by rejecting the risk-communication slogan that “perception is everything”: “This is, of course, strictly speaking, not true. No matter how much you fear flying, it isn’t going to affect the likelihood of a crash” (FR p.154). This is a realist anchor for his “risk is social” commitment (C5). Probability of harm stays real. Perception governs decisions, and the task is to “recognize when our perceptions diverge from reality” (FR p.155). - Instinct (ch.10): “This is not to say that feelings-based decisions are wrong—far from it. But when they run counter to evidence…” (FR p.56). Evidence keeps its authority over feeling. Feeling is not dismissed.

Bearing on Note 2 (the hubris of risk assessment, and humility). The humility behind his reluctance to quantify AI risk is stated, in general form, in 2020. - Complexity (ch.39): “The more precise we try to be with our predictions of the future, the less likely they are to be accurate”, and “The danger, of course, is that we become so enamored with our brilliance that we choose to overlook this, and act as if the future is something we can fully control” (FR p.148). This is the most direct statement in his own prose of the danger Andrew names: taking comfort in precise-looking methods that cannot capture a complex system. - Hubris (ch.40) is a whole chapter. The war on cancer is his case. “Driven by the hubris of believing that, if only we understand how cancer works, we can fix it, we’ve learned the hard way that we don’t have the ability to find solutions to every problem” (FR p.150). Hubris “encourages belief in the absence of evidence” and “blinds us to thinking critically about what might go wrong”. Certainty that ideas “are going to solve the world’s problems” is “naive myopia” that “too often ends up causing a trail of destruction” (FR p.151). He also keeps hubris’s productive side: “Our hubris pushes us forward to take small but important steps” (FR p.150). - Delusion (ch.41): “There are even indications that the smarter we are, the better we are at justifying our beliefs in spite of evidence to the contrary”. Our own delusions are the hardest to see, the “log sticking out of our own” eye (FR pp.152–153). Humility applies to experts, and to the author himself. - Reason (ch.17): “reason can be blind to the future” when causal threads “are so complex and convoluted that they defy analysis” (FR pp.76–77). - Afterword: “a gnawing worry that our growing ability to mold and change the future continues to exceed our understanding of how to do this responsibly” (FR p.215). This is the “sheer lack of understanding” of Note 2, stated as a general condition of the age, six years before his 2026 writing on AI. - Humility does not mean inaction. The same book insists on grappling: - “almost unimaginable levels of responsibility as, together, we grapple with what we want it to look like, and how to avoid costly and potentially catastrophic mistakes” (FR p.26); - “if we don’t know what we’re doing, the chances of breaking something along the way are pretty high” (FR p.72); - intention without understanding “can easily do more harm than good, no matter how intentional we are” (FR p.67). - My reading: the book gives a textual basis for Andrew’s own account. The absence of AI risk numbers in his later work is a principled position, humility about precision in complex, poorly understood systems. It is not a gap in his risk-science toolkit. The book does not discuss quantifying AI risk. The link is an inference, but a close one.

New ideas, or earlier origins of ideas the map dates later. - Evolutionary mismatch (2020, not 2026). “Instinct relies on the future being similar to the past”. We have learned “to change the future faster than any evolutionary process can accommodate” and “created a world that our instincts are increasingly poorly equipped to handle” (FR pp.55–56). The map dates the evolutionary-mismatch framing of the cognitive Trojan horse to 2026-01-10 (concept index §1, 2026 entry; C14). The ch.10 quotation is in the corpus (2024-09-08), but the map does not connect it. It is the earliest statement in his own prose of the mechanism behind the 2026 thesis. - The Intelligent User Trap (2020). “the smarter we are, the better we are at justifying our beliefs” (FR p.153) anticipates the map’s lens 11 (“Do not assume the clever or informed are immune”) and the 2026 Intelligent User Trap. - Deepfakes and “fake future” artists (2020). Ch.43 is a full treatment. “Con artists, marketers, politicians, sociopaths—they all depend on playing to our limitations”. “Thankfully, most of us have a finely tuned antenna for spotting deceptions. However, technology is beginning to challenge this” (FR p.156). He asks about fakes that “so incense or enamor us that our ‘fake-o-meter’ simply doesn’t kick in” (FR p.157). He ends: “The hope is that we learn to inoculate ourselves against such fakes… Yet even with these precautions, deceptions are a growing part” of the threat (FR p.158). - This qualifies the map’s deepfake narrative (§5.7; T8). There, a 2019 “passing” aside is followed by a 2024 change of mind from trusting common sense to “far less sure”. In 2020 he already hedged hope with doubt. So 2024 is a sharpening, not a reversal. - It also fills the 2018→2023 gap in the manipulation thread (C14, §2). Between the embodied manipulator of 2018 and the language turn of 2023 there is a 2020 stage. Here manipulation works by faking our perception of the present and by emotional content that disables vigilance. This is a direct precursor of the 2026 idea that fluent AI slips past epistemic vigilance. My reading. - AI in 2020: redundancy and inspiration. “if we crack artificial intelligence, we could be heading toward designing a future in which we are, for all of our capabilities, redundant” (FR p.191). Among possible sources of a new “Earthrise”, he lists “perhaps the emergence of true artificial intelligence” (FR pp.212–213). The Introduction speaks of being “on the cusp of… creating machines that think” (FR p.17). AI is still one converging strand, as the map says. But the specific worry about human redundancy appears here, five years before AI and the Art of Being Human and his “what we do / who we are” model. So does a hopeful reading of “true” AI. - Thriving is named in 2020, not first in 2025. - “everyone has the right to thrive, as long as in doing so, they don’t deprive others of this selfsame right” (FR p.192); - “if we’re to thrive as a species” (Excerpts.pdf p.2). - The map says flourishing and thriving are “named explicitly… from 2025” (C1 firmness line; §5.9). That should be qualified. The language of thriving is present in 2020. What 2025 adds is its central place. - The inclusive “we” as self-correction. “Up to now… I’ve been rather loose with the term ‘we’.” Excluding others from the “we” is “a deeply selfish and destructive approach to building the future” (FR pp.191–192). This is early evidence for the justice thread (C11) and for self-implication (method note), in the same register as his 2024 public correction on disability language (C16). - Could versus should, 2018→2020→2025. “there is a worryingly large gap emerging between what we can do, and what we should be doing” (FR p.210). This links the FFTF touchstone (§5.2) to the Afterword (FR p.215) and to the 2025 book passage attributed to him (§2 below). - Anti-fatalism in 2020. “Dire as the outlook seems, it is not inevitable” (FR p.17). He asks what vision will help us become “what we can be, rather than being resigned to what we assume is inevitable” (FR p.212). This bears on the inevitability tension (§8, tension 5; C6). - In 2020 what is not inevitable is a bad future. In 2025–26 what is inevitable is AI’s momentum (“the boat has already left the harbor”). - The two fit the map’s reconciliation: the trajectory is fixed, the destination and designs are open. But the 2020 text shows the emphasis moving from contesting inevitability to accepting it as a working assumption. My reading. - A seed of the reversibility test (2020). Learning (ch.13) praises trial and error: poison ivy, Edison, “fail fast, fail forward”. It then adds that “to be effective, it needs to be accompanied by the ability to predict what might happen in the future, based on previous experience” (FR p.63). Knowledge (ch.16) rejects “let’s just try it and see” when lives depend on the outcome (FR p.72). My reading: experiment freely where outcomes are knowable and recoverable, but not where they are not. This is the 2025-03-02 reversibility footnote in embryo. - Intelligence depends on values. “our notions of intelligence are deeply tied to our personal visions of the future”. People who value a technology-enabled future define intelligence as logic and rationality. People who value sustainability or happiness define it as “empathetic, artistic, and inspirational traits” (FR p.70). This is an early source for his later scepticism of intelligence-centric AI framings (“intelligence a term of convenience”; “Being smart doesn’t make you good”).

Corrections and qualifications. - Plausibility is less absolute than the map implies (C7). In 2020 he defends belief in the implausible as a generative human faculty. We are “predisposed to believing in futures that we cannot say with certainty are plausible. And, truth be told, the world, and the future we strive for, are all the richer for this” (FR p.86). “Even when it strays into fantasy, our imagination enables us to construct different future possibilities” (FR p.88). Plausibility is a filter for building (“Knowledge and reason help us sift out those that are plausible from those that are mere fantasy”, FR p.93), not a gate on imagining. Art reveals “the consequences of our actions in ways that transcend rational analysis” (FR p.97). This bears on Note 2 as well. He values non-quantitative ways of knowing the future. - Pluralism about outcomes (§8, tension 4, “Whose value?”). “we’re committed to a future where someone, somewhere, is not going to be happy” (FR p.197). Societies manage this through a shared “minimum viable product” of norms (FR p.197). He offers no procedure for resolving conflicts between values, which the map notes. But he does accept that such conflict cannot be removed. - Mastery is not benign. “We should never assume that our mastery of cause and effect is, by default, benign” (FR p.58). Humans are also, following Jeremy England, entropy accelerators, which suggests “our tendency to cause chaos in the name of progress” (FR p.47). This is a darker frame than the map’s picture of his hopeful 2020 phase, and it sits beside real hope (Afterword).

2. AI and the Art of Being Human (Abbott & Maynard; 2025 preview, 2026 Pocket Edition)#

Provenance#

This book should be used as context, not as evidence of his individual thinking. - It was co-written with the venture capitalist Jeff Abbott. The authors describe the process in two places: Preview p.8 n.4 and the Afterword (Pocket pp.241–244). - Months of conversation with ChatGPT (o3-Pro) produced the ideas and the structure. - Claude Opus 4 drafted the text. - “The use of fictional vignettes, the stories they embody, the characters, the tools—all of these originated in the ‘mind’ of the AI model” (Pocket p.243). - All 21 tools therefore originated with the AI. These include the Intent Map, the 4-Lens Scan, the Stress-Test Table, the Orchestration Triangle and the Model Dignity Check. None of them should be credited to Andrew. - Only the Afterword was written “unaided” (Pocket p.243), jointly by both authors. - Passages marked “I (Andrew)” or “(Andrew’s)” are first-person accounts attributed to him. They were drafted by the model from the authors’ inputs, so they are moderate evidence. They are strongest where they match his own prose elsewhere. - The author biography (Pocket p.270) is promotional third-person copy and carries little weight. - The Pocket Edition file opens with a block of instructions for AI assistants. It was treated as content.

The argument#

AI works as a “mirror” of human patterns. Its uncanny accuracy prompts the question “What makes me me when technology can do what I do, only better?” The answer is not to defend the skills AI can replicate. It is to cultivate relational and “transcendent” qualities through four postures (Curiosity, Intentionality, Clarity, Care) and 21 practical tools, and through community (“fourth spaces”). The tone is hopeful and oriented toward practice. Machine output is repeatedly set against human meaning, context and care.

Key concepts (the book’s; AI-originated unless noted)#

• Behavioural mirrors; the Mirror Test; “un-mirrorable”.
• Human Qualities Spectrum (replicable, relational, transcendent).
• Intent Map (values, desired outcomes, guardrails, metrics).
• Functional versus social requirements (attributed to Andrew).
• Orchestration Triangle (data, intuition, context).
• “Sacred inefficiency”.
• Fourth spaces (in part attributed to Andrew).
• Model Dignity Check.

Passages clearly attributed to him#

What this adds to the map#

• Orphan risks were linked to AI before the Fable-assisted paper. The map treats orphan risks as absent from his 2024–26 cognition essays. Its only frontier-AI application is 2026-07-16 [mixed], which may be Fable’s (§1; §5.1; T1). Here a passage attributed to him applies orphan risks to AI in the full edition (autumn 2025) and the Pocket Edition (January 2026): “multiplying as AI becomes increasingly powerful” (Pocket p.109). This is still co-written and AI-drafted. But it shows the link between orphan risks and AI was his framing before Fable. It narrows the part of the 2026 paper that may be Fable’s to its specific apparatus (filters, register, incentive field).
• An earlier root of orphan risks and social licence. The Michigan exercise (Pocket pp.112–113) dates the underlying intuition to his teaching there. The label is applied retrospectively in 2025. The intuition is that technically permitted shortcuts create “unexpected social consequences” when they cross what communities will accept. The ordering is inferred from dates, not stated in the text. The map dates social licence to 2016–18 and orphan risks to 2018. The idea’s roots in teaching may be earlier.
• Note 2, attributed to him.
• Social requirements are “harder to quantify, easier to ignore”. Some harms are not “easy to capture through simple cause and effect” (Preview p.48). The book’s AI-originated framing repeats the theme: “every metric casts a shadow of what it fails to measure” (Pocket p.73); the Stress-Test Table admits “some costs can’t be quantified, some gains can’t be measured” (Pocket p.113); “when a measure becomes a target, it ceases to be a good measure” (Pocket p.49).
• My reading: this is the ethics-to-risk move (C3) turned against metric solace. What matters most in AI deployment is least measurable, so relying on the metrics is a form of hubris.
• Note 1, qualifying Note 2. Metrics are kept, not discarded. The tools consistently keep measurement as a partner, not a master.
• “Values without metrics become empty rhetoric… Metrics without values optimize for the wrong outcomes” (Pocket p.37).
• Data is “Essential but incomplete”. “The sharpest intuition can miss systematic patterns” (Pocket pp.154–155).
• The fictional teams build “portfolio” metrics for dignity and trust (Pocket pp.39–40).
• These tools are AI-originated, so they are context only. But the passage attributed to him (“The data never speaks for itself”, Pocket p.152) takes the same position. Data is necessary, it is never self-interpreting, and its apparent objectivity is the danger.
• Formation and constitutive change, earlier and more positive than the map dates.
• The co-written text of 2025 says AI mirrors “actively construct versions of us, reshaping how we see ourselves and proposing subtle identity shifts” (Preview p.6).
• In the unaided Afterword the two authors report that “we found those reflections changing us—a human-AI virtuous cycle of discovery, insight, and growth” (Pocket p.244).
• The map treats relational and formative change as rising in 2026 (C14, C16; §5.8, “Rising (2026)”). This is an August/September 2025 statement of the same phenomenon, from their own experience, with positive valence.
• My reading: between August 2025 (“virtuous cycle”) and July 2026 (AIs “beginning to train us to think like them”) his evaluation of one and the same process moves from celebration to concern. That is a clearer turn than the map currently records under “What changes” (§7).
• A signal of the concern behind the book (Afterword, unaided): “important perspectives that were being overlooked in the headlong rush to create an AI-driven future” (Pocket p.241). This is consistent with C10 and the critique of speed.
• Plausibility hedging in co-written text: “this is a fictional example—AI may not be quite at this stage, yet…” (Pocket p.8). Also “not answers (those would be obsolete by publication)” (Pocket p.12). And “doomsday determinism, or the lure of breathless optimism” (Pocket p.9). These are consistent with C7 and with his refusal of the binary.
• For the map’s provenance section (§1). The map already treats the book as weaker co-written evidence. It should add the stronger reason: by the authors’ own account, the stories and the tools originated with the AI models. Any lens or concept drawn from the book should be attributed to the book, not to him.

3. Future Travel Foresight Catalyst (Maynard & Leahy, USDOT/TBD final report, August 2025)#

Provenance#

Co-authored with Sean Leahy. It was drafted with AI assistance, disclosed on p.i, and the authors claim the ideas as their own. It is a grant report, so it is partly self-presentation to a funder. Moderate weight.

The argument#

Traditional research and outreach cannot keep up with fast-changing technology or with falling trust in experts. So the project built a “foresight catalyst” on three pillars: transdisciplinary blurring of expert and public, parasocial relationship-building (the Substack, the Modem Futura podcast, the “Pizza and a Slice of Future” class) and futures-thinking frameworks. It treats audience relationships and trust as the primary outcome. It argues that short-term quantitative metrics cannot capture that impact.

Key concepts#

Foresight catalyst; parasocial relationship-building; relational versus transactional communication; trust over clicks; relational capital; engagement as “infrastructure”; knowledge mobilisation.

What this adds to the map#

• Note 1 and Note 2 in a methods statement.
• “we must complement rigorous research with imaginative cross-sector and cross-stakeholder conversations”; “breakthroughs will not come solely from number-crunching” (p.2). This is complement, not replace.
• “The pace of innovation has accelerated beyond the reach of our usual tools for studying and understanding emerging landscapes” (p.2). This is Note 2’s “sheer lack of understanding” as a statement about method.
• Futures work is valuable “less about predicting accurately and more about expanding the imagination” (p.3). “we do this by asking questions, not predicting” (p.5).
• On evaluation: “numbers are easy to report, they often do not correlate well with effective reach and impact” (p.6); “hard quantitative KPIs would be both difficult and inadequate” (p.9).
• Humility as a practice of trust. “drop the veneer of unassailability that experts often wear. Show your curiosity, even your doubts” (pp.13–14). “approachability with authority” is “a craft and an art” (p.5). This deepens the map’s method note (humility tested in public) and tension 12 (humility and authority). He treats displayed humility as a way of earning trust.
• Tension 11 sharpens (persuasion he practises and persuasion he fears). The report is a funded, formal statement of his parasocial method. Trust “gives us permission to talk about what we know” and lets the hosts “explore complex or speculative territory and listeners will come along because the relationship foundation is there” (p.4). My reading: five months later (2026-01-10) he identifies the same mechanism as the core AI risk: trust built through relationship and fluency lowering critical scrutiny. The map already notes the tension (§8, tension 11; 2025-05-25; 2025-11-19 “read the tea leaves wrong”). This report marks August 2025 as the high point of his parasocial optimism and makes the tension concrete.
• Tension 9 (engagement as principle, thin as mechanism) is partly a corpus artefact. The report shows that his operational mechanisms of engagement exist, are funded and are cited as grounded in participatory technology assessment and responsible innovation (p.2). They are relational media at scale plus an intergenerational discussion class. They are not deliberative instruments such as consensus conferences. My reading: the map is right that he never works a deliberative mechanism through for AI. It should add that his chosen mechanism is relationship-based communication, which is exactly the channel he later fears AI exploits.
• Symmetry in autonomous vehicles. His TV commentary on a Waymo incident emphasised “both the technology’s promise and the need for public patience” (p.8). This is consistent with C6 and 2023-11-09.
• AI-augmented foresight. The report plans “Tech Futures Briefs” to be “generated through an advanced AI-assisted methodology” (p.15). This is further evidence that AI is built into his method (tension 13).

4. Biopreservation and research-ethics papers (shared positions; skimmed)#

All four come from the Ethics & Public Policy group of the NSF ATP-Bio Engineering Research Center, where Maynard is one of many co-authors. The companion first-authored paper (2024_Risk-Innovation-ATP-Bio_JLME.pdf) is outside this share. The map knows this work only through the 2024-12-17 post. These papers are shared positions and are not evidence of his individual thinking. They show which positions he has signed.

Wolf et al. 2024, “Anticipating Biopreservation Technologies that Pause Biological Time” (JLME 52: 534–552) - Argument. Advanced biopreservation is a multi-use platform technology. It suffers the “pacing problem” and fragmented jurisdiction (p.537). Governance should therefore rest on four things (Table 2, p.546): harmonised terminology and standards; early and continuing engagement; sustained dialogue on oversight, using soft and hard law dynamically; and cross-agency coordination. - What it adds. - It cites his definition as a governance frame. “Maynard’s conceptualization of risk innovation ‘frames risk as a threat to existing or future ‘value’’” sits beside responsible innovation, anticipatory governance and adaptive governance (p.538). This is independent confirmation that peers treat threat-to-value as a stable, citable concept (C2). - Note 1: risk assessment sequenced, not dropped. “This kind of technology anticipation and assessment is prior to and broader than risk assessment in product regulation. Early in technology development, risks may be unclear”. Technology assessment “can help build the foundation for later risk specification and regulatory response” (p.538). It endorses a “dynamic oversight approach” that moves between soft and hard law “dynamically as data become available” (p.546). My reading: this is the clearest statement in the whole share of how Note 1 and Note 2 fit together. Quantitative risk assessment is the right tool once hazards can be specified. Under deep early uncertainty, anticipation and engagement come first, and formal assessment follows as evidence accrues. It is a shared position, but it matches his own FR ch.46. - The GMO lesson (consistent with his GMO failure case, T2): “Despite research showing the safety of genetically modified crops, opposition to GMO foods has been substantial”. Outreach should address “not only facts… but also cognitive and emotional issues” (p.544). - Urgency: “the time to address those risks and benefits is now” (p.547).

Hyun et al. 2024, “The Need for Early Engagement with Interested Groups on Advanced Biopreservation” (JLME 52: 585–594) - Argument. Early engagement is both instrumentally useful and “an ethical imperative” (p.587). It is justified intrinsically by respect for persons and communities (p.591). It should be iterative, share power and involve “cultural humility” (p.588). The paper’s precedent is Cambridge City Council’s open hearings on recombinant DNA in 1976 (p.588). - What it adds. - A concrete engagement mechanism (tension 9). Funders and companies should require and pay for engagement. But relying on them creates “disincentives to change course or to halt research trajectories altogether if engagement… yields conclusions that are unfavorable” (p.591). So engagement should run through trusted intermediaries, in this case science museums, which surveys show to be more trusted than scientists (pp.591–592). This is a shared position. It answers “who convenes, and how is capture avoided?” more concretely than anything in the map’s corpus. - Engagement despite cost. Engagement “could be slow, delay progress, expose and fuel objections… or possibly even mobilize people against it”, yet it “remains an imperative” (p.589). This matches C5 and C9. It also carries the caution that “speakers may overstate the risks or harms” (p.589), a symmetric-risk note (C6). - A different reference case from Asilomar: participatory municipal governance (Cambridge), not scientists’ self-governance.

Pruett et al. 2025, “Governing new technologies that stop biological time” (Am. J. Transplant. 25: 269–276; author manuscript) - Argument. Prolonged organ biopreservation “transforms an organ into a new, manufactured product not found in nature”, a “product of human artifice”. So the FDA’s 1983-era exclusion of vascularised organs from oversight should be revisited (section “Establishing oversight authority and standards”). The authors also call for new facilities, allocation policy and attention to equity, commercialisation and global organ trafficking. “That process should begin now” (Abstract). - What it adds. This is a co-signed case for hard-law regulatory authority when a process changes a technology’s category. It bears on C12, which the map describes as firm on the portfolio but hedged on instruments. It also mirrors his rule of behaviour, not labels (C13): the transformed product no longer fits the old category. It keeps quantitative tools central: “predictive analytics to identify organs with suitable function for recipients must continue to evolve” (Conclusion). That is Note 1 in a clinical setting.

Wolf et al. 2026, “Filling the Network Gap in Research Ethics” (Hastings Center Report 56(4): 32–43) - Argument. Research ethics addresses individual researchers (micro) and society-wide impacts (macro). It neglects the “mesolevel” of large multi-institution research networks, where authorship, data sharing, misconduct detection and societal engagement must be governed at network level. - What it adds (minor). Table 1 (p.34) lists “risk analysis, horizon forecasting, anticipatory governance, and responsible innovation” together as macro-level methods, so risk analysis is kept within the toolkit (Note 1). It notes NSF’s expectation that ERCs “build the capacity for upstream stakeholder engagement to codevelop emerging technologies” (p.38). And it says networks developing technologies of “tremendous benefit as well as profound disruption have a duty to pursue their innovative research and technology development responsibly” (p.39). My reading: locating responsibility at the level of the organisation, not the individual researcher, fits his structural account of sincere actors inside incentive systems (C10). But the paper does not mention AI, and his individual contribution cannot be identified.

5. Consolidated evidence on Andrew’s two notes#

Note 1: quantitative risk assessment is a foundation that has been built on. The evidence runs in one direction and is strongest in his own 2020 prose. - Mathematics and science of change “help identify impending dangers, as change threatens to take away what we value” (FR pp.166–167). - Apollo’s predictive knowledge is the model, and “try it and see” is rejected (FR pp.72–73). - “there remains a lot that we can predict” (FR p.148). - Perception is “strictly speaking” not everything: crash probabilities do not bend to fear (FR p.154). - Evidence outranks feeling when they conflict (FR p.56). - In co-signed papers, risk assessment comes after anticipation, “as data become available” (Wolf 2024, pp.538, 546). Risk analysis is listed alongside anticipatory governance (Wolf 2026, p.34). Predictive analytics are expected to keep developing (Pruett 2025). - In the co-written book, a passage attributed to him says data is necessary but “never speaks for itself” (Pocket p.152). - Implication for the map. C3 (“necessary but no longer sufficient”) and C4 are right. The map’s phrasing in §2 (“impatience with it”) and in §8 “Missing methods” reads more like abandonment than the record supports. The better frame is layering: probability and evidence remain the base, threat-to-value widens what counts, and anticipation governs the phase before hazards can be specified.

Note 2: the absence of AI quantification reflects humility about hubris, while still grappling. In general form, the 2020 book states the underlying position in his own words. - Precise predictions in complex systems are “less likely… to be accurate”. The danger is being “so enamored with our brilliance” that we act as if we can “fully control” the future (FR p.148). - Hubris “encourages belief in the absence of evidence” and breeds “naive myopia” (FR pp.150–151). - The clever are better at self-justification (FR p.153). - Capability “continues to exceed our understanding of how to do this responsibly” (FR p.215). - The same humility appears in the co-written book: social harms are “harder to quantify, easier to ignore” (Preview p.48); the gap between capability and understanding “widens daily” as AI advances (Pocket p.109). It appears in the USDOT report too: “usual tools” are outpaced (p.2). - Humility is paired with a duty to act. “Grapple with… how to avoid costly and potentially catastrophic mistakes” (FR p.26). Engagement “remains an imperative” (Hyun p.589). “The time to address those risks and benefits is now” (Wolf 2024, p.547). - Caveat. None of these texts says in terms: “I do not quantify AI risk because quantification would be hubristic”. That link is Andrew’s own account, now supported by a 2020 textual basis. It is not a quotation.

6. Suggested adjustments to the map (for the reviser)#

1. §1 Sources. Future Rising is now partly available in full text (33 of 60 chapters). Note the README error about the “Subscriber Excerpts” file.
2. §1 Provenance, and C1 sources. For AI and the Art of Being Human, add the authors’ own statement that stories and tools originated with ChatGPT/Claude, and that only the Afterword is unaided (Pocket p.243).
3. §2 and §8 “Missing methods”. Reframe along the lines of Andrew’s notes. His AI writing has little quantitative treatment. But he explicitly values mathematical and predictive methods elsewhere (FR pp.72–73, 148, 166–167) and warns in 2020 against false precision and the hubris of control (FR pp.148–151). Mark the link to AI as Andrew’s own account.
4. C1 firmness and §5.9. Thriving language dates from 2020 (FR p.192; Excerpts.pdf p.2), not first from 2025.
5. C7 and §5.5. Plausibility governs building, not imagining. Add FR pp.86, 88, 93.
6. §5.2 Hubris. Add FR ch.39–40 (2020): the hubris of prediction and control, which fills the 2018→2023 gap.
7. §5.7 Deepfakes; T8. 2020 already hedged hope with doubt (FR pp.156–158). The 2024 shift is a sharpening, not a reversal from trust.
8. C14 and §5.8. Add the 2020 stages: evolutionary mismatch (FR pp.55–56), the clever-are-not-immune point (FR p.153) and emotionally disabled fake-detection (FR p.157). Add the 2025 co-written “virtuous cycle” (Pocket p.244) as an earlier and positive statement of AI changing its users, and record the evaluative turn by 2026 in §7.
9. §5.1 Orphan risks; T1. A passage attributed to him links orphan risks to AI in 2025 (Pocket p.109), before the Fable-assisted paper. The Michigan teaching is a retrospective earlier root (Pocket pp.112–113).
10. §8 tensions 9 and 11. Add the USDOT report (parasocial trust as a funded method, p.4) and Hyun 2024 (a funder-mandated, intermediary-run engagement mechanism, pp.591–592, shared).
11. §8 tension 5. Add the 2020 anti-fatalism (FR pp.17, 212) as the starting point of the shift in emphasis on inevitability.
12. C12. Co-signed support for hard-law FDA authority when processing changes a product’s category (Pruett 2025), and for dynamic movement between soft and hard law (Wolf 2024, p.546).

7. Digest: the most important additions from S7#

The most valuable item in this share is Future Rising (2020): sole-authored, pre-ChatGPT, and now available for 33 of its 60 chapters. The map previously knew it only through one-line quotations in a 2024 post. The full chapters bear directly on both of Andrew’s notes.

On Note 1, quantitative and predictive science remains a foundation he values, not something he has left behind. - In “Change” (ch.46) the mathematics of change lets us “get a sense of where observed trends and trajectories are heading”, and the same science and math “help identify impending dangers, as change threatens to take away what we value” (FR pp.166–167). That sentence joins the probabilistic and threat-to-value senses of risk that the map treats as unintegrated. - “Knowledge” (ch.16) holds up Apollo’s predictive accuracy as the model and rejects “let’s just try it and see” (FR pp.72–73). - “Perception” (ch.42) says the slogan that perception is everything is “strictly speaking, not true”: fear does not change the likelihood of a crash (FR p.154). - Co-signed papers make the relationship explicit. Anticipation is “prior to and broader than risk assessment”, and formal risk specification follows “as data become available” (Wolf et al. 2024, pp.538, 546). The right frame for the map is layering, not replacement.

On Note 2, the book states the humility behind his reluctance to quantify AI risk, in general form and in his own words, six years early. - “The more precise we try to be with our predictions of the future, the less likely they are to be accurate”. The danger is becoming “so enamored with our brilliance” that we act as if the future “is something we can fully control” (FR p.148). - A whole chapter on hubris warns that it “encourages belief in the absence of evidence” (FR p.151). - The Afterword names the condition: our ability to change the future “continues to exceed our understanding of how to do this responsibly” (FR p.215). - Humility is never a reason to stop: we must “grapple with… how to avoid costly and potentially catastrophic mistakes” (FR p.26). - My reading: this makes the absence of AI risk numbers look principled, not a gap in method. The explicit link to AI is Andrew’s own account, not a quotation.

Future Rising also moves several ideas earlier than the map dates them. - Evolutionary mismatch: instincts “increasingly poorly equipped” for a world we change faster than evolution can follow (FR pp.55–56). The map dates this to 2026. - The clever are not immune: “the smarter we are, the better we are at justifying our beliefs” (FR p.153), a precursor of the Intelligent User Trap. - Deepfakes: a full 2020 chapter already hedges hope with doubt (FR pp.156–158), so the 2024 “change of mind” is a sharpening. It also fills the manipulation thread’s 2018–2023 gap. - Thriving is named in 2020 (FR p.192), not first in 2025. - AI: cracking it could make us “redundant” (FR p.191), yet “true artificial intelligence” might inspire (FR pp.212–213). - Imagination: faith and fantasy are honoured as generative (FR pp.86, 88). Plausibility filters building, not imagining, which qualifies the map’s picture of an absolute discipline. - Inevitability: he resists being “resigned to what we assume is inevitable” (FR p.212), the baseline from which his 2025–26 language moved.

AI and the Art of Being Human matters mainly as a provenance warning. By the authors’ account, its stories and all 21 tools “originated in the ‘mind’ of the AI model”; only the Afterword is unaided (Pocket p.243). None of its frameworks should be credited to him. Three things still stand out. - Orphan risks and AI: a passage attributed to him says orphan risks are “multiplying as AI becomes increasingly powerful” (Pocket p.109). This predates the Fable-assisted paper, and he roots the idea in his Michigan teaching. - Metrics: “The data never speaks for itself”, and metrics “feel objective” (Pocket p.152); social harms are “harder to quantify, easier to ignore” (Preview p.48). Both notes in miniature. - AI changing its users: the unaided Afterword calls it “a human-AI virtuous cycle” (Pocket p.244). That August 2025 positive view of formative influence becomes a risk by mid-2026, an evaluative turn the map should record.

The USDOT report (August 2025, co-written, AI-assisted) shows his working engagement mechanism: parasocial, relationship-based media, through which audiences “will come along” into speculative territory because of trust (p.4). That sharpens the tension between the persuasion he practises and the AI persuasion he fears, and shows “thin mechanism” to be partly a corpus artefact. The co-signed biopreservation papers add a concrete engagement design (funder-mandated, run through trusted intermediaries, alert to capture; Hyun 2024) and support for hard-law oversight when a technology changes a product’s category (Pruett 2025).