LL2-28 — Ch28 In conclusion#

Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013), Part E “Implications for science and governance”. Report pp. 670–684 (PDF pp. 672–686). The chapter text runs pp. 670–680, references pp. 680–682, then two memorial pages: Masazumi Harada (p. 683) and Poul Harremoës (p. 684).

Read in full from the text extract (every page marker, through report p. 684). All fifteen PDF pages were also rendered and checked visually. The chapter has no boxes, tables, figures or panels, and the extraction is faithful. Several of the chapter’s cross-references were checked against the underlying chapters in the same PDF (Ch 2, 5, 15, 18, 19, 21, 22, 24, 25, 26 and 27), plus the contents, Acknowledgements and Introduction (pp. 4–5, 9). Where those checks bear on the strength of a claim, the notes say so. (An audit re-verified these page references against the PDF text; see the Audit log.)

Authors and standpoint#

No named authors. Unlike the case-study chapters, Ch 28 has no byline. The table of contents (p. 4) lists it without authors. The Acknowledgements (p. 5) credit an EEA editorial team: David Gee, Philippe Grandjean, Steffen Foss Hansen, Sybille van den Hove, Malcolm MacGarvin, Jock Martin, Gitte Nielsen, David Quist and David Stanners. They name David Gee as “Originator” of the project. The Introduction (p. 9) says the volume combines external case studies with “chapters written by members of the report’s editorial team”. Ch 28 is best read as the EEA’s institutional synthesis: the report’s own voice, not an individual scholar’s argument. The chapter writes as “we” (“we see the destruction…”, p. 670; “as we saw above”, p. 679).

Inferred influences (my inference, not stated): much of the argument condenses Ch 26 (Grandjean, “Science for precautionary decision-making”) and Ch 27 (Gee, “More or less precaution?”). Examples are the evidence-strength spectrum from “scientific suspicion” to “beyond all reasonable doubt” (Ch 27, pp. 656–658), the research-neglect argument (Ch 26, pp. 625–629), the research-funding imbalance between product development and hazards (Ch 27, p. 646), the claim that inconsistent results are “to be expected” under multicausality (Ch 27, p. 652) and the monopoly/low-price argument (Ch 27, p. 659: “late actions have consolidated technological monopolies for products, at unrealistically low prices”). The forgetting-cycle passage (p. 680) closely paraphrases Ch 15 (Kundzewicz, p. 360). The “innovation as a means, not an end” framing (p. 678) and the GMO/nanotech monopoly claim (p. 673) are explicitly cited to van den Hove, McGlade, Mottet and Depledge (2012). The closing heading “governance of innovation and innovation in governance” (p. 680) carries no citation, but is in the same vein. McGlade was then EEA Executive Director and wrote the Preface; van den Hove was on the editorial team. EEA’s own reports supply much of the economic argument: EEA 2011a–d on megatrends, environmental tax reform and ecosystem capital accounting.

Evident stance. The chapter argues openly for the precautionary principle. It is framed within ecological economics and sustainability: economies exist to serve well-being and depend on natural capital (p. 670, citing Passet 2001). It is critical of corporate influence on science and favours participatory, transparent, systems-oriented governance. It is also optimistic that precaution fosters innovation. It is written as a set of findings plus recommendations addressed to policymakers, business, scientists and citizens. Its tone is advocacy grounded in the cases, not neutral review.

Panels / commentaries: none. This chapter contains no panels, no industry or regulator responses, and no dissent. The only “other voices” are the cited literature. That is a structural difference from many case chapters and matters for the bias check below.

Memorial pages (pp. 683–684). These are not analytical but frame the report’s values: - Masazumi Harada (1935–2012), co-author of Ch 5 (Minamata) (per the contents page and the Acknowledgements, pp. 4–5; the memorial itself does not say so). He first examined Minamata patients in summer 1961 as a graduate student at Kumamoto University. His 1964 thesis on congenital Minamata disease “disproved the conventional belief at the time that the placenta does not pass poisons” (p. 683) and won an award from the Japanese Society of Psychiatry and Neurology in 1965. He founded the Open Research Center for Minamata Studies in 2005 and visited Brazil, China and “native Indian communities in Canada” to find suspected cases. He first came to the Grassy Narrows and White Dog First Nations “in the early 1970s”. His final report for Grassy Narrows was released on 4 June 2012 “after 30 years of research”. It showed that mercury from the Dryden paper mill affected people “not yet born when the dumping ceased” (p. 683). He died in June 2012, aged 77. - Poul Harremoës (1934–2003), chairman of the editorial team for Volume 1 (2001). He was a Danish civil and environmental engineer (biofilm kinetics, sewer design, then risk analysis and the precautionary principle), a member of the Danish Pollution Council, and on the first EEA Scientific Committee from 1995. He won the Heineken Prize for Environmental Sciences in 2000 (p. 684). - Note: Harada’s placenta finding is the scientific counterpart of the p. 675 example of a mother’s hypothesis dismissed by experts who assumed the placenta was protective. The memorial reinforces the chapter’s “early warner” theme. [Cross-check, Ch 5, p. 105: Ch 5 quotes Harada (2005) recalling the 1961 encounter. It shows that Harada himself was at first one of the experts who held the placenta assumption. He took the mother’s view for “the fancy of an amateur”, and “time proved that she was correct”. He illustrates both the dismissal and its later correction, not only the correction.]

Section-by-section notes#

The chapter has three bold-headed sections. The middle one contains six italic sub-headed “findings”.

1. “Since 2001: many changes, crises and lessons relearnt” (pp. 670–671)#

Context-setting (p. 670). The world since 2001 is described as “larger in population but smaller in interconnectivity; faster in terms of technology adoption but slower in terms of policy action”; “more volatile” economically and environmentally “yet more static in terms of political reflexivity and adaptations in governance”. Named long-term systemic challenges: resource depletion, climate change, “a 2‑billion person increase in the world population by 2050”, and diminishing ecosystem resilience (EEA 2011a; OECD 2012; WEF 2012).

Two “realities” (p. 670): 1. Governance systems “misrepresent the socio‑ecological system, making societies and the environment subordinate to the economy”. This ignores that civilisation depends on ecological and social foundations and that economies exist to serve well-being (Passet 2001). 2. “the scale, interconnectedness and sheer complexity of feedbacks between nature and human interventions have outstripped society’s capacity to understand, recognise and respond to these effects” (p. 670).

Progress since Volume 1 (p. 670): - Innovation: Volume 1 recognised the difficulty of balancing precaution and innovation. Now there is “increasing evidence that precautionary measures do not stifle innovation, but can encourage it, in particular when supported by smart regulation or well‑designed tax changes” (citing EEA 2011b, 2011c; Ambec et al. 2011; Ashford and Hall 2011). [Check: these sources concern environmental regulation and environmental tax reform in general, not precautionary action under uncertainty specifically. See the bias check.] - Improved: using a “richer body of information from more diverse sources” and bridging the separate, polarised fields of public health and environment. Also involving wider stakeholders to “improve public trust in society’s capacity to control hazards, without stifling innovation or compromising science”. All are “areas where improvements have been made since 2001”. No evidence is given for these improvements. - Less progress: Volume 1’s call to “identify and reduce institutional obstacles to learning and action”. Political and scientific “bureaucratic silos” persist despite calls for policy integration (Hamdouch and Depret 2010; Phoenix et al. 2012).

Worsening picture (p. 670). “Worryingly, warnings of impending hazards are, in many areas, still not being heeded.” Damage is described as more widespread geographically, across species and into future generations, especially from “current energy systems, chemicals and technologies”. “Damage is now shown to be occurring at increasingly lower levels of exposure to pollution”. Polluters “for the most part” do not pay full costs, “partly because of a lack of incentives”. Natural capital stocks are being destroyed, and dependence on them is easy to forget “especially in times of economic crises”. No specific evidence is given in this paragraph.

Volume 1’s key message restated (pp. 670–671). “‘the growing innovative powers of science seem to be outstripping its ability to predict the consequences of its applications…’” This “is happening at an ever‑greater pace, with globalised industries racing to introduce new technologies but with limited understanding of what their impacts might be. National governments now have less control over globalised technologies” (pp. 670–671).

Positive developments (p. 671). “new transformative approaches are emerging for managing the systemic and interconnected challenges” (Gladwell 2012, a Nielsen blog post; Stirling 2008). They build on citizens’, consumers’ and shareholders’ use of “the power of the internet and social media to demand and foster increased participation, responsibility, accountability and transparency”. - These need longer-term perspectives. Complexity, uncertainty, “scientific ignorance”, broader risks and irreversibility “necessitate the increased use of long‑term scenarios and strategy analysis by citizens, governments and corporations alike” (EEA 2011a). - Society’s long-term interests, “distinct from the partial interest of particular stakeholders and individuals”, require “new political and financial institutions that can help overcome the short termism of most politics and much finance” (Ward 2012; Roderick 2010; Mainelli and Giffords 2009; RMNO 2009). - The Santayana epigram closes the section: “‘those who cannot remember the past are condemned to repeat it’” (p. 671).

2. “2001–2013: what new insights emerge?” (pp. 671–680)#

Framing (p. 671). Many cases “reveal similar lessons to those in the 2001 report”, and some insights “have been strengthened” as evidence and understanding of ecological and biological systems improved. The cases are diverse and each unique (type of innovation, nature of hazard, policy approach, culture). The common features named are: - “key decisions on innovation pathways made by a few people on behalf of many”; - “a lack of institutional and other mechanisms to respond to early warning signals”; - “misleading market prices that do not properly reflect all costs and risks to society and nature”; - “inadequate accounting for assets and liabilities across different types of capital” (p. 671).

Questions for the future (p. 671): - How can innovations driving knowledge economies, “such as nanotechnologies”, be developed without repeating past mistakes? - How can “wider and wiser” use of the precautionary principle support decisions in complex systems “where ‘surprises’ are inevitable”? - How can we ensure “that the lack of ‘perfect’ knowledge is not a justification for inaction in the face of ‘plausible’ evidence of serious harm”? - How can conflicting public and private interests be balanced across the development, use and impact phases? - How can the distribution of costs and benefits over time be made more equitable?

Actors (p. 671). Governments, policymakers, businesses, entrepreneurs, scientists, civil society, citizens and media each bring “different and often conflicting knowledge, perceptions, interests and priorities”. Balancing these “numerous and often antagonistic positions should be seen as a prelude to making decisions on those innovations that have broad societal implications.”

The three main “opportunities” (pp. 671–672): 1. “to correct the prioritisation of economic and financial capital over social, human and natural capitals”, through broader use of the principles of precaution, prevention and polluter-pays, and improved accounting across government and business. 2. “to broaden the nature of evidence and public engagement in choices about crucial innovation pathways”. This means rebalancing science “towards dealing with complex, systemic challenges and unknowns” and complementing it “with lay, local and traditional knowledge”. 3. “to build greater adaptability and resilience in governance systems to deal with multiple systemic threats and surprises”. This is to be done through stronger institutions and “deploying information technologies in support of the concept of responsible information and dialogues” (p. 672).

The six findings that follow “supplement[] the conclusions of Volume 1” (p. 672).

2a. Reduce delays between early warnings and actions (pp. 672–673)#

Headline claim (p. 672): “Most of the case studies in both volumes… illustrate that if the precautionary principle had been applied on the basis of early warnings, many lives would have been saved and much morbidity and damage to ecosystems would have been avoided.” This is a counterfactual and is not quantified here.

Three contemporary factors that worsen delay (p. 672): 1. Moving target: “by the time evidence of harm is confirmed, the technology has often changed, leading to assumptions that, unlike yesterday’s technology, today’s technology is now safe.” 2. Sunk-cost lock-in: for capital-intensive technologies (e.g. “broad‑scale energy production systems or chemical plants”), “the huge initial investments mean that yesterday’s investments will be redeemed before any serious risk reduction is implemented, creating de facto technological lock‑ins.” 3. Scale overwhelms monitoring: the scale of development “puts very difficult demands on those attempting to monitor and respond to the risks before they have become serious, widespread and irreversible.”

These “further strengthen the case for taking early warning signals more seriously and acting on lower strengths of evidence than those normally used to adduce ‘scientific causality’” (p. 672).

The main barriers to timely action, according to the case studies (p. 672): - “the short‑term nature of many political and financial horizons”; - the novelty of the technologies and the scientific problems arising from interactions with complex biological, ecological and social systems; - “the conservative nature of much environment and health science”; - how scientific and other evidence is evaluated; - different stakeholder perspectives “and the vested interests of some powerful ones”; - cultural and institutional circumstances “that often favour the status quo”.

An acknowledged limit (p. 672): “the unequal distribution of political power between citizens, business and financial actors, and governments is a persistent problem of politics, which has increased through globalisation and the rise of multinational corporations, yet it is an issue that is well beyond the scope of this report.” The chapter adds that “Some of the other causes of delay are more amenable to change and these are addressed in the rest of this section.” This candid admission matters: a structural driver of delay that the chapter itself names is explicitly left unaddressed. (The chapter does not rank it as the most important driver. It offers it as an example of why “tackling them is not easy”.)

“Harm expansion” (p. 672). Harms “turned out to be more diverse and widespread than anticipated” and often occur “at exposures lower than initially considered dangerous”. The examples, with the dates as stated in Ch 28: - asbestos: asbestosis (1906–1929), lung cancer (1955), mesothelioma (1960); - smoking: lung cancer (1951), later “a wide range of cancers, heart disease and foetal damage”; - PCBs: eagle reproduction (1960s), later neurological problems in children and cancer; - lead: first children’s IQ, “now known to cause heart disease in adults”; - radiation: “a similar expansion of known hazards”.

The chapter introduces the term: “This phenomenon of ‘harm expansion’ is rendered more problematic by the discoveries that harm from all of the above agents has been found to occur at lower and lower levels, such that, more often than not, no ‘safe’ threshold of exposure can be identified” (p. 672). The report index lists “harm expansion” only at p. 672, and a text search of the full report finds it nowhere else. Within the report, then, the term is the conclusion’s own synthesis; whether it was used earlier in the wider literature was not checked. Recommendation: “Continuous, anticipatory reductions in exposures to emerging hazards could help to avoid repeating these histories of harm expansion.”

Costs of action and inaction (pp. 672–673). Better prospective and retrospective analyses “across the full lifecycle of a technology” would show the value of precaution. This is especially true of “‘secondary benefits and costs’”, such as the health benefits of cutting fossil fuel use for climate reasons. Such analyses should also count the psychological and societal costs of false alarms, e.g. “the over‑reaction to swine flu in the US in the 1970s”. They should equally count misplaced reassurances, e.g. “the downplaying of risks associated to nuclear power plants by Japanese authorities and utilities” (p. 673). Key normative point: “Such pro and con analyses should be independent of interested parties, both commercial and political, as they often have a ‘natural’ tendency to exaggerate costs of hazard reduction and to underestimate the benefits of action” (p. 673).

Precaution, diversity and innovation (p. 673). “the timely use of the precautionary principle can often stimulate rather than hamper innovation, in part by promoting a diversity of technologies”. Diversity increases resilience to surprises. “Keeping options open and following multiple paths” lets a risky option be dropped. It also avoids “technological monopolies” like those in asbestos, CFCs and PCBs. - “In contrast, technological monopolies hamper innovation.” The “monopolies of lead in petrol, asbestos, CFCs and PCBs… both prolonged the harms they caused and made those harms widespread”. They produced technological, “institutional and ideological lock‑ins” that hampered alternatives. - The products were “‘cheap’ in the market place, bearing little relation to their real costs”. “These artificially low market prices in turn helped to stifle the development of smart substitutes.” - Applied forward: GMOs and nanotechnologies show “already signs of technological monopolies, driven by the high costs of research, development and production involved and the patent protections for developers” (Stirling 2007; van den Hove et al. 2012). [Cross-check: the report’s own nanotechnology chapter, Ch 22, does not make a “monopoly” argument in these terms. The forward-looking nanotech/GMO monopoly claim rests on external sources, not the report’s case evidence. The backward-looking claim about lead, asbestos, CFCs and PCBs does have internal support: Ch 27 (Gee, p. 659) says that “late actions have consolidated technological monopolies for products, at unrealistically low prices, which served to keep smarter substitutes out of the markets for many years”.]

Ethics of precaution (p. 673). The precautionary principle involves “not only scientific issues… but also ethical choices” about: - the appropriate strength of evidence; - equity in the costs and benefits of action and inaction; - “the appropriate balance between generating false negatives and false positives”; - “the social necessity of large‑scale innovations”.

False alarms (p. 673). Acting on weaker evidence “will sometimes increase the number of false alarms — although the review of 88 cases of alleged false positives in Volume 2… confirmed just four actual cases, suggesting that the risks are considerably less than sometimes claimed.” [Cross-check against Ch 2 (Hansen and Tickner): the four are Southern corn leaf blight (1971), saccharin labelling (1977), swine flu immunisation (1976) and food irradiation (p. 25). About a third of the 88 were classed as “real risks” and about a third as “the jury is still out” (pp. 20–21). Ch 2 acknowledges that classification “includes some level of subjectivity” and that others “might come up with slightly different categorisations” (p. 33). Ch 28’s “just four actual cases” is accurate as a count but drops these qualifications. The 88 cases came from a literature search for claimed false positives (“over-regulation”, “health scares” and similar), most of them drawn from a few sources (Mazur 2004; Wildavsky 1995; Lieberman and Kwon 1998; Milloy 2001) (p. 19). They are not a sample of all precautionary actions. In fairness, Ch 2 also says its authors “feel confident in our core findings” that genuine false positives are few (p. 33).]

Asymmetry argument (p. 673). Where damage is long-lasting and “may irreversibly alter the system”, there is “a fundamental asymmetry between the competing policy and scientific options of avoiding false negatives and avoiding false positives”. The examples of irreversibility are climate change, “modification to the genetic make‑up of humans or other species”, persistent chemical or radioactive contamination, and species loss. - The logic is spelled out. If an early warning triggers “a double reaction of precautionary policy measures and more intensive research”, then a false alarm costs only a delay in benefits (or mitigation costs) “but the system will not be irreversibly altered”. - If no action is taken and research later confirms harm, “irreversible systemic damages will already have taken place”. “Acknowledging this asymmetry is central to understanding when precautionary and preventive approaches are best deployed.” - Normative conclusion: “Tipping the overall balance of public policy towards avoiding harm, even at the cost of more false alarms, would seem to be a price that is well worth paying”. It is “one of the strategic and ethical societal choices, similar to the choice of strengths of evidence to be used in civil or criminal court cases, that needs to be openly debated” (p. 673).

2b. Acknowledge complexity when dealing with multiple effects and thresholds (pp. 673–675)#

State of harm (pp. 673–674). The world is “drawing down its natural capital through an over‑reliance on fossil‑fuel‑based, synthetic chemicals”. These are compromising ecosystems and “key organisms such as fish and bees” in combination with climate change and invasive species. “There is also evidence that some types of genetically modified crops (and the agrochemical substances used alongside them)… present a threat to human health, some species and ecosystems, and food security” (p. 674). - [Cross-check: the citations attached to this paragraph (Barouki et al. 2012; EEA 2012; Kortenkamp et al. 2011) concern developmental origins of disease and endocrine disruptors, not GM crops. Ch 19 (Quist et al.) is more cautious on human health, though not wholly so. It says indications of harm from glyphosate “remains a concern” (p. 462), citing Greenpeace 2009 and Séralini 2012. It says the literature is “accumulating indicators both of inflated benefit claims and of evidence of adverse effects” (p. 468). But its main argument is that GM safety is “presumed when there is a lack of evidence of harm” (p. 468), with no mandatory toxicity testing required (p. 469), and it concedes that “it is plausible that there simply are no effects to be found” (p. 468). Ch 28’s flat statement that there is “evidence that some types of genetically modified crops… present a threat to human health” is stronger than Ch 19’s framing of human-health effects as unresolved.] - Human health is compromised “by chemicals that threaten health from before birth, through childhood and into adulthood” (p. 674). - Such exposures “appear to contribute to increases of many types of cancers, birth defects, male infertility, and cardiovascular, neurological and immunological dysfunctions”. Added to these are unhealthy diets and lifestyles (“epidemics such as diabetes and obesity”). “Taken together, these multiple stressors have profound public health significance” (p. 674). The hedged verb (“appear to contribute”) is worth noticing.

Anatomy of complexity (p. 674). The section opens with a confident framing: “Growing scientific knowledge clearly shows that the causal links between stressors and harm are more complex than was previously thought”. - Co-causality: much harm (cancers, species decline) “is caused by several co‑causal factors”. Examples: reduced child intelligence linked to lead in petrol, mercury, PCBs and socio-economic factors; “bee colony collapse can be linked to viruses, climate change and nicotinoid pesticides”; climate change itself. - Timing: “In some cases, such as foetal or fish exposures, it is the timing of the exposure to a stressor that causes the harm, not necessarily the amount”, and sequences of stressors matter. - Non-monotonic dose: “In other cases, such as chemicals like BPA, low exposures can be more harmful than high exposures”. - Mixtures: “asbestos with tobacco, and some endocrine disrupting substances”: combined effects greater than separate effects. - Variable susceptibility: by genetics, epigenetics and pre-existing stress, giving different thresholds or “tipping point exposures” in different groups. - Emergent, system-level effects: “some harmful effects that occur only at the level of the system, such as a bee colony, which cannot be predicted from analysing a single part of the system, such as an individual bee.” - Long-range transport: PCBs and DDT move globally and accumulate “many thousands of kilometres away”.

Three practical implications (p. 674): 1. Strong single-cause evidence is “very difficult” to establish; “in many cases only reasonable evidence of co‑causality will be available.” 2. “a lack of consistency between research results is not a strong reason for dismissing possible causal links: inconsistency is to be expected from complexity.” [Source: Ch 27 (Gee, p. 652) makes the same argument in its reappraisal of the Bradford Hill “criteria”. It grounds it in the lead literature (Needleman 1995; Bellinger 2007: most lead studies “explain” only 30–40 % of the variance).] 3. Reducing one co-cause may not greatly reduce total harm, but “in some cases the removal of just one link in the chain of multi‑causality could reduce much harm.”

Recommendations (p. 674): “A more holistic and multi‑disciplinary systems science” is needed. It should explore “much earlier and more systematically” multiple, cumulative, metabolite and mixture effects, and low-dose effects in susceptible subgroups, with “more biological monitoring” to detect “precursors of disease”.

“No evidence of harm” ≠ “evidence of no harm” (p. 674). Several cases show “‘no evidence of harm’” interpreted as “‘evidence of no harm’”, which “may not be the case if appropriate research over relevant time periods is missing”. Examples: leaded petrol, 1920s–60s; “risks to children from mobile phones before 2011, when the first study on children was published”. [Ch 21, p. 514: the CEFALO study (Aydin et al. 2011). Ch 21’s authors (Hardell group) dispute its reassuring interpretation.] “Such authoritative but unsubstantiated assertions of safety have led to much harm, for example, in cases such as asbestos, tobacco, lead and mercury” (p. 674).

Uncertainty and ignorance (p. 675). Acknowledging them is “particularly important where the science is relatively immature, as with such emerging technologies as GM crops, mobile phones, nanotechnology and invasive alien species and where exposures are widespread”. Recognising uncertainty “helps to avoid putting too much reliance on simple models of complex systems”: floods, nuclear accidents, climate change, ecosystem resilience and multi-pollutant exposures.

Uncertainty as a weapon (p. 675). “Uncertainty, though, can be a two‑edged sword, being used as the basis for challenging both assurances of safety and evidence of a hazard.” “In particular, uncertainty has been misused, exaggerated, or even ‘manufactured’ in order to delay and undermine regulatory measures”. Examples: climate change, tobacco, lead, honeybees and beryllium (Michaels 2008; Oreskes and Conway 2010). Note the structure. The “two edges” are uses of uncertainty against safety assurances and against hazard evidence. The chapter treats the first as legitimate: it does this itself in the preceding paragraphs on “no evidence of harm”. Only the second is described as misuse (“misused, exaggerated, or even ‘manufactured’”). The chapter gives no example of uncertainty or alarm being exaggerated to block a product or technology.

Asymmetric burden of proof (p. 675). “There is also an asymmetry between the high levels of proof of harm demanded by proponents of a technology as sufficient to justify remedial or preventive actions compared to the level of evidence they deem sufficient to claim that their products are ‘safe’.” “Waiting for high levels of proof of harm before acting not only leads to much harm but also to a stifling of innovation”, citing asbestos, lead, mercury, PCBs and CFCs.

2c. Rethink and enrich environment and health research (pp. 675–676)#

Research-agenda bias (p. 675). Research should focus more on hazards of emerging technologies, not just product applications, and on “emerging hazards rather than on well‑known risks”. Evidence comes from Grandjean et al. 2011 (bibliometrics; developed in Ch 26, pp. 625–629). “The top ten substances studied are all metals such as copper, lead, zinc and cadmium.” These “account for approximately half of all the journal articles on impacts of chemical substances of the last ten years”. - [Cross-check, Ch 26, p. 626: Ch 26 confirms that “All of the top‑10 substances are metals (including arsenic, which is regarded as a semimetal)”. But the figures it reports from the same study (Grandjean et al. 2011) do not show “approximately half”. The top-20 substances account for about 12 % of all CAS-number links, which Ch 26 assumes means about 12 % of articles. The top-100 account for 180,822 of 760,056 links, about 24 %. Ch 28’s “approximately half” may use a different metric, such as the Web of Science title counts for “early warnings” substances in Table 26.1. As stated, it is not supported by Ch 26 and should be checked against the original paper. The direction (heavy concentration on a few well-known substances) is well supported.] - This “has crowded out research” into endocrine disruptors and other less-understood hazards. That holds “despite over EUR 100 million of EU research funding on endocrine disrupting compounds in the last decade”, while evidence of widespread impacts grows (EEA 2012; Kortenkamp et al. 2011).

Why the imbalance (p. 675), hedged (“may relate”, “likely”): - “the prevailing regulatory science paradigm, where solid conclusions depend on replication and verification”; - “the effective use of costly infrastructure to ensure value for money”; - “the desire of policymakers for more certainty from science regarding politically difficult choices”; - “the tendency of funding agencies to be conservative in their research strategies”.

Long-term monitoring (p. 675). Needed to catch hazards “that may only appear over decades”, focused on “‘surprise sensitive’ parameters such as bees, amphibians, invertebrates, foetuses etc.” It is also needed to evaluate whether measures work, and can be supported by citizen scientists using GIS and monitoring technologies.

Lay and local knowledge (p. 675). - Key example: “when a mother hypothesised that neurological signs observed in her son were due to exposure to mercury in her womb, this was dismissed by experts who did not question their assumption that the placenta provided protection” (Ch 5, Minamata). - Other non-scientist early warners: “Patients, fishers, wives (e.g. in the sperm damaging, described in Chapter 9 on DBCP), mothers (… DES in Volume 1…), factory workers, and bee keepers, as well as clinicians and factory inspectors are amongst those non‑scientists who have reliably provided early warnings”. - Clinicians and factory inspectors are professionals but not research scientists. The category is “outside the research establishment” rather than strictly lay.

Scientists’ asymmetric caution (pp. 675–676). “in their search for ‘certainty’ scientists are cautious in attributing causation to an agent while some scientists may sometimes be less cautious when asserting ‘safety’.” - There were “premature assertions of safety based on inadequate scientific methods, such as an over‑reliance on studies that were conducted over too short a period to reveal long‑term effects”. - Evidence of harm “has often had to reach the high standard of ‘causality’ as is the standard for less complex situations, rather than precautionary strengths of evidence based on plausible association” (pp. 675–676). - The strength of evidence chosen “can range from ‘a scientific suspicion’ of harm to ‘beyond all reasonable doubt’, depending on the complexity of the system, the level of protection required and the pros and cons of being wrong” (p. 676). This is elaborated in Ch 27, Table 27.2, pp. 657–658.

Use of animal and alternative data (p. 676). “data from animal or other species and methods (ECVAM), should be more widely used to justify precautionary action”, especially where damage is irreversible (some cancers, species and ecosystem loss, reproductive or developmental effects).

Regrettable substitution (p. 676). “Research, precaution, and exposure control also need to be applied to the substitutes or alternatives to hazardous agents.” - Cases: perchloroethylene (Ch 4), leaded petrol (Ch 3), DDT (Ch 11), booster biocides (Ch 12), and CFCs and MTBE (Volume 1). They “illustrate the hazards that some alternatives have brought in the wake of banned substances, especially when the alternatives are chemically very similar (e.g. HFCs for CFCs).” - Remedies: avoid persistence, bioaccumulation and large spatial range; hazard screening of alternatives; “smarter and greener chemistry and technology”. - The chapter frames this as a reason to extend precaution to alternatives, not as a limit on precaution. Read another way, it is also an implicit concession (this is the notes’ reading) that bans can shift harm elsewhere when alternatives go unscrutinised.

Humility and framing (p. 676). Complexity “underlines the need for greater humility about what science can and cannot tell us”. “Framing issues as purely scientific and technical inappropriately places scientific perspectives above equally valid social and ethical contributions”. A shift to “more explicitly integrative environmental science” has “started to take place in discourses but often not in practices”.

Narrowness of regulatory science (p. 676). Volume 2’s cases show “regulatory health and environmental science is still defined in very narrow terms”. This obstructs identification of multifactorial stresses. The remedy is to balance disciplinary focus with “holistic cross‑disciplinary scientific research, thereby complementing precision with relevance and comprehensiveness” (Phoenix et al. 2012), using “longer timescales, more end‑points, and multi‑causality”. Emerging fields named: “‘sustainability science’, ‘systems biology’ or ‘futures research’” (Kates 2011).

Protecting early warners (p. 676). “early warners — scientists and others — have often been harassed for their pioneering work which threatened economic interests and often challenged conventional scientific paradigms.” - Forms: “bans on speaking out or publishing; loss of funding; legal or other threats; demotion; transfer to other work and character assassination in scientific and other media” (McCulloch and Tweedale 2007, on Selikoff; Martin 1999, 2008; UCS 2012). - Remedies: extend whistleblowing and discrimination laws, more active support from scientific societies, and awards.

2d. Improve the quality and value of risk assessments (pp. 676–678)#

Progress acknowledged (p. 676). Volume 1 distinguished risk, uncertainty and ignorance. “Since 2001, some considerable progress has been made in characterising uncertainties in risk assessments”: the food sector (EFSA 2006, 2013), emerging risks (SCENIHR 2012) and climate (IPCC 2010, cited as “IPPC” in the references). This matters most where modelling dominates (climate, invasive species, exposure assessment).

Critique of narrow “risk” (pp. 676–677). “The majority of case studies indicate that it is often inappropriate to use a narrow conception of ‘risk’ to manage the complex issues at hand with their inevitable features of ignorance, indeterminacy and contingency.” Specific methodological defaults criticised (p. 677): - “linear dose response curves can be inappropriate when low doses are more harmful than high doses, as in the BPA story”; - the dose “‘makes the poison’” dictum “is inaccurate when it is the timing of the dose that makes the dose harmful, as in the TBT and DES cases”; - “assuming uni‑causality is too simplistic when multi‑causality is the reality” (lead; fisheries); - “testing for single substances is inadequate when mixtures are present as in all cases of chemical exposures”; - “over‑reliance on statistical significance when use of confidence limits would be more appropriate”.

Technological risk assessment: Fukushima (p. 677). “As the Fukushima Investigation Committee (NAIIC, 2012) concluded, ‘the accidents present us with crucial lessons on how we should be prepared for ‘incidents beyond assumptions’. With its failure to plan for the cascade effects beyond design–base accidents ‘the regulatory emphasis on risk based probabilistic risk assessment has proven very limited’.” - [Attribution problem, verified in Ch 18: the “incidents beyond assumptions” quotation is attributed in Ch 18 (p. 448) to the “Fukushima Investigation Committee (2011, p. 22)”. That is the government’s Investigation Committee interim report of December 2011 (p. 441), not the Diet’s NAIIC (2012). - The second quotation (“the regulatory emphasis on risk-based probabilistic assessment has proven very limited”) is Ch 18’s authors’ own judgement (Dorfman, Fucic and Thomas), in their chapter summary (p. 432) and conclusion (p. 449). It is not a finding of either official commission. Ch 28 presents it as though it continues the committee’s conclusion.]

Consequence (p. 677). “narrow risk assessment approaches are now outstripped by the realities that they cannot address, recognise and communicate. Too often this contributes to effective denial of those risks that do not fit the risk assessment frame.” Therefore it is “urgent to transform risk assessment practices to make them broader‑based, more inclusive, transparent and accountable”. That includes transparent communication of diverse scientific views where “genuine differences of scientific interpretations are likely, desirable, and defensible” (Stirling 2010).

Practical reforms (p. 677): - include a wider range of stakeholders “when framing the scientific risk agenda”; - make all evidence readily accessible; - broaden the scope and membership of risk evaluation committees; - make committee approaches more transparent and consistent; - ensure “independence of vested interests”.

Cited as progress: EFSA’s January 2013 announcement that it wants all data submitted for product authorisation made public (EFSA 2013).

Divergent committees (p. 677). The cases on mercury, nuclear accidents, leaded petrol, mobile phones, BPA and bees “have shown that there can be significant divergence in the evaluations of the same, or very similar, scientific evidence by different risk assessment committees”, often without explanation. - Each report should explain its “choice of paradigms, assumptions, criteria for accepting evidence, weights placed on different types of evidence, and how uncertainties were handled”. - This would help users distinguish “‘settled fact, majority opinion, legitimate minority view, and unsubstantiated assertions’” (Weiss 2002).

Funding bias (p. 677). The sources of research finance should be made explicit because of “‘funding bias’” observed for tobacco, pharmaceuticals, food, BPA, GM products and mobile phones.

Committee composition (p. 677). The cases on bees, lead, BPA and nuclear accident risks show committees “too narrow, and… sometimes… dominated by one discipline or paradigm with shared assumptions which are not therefore questioned”. Example remedy: bring endocrinology in to complement conventional toxicology, and use “recent yet reliable scientific knowledge emerging from academic research fields”.

Test relevance (p. 677). “toxicity tests designed for acute effects are unlikely to be relevant to chronic effects”. Novel technologies “such as systemic pesticides that replace sprayed pesticides or new chemical compounds replacing earlier ones, usually need novel risk assessments.”

On consensus (pp. 677–678). “Scientific conclusions should not be portrayed as if there is consensus when there is not.” “Several cases show that disagreement can be helpful to decision‑makers with a broader picture of the alternative directions and options available before making a decision.” The whole risk analysis process (assessment, management, communication) would benefit from stakeholder involvement, “particularly when framing the risk assessment and identifying options for risk management” (p. 678; see Ch 27).

2e. Foster cooperation between business, government and citizens (pp. 678–679)#

Innovation as means (p. 678). “An element that is often missing from innovation policies and practice is the recognition that innovation should be considered as a means, not an end in itself, and desirable to the extent that it improves human health and well‑being while maintaining ecological resilience.” The concept should include “non‑technological, social, institutional, organisational and behavioural innovation” (van den Hove et al. 2012).

Government roles (p. 678). The chapter says governments have “at least three roles”: 1. “providing direction by putting in place smart regulations and consistent market signals”; 2. “ensuring that the distributional consequences of innovations are balanced between risks and rewards across society”; 3. “fostering a diversity of innovations so that the wider interests of society take precedence over narrower interests”.

Business (p. 678). “Numerous case studies show that decisions to act without precaution often come from businesses.” The impediments are a “fundamental economic focus on creating and increasing short‑term economic value for shareholders” and psychological factors: “‘ethical blindness’ or a ‘self‑serving bias’ whereby people largely (and often unconsciously) tend to interpret ambiguous situations in their own interests”. This draws on Ch 25 (Le Menestrel and Rode, p. 614). Ch 25 is careful to note that some business actors “may consider in good faith that the early warning signals are not strong enough” (p. 616). Ch 28 omits that nuance.

Concentration of decision power (p. 678). Parallels are drawn between older cases and “fast emerging issues such as nanotechnologies, genetically modified crops, new chemicals, and the possible link between brain tumours and non‑ionising radiation from mobile phones”. The mobile-phone link is carefully termed “possible”. - “only a very small number of actors were involved in making strategic decisions about lead in petrol in the USA in 1925 yet the technology spread all around the world before being phased out some 60 years later.” - With GMOs and low-carbon energy options, “only a relatively few actors are involved in choosing innovation pathways that will shape the future of agriculture and energy supply and use for many decades.”

Disclosure of value conflicts (p. 678). Governments and businesses “could collaborate more with citizens and civil society on publicly disclosing and analysing the potential value conflicts entailed in acting on early warning signals”. Transparency “can in turn promote positive business attitudes and innovations”. Because “accurate determination of risk is difficult and open to disagreement”, engagement, openness and transparency are “all the more important” (p. 678).

Public involvement (p. 678) can help with: - choosing innovation pathways (WBCSD 2010; EC 2011; WBGU 2012); - prioritising public research; - supplying data for monitoring and early warning; - improving risk assessments; - identifying alternatives and unintended consequences “of both actions and inactions”; - striking trade-offs between innovations and plausible harms; - “risk‑risk trade‑offs, such as the health benefits of consuming fish which contains mercury and PCBs”.

“In particular, a feature of the studies is the top‑down nature of innovations — the history of antibiotics in animal feed and lead in petrol, for example, show how a very small number of people can take decisions which have a major impact on millions.” “The public should help shape the future, including helping to choose strategic innovation pathways, for example, to sustainable agriculture and low impact renewable energy systems, by 2050” (p. 678).

Access to private research (p. 678). There is “often a lack of public accountability and access to the private research on which public protection authorities rely”. Access would aid independent verification of licensing data “and would increase public trust in the regulatory authorities at a time when such trust in elites is very low.”

ICT, double-edged (pp. 678–679). ICT has spawned collaborative tools that “are already transforming the dynamics of governance and innovation, fostering two‑way interactions”. “Less positively”, it may foster “more hectic interactions and competition in the pursuit of enhanced productivity, less face‑to‑face contact, and less space for thinking through possible solutions to complex realities”. Creating “space for more deliberative thinking” is recommended (p. 679).

Conditions for effective engagement (p. 679): - procedures to identify stakeholders; - “adequate educational and financial resources” for public interest groups; - simpler access to data, building on the Aarhus Convention and FOI laws.

“Business concerns about confidentiality and competitiveness can be overcome through judicious use of information technologies to manage access rights while maintaining transparency”. That is asserted, with no example.

Research funding imbalance (p. 679). “Today there are large imbalances within publicly financed research between product development and the study of potential hazards, an imbalance that seems to repeat the histories of better‑known hazards.” “In Europe for example, in the period 2002–2013, about 1 % of the total amount that the EU Framework Programmes… allocated to developing products from nanotechnologies, biotechnologies and ICT was spent researching their potential hazards.” Private research “may well show a similar imbalance, but data is not easy to obtain.” - [No source is given. Internal inconsistency: Ch 27 (Gee, p. 646) says “over the past two decades… only about 3 % of the EUR 28.5 billion budget” went on hazards for the same three technology areas. Ch 27 also cites USDA biotech research 1992–2002, where “just 1 %” went to risk-related research (Mellon 2003, via Ch 19). The 1% figure in Ch 28 may conflate the US GM figure with the EU FP figure, or reflect a different calculation. The direction (a very small share) is consistent. The magnitude needs verification.]

Correcting the imbalance, with hazard research “throughout their life cycle (production, use, recycling and disposal)”, can “help avoid unequal distribution of costs and benefits further down the line and support a better public acceptability of such technologies” (p. 679).

2f. Correct market failures using the polluter pays and prevention principles (pp. 679–680)#

Internalisation (p. 679). “When evidence of initial harm emerges, the costs of such harm need to be internalised into the prices of polluting products, via taxes and charges”. Revenues would go “partly to stimulating research into less hazardous alternatives, as was the case in the US with CFCs, and partly to reducing taxes and charges on labour.” - [The US CFC example is unsourced. No other chapter I searched mentions it. My understanding is that the US excise tax on ozone-depleting chemicals (from 1990) went to general revenue rather than being earmarked for research. This needs checking.] - Taxes “would rise or fall in line with knowledge about increasing/decreasing harm”. This is an adaptive, evidence-indexed instrument that would “level the market playing field” for alternatives otherwise disadvantaged by incumbents’ unpriced external costs. - Tax shifts from labour to pollution bring “increased employment, a stimulus to innovation and a more efficient tax system” (EEA 2011b, 2011c: the EEA’s own environmental tax reform reports).

Accounting (p. 679). Firms and governments should extend accounting “beyond economic and financial capital considerations to incorporate the full human and natural capital impacts” (UN SEEA 2012; EEA 2011d; Puma 2011 environmental profit-and-loss).

Justice delayed (p. 679). “Many case studies also demonstrate the long time lags between evidence of harm and the additional injustice and time of forcing victims to pursue their case through civil compensation claims. In the case of Minamata this took over 50 years.” This is consistent with Ch 5: the disease was “officially identified in 1956” (p. 92), the Supreme Court invalidated the government’s diagnostic criteria in 2004, and many claims are still unresolved (pp. 92, 94).

No-fault compensation (p. 679). “Prompt and anticipatory no‑fault compensation schemes for victims of harm and damage to ecosystems could be set up and financed in advance of potential harm by the industries that are producing novel and large‑scale technologies”. Claimed effect: they “increase incentives for innovating companies to carry out more a priori research into the identification and elimination of hazards.” - Precedents: nuclear accidents, oil spills, some radiation exposures, some environmental liability laws “including contamination by GM crops of adjacent non‑GM farms”. - Such schemes need “provision for penalising gross negligence, which under a tort system justifies punitive damages”. - Also: “anticipatory liability bonds by innovating companies”, and a role for re-insurance “in helping to anticipate long tail liabilities” (pp. 679–680). - [Tension with Ch 24 (Cranor): Ch 24 notes that no-fault schemes “appear[] to lack deterrents” (p. 600) and that payouts alone “would lack deterrence value” (p. 599). Ch 28’s claim that such schemes raise incentives for up-front hazard research depends on how they are financed (e.g. risk-rated levies). Ch 28 does not specify this; the gross-negligence provision partly addresses it.]

Document discovery (p. 680). “Attributing responsibility and sometimes negligence to corporations and others active in the history of hazards has relied mainly upon evidence uncovered by the legal processes of document discovery in civil compensation cases.” FOI laws and the Aarhus Convention could be quicker. This “will be even more necessary if no‑fault administrative schemes replace some civil compensation cases”. That is an implicit acknowledgement that no-fault schemes would weaken a key route to historical accountability.

3. “Governance of innovation and innovation in governance” (p. 680)#

Diagnosis (p. 680). Change brings benefits but “also exposes them to more shocks and surprises”. - “Scientific and technological innovations proceed apace, more often than not on trajectories that exacerbate risks and threats.” This is a sweeping, unevidenced claim. - “those researching and developing technological innovations often fail to acquire relevant existing knowledge from other disciplines.” - “Governments tend to use structures and methods from the past to monitor the potential hazards of future technologies, rather than implementing more advanced, flexible and relevant approaches.”

The forgetting cycle (p. 680). “it appears that memories fade quickly”. A hazardous event generates urgency, preparedness, research, monitoring and heavy spending, “but lessons are soon forgotten. After some time without adverse events, willingness to invest in risk research, long‑term monitoring etc. decreases sharply and projects are downscaled or suspended. Chernobyl and Fukushima are cases in point.” - In Ch 15 this is the “‘hydro‑illogical cycle’”, a concept Ch 15 says Wilhite introduced for drought in the mid-1980s (Ch 15, Kundzewicz, p. 360). Ch 28’s description of the cycle closely paraphrases Ch 15’s flood passage (p. 360). Ch 15 already calls the cycle “a general principle, valid across different political and economic systems”. Ch 28 generalises it further, proposing “‘homo‑illogical cycle’ as it seems to be a recurrent pattern for humankind, which is found across many cultural, political, social and economic systems”. The extension from floods to all hazards (Chernobyl, Fukushima) is asserted, not demonstrated. - It adds: “this pattern need not be inescapable. Humans can learn, change and transform”. There is “enormous potential in human creativity” for “cultural, social, political, institutional, organisational and behavioural innovation, beyond ‘mere’ technological innovation”. - Plato: necessity as the mother of invention; “the crises we are facing create a level of necessity that will hopefully engender the needed innovations.” - Ch 15 (p. 361) notes that “codifying preparedness in legislation helps overcome” the cycle “in some countries” (the EU Floods Directive). Ch 28 does not carry that concrete remedy across.

Value conflicts (p. 680). “governance systems also need to better recognise the value conflicts that are underpinning all societal and environmental issues. They are unavoidable and are even desirable as they are constitutive of the human condition.” - What is missing is “the institutional space to have a much more systematic, and non‑judgmental, analysis of such conflicts so that they can be made explicit”. - Some parliamentary commissions and NGOs do this “(in part)” but not systematically. “There could be merit in establishing a place in formal institutional frameworks where such value conflicts (and consequent conflicts of interests) could be analysed and proposals offered for their resolution.”

Closing (p. 680). “The ideas for the governance of innovation and innovations in governance presented in this chapter will remain at the level of good intentions unless they are translated into institutional arrangements and practices. This is the task that lies ahead.”

References (pp. 680–682)#

About 45 references. Heavy reliance on EEA’s own outputs (EEA 2001, 2011a–d, 2012), Stirling (2007, 2008, 2010), and the doubt-manufacture literature (Michaels 2008; Oreskes and Conway 2010; UCS 2012). Some sources are unusual for structural claims (Gladwell 2012, a Nielsen blog post; Puma 2011 corporate EP&L). There are minor citation errors: “IPPC, 2010”; “Levidov” for Levidow; “Rodercick”/”Roderick”; Diedrich et al. page range “965–939”. The EEA 2001 entry gives Volume 1’s subtitle as “the precautionary principle 1986–2000” instead of 1896–2000. The Harada memorial dates the Dryden mercury dumping to “the 1970s” (p. 683); it is usually dated to 1962–1970, so this is a minor point to verify. Several key claims are uncited: the 1% funding figure, the US CFC tax revenue, and the GM human-health threat. One cited figure (top-studied substances as “approximately half” of articles, citing Grandjean et al. 2011) does not match the numbers Ch 26 reports from the same study (p. 626).

Case timeline (cross-case dates cited in the conclusion)#

This chapter is a synthesis, not a case study. The table collects the dated examples it uses, as stated in Ch 28, with notes.

Lag between warning and action: the chapter asserts delay as a general pattern (p. 672). It only quantifies it indirectly: lead in petrol ~60 years (p. 678), Minamata compensation >50 years (p. 679), and the asbestos harm chronology (p. 672).

The authors’ own lessons and conclusions#

A. Lessons the chapter derives from the case evidence (descriptive/analytical)#

1. Applying precaution on the basis of early warnings would have saved many lives and avoided much morbidity and ecosystem damage in most cases (p. 672). This is counterfactual.
2. Delay is driven by short-termism, novelty and complexity, conservative science, evidence-evaluation practices, stakeholder and vested interests, and status-quo institutional culture (p. 672). Power asymmetry is also a driver but is placed out of scope (p. 672).
3. Contemporary speed and scale worsen delay through moving-target technologies, sunk-investment lock-in and overwhelmed monitoring (p. 672).
4. Harm expansion: confirmed hazards usually turn out broader and to act at lower doses, often with no safe threshold (p. 672).
5. False positives are few: 4 of 88 alleged cases (p. 673).
6. Technological monopolies prolong and spread harm and cause technological, institutional and ideological lock-in. Artificially cheap prices stifle substitutes (p. 673).
7. Causation in complex systems is multi-causal, timing-dependent, sometimes non-monotonic, mixture-sensitive, variably susceptible and system-emergent, so single-cause proof is rarely available (p. 674).
8. “No evidence of harm” has repeatedly been misread as “evidence of no harm” (p. 674).
9. Uncertainty has been manufactured to delay regulation (p. 675), and proponents apply asymmetric evidentiary standards (p. 675).
10. Research concentrates on known hazards and neglects emerging ones (p. 675). Public research funding is heavily skewed to product development over hazards (p. 679).
11. Non-scientists have “reliably” provided early warnings (p. 675).
12. Safety has been asserted prematurely on the basis of too-short studies (p. 675).
13. Substitutes for banned agents have often been hazardous (p. 676).
14. Early warners have been harassed (p. 676).
15. Narrow risk assessment methods contribute to “effective denial” (p. 677). Committees diverge without explaining why, can be discipline-captured, and funding bias exists (p. 677).
16. Decisions to proceed without precaution often come from business, shaped by short-term shareholder value and self-serving bias (p. 678). A few actors decide for many (p. 678).
17. Regulators rely on private research that the public cannot access (p. 678).
18. Victims wait decades for compensation, and responsibility is mostly established through litigation discovery (pp. 679–680).
19. Attention to hazards decays after crises (the “homo-illogical cycle”, p. 680). Governments monitor future technologies with past structures (p. 680).

B. Recommendations and advocacy (normative)#

• Act on lower strengths of evidence than “scientific causality” (p. 672). Choose the strength of evidence case by case (“scientific suspicion” to “beyond reasonable doubt”, p. 676). Tip policy toward avoiding harm “even at the cost of more false alarms” (p. 673).
• Make continuous, anticipatory reductions in exposure to emerging hazards (p. 672).
• Carry out lifecycle, prospective and retrospective cost-of-action and cost-of-inaction analyses, including secondary benefits, false-alarm costs and misplaced reassurance, done independently of interested parties (pp. 672–673).
• Promote technological diversity and avoid monopolies (p. 673), including in GMOs and nanotechnology (p. 673).
• Openly debate the ethical choices in precaution (p. 673).
• Build holistic systems science: mixtures, low doses, susceptible groups, biomonitoring (p. 674).
• Refocus research on emerging hazards (p. 675). Set up long-term monitoring of “surprise-sensitive” indicators with citizen scientists (p. 675). Integrate lay and local knowledge (p. 675).
• Use animal and alternative-method data to justify precaution, especially for irreversible harms (p. 676). Apply precaution to substitutes (p. 676).
• Practise humility, integrative science, and avoid framing issues as purely technical (p. 676).
• Protect early warners through whistleblowing and discrimination law, scientific-society support and awards (p. 676).
• Transform risk assessment to be broader, inclusive, transparent and accountable. Explain committee reasoning, disclose funding sources, broaden membership, add new disciplines (endocrinology), design novel assessments for novel technologies, and do not fake consensus (pp. 677–678). Involve stakeholders in framing (p. 678).
• Treat innovation as a means, not an end, and broaden it to social and institutional innovation (p. 678). Governments have three roles: direction, distributional balance, diversity (p. 678).
• Disclose value conflicts publicly (p. 678). Involve the public in choosing strategic innovation pathways, “for example, to sustainable agriculture and low impact renewable energy systems, by 2050” (p. 678). Here “by 2050” dates the destination pathways, not the public involvement. Open access to the private research regulators rely on (p. 678). Resource public engagement and use Aarhus/FOI (p. 679).
• Rebalance public research funding toward hazard research across the lifecycle (p. 679).
• Use pollution taxes indexed to evidence of harm, with revenue recycling to alternatives research and labour-tax cuts, and a tax shift from labour to pollution (p. 679). Extend accounting to natural and human capital (p. 679).
• Set up anticipatory no-fault compensation financed by industry, with penalties for gross negligence, liability bonds and re-insurance (pp. 679–680). Use FOI to speed access to documentary history (p. 680).
• Create a formal institutional place for the systematic, non-judgemental analysis of value conflicts (p. 680).
• Translate all of this into “institutional arrangements and practices” (p. 680).

Mechanisms and dynamics#

1. How warnings arise, and from whom. Warnings often come from outside the research establishment: mothers, wives, fishers, workers, beekeepers, clinicians and inspectors (p. 675). Research scientists’ own conventions (replication, causality standards, statistical significance) make them slow to confirm harm (pp. 675–677). The mechanism is an epistemic mismatch. Those closest to the exposure see patterns first, but their observations lack the form that the evidentiary system recognises. Experts’ unexamined background assumptions (e.g. the placenta as barrier, p. 675) cause those observations to be dismissed. The Harada case (memorial, p. 683, read with Ch 5, p. 105) shows both halves of this. Harada was at first one of those who dismissed the mother’s view as “the fancy of an amateur”, because he shared the placenta assumption. Congenital cases then accumulated, and his 1964 thesis overturned the assumption. The correction came when later clinical evidence caught up with the lay observation, not because the observation was taken seriously at the time. (Linking p. 675 to p. 683 is the notes’ reading; Ch 28 does not draw the connection.)

2. How warnings are contested, ignored or suppressed. - Evidentiary asymmetry: proponents require strong proof of harm but weak proof of safety (p. 675). Scientists are cautious about causation but sometimes less so about safety (p. 675). Short-duration studies underpin premature assurances (p. 675). - Logical slippage: “no evidence of harm” is converted into “evidence of no harm” (p. 674). The chapter calls such statements “authoritative but unsubstantiated assertions of safety”. - Manufactured doubt: uncertainty is deliberately exaggerated to delay regulation (p. 675). - Suppression and retaliation: publication bans, defunding, legal threats, demotion, transfer and character assassination of early warners (p. 676). - Divergent and opaque expert evaluation: committees reach different conclusions on the same evidence without explaining why (p. 677). Membership captured by one discipline or paradigm leaves shared assumptions unquestioned (p. 677). Funding bias shapes the evidence base (p. 677). - Framing: treating issues as “purely scientific and technical” subordinates social and ethical considerations (p. 676). Narrow risk frames cause “effective denial” of risks that do not fit (p. 677).

3. Burden and standard of proof. This is the chapter’s central lever. It argues that the standard of evidence for action is a societal and ethical choice, not a scientific given (pp. 673, 676). It uses the courtroom analogy of civil versus criminal standards (p. 673). Where harm is irreversible, the asymmetric costs of error favour a lower trigger for action (p. 673). It pairs the lower trigger with an explicit research commitment (the “double reaction”, p. 673), which makes precaution provisional and correctable rather than permanent.

4. Incentives, interests and cognition. - Businesses focus on short-term shareholder value (p. 678). Interested parties have a “‘natural’ tendency to exaggerate costs of hazard reduction and to underestimate the benefits of action” (p. 673). “Self-serving bias” and “ethical blindness” lead people “often unconsciously” to read ambiguity in their own favour (p. 678). - Note that the psychological mechanism does not require bad faith. It predicts that sincere decision-makers with a stake will systematically discount warnings. - Policymakers want certainty from science for politically difficult choices (p. 675). Funders are conservative (p. 675). Political and financial horizons are short (pp. 671, 672).

5. Lock-in and path dependence. - Capital lock-in: investments must be “redeemed” before risk reduction (p. 672). - Monopoly lock-in: dominant technologies create technological, institutional and ideological lock-in (p. 673). - Price lock-in: unpriced externalities make incumbents “cheap” and starve substitutes (p. 673). - Decision lock-in: a few actors’ early choices spread globally and persist for decades, as with lead in petrol, 1925 to about 60 years later (p. 678). - The chapter’s antidote is diversity: “Keeping options open and following multiple paths” (p. 673).

6. The moving-target dynamic. Evidence of harm necessarily concerns an earlier version of the technology. By the time it is confirmed, “the technology has often changed, leading to assumptions that, unlike yesterday’s technology, today’s technology is now safe” (p. 672). The chapter does not say whose assumptions these are. This structurally disadvantages retrospective evidence when technology changes quickly. Together with scale, which “puts very difficult demands” on monitoring (p. 672), it means monitoring capacity tends to lag. The chapter gives no worked example of this dynamic.

7. Substitutes and alternatives. Alternatives are both the escape from lock-in (p. 673) and a source of new hazard: regrettable substitution, especially with chemically similar replacements (HFCs for CFCs, p. 676). The chapter’s answer is hazard screening of alternatives and design criteria (avoid persistence, bioaccumulation, long-range transport) (p. 676).

8. Time lags and irreversibility. Harm often shows only over decades, hence “surprise-sensitive” long-term monitoring (p. 675). Irreversibility grounds the asymmetry argument (p. 673) and the call for using animal data (p. 676). Justice lags too: over 50 years for Minamata (p. 679).

9. Distribution of costs, benefits and risks. A few decide for many (pp. 671, 678). Polluters do not pay full costs (p. 670). Victims bear delays in compensation (p. 679). Future generations bear the harms of current energy systems and chemicals (p. 670). Distribution is framed as an ethical dimension of precaution (p. 673). It is also framed as a government role: “ensuring that the distributional consequences of innovations are balanced between risks and rewards across society” (p. 678).

10. Innovation effects. The chapter claims: - precaution can stimulate innovation via diversity (p. 673); - waiting for high levels of proof of harm stifles innovation, as the asbestos, lead, mercury, PCB and CFC cases “illustrate” (p. 675). The mechanism is not spelled out there; the link to lock-in blocking alternatives is the notes’ reading, via p. 673; - monopolies hamper innovation (p. 673); - smart regulation and tax changes encourage innovation (p. 670); - pollution taxes level the field for alternatives (p. 679).

It also argues that innovation should be judged as a means to well-being and resilience (p. 678) and broadened to social and institutional forms (pp. 678, 680).

11. Complex-systems dynamics. Multi-causality, timing windows, non-monotonic dose-response, mixture effects, variable susceptibility and thresholds, emergent colony-level effects, and long-range transport (p. 674). These have epistemic consequences: single-cause proof is rare, inconsistency is expected, and removing one link may still help (p. 674). Simple models of complex systems (floods, nuclear accidents, climate) invite over-reliance (p. 675).

12. Institutional behaviour and culture. - Bureaucratic silos persist (p. 670). Institutions favour the status quo (p. 672). - Governments monitor future technologies with past structures (p. 680). Innovators fail to learn from other disciplines (p. 680). - Organisational forgetting: post-crisis enthusiasm then decay (“homo-illogical cycle”, p. 680). - Integrative science has advanced “in discourses but often not in practices” (p. 676). This is a notable observation about the gap between institutional rhetoric and practice.

13. Mental models of proponents, experts and regulators. As the chapter depicts them: - someone assumes “today’s technology is now safe” (p. 672). The chapter does not say who: proponents, regulators or the public; - experts hold unexamined physiological assumptions (the placenta, p. 675); - toxicologists hold default models (linear dose-response, “the dose makes the poison”, single-substance testing, p. 677); - nuclear regulators rely on probabilistic risk assessment that excludes cascading beyond-design events (p. 677); - Japanese authorities and utilities offered “misplaced reassurances” (p. 673); - business actors are subject to self-serving bias (p. 678).

The pattern throughout is confidence built on frames that exclude the relevant hazard.

14. Framing and language. The chapter pays attention to loaded terms and often puts them in scare quotes: “‘safe’” (pp. 672, 675), “‘cheap’” (p. 673), “‘causality’” (pp. 672, 675), “‘certainty’” (p. 675), “‘no evidence of harm’” vs “‘evidence of no harm’” (p. 674), “‘manufactured’” uncertainty (p. 675), “misplaced reassurances” (p. 673), “authoritative but unsubstantiated assertions of safety” (p. 674). It also coins or adopts its own frames: “harm expansion” (p. 672), “surprise sensitive” (p. 675), “homo-illogical cycle” (p. 680), “innovation… as a means, not an end” (p. 678), “‘mere’ technological innovation” (p. 680). Its own framing is also loaded: “smart regulation”, “smart substitutes” and “responsible information”.

15. Law, trade, media and globalisation. - Globalisation reduces national governments’ control over technologies (p. 671) and increases corporate power (p. 672). - Law is described as slow for victims (p. 679) but the main route to discovering corporate documents (p. 680). - The Aarhus Convention and FOI are proposed as faster routes (pp. 679–680). - Media appear as an actor (p. 671) and a venue for character assassination (p. 676). Internet and social media appear as enablers of accountability (p. 671). ICT is double-edged (p. 679). - Trade is not discussed in this chapter.

Transferable insights (technology-neutral)#

Ratings reflect the support available in this section, informed by the cross-checks of underlying chapters noted above. Strong = well evidenced across cases and consistent with wider literature. Moderate = supported by several cases but with gaps or selection effects. Suggestive = plausible, thinly evidenced here. Asserted = stated without evidence in the section.

1. The moving-target problem. Evidence of harm matures slowly while the technology changes quickly. Confirmed harm can then be dismissed as belonging to “yesterday’s” version (p. 672). Suggestive: a clear and plausible mechanism, but asserted without a worked example or case citation in this section.

2. Sunk investment defers risk reduction. Large up-front capital creates pressure to recover investment before costly fixes, a de facto lock-in (p. 672). Suggestive: standard lock-in logic, but illustrated only generically (energy production systems, chemical plants), with no case evidence in this section.

3. Dominance breeds multi-layer lock-in. When one technology dominates, lock-in becomes institutional and ideological as well as technical. Harms persist longer and spread wider, and alternatives are starved (p. 673; also p. 678 on lead). Moderate: four historical examples named and developed elsewhere in the report. It blends market concentration with technological dominance, and the forward claim about newer technologies rests on external sources.

4. Unpriced harm makes incumbents artificially cheap and suppresses substitutes. Internalising costs, ideally indexed to evolving evidence of harm, levels the field (pp. 673, 679). Strong in economic logic. Moderate as demonstrated here, since the empirical examples are assertions in this section.

5. Harm expansion. For hazards that turn out real, the range of effects tends to widen and the doses of concern to fall, often with no identifiable safe threshold (p. 672). Strong for the named agents (asbestos, tobacco, PCBs, lead). Only moderate/suggestive as a general prior for new agents, because the pattern is drawn from agents already confirmed as hazardous (selection).

6. The costs of errors are asymmetric under irreversibility. If a protective measure is reversible and the potential harm is not, the expected cost of wrongly waiting exceeds the cost of wrongly acting. Hence a lower trigger for provisional action, paired with research (p. 673). Moderate: logically sound as a conditional. It depends on premises the section does not test: that forgone benefits are recoverable, that the measure causes no irreversible harm of its own (risk-risk trade-offs), and that research continues after action.

7. False alarms are rarer than critics claim. Of 88 alleged over-regulation cases, 4 were confirmed false positives (p. 673). Moderate: a systematic review (Ch 2), but classification is admittedly subjective, about a third remained undecided, and the denominator is critics’ lists, not all precautionary decisions.

8. Evidentiary asymmetry between harm and safety. Those promoting a product tend to demand strong proof of harm while accepting weak proof of safety. Absence of evidence is treated as evidence of absence (pp. 674–675). Strong: recurrent across multiple case chapters (leaded petrol, asbestos, tobacco, mercury), stated here as a synthesis.

9. Uncertainty can be weaponised. Actors with a stake can exaggerate or manufacture doubt to delay protective action. Uncertainty is two-edged and can also be used against assurances of safety (p. 675). Strong for tobacco, lead and beryllium (documented in underlying chapters and cited literature). Less established for some other cases named.

10. Complexity defeats single-cause proof. Multi-causality, timing, non-linear dose-response, mixtures and emergent system-level effects make strong single-agent causal evidence rare. Inconsistent findings are expected, not exculpatory. Removing one co-cause can still prevent much harm (p. 674). Moderate: sound in principle and illustrated by examples. Some elements (non-monotonic dose-response) are contested. The “inconsistency is expected” principle, applied loosely, can also shield weak hypotheses.

11. Research attention is path-dependent. Science keeps studying well-known hazards (a “Matthew effect”). Emerging hazards are neglected, driven by replication norms, infrastructure utilisation, policymakers’ desire for certainty and conservative funders (p. 675). Moderate–strong on direction: bibliometric evidence (Grandjean et al. 2011; Ch 26, pp. 626–629). The “approximately half of all the journal articles” magnitude does not match Ch 26’s reported figures (about 12 % of links for the top-20; p. 626). The causal explanations are explicitly hedged.

12. Development outpaces hazard research. Public funding for developing technologies vastly exceeds funding for studying their hazards (p. 679). Moderate on direction, suggestive on magnitude: the section’s 1% figure is unsourced and conflicts with the 3% in Ch 27 (p. 646).

13. Early warnings often come from those closest to exposure, outside the research establishment, and are dismissed when experts’ unexamined assumptions contradict them (p. 675). Moderate: several documented cases (Minamata, DBCP, DES). “Reliably” is not quantified, and the false-alarm rate of lay warnings is not examined.

14. Early warners face retaliation. Defunding, gagging, legal threats, demotion and reputational attack (p. 676). Moderate: documented examples in cited literature. Prevalence is not established.

15. Assessment frames create blind spots. Default methods such as monotonic dose-response, single-substance testing, acute-toxicity tests for chronic hazards, reliance on statistical significance and probabilistic models that omit cascades exclude classes of hazard. The result is “effective denial” (p. 677). Moderate: concrete examples for each default. Some are contested (BPA). The Fukushima quotation supporting the probabilistic-assessment point is misattributed.

16. Opaque expert divergence erodes usable knowledge. Different expert bodies reach different conclusions from the same evidence without explaining why. Discipline-captured committees leave assumptions unexamined. Funding sources shape findings (p. 677). Moderate–strong: divergence is documented in several underlying chapters. The funding-bias literature is substantial.

17. Regrettable substitution. Replacing a banned agent with a close chemical or functional analogue often reproduces or shifts the hazard. Alternatives need the same scrutiny (p. 676). Strong: multiple independent cases across both volumes.

18. A few decide for many. Strategic pathway choices with global, multi-decade consequences are often made by a very small number of actors (pp. 671, 678). Moderate: well illustrated (lead in petrol 1925; antibiotics in feed). The extension to current pathway choices is asserted.

19. Interest shapes perception without bad faith. Stakeholders, sincerely and often unconsciously, read ambiguous warnings in their own favour. Interested parties systematically overstate the costs of hazard reduction (pp. 673, 678). Moderate: grounded in behavioural literature (via Ch 25). The claim that decisions to act without precaution “often come from businesses” reflects case selection.

20. Justice lags harm. Victims wait decades for compensation, and accountability depends on litigation discovery. Anticipatory, pre-funded no-fault schemes are proposed, but they may weaken both deterrence and discovery unless designed for it (pp. 679–680). Strong on the lag (Minamata). Suggestive on the remedy’s incentive effects (contrast Ch 24, pp. 599–600).

21. Institutional memory decays after crises. Post-disaster surges in investment in research and monitoring fade after quiet periods (p. 680). Moderate: drawn from the floods literature (Ch 15). Chernobyl and Fukushima are cited but not demonstrated here.

22. Governance lags the governed. Oversight uses structures designed for past technologies. Developers do not absorb relevant knowledge from other disciplines (p. 680). Asserted.

23. Treating value conflicts as technical questions hides them. They need explicit, institutionalised analysis (pp. 676, 680). Asserted (normative), though consistent with the divergent-committee evidence.

24. Diversity is a hedge. Maintaining multiple technological options preserves the ability to abandon one that proves hazardous and increases resilience to surprise (p. 673). Moderate: sound portfolio logic with historical illustrations. The costs of maintaining diversity are not discussed.

Limitations, contestation and bias check#

Advocacy vs analysis. - This is the report’s closing argument, written in the institution’s voice with no named author, no peer-review note (unlike case chapters, see p. 5), no panels and no dissent. - Large parts are recommendation, not demonstration. Normative conclusions are presented as following from the cases: “would seem to be a price that is well worth paying” (p. 673); “It is therefore urgent to transform risk assessment” (p. 677). - Several sweeping claims are unsupported in the text. Examples: innovations proceed “more often than not on trajectories that exacerbate risks and threats” (p. 680); warnings “in many areas, still not being heeded” (p. 670).

Case selection and hindsight. - Both volumes are predominantly “false negative” cases, as the Introduction acknowledges (p. 9). Generalisations such as harm expansion (p. 672), “decisions to act without precaution often come from businesses” (p. 678) and monopolies prolonging harm (p. 673) are drawn from agents already known to be harmful. Agents that were suspected and turned out benign, or technologies whose dominance did no harm, are not in the sample. The 88-case false-positive review is the counterweight, with the caveats below. - Early warnings are identified with hindsight. The chapter does not address the prospective problem: how to tell, at the time, which of many weak signals deserve action. Its answer is to lower the evidentiary threshold and accept more false alarms (p. 673), which moves rather than solves the discrimination problem.

The false-positive count (p. 673). Accurate as a count but stripped of Ch 2’s own qualifications: - subjectivity of classification (p. 33); - about one third “jury still out” (p. 21); - a denominator drawn from critics’ lists (p. 19).

The inference “risks are considerably less than sometimes claimed” is reasonable but not as strong as the wording implies.

Costs of precaution are under-weighted. - The asymmetry argument treats a false positive’s cost as “a delay in economic and social benefits” and says “the system will not be irreversibly altered” (p. 673). Forgone benefits can themselves be irreversible (e.g. health gains not realised during the delay), and precautionary measures can create risk-risk trade-offs. - The chapter touches on risk-risk trade-offs in three places: the explicit fish-consumption example (p. 678), regrettable substitution (p. 676), and “unintended consequences of both actions and inactions” (p. 678). It gives one concrete false-alarm cost (swine flu, p. 673). It does not integrate either into the asymmetry argument. - It also assumes that after precautionary action, research continues and can lift the measure. Its own “homo-illogical cycle” (p. 680) suggests attention and funding may lapse once a threat seems managed.

The innovation claim is conflated. - The p. 670 claim that “precautionary measures do not stifle innovation, but can encourage it” cites the Porter-hypothesis literature (Ambec et al. 2011) and environmental tax reform. That literature concerns well-designed environmental regulation and market instruments generally, not precautionary action under uncertainty. - Ambec et al. found reasonable support for the “weak” version (regulation spurs some innovation) and mixed evidence for the “strong” version (net competitiveness gains). This is based on my knowledge of that paper and should be verified. - The chapter’s own mechanism for precaution driving innovation is diversity and anti-lock-in (p. 673). That is plausible but not empirically demonstrated here.

Uncited or inaccurate specifics: - The 1% EU research funding figure (p. 679) is unsourced and inconsistent with Ch 27’s “about 3 % of the EUR 28.5 billion” (p. 646). - The Fukushima quotations (p. 677) are misattributed. The first is from the government Investigation Committee’s 2011 interim report, not the NAIIC 2012 (per Ch 18, pp. 441, 448). The second is Ch 18’s authors’ own judgement (pp. 432, 449), presented as though it were an official finding. - GM crops as “a threat to human health” (p. 674) is stronger than the underlying chapter (Ch 19, pp. 462, 468–469), which frames health concerns as unresolved and cites Séralini 2012 among others. No GM-specific source is attached in Ch 28. - The US CFC tax revenue funding alternatives research (p. 679) is unsourced and not found elsewhere in the report. - The nanotechnology “signs of technological monopolies” claim (p. 673) is sourced externally (Stirling 2007; van den Hove et al. 2012), not from the report’s nanotechnology chapter (Ch 22). Ch 22 does raise a related point: the conflict between promoting and assessing nanotechnology within one programme (NAS 2012, p. 546). - The no-fault compensation incentive claim (p. 679) sits in tension with Ch 24’s own caveats on deterrence (pp. 599–600). - The “approximately half of all the journal articles” figure (p. 675) is cited to Grandjean et al. 2011. But Ch 26, reporting the same study, gives the top-20 substances about 12 % of CAS links and the top-100 about 24 % (p. 626). - Minor: smoking and lung cancer dated 1951 (p. 672); Dryden mercury “in the 1970s” (p. 683); several reference-list errors, including Volume 1’s subtitle misdated “1986–2000” (p. 681).

Complexity arguments can cut both ways. “Inconsistency is to be expected from complexity” (p. 674) is epistemically correct under genuine multi-causality. Without criteria for distinguishing complexity-driven inconsistency from a null effect, it can protect any hypothesis from disconfirmation. The chapter does not offer such criteria. Ch 27 reappraises the Bradford Hill criteria under multicausality (pp. 651–654). It argues that “the presence of the criteria can be robust evidence for a causal association, whereas the absence of the criteria is not robust” evidence against it (p. 653).

Uncertainty’s two edges. The chapter acknowledges that uncertainty is “a two-edged sword” (p. 675): it can be used to challenge both safety assurances and hazard evidence. But it treats only the second use as misuse, and its examples of misuse are all of uncertainty deployed against regulation. It gives none of uncertainty or alarm being exaggerated to block technologies or products. The chapter does not claim that misuse runs both ways. The asymmetry lies in which uses it examines, not in a stated symmetry left unexplored.

Fairness in the other direction. The chapter is more self-aware than a pure polemic: - It concedes that the power imbalance is “well beyond the scope of this report” (p. 672). - It accepts that lower evidentiary thresholds mean more false alarms (p. 673). - It counts false-alarm costs (p. 673) and insists that cost analyses be independent of political as well as commercial interests (p. 673). - It warns of regrettable substitution (p. 676) and notes ICT’s downsides (p. 679). - It credits real progress: EFSA, SCENIHR, IPCC (p. 676); EFSA data transparency (p. 677); improvements since 2001 (p. 670). - It treats business confidentiality as a legitimate concern to be managed (p. 679). - It treats value conflicts as “desirable” rather than as obstacles (p. 680). - It ends by admitting that everything proposed remains “good intentions” until institutionalised (p. 680). - Its hedging is sometimes careful: the mobile-phone link is only “possible” (p. 678); chemical exposures “appear to contribute” to disease (p. 674).

Institutional position. The EEA is an EU agency. Many recommendations promote EEA’s own programmes (environmental tax reform, ecosystem capital accounting, EEA 2011a–d). That does not make them wrong, but the evidence base is partly self-referential.

Notable quotes#

1. “by the time evidence of harm is confirmed, the technology has often changed, leading to assumptions that, unlike yesterday’s technology, today’s technology is now safe.” (p. 672)
2. “This phenomenon of ‘harm expansion’ is rendered more problematic by the discoveries that harm from all of the above agents has been found to occur at lower and lower levels” (p. 672)
3. “Such pro and con analyses should be independent of interested parties, both commercial and political, as they often have a ‘natural’ tendency to exaggerate costs of hazard reduction and to underestimate the benefits of action.” (p. 673)
4. “These artificially low market prices in turn helped to stifle the development of smart substitutes.” (p. 673)
5. “Tipping the overall balance of public policy towards avoiding harm, even at the cost of more false alarms, would seem to be a price that is well worth paying” (p. 673)
6. “a lack of consistency between research results is not a strong reason for dismissing possible causal links: inconsistency is to be expected from complexity.” (p. 674)
7. “in their search for ‘certainty’ scientists are cautious in attributing causation to an agent while some scientists may sometimes be less cautious when asserting ‘safety’.” (p. 675)
8. “narrow risk assessment approaches are now outstripped by the realities that they cannot address, recognise and communicate. Too often this contributes to effective denial of those risks that do not fit the risk assessment frame.” (p. 677)
9. “innovation should be considered as a means, not an end in itself, and desirable to the extent that it improves human health and well‑being while maintaining ecological resilience.” (p. 678)
10. “could perhaps be called the ‘homo‑illogical cycle’ as it seems to be a recurrent pattern for humankind” (p. 680)

Open questions#

1. Prospective discrimination. If thresholds for action are lowered, what criteria separate warnings worth acting on from noise, before hindsight is available? The chapter’s answer (case-by-case strength of evidence weighted by irreversibility and protection level, p. 676) states the factors but not how to weigh them.
2. Does precautionary action actually trigger sustained research? The asymmetry argument depends on a “double reaction” of action plus research (p. 673). What happens to research funding once a precautionary measure is in place, given the attention cycle (p. 680)? Ch 2’s four false positives did generate research (p. 33); is that typical?
3. What was the real EU hazard-research share, 1% or 3%? How did it change under later Framework Programmes?
4. Monopoly vs. dominance. Is the harm mechanism market concentration (few firms), technological monoculture (one design), or both? The policy implications differ (competition policy vs. portfolio diversity).
5. Do anticipatory no-fault schemes improve safety incentives, or mainly speed compensation? How should levies be risk-rated to preserve deterrence, and how is documentary accountability preserved without litigation discovery (p. 680)?
6. How should forgone benefits and risk-risk trade-offs enter the asymmetry calculus when they are themselves irreversible?
7. Who decides the “strategic innovation pathways” that the public should help choose, e.g. toward sustainable agriculture and low-impact renewables “by 2050” (p. 678)? Through what institutions? The chapter’s closing admission (p. 680) leaves this open.
8. Is “harm expansion” a general law or a selection artefact? How often do suspected hazards contract rather than expand as evidence accumulates?
9. What share of chemical-impact articles do the most-studied substances actually account for? Ch 28 says “approximately half” (p. 675), but Ch 26’s figures from the same study suggest far less (p. 626). This needs checking against Grandjean et al. 2011.

Pointers for the hindsight strand. These are outside the report and must be verified there. - Séralini et al. 2012, cited in Ch 19, was retracted by the journal in late 2013 and republished elsewhere in 2014. Post-2013 consensus assessments of GM food safety (e.g. the US National Academies, 2016) bear on p. 674. - EU restrictions on three neonicotinoids (2013) and outdoor bans (2018) bear on the bee-related claims. - The Kigali Amendment (2016) on HFCs bears on regrettable substitution (p. 676). - The EU Whistleblower Directive (2019) bears on early-warner protection (p. 676). - The EU Transparency Regulation for food-chain risk assessment (2019) bears on data access (pp. 677–678). - EFSA’s much-reduced BPA tolerable daily intake (2023) and subsequent EU restrictions bear on BPA and divergent committees (pp. 674, 677). - Later large studies and systematic reviews on mobile phones and brain tumours bear on pp. 674 and 678. - The emergence of PFAS as a major regulatory issue bears on the “emerging hazards neglected” claim (p. 675; Ch 26 flagged perfluorinated compounds as under-studied, p. 628). - Trends in environmental tax shares and natural capital accounting (UN SEEA Ecosystem Accounting, 2021) bear on p. 679.

Audit log#

Independent audit against the full text extract (report pp. 670–684) and the PDF (text re-extracted; word counts match the extract on all 15 pages; no tables or figures, and images only on the memorial pages). Cross-references were re-checked in the same PDF.

Verified, no change needed: all main quotations spot-checked as verbatim. Ch 18 Fukushima misattribution confirmed (pp. 432, 441, 448, 449). Ch 27 “about 3 % of the EUR 28.5 billion” confirmed (p. 646). Ch 2’s four false positives, “one third” categories and subjectivity caveat confirmed (pp. 20–21, 25, 33). Ch 19 pp. 462, 468–469; Ch 21 p. 514 (CEFALO); Ch 22 p. 546 (NAS 2012); Ch 24 pp. 599–600; Ch 25 pp. 614, 616; Ch 5 p. 92; Ch 15 pp. 360–361; Introduction p. 9 (“false negatives”); Acknowledgements p. 5 (editorial team; case chapters peer reviewed); index (“harm expansion” only at p. 672). The CFC-tax example appears nowhere else in the report.

Changes to notes: 1. Inferred influences: relabelled the Ch 27 p. 646 point as research-funding imbalance (it is not “funding bias”). Added Ch 27 pp. 652 and 659 and Ch 15 p. 360 as evident sources. Limited “explicitly” drawing on van den Hove et al. to pp. 673 and 678; the closing heading on p. 680 is uncited. 2. Harada memorial: noted that his Ch 5 co-authorship comes from pp. 4–5, not the memorial. Added his early-1970s first visit to Grassy Narrows and White Dog. Changed “worked in” to “visited” to match the source. 3. Harada note: added Ch 5 p. 105. Harada himself at first dismissed the mother’s hypothesis as “the fancy of an amateur”. 4. Power-asymmetry limit: removed the unsupported “deepest driver” characterisation and added the omitted “more amenable to change” sentence. 5. “Harm expansion”: changed “coins” to “introduces”; a full-report text search confirms the term appears nowhere else. 6. Monopoly cross-check: added Ch 27 p. 659 as internal support for the historical (not the forward-looking) monopoly claim. 7. False-positive cross-check: replaced “critics’ lists” with Ch 2’s actual method and main sources (p. 19). Added Ch 2’s own statement of confidence in its core finding (p. 33). 8. GM cross-check: added Ch 19 p. 468 (“accumulating… evidence of adverse effects”; “plausible that there simply are no effects”). The quotation is now verbatim, and the “stronger than Ch 19” judgement is narrowed to human health. 9. Complexity: added the chapter’s framing sentence “Growing scientific knowledge clearly shows…” (p. 674). 10. “Inconsistency is to be expected”: traced to Ch 27 p. 652 (Needleman 1995; Bellinger 2007). 11. Uncertainty as a two-edged sword (section notes): clarified that the chapter treats challenges to safety claims as legitimate and labels only anti-regulatory uses as misuse. 12. Research agenda: flagged that “approximately half of all the journal articles” does not match Ch 26 p. 626. There the top-20 substances have about 12 % of CAS links and the top-100 about 24 %. Also noted Ch 26’s arsenic-as-semimetal caveat. 13. Regrettable substitution: reframed “important moment of self-limitation” as the notes’ reading; the chapter frames it as extending precaution to alternatives. 14. Government roles: restored “at least three roles”. 15. Value-conflict disclosure: added the omitted sentence on risk being “difficult and open to disagreement” (p. 678). 16. Funding imbalance: added the omitted clause “an imbalance that seems to repeat the histories of better-known hazards” (p. 679). 17. Document discovery: “an acknowledgement… would remove” is now “an implicit acknowledgement… would weaken”. 18. Forgetting cycle: noted that Ch 28 closely paraphrases Ch 15 p. 360 and that Ch 15 already generalises the cycle. Corrected “can break the cycle” to Ch 15’s “helps overcome… in some countries” (p. 361). 19. References paragraph: added the EEA 2001 entry’s misdated subtitle (“1986–2000”). Changed “numerical claims” to “claims”, since the GM claim is not numerical. Added the “approximately half” mismatch. 20. Recommendations (B) and open question 7: corrected a misreading of “by 2050”. It dates the destination pathways (sustainable agriculture, renewables), not public involvement. 21. Mechanism 1: corrected the claim that the Harada memorial shows a physician taking the lay observation seriously. Per Ch 5 p. 105 he first dismissed it, and later evidence vindicated it. The p. 675 to p. 683 link is now marked as the notes’ reading. 22. Mechanism 6 and mechanism 13 (mental models): removed the attribution of “today’s technology is now safe” to proponents, since the chapter does not say whose assumption it is. Softened “always behind” to “tends to lag” and noted that there is no worked example. 23. Mechanism 10: marked the lock-in explanation of the p. 675 innovation claim as the notes’ inference. 24. Transferable insights 1 (moving target) and 2 (sunk investment): downgraded from Moderate to Suggestive under the notes’ own rating definitions. Neither has case evidence in this section. 25. Transferable insight 11: added the magnitude caveat on the “approximately half” figure. The rating stays Moderate–strong on direction. 26. Bias check (costs of precaution): corrected “mentions risk-risk trade-offs once”. The chapter also touches on them through regrettable substitution (p. 676) and “unintended consequences of both actions and inactions” (p. 678). 27. Uncited/inaccurate specifics: added the “approximately half” mismatch and the Volume 1 subtitle error. 28. Complexity cuts both ways: corrected the Ch 27 reference to pp. 651–654 and quoted the asymmetry statement (p. 653). 29. Uncertainty’s two edges (bias check): reworded. The chapter does not claim that misuse runs both ways; the one-sidedness is in which uses it examines. 30. Open questions: added question 9 on the true bibliometric share. 31. Header: the list of cross-checked material now includes the contents, Acknowledgements and Introduction (pp. 4–5, 9).

Changes to digest (logged here): 32. Header: “condensing” Ch 26/27 is now marked as inference, with Ch 15 added. The memorials “follow the chapter” rather than “close the volume” (annexes and the index follow). 33. Delay barriers: added the omitted novelty/complexity barrier. 34. False alarms: denominator description made precise, with Ch 2 page references. 35. Epistemic failures: flagged the “approximately half” mismatch. 36. Insight 1 (moving target): downgraded to Suggestive. Insight 11 now carries the Harada nuance. Insight 14 now notes the Ch 15 source and that the generalisation is asserted. 37. Caveats: replaced “no method… beyond lowering thresholds” with an accurate statement. The chapter names threshold-setting factors (p. 676) but no way to weigh them. 38. Accuracy problems: added the “approximately half” mismatch and the Volume 1 date error in the reference list.