T08 — Actors, mindsets, framing and narratives#

Cross-cutting thematic synthesis across Late lessons from early warnings (EEA 2001, “LL1”) and Late lessons from early warnings: science, precaution, innovation (EEA 2013, “LL2”). Strand A working document, 26 September 2026.

How to read this document#

Scope. How proponents, firms, experts, scientists and regulators thought about the technologies and hazards in the two reports; the language that carried their thinking (“magic”, “safe”, “normal”, “no evidence of harm”, “controlled use”); who counted as an expert; and whether harm came from sincere belief, motivated reasoning or knowing bad faith.

Sources. All 47 digests; notes for the most central sections (LL1-15, LL2-06, LL2-25) in full and the relevant sections of about 30 others; keyword searches across all notes; hindsight files for sections relied on; external/critiques.md. Quotations and page numbers were re-checked against working/text/chunks/.

Citation form. Section id plus report page, e.g. “LL1-15, p. 161”. Hindsight material is cited as “(hindsight LL1-15)”.

Three voices, kept apart. - Reports say: what the chapter or panel claims. - Evidence and hindsight: what the documents cited, the audited notes and the post-publication record show. - Analysis: my own synthesis, inference or judgement.

Strength ratings. - Strong: documented with contemporaneous or primary evidence in three or more independent cases, or confirmed by later evidence that does not come from the chapter authors. - Moderate: documented in several cases but resting partly on protagonist or secondary sources, or needing an inferential step. - Suggestive: one or two cases, or mainly inference. - Asserted: stated in the reports without supporting evidence.

1. Summary of findings#

1. Two unreconciled theories of failure. One is cognitive: “misplaced certainty”, blind spots, “hubris” (LL1-00, p. 4; LL1-16, p. 170). The other is interest-based: product defence, manufactured doubt, “irresponsible corporations” (LL2-00, p. 11; LL2-27, pp. 645–646). Each fits different cases. The best bridge is LL2-25 (pp. 613–617) and LL2-28 (p. 678): interest shapes perception, often without bad faith, and the sharper line is between “business actions” and “political actions” that change the rules and the evidence.

2. Sincere belief did at least as much damage as bad faith. The reports attribute no bad faith to DES prescribers (LL1-08, p. 88), radiation pioneers (LL1-03), the CFC inventors (LL1-07), TBT regulators (LL1-13), the antimicrobial reassurances (LL1-09), the PCE pipe-liner maker (“merely indifferent”, LL2-04, p. 84) or fisheries scientists (LL1-02; LL2-17). The BSE inquiry found ministers sincerely believed the risk remote (hindsight LL1-15). The swine-flu false positive was also sincere, confident belief (LL2-02, p. 28).

3. Documented bad faith exists in a minority of cases, and later evidence has strengthened it: second-hand smoke, vinyl chloride, Minamata, beryllium, PCBs, leaded petrol (LL2-07; LL2-08; LL2-05; LL2-06; LL1-06; LL2-03; hindsight for each and LL2-25).

4. Recurring mindsets are well evidenced and technology-neutral: celebrating the property that later proves hazardous; confidence built on not having looked; models of harm fixed on the first visible endpoint; faith that systems perform to specification; seeing the public as prone to “hysteria”; defining “safe” by what can be measured; commitment that hardens once a public position is taken.

5. Language did real work. “Magic mineral”, “gift of God”, “modern and scientific”, “normal/natural”, “perfectly safe”, “controlled use”, “sound science”, “public relations problem” and “residual risk” turned contested judgements into apparent facts (LL1-05, p. 53; LL2-03, pp. 53, 56–58; LL1-08, p. 88; LL1-15, p. 161; LL2-06, p. 133; LL2-18, p. 448).

6. Expertise was institutionally situated. The same evidence produced different verdicts depending on the body, its admissibility rules and its remit (LL2-04, pp. 84–85; LL2-10, pp. 221–223; LL1-11, p. 113; LL1-16, p. 181), a pattern reinforced since 2013 (hindsight LL2-10, LL2-04, LL2-28).

7. The same mindsets appear among the reports’ authors. Protagonist chapters apply asymmetric scrutiny, and those most driven by conviction (mobile phones, GM health, Chernobyl numbers) fared worst in hindsight (hindsight LL2-21, LL2-19, LL2-28, LL2-18). A lens must question warners as well as proponents.

2. The reports’ own account of actors#

Reports say. - LL1’s Preface blames “misplaced ‘certainty’ about the absence of harm” in “most” cases but rates lack of political will “an even more important factor” (LL1-00, p. 4). - The LL1 editors separate “institutional ignorance” from warnings that were “discounted”, sometimes for “short-term economic and political” reasons (LL1-16, pp. 168, 171), and call for “more humility and less hubris” (p. 170, quoting LL1-09, p. 98). - LL2 frames delay as “the relationship between knowledge and power” (LL2-00, p. 7), names “corporate and scientific establishments” and “irresponsible corporations” (pp. 10–11), but also causes involving no bad faith: novelty and poorly evaluated information (pp. 10–11). - LL2’s business chapter gives the fullest model. Firms count social harm only when it returns through liability, regulation or reputation (LL2-25, p. 608). Uncertainty is both genuinely hard and a “welcome ‘excuse’” (p. 614). “Good people” build cultures where problems get buried (p. 615). “Political actions” that change rules and evidence differ from “business actions” within them (p. 615), and secrecy about them “can be seen as a signal” of bad faith (p. 617). - LL2’s conclusion adds that “self-serving bias” and “ethical blindness” lead people, “often unconsciously”, to read ambiguity in their own favour (LL2-28, p. 678).

Analysis. The reports hold three positions at once: a charitable, cognitive account (LL1-16; LL2-25 on culture; Guidotti in LL2-06), a prosecutorial, interest-based account (LL2-00, p. 11; LL2-27, pp. 645–646; several case chapters) and a structural account in which no one needs to act in bad faith (LL2-26, pp. 624–629; LL1-17, p. 193). LL2-25 gives no way to tell these apart in a live case (notes LL2-25); the rest of this document tries to do so from the case evidence.

A standing caveat. Case authors were “for the most part active participants” in their histories (LL1-00, p. 12), and LL2 chose authors for “substantial involvement” (LL2-00, pp. 9–10). Many attributions of motive in the reports are made by adversaries of the people described. That does not make them wrong, but it matters most on exactly this theme.

3. Pattern A — Enthusiasm for benefits, and the property that is both virtue and hazard#

Reports say. - Radiation: the excitement of discovery and real medical value meant “caution tended to be thrown away” (LL1-03, p. 31). Signs of injury were even taken up as therapy (pp. 31, 33). - Asbestos was the “magic mineral”, with more than 100 products within 20 years (LL1-05, p. 53). In 1967 the Lancet called it “often irreplaceable” and said it could “save more lives than it can possibly endanger” (p. 58). - DES prescribers “undoubtedly… believed it was safe and effective, and both ‘modern and scientific’”, and refusing it took “a courageous physician” (LL1-08, p. 88). - Tetraethyl lead was an “apparent gift of God” (LL2-03, p. 53). For EE2 “there seemed no limit” to medical uses of artificial oestrogens (LL2-13, p. 280). The 1948 DDT Nobel presentation treated resistance as “an opportunity to develop new chemical pesticides” (LL2-11, p. 241). - CFCs were “efficient, non-toxic and non-flammable” (LL1-07, p. 79); PCBs “thought to be safer” (LL1-06, p. 64); TBT “extremely effective and relatively economical” (LL1-13, p. 135); MTBE could be made at the refinery and moved through existing systems (LL1-11, p. 110). - Emerging issues: “wonder technologies” and “nano-fever” (LL2-22, pp. 530, 546); Gaucho as “reducing aerial pollution” (LL2-16, p. 371); GM promises narrowing to two traits (LL2-19, pp. 463, 466). - Farman’s general point: short-term safety “appears to demand” inertness, which means persistence (LL1-07, p. 83).

Evidence and hindsight. - The quotations are verified; they show a mindset, not a measured effect on decisions. - In several cases the celebrated property was the hazard: persistence for DDT, PCBs and CFCs; stability for EE2 (my inference from LL2-13, p. 292); resistance to degradation for MTBE. The lesson has since entered law through the EU’s 2023 persistent-and-mobile hazard classes (hindsight LL1-11, LL1-16, LL1-17), and later substitute cycles repeated the pattern (hindsight LL2-12).

Analysis. - Supporting cases: about 13 (radiation, asbestos, DES, lead, EE2, DDT, CFCs, PCBs, TBT, MTBE, nano, neonicotinoids, GM). - Strength: moderate for “benefit salience crowds out attention to slow, invisible harm”. The mindset is well documented across many cases. The causal step from enthusiasm to neglect is inferred, and the cases were chosen because harm followed. - Strength: strong for the narrower mechanism that the property prized for performance can be the source of lasting harm. It recurs in at least five independent cases and has since been institutionalised. - Cases that cut against the pattern: - Several enthusiasms were for real safety gains. PCBs cut fire risk, CFCs replaced toxic refrigerants, and asbestos had fire-safety value. The chapters barely weigh this (LL1-06 caveats; LL1-05, p. 58). - The Lancet’s benefit argument was dismissed on competence grounds rather than engaged (notes LL1-05). - Enthusiasm also runs on the other side. Hindsight finds precautionary advocates oversold alternatives: alcohol as “equally effective” (hindsight LL2-03) and agroecology yields (hindsight LL2-19).

4. Pattern B — Mental models that fixed what could count as harm#

This is the clearest case of sincere, paradigm-bound error, and the reports mostly present it as such.

Reports say. Six recurring models appear.

1. Acute, visible, threshold harm. - Early radiation limits were calibrated to skin burns and hair loss, with “no realisation” of latent cancer. The 1925 tolerance dose was about 700 mSv a year, against 20 mSv now (LL1-03, p. 33). - Kehoe’s school held that lead poisoning occurred “only at high doses with obvious signs of severe illness” (LL2-03, p. 58). - The beryllium limit was calibrated to the acute disease. Chronic cases in workers exposed below the limit, and in community residents, were “viewed as anomalous” (LL2-06, pp. 133–134). - The TBT target of 20 ng/l rested on acute toxicity data. Dogwhelks were sterilised at 3–5 ng/l (LL1-13, pp. 136–137, 141). - The DBCP limit was set below the lowest dose tested, an assumption the chapter calls “mistakenly assumed to be safe” (LL2-09, pp. 205, 210). - Hormone residues were judged against pill users and pregnant women rather than children (LL1-14, p. 150).
2. Protective barriers and invisibility. Harm was assumed impossible from an agent “that could not be appreciated by the senses” (LL1-03, p. 31). The placenta was believed to protect the foetus (LL1-08, p. 84; LL2-05, p. 105).
3. Dilution and dispersal as disposal. The electricity industry was “confident” emissions could be dispersed to harmless levels (LL1-10, p. 102). PCB discharges were “considered acceptable, legal, and hazard-free” (LL1-06, p. 70). Large ships were exempted from TBT controls on an “open seas” model (LL1-13, p. 139).
4. Systems perform to specification. “Controlled use” of asbestos, “closed” PCB systems, leak-free MTBE tanks and compliant slaughterhouses (LL1-16, pp. 174–175; LL1-05, p. 57; LL1-11, p. 115; LL1-15, p. 162). Nanomaterials were assumed to be handled in small, sealed quantities (LL2-22, pp. 544–545). Levees built to a “design flood” create a “false feeling of security” (LL2-15, p. 356). Nuclear design bases were assumed to bound credible events (LL2-18, pp. 437–438, 447–448).
5. The model as object of loyalty. - Fisheries science: more mathematics brought “optimism that past mistakes could be avoided”, and failed predictions were met with added complexity (LL1-02, pp. 20–22). - Harris (1990) found scientists “lulled by false data signals” and emotionally committed to their management strategy (LL2-17, p. 413). - Antimicrobials: successive assumptions that resistance would stay confined were overturned, and two had been used to reassure about farm use. The authors diagnose reliance on “what was precisely known rather than… what was not known” (LL1-09, pp. 94–95, 98). - BPA: expectations from its weak receptor binding, plus monotonic dose-response and GLP-only evidence, led influential scientists to call low-dose findings “neither credible nor plausible” (LL2-10, pp. 219–220). - Mobile phones: an association was “biologically bizarre” under the thermal paradigm (LL2-21, p. 513).
6. Stationarity and simple systems. In 1900 a “comfortable feeling” that the atmosphere was simple, which Farman calls complacency (LL1-07, p. 78). Design codes assume “the past is the key to the future” (LL2-15, p. 355). The 1985 Villach statement warned that long-term decisions assumed past climate data were “a reliable guide for the future” (LL2-14, p. 316).

Evidence and hindsight. - The chapters describing these models mostly decline to attribute bad faith: TBT regulators’ errors are presented as paradigm-bound (notes LL1-13), 1930s–60s PCB science as within the normal limits of knowledge at the time (notes LL1-06), the 1929 CFC choice as ignorance (LL1-07, p. 82). - Later evidence confirms several of these errors independently of the authors: - LNT-based radiation risk at low doses (hindsight LL1-03); - TBT thresholds wrong by about 100x (hindsight LL1-13); - no threshold for lead neurotoxicity (hindsight LL2-03); - the beryllium limit cut tenfold by OSHA in 2017 and the EU in 2019 (hindsight LL2-06); - the IAEA’s 2015 finding that Japan’s “widespread assumption” that an accident of this magnitude was “unthinkable” went unchallenged by regulators (hindsight LL2-18). - But paradigm-based scepticism was sometimes right. The “biologically bizarre” prior on mobile phones fits later large cohorts and a WHO-commissioned review (hindsight LL2-21); the dominant view on food irradiation was upheld (LL2-02, pp. 29–31; hindsight LL2-02); BPA remains contested among major assessors (hindsight LL2-10).

Analysis. - Supporting cases: at least 17 across both volumes. - Strength: strong that experts’ mental models fixed what counted as harm, and that this produced long-lived false reassurance. The evidence is contemporaneous statements, and in most cases later independent confirmation. - Strength: moderate for the claim that paradigm commitment, rather than interest, was the main driver. In lead, beryllium and Minamata, paradigm and interest are intertwined. - Cases that cut against the pattern: mobile phones, irradiation and, in part, BPA show that “paradigm defence” is not by itself evidence of error. What distinguishes the harmful cases is not holding a prior but: - never testing it against an independent baseline (Kehoe’s contaminated controls, LL2-03, pp. 56–57); - treating the edge of current knowledge as the edge of risk (LL1-09, p. 98); - refusing to specify what evidence would change the view.

5. Pattern C — Confidence built on absence: “safe”, “no evidence of harm”, and the reassurance trap#

Reports say. - The fallacy that “no evidence of harm” means “evidence of no harm” is named for asbestos (LL1-05, p. 53), BSE, where evidence was not “actually being sought” (LL1-16, p. 172), lead (LL2-03, pp. 50, 70), beryllium (LL2-06, pp. 139–140), cod (LL2-17, p. 411), GM assessment (LL2-19, p. 468), mobile phones (LL2-21, p. 511), and science generally as “the greatest error” (LL2-26, p. 631; LL2-27, pp. 657–658; LL2-28, p. 674). - The absence was often produced, not found. BSE: the national surveillance body was excluded, and there was no random survey of food-chain cattle until 1999 (LL1-15, pp. 163–164). Gaucho: public researchers were told not to go below a 10 ppb detection limit that matched the manufacturer’s method (LL2-16, p. 373). Beryllium: detection limits and a blood test not yet available (LL2-06, pp. 135, 139). Asbestos: short follow-up and survivor bias, the “pensioners’ party” (LL1-05, pp. 55, 60). - Categorical safety language: “perfectly safe” beef (LL1-15, p. 161); “The standard is safe” for beryllium (LL2-06, p. 137); Dodds’s DES “non-toxic” (LL1-08, p. 85); “no reason to be concerned” in fisheries (LL1-02, p. 20); enough information “to discount any possibility of risk” from oestrogens in drinking water, on a small data set (LL2-13, p. 283); “authoritative assertion but without evidence” (LL2-09, p. 211; LL2-03, p. 50). - The reassurance trap (LL1-15, pp. 161–162). Once beef had been declared safe, any further measure, “no matter how useful or cheap”, implied the reassurance had been false. Cheap measures were then refused for presentational reasons: the mechanically recovered meat issue was “agreed not to raise”; a head-meat ban was blocked because “Nor would it end with spinal cords”. Rules that were adopted were described as scientifically unnecessary, and about 48% of abattoirs visited in 1995 were not complying, which an enforcer called “a bit of window dressing” (p. 162). - The safety language also rested on proxies. Odour was treated as a warning when the odour threshold was far above the hazardous level: DBCP at 1.7 ppm against a 1 ppm limit (LL2-09, p. 205); vinyl chloride at 4,000 ppm, not 400 (LL2-08, p. 184). A pipe lining that was leaching solvent was said to be not “cured” completely (LL2-04, p. 77).

Evidence and hindsight. - The BSE mechanism is the best documented in either report, resting on contemporaneous minutes. The Phillips Inquiry adds that the reassurance campaign’s object was “sedation”, not deception, and that it contributed to poor enforcement (hindsight LL1-15, citing paras 1179, 1186). - Hindsight also shows the premise was false in practice. Mechanically recovered meat remained a route of infection until December 1995, and spinal cord was found in carcasses that had passed inspection (hindsight LL1-15). - Crucially, the Department of Health, which had no producer interest, was “as eager as MAFF” to avoid alarm (hindsight LL1-15, para 1189). The trap operated without a sponsorship conflict and without lying. - Analogues: the beryllium limit as “fundamental to our product liability defense” (LL2-06, p. 137); Japan’s nuclear “safety myth” (hindsight LL2-18); DFO’s reassurance (LL2-17, p. 414).

Analysis. - Supporting cases: about 14 name the fallacy and about 8 show it produced by not looking. - Strength: strong for the fallacy and its recurrence. - Strength: strong for the reassurance trap in BSE. Categorical reassurance makes graded precaution unpresentable and erodes enforcement. - Strength: moderate for the trap as a general dynamic, with clear analogues in three other cases. - Complications. Several independent, well-powered null lines, followed long enough, do become evidence of absence; mobile phones are the clearest case (hindsight LL2-21). Over-confidence also runs the precautionary way: swine-flu advisers thought the vaccine safe, with “little to lose” (LL2-02, p. 28), applying precaution to the disease but not the intervention (notes LL2-02).

6. Pattern D — How decision-makers saw the public, lay observers and critics#

Reports say. - The public as prone to panic. BSE: the Chief Veterinary Officer warned that publicity might produce “hysterical demands” for “draconian” measures (LL1-15, p. 159); Phillips described information policy as aimed at “sedation” (p. 161). PCBs: the 1937 meeting on worker poisoning closed with a plea to avoid “mob hysteria” among the workmen (LL1-06, p. 64). Radiation: a professional society dismissed press critics as better at “lurid journalise” than science (LL1-03, p. 33). Swine flu: officials managed the media to avoid “gloom and doom” (LL2-02, p. 27). Agencies fear being accused of “hysteria” (LL2-02, p. 34). - Lay knowledge dismissed. At Minamata, lay observations were “not scientific” (LL2-05, p. 96), and Harada recalls dismissing a mother’s reasoning about the placenta as “the fancy of an amateur”; “time proved that she was correct” (p. 105). The inshore cod fishers’ report was “biased pseudoscience” (LL1-02, p. 21) and inshore data “messy and often anecdotal” (LL2-17, p. 414). Environmental health officers’ head-meat warning was blocked (LL1-15, p. 162). DBCP workers found their own sterility by comparing notes at lunch (LL2-09, p. 204); beekeepers saw Gaucho’s effects first (LL2-16, pp. 372–373). - The editors’ view. Lay knowledge is “complementary” and may rest on different assumptions about what is salient, while specialists “may simply respond to granted authority” (LL1-16, p. 177). Institutions presume mistrust stems from the public’s “supposed misunderstandings of science, risk and uncertainty”, whereas the public grasps ignorance “rather better than does institutionalised science itself” (p. 185).

Evidence and hindsight. - The “mob hysteria” remark is verified word for word in the published symposium record (hindsight LL1-06). - Flint (2014–15) repeated the pattern. The official task force found “aggressive dismissal, belittlement, and attempts to discredit” independent testers and a paediatrician (hindsight LL2-03). - The editors’ claim about publics rests on two focus-group studies, one co-authored by an editor (LL1-16 caveats). Global survey data do not support LL1’s claim of “very low” public trust in scientists (hindsight LL1-00; hindsight LL2-00, Cologna et al. 2025).

Analysis. - Supporting cases: about 10. - Strength: moderate that seeing the public as irrational shaped communication and suppressed precaution: well documented in BSE, PCBs and radiation, less directly elsewhere. - Strength: moderate that lay and frontline observers often saw harm first. This is documented in at least six cases, but the cases were chosen because a warning proved right. - Strength: asserted that publics generally understand uncertainty better than institutions. - Cases that cut against the pattern. Lay knowledge was also wrong: asbestos workers took healthy retirees at a party as proof of safety (LL1-05, p. 60; LL1-16, p. 178), and offshore fishers resisted cuts (LL1-02, p. 21). The MMR alarm caused large, lasting harm (hindsight LL2-02). LL2-25 concedes publics can over-rate risk after vivid events, citing Germany’s post-Fukushima phase-out (p. 613), whose social cost one estimate puts at about USD 12bn a year (hindsight LL2-25). LL1 cites public rejection of irradiated food only as a “costs of failure” example (LL1-16, p. 188), without asking whether the rejection was well founded. What is documented as harmful is dismissing lay reports without testing them, not lay knowledge being right as such.

7. Pattern E — Expertise, authority and who counts as an expert#

Reports say. - Dominant authorities built on industry funding. Kehoe, the industry-funded authority on lead, treated contaminated “controls” as “normal” and “natural”, claimed “greater responsibility” than anyone for the evidence, and wrote that a critic’s paper “must be faced and demolished” (LL2-03, pp. 56–59). For about 40 years TEL research was industry-funded (p. 56). The government research agency accepted sponsor control of wording (p. 50). A Public Health Service expert agreed privately with the critics but set this aside because “human progress cannot go on under such restrictions” (p. 53). - Borrowed credibility. A 1987 beryllium memo planned a textbook under not-for-profit or academic auspices “to be fully acceptable and credible” (LL2-06, p. 136). The tobacco industry used funding bodies, non-peer-reviewed symposia and a “sponsored agent” (LL2-07, pp. 156–159). Chisso sponsored a rival committee and alternative theories (LL2-05, pp. 102–103). - Curated or self-censoring advice. BSE advisers were “carefully selected” (LL1-15, p. 162). The Southwood committee judged a brain ban “a no-goer” politically (p. 160). Mixed scientific and non-scientific advice was presented as “purely scientific”, which “flattered the scientists” (p. 165). Minamata’s closed “expert opinions” were later found to be government positions (LL2-05, pp. 109–110). A trade-association committee voted 7–3 against accepting a vinyl chloride report until it stated that the “etiologic agent of the disease is unknown” (LL2-08, p. 184). - The “ignorant expert.” Doctors pronounced on engineering and substitutability (LL1-05, pp. 58, 61). Disciplinary capture meant clinicians looked at acute effects, veterinarians at animal health and chemists at identity (LL1-16, p. 174; LL2-22, p. 543). Drug regulators lacked environmental expertise (LL1-14, p. 152; LL2-13, p. 284). At Minamata, neurology outranked epidemiology and Tokyo outranked “hick” Kumamoto (LL2-05, pp. 103, 112–114). - Same evidence, different verdicts: TCE and PCE assessments split by institutional type and by weighting of largely the same studies (LL2-04, pp. 84–85); MTBE carcinogenicity verdicts ranging from “not classifiable” to “probable” across at least seven bodies and reviewers (LL1-11, p. 113); BPA guidance values orders of magnitude apart, depending on whether non-guideline studies are admitted (LL2-10, pp. 221–223); US and UK decisions on the same scrapie evidence, and national DES decisions (LL1-16, p. 181). - Outsiders who broke paradigms and paid for it: Patterson, attacked for “hubris in stepping outside his field” (LL2-03, pp. 57–58); Stewart, “disbelieved” until replicated (LL1-03, p. 34); the Keats and Memorial University cod reanalyses (LL1-02, p. 21; LL2-17, p. 412). - Professional cultures: DFO scientists silent “once it had become policy” (LL2-17, p. 415); cod scientists asserting more precision than their internal assessments supported, then losing credibility (LL1-02, p. 22); clinical research at Kumamoto barred as “work conducted by social activist or Prefecture Government” (LL2-05, p. 104); researchers’ “desire” to expand research matching polluters’ insistence on solid evidence (Grandjean, LL2-05, p. 123), with Grandjean counting himself “part of the inertia” (LL2-26, p. 628).

Evidence and hindsight. - Divergence among expert bodies persisted and widened after 2013: PCE is still split five ways; BPA guidance values span about 250,000-fold; titanium dioxide, glyphosate and aspartame show the same (hindsight LL2-04, LL2-10, LL2-28). - CLARITY-BPA, jointly designed and publicly funded, still reproduced the split. That weakens funding as the explanation and strengthens the “different languages” diagnosis, in which guideline toxicology and academic endocrinology speak past each other (LL2-10, p. 229; hindsight LL2-10). - Ghost-authorship was confirmed by a December 2025 retraction of a herbicide safety review (hindsight LL2-25, LL2-26). - Against blanket discounting: industry-linked beryllium researchers produced valid adverse findings, and OSHA weighed industry studies on their methods rather than their funding (hindsight LL2-06; LL2-06, pp. 138–139, 148).

Analysis. - Supporting cases: divergent verdicts in about 7 cases; curated or borrowed authority in about 6. - Strength: strong that who counts as an expert, and which evidence is admissible, is an institutional choice that moves verdicts by orders of magnitude. - Strength: strong for curated advice in BSE and borrowed credibility in beryllium and tobacco, which rest on primary documents. - Strength: moderate that dominant expert communities resist outsiders’ findings as a general pattern. It is well illustrated, but the outsiders counted are those later proved right. - Guidotti’s counterpoint deserves weight. “Professional and personal incentives, which may be as powerful as financial incentives, attach to everyone involved”, because once a position is public it is “human nature to be emotionally invested in defending it” (LL2-06, p. 148). That applies to academic warners as much as to industry scientists.

8. Pattern F — Sincere belief, motivated reasoning and bad faith: sorting the cases#

The theme asks for care in telling these apart. The table below classifies cases by the evidence of mental state the reports and later records provide, not by how bad the outcome was.

Reports say, on how to tell the difference. - Minamata’s authors propose asymmetric standards of proof as a marker of manufactured doubt: “high levels of proof” demanded for results that call for action, “low levels of proof” accepted for the actor’s own alternative hypothesis (LL2-05, p. 112). - LL2-25 proposes secrecy about political actions as a signal of bad faith (p. 617). - The beryllium chapter points to rationales that shift while the conclusion stays fixed (LL2-06, pp. 137–138).

Evidence and hindsight. - Where bad faith is alleged on internal documents, later records have corroborated it in every case checked: tobacco, vinyl chloride, PCBs, lead, beryllium (hindsight LL2-07, LL2-08, LL1-06, LL2-03, LL2-06). - Since 2013 the private-knowledge, public-doubt pattern has been documented afresh for climate communications and fluorochemicals (hindsight LL2-25). - Where motive was inferred from timing or outcome, hindsight has usually weakened the attribution: BSE’s “covertly subordinated” (LL1-15, p. 164), PCE’s uncited “purposefully creating doubt” (LL2-04, p. 88), the mobile-phone “spinning machine” (LL2-21, p. 521), and LL2’s “irresponsible corporations”, given harms driven by public authorities such as Flint and Camp Lejeune (hindsight LL1-15, LL2-04, LL2-21, LL2-00).

Analysis. 1. Supporting cases. About seven cases show documented private-public divergence or concealment, and about ten show sincere or paradigm-bound error. The rest are mixed. 2. Strength: strong for documented bad faith in a subset of cases: contemporaneous internal documents plus later independent corroboration. 3. Strength: strong that sincere belief also caused serious harm. The reports’ own verdicts say so for DES, radiation, CFCs and TBT, and the official inquiry says so for BSE. 4. Strength: moderate that motivated reasoning, sincere but shaped by stakes, is the common middle mechanism (psychology cited in LL2-25, pp. 613–614; Guidotti’s panel; Phillips on BSE). 5. The markers are only partly diagnostic. - Shifting rationales also appear in sincere cases: the antimicrobial reassurances moved ground as each assumption fell, with no bad faith alleged (LL1-09, pp. 94–95). - Asymmetric scepticism appears among warners too (Pattern K). - The most reliable marker in the corpus is a documented gap between private and public positions. But it is observable only where litigation or archives opened the record (LL2-07, pp. 152, 169; hindsight LL2-08), so its absence proves nothing. 6. Sincerity does not reduce the harm. DES, radiation and swine flu show confident, well-meant actors causing harm. The more useful question for a lens is not “are they lying?” but “what is their reasoning insulated from?”: feedback from harm, independent baselines, dissent, and the costs borne by others.

9. Pattern G — Commitment, liability and escalation: how mindsets harden#

Reports say. - BSE. An early public position raised the political price of every correction: a “house of cards” and a “policy edifice” (LL1-15, pp. 161, 164). - Beryllium. Guidotti: as uncertainty fell, the stakes rose, “perhaps exponentially”, and a late admission would have been “tantamount to admitting” the company had stalled (LL2-06, p. 149). Management “believed that they understood the problem best” while working “in a social vacuum” (p. 147). His remedy: “there must be room for them to turn around” (p. 150). - Fisheries. An agency whose authority rested on its science resisted revision: “credibility… not questioned” (LL1-02, p. 21; LL2-17, p. 414). Ministers used earlier scientific error to discount a correct warning (LL2-17, p. 413). - Minamata. Case definitions narrowed defensively under attack and hardened into exclusion (LL2-05, pp. 104, 109). Admitting error meant “loss of face”, and official acceptance came four months after the plant stopped production as “no longer necessary” (p. 105). - Finland’s nuclear programme: reversing course would mean “loss of face and identity” (LL2-18, p. 445).

Evidence and hindsight. - PCBs. Among other reasons, Monsanto’s 1969 plan reasoned that stopping would mean “admitting guilt by our actions” (hindsight LL1-06, primary text). - Beryllium. Guidotti’s exit-route thesis “partly held up”: the producer backed a tenfold-lower limit after legacy liability was extinguished, partly by a compensation statute that channels claims away from vendors, but it already met the limit it backed, so interest-alignment explains the turn equally well (hindsight LL2-06). - Northern cod. Its 2023 status upgrade came from a lowered limit reference point, not more fish (hindsight LL1-02). Institutions can protect their past judgements by redefining the yardstick.

Analysis. - Supporting cases: 6–7. - Strength: moderate to strong. Strong within BSE and beryllium, where contemporaneous or insider sources exist; moderate as a generalisation. - Cases that cut against the pattern: - Organisations did reverse. The beryllium producer co-drafted the stricter limit (hindsight LL2-06). Danish farmers and feed makers stopped avoparcin voluntarily (LL1-09, p. 96). Swedish farmers asked for a ban (p. 95). Downstream firms moved ahead of the BSE regulator (LL1-15, p. 160). - LL2-27 notes that the few examples of responsible corporate behaviour in the historical cases mostly come from firms that used or sold a product rather than made it (p. 647). My inference: actors with less sunk commitment to the product change their minds more easily.

10. Pattern H — Framing contests: who names the problem, sets the question and defines “safe”#

Reports say. - Setting the question sets the answer. The UK asked its research council to “prove” obvious harm rather than weigh probabilities (LL2-03, p. 68). The French pesticide committee answered whether Gaucho was solely responsible for all bee losses everywhere, “a question that had never been asked” (LL2-16, p. 379). The Japanese health ministry wanted evidence that “all fish and all shellfish” were poisoned (LL2-05, p. 99). Hormone committees were confined to authorised use, single substances and manufacturers’ data, and never asked to characterise uncertainty (LL1-14, pp. 150, 154). A 19th-century fisheries inquiry required evidence that “decisively settled” the issue (LL1-02, p. 18). - “Safe” is a value judgement. EU product law defines as safe any product that does not present “unacceptable risks” (LL1-00, p. 5); the level of proof for action is a political and ethical choice (LL1-17, p. 193); “dangerous” and “tolerable” in climate policy are value-laden (LL2-14, pp. 322, 327). For EE2, as evidence grew, industry and some scientists asked “so what?” (LL2-13, p. 295). - Harm recast as a communications problem. The 1947 AEC feared public indignation, and beryllium was a “public relations problem” (LL2-06, pp. 131, 133). Its PR firm treated concern as “myths and misinformation” (p. 136). BSE measures were judged by what they would signal (LL1-15, pp. 159, 162–163). Critics’ vocabulary casts concern as “scares”, “hoaxes” and “hysteria” (LL2-02, pp. 18, 22, 34). - Progress, necessity and the national interest: “survive among the nations” (LL2-03, p. 53); “Never stop it!” (LL2-05, p. 99); a sardine catch increase to “assist the war-effort”, and “wise use” versus “hoarding” (LL1-02, p. 20); beryllium as a Cold War “critical industry” (LL2-06, p. 147); nuclear power “reframed” as a climate response (LL2-18, p. 433); nanotechnology as a race to be “at the forefront” (LL2-22, pp. 545–546). - Naming: “Ethyl” rather than lead (LL2-03, p. 50); “sympathy money” and “relief money (not compensation)” (LL2-05, pp. 107–110); a “hygiene standard” covering one disease (LL1-05, p. 56); “emerging” substance, called euphemistic (LL2-10, p. 216); “normal” and “natural” for lead body burdens (LL2-03, pp. 56–58); a disputed “physiological range” for hormone residues (LL1-14, p. 152); “sound science” used against unwelcome findings (LL2-27, p. 646; LL2-26, p. 632). - Precautionary vocabulary without precautionary practice: Canada’s “deliberately conservative” catch target and the EU’s “precautionary gloss” (LL1-02, pp. 20, 24); UK ministers’ retrospective claim of an ultra-precautionary approach (LL1-15, p. 157); numerical nuclear “residual risk” as reassurance, with scientific uncertainty turned into “the language of certainty” (LL2-18, pp. 447–448).

Evidence and hindsight. - Question-setting and evidence-admission rules proved decisive after 2013: - EFSA’s 2023 tolerable daily intake for BPA rests on the kind of non-guideline study the chapter said was being excluded (hindsight LL2-10); - the EU mandated effect-based oestrogen monitoring because substance-level measurement was too hard (hindsight LL2-13); - courts reframed antimicrobial precaution as requiring a scientifically grounded risk, not mere uncertainty (hindsight LL1-09). - Precautionary labels continued to coexist with unchanged practice: North Sea cod catch limits set above zero-catch advice in 2026 (hindsight LL1-02); the Great Lakes “virtual elimination” goal qualified “as appropriate” in 2012 (hindsight LL1-12).

Analysis. - Supporting cases: question-setting in about 6 cases; progress and national-interest framing in about 7; precautionary relabelling in about 4. - Strength: strong that the question put to experts, and the evidence they may admit, decides the verdict. It is documented in several cases and reinforced since. - Strength: moderate that rhetorical frames such as “gift of God” or “Never stop it!” shaped decisions. The quotations are verified, but their causal weight is inferred. - Strength: moderate that precautionary vocabulary can relabel unchanged practice. - Warners framed strategically too. Odén announced acidification in a newspaper with “sweeping statements”, and Sweden’s “creative presentation” at Stockholm worked “as intended” (LL1-10, p. 102). The chapters’ own titles and metaphors (“time bomb”, “saga of secrecy”, “last hunters”; LL1-08, LL2-08, LL2-17) are persuasive devices. Framing is a contest with two sides, and the reports describe mainly one.

11. Pattern I — Institutions that shape mindsets: promoters as protectors#

Reports say. - The UK agriculture ministry was “responsible first to the industry and only second to consumers” (LL1-16, p. 179; LL1-15, pp. 157, 164). - In the US weapons programme worker safety was the “last priority” (LL2-06, p. 132), and the AEC framed harm partly as a threat to its mission (p. 133). - Japan’s trade ministry acted as the polluter’s advocate (LL2-05, pp. 99, 102). USDA “tended to align itself” with manufacturers and farmers (LL1-12, p. 128). - DFO’s authority rested on “a certain control of science” and survived the collapse (LL2-17, p. 419). - Japan’s nuclear regulators had industry ties, and the parliamentary inquiry named “regulatory capture” (LL2-18, pp. 441–443). - The US nanotechnology initiative both promoted the technology and oversaw its risks (LL2-22, pp. 546–547). - Radiation standard-setters faced the problem of setting limits that did not “restrict the expansion” of nuclear industries (LL1-03, p. 34). This is asserted, not documented.

Evidence and hindsight. - Phillips on BSE. The Department of Health, with no sponsorship conflict, was as keen to reassure as the agriculture ministry (hindsight LL1-15). Separating promotion from protection is probably necessary but not sufficient. - Ireland versus Austria. The claim that institutional design made one less precautionary is confounded by exposure: Austria had 5 classical cases, Ireland 1,656 (hindsight LL1-15). - US nanotechnology. The initiative’s risk-research share fell from about 10% (2016) to about 1% in the FY2026 request (hindsight LL2-22).

Analysis. - Supporting cases: about 7. - Strength: moderate. The link between a promotional mandate and reassurance-seeking is documented in several cases, but the one comparative test offered fails. Mindsets of reassurance also appear in bodies without a promotional mandate.

12. Pattern J — Warners, messengers and the treatment of dissent#

Reports say. - Warnings came from the margins. They came from women factory inspectors (LL1-05, p. 53), a dentist, a journalist and a local pathologist (LL1-03, pp. 31–34), wildlife observers (LL1-12, p. 126), sterile workers (LL2-09, p. 204), beekeepers (LL2-16, p. 372) and a Minamata mother (LL2-05, p. 105). - They also came from insiders: a DBCP consultant (LL2-09, p. 205); Dow’s toxicologist and Union Carbide’s consultant on vinyl chloride (LL2-08, p. 182); Chisso’s company doctor (LL2-05, pp. 100–101); DES proponents who flagged latent effects (LL1-08, p. 88); BSE officials who urged compensated slaughter (LL1-15, p. 159); the beryllium limit’s own author (LL2-06, p. 136). - “Shooting the messenger” “rarely, if ever, promotes societal welfare” (LL1-16, p. 179). Instances include: - asbestos investigators called a “disturbing sore thumb” (LL1-05, p. 55); - Patterson and Needleman attacked (LL2-03, pp. 58, 61, “character assassination” undescribed); - legal threats and cancelled programmes in France (LL2-16, p. 378); - Bayer suing three beekeeper leaders, and losing (p. 380); - harassment of warners generally (LL2-26, p. 637; LL2-24, pp. 582–585). - Reformers also gatekeep. Legge, a former chief medical inspector then advising the TUC, wrote that the TUC could be “quit of” an investigator by paying him a small fee (LL1-05, p. 55; notes LL1-05). - People whose own work is challenged are conflicted judges of the challenge, which the chapter calls “intellectual bias” (LL2-24, p. 585).

Evidence and hindsight. - Recurrence. Flint repeated the dismissal of warners (hindsight LL2-03). - Protection adopted, but narrowly. The EU Whistleblower Directive covers breaches of law, not warnings about lawful products, and France’s health and environment alert commission was abolished in 2026 (hindsight LL2-24, LL2-28). - Warners wrong. Several of the reports’ own warners were not vindicated. The swine-flu warning fitted three theories of flu cycles, and “perhaps too much faith was placed” on science’s ability to foresee the outbreak (LL2-02, p. 31). The GM health “early indications” rested on a study since retracted (hindsight LL2-19). The mobile-phone case-control findings have not been borne out (hindsight LL2-21). LL2-24’s model early-warner prize honoured a retracted study (hindsight LL2-24).

Analysis. - Supporting cases: 10 or more for dismissal of warners. - Strength: moderate. The pattern is widely documented, but the corpus selects for warners who were right, and unvindicated warners are largely absent. The better-supported lesson is procedural: protection should turn on good faith and reasonable belief, not on being proved right, which means accepting some false alarms (LL2-24, p. 584). - The insider warner is as common as the outsider. In about six cases the first protective advice came from inside the firm or agency and was overridden or diluted. My inference: the failure was less “nobody inside knew” than “knowledge inside did not govern the decision”.

13. Pattern K — The reports’ own mindsets#

A fair synthesis must apply the theme to the reports too.

Reports say. - The editors acknowledge case authors were “not without strong views” (LL1-00, p. 12); LL2-25 warns that blaming business “with hindsight” “may not always be constructive” (p. 616); Grandjean includes himself in the inertia he criticises (LL2-26, p. 628).

Evidence and hindsight. - Asymmetric scrutiny by protagonists. - The mobile-phone chapter uses long latency to discount null studies but accepts short-latency positive findings (LL2-21, p. 514); it scrutinises the rival study’s industry funding while leaving its own telecom-operator funding unremarked (p. 517); and it ends a quotation just before a clause favouring the rival study (notes LL2-21). Its own lesson that participants scrutinise contrary evidence harder is “illustrated by the chapter itself” (hindsight LL2-21). - The GM chapter holds GM to strict evidential standards while accepting uncontrolled agroecology syntheses (notes LL2-19). - The neonicotinoid chapter faults national-scale framing of bee losses but uses national yields against farmers’ claims (notes LL2-16). - Loaded framing of critics. LL1’s Preface says critics “fear or imagine” (LL1-00, p. 4). LL2’s Preface opens “There is something profoundly wrong with the way we are living today” (LL2-00, p. 6). - Where conviction ran furthest, hindsight was harshest. The chapters most shaped by authorial conviction have fared worst on specifics: mobile phones, the GM health sentence in Ch 28, the Chernobyl mortality figures, the PCB paediatric attributions and Annex 3’s single-laboratory claims (hindsight LL2-21, LL2-28, LL2-18, LL1-06, LL2-A3). By contrast, the reports’ analyses of other actors’ mindsets (reassurance, paradigm defence, private-public divergence) have mostly been reinforced.

Analysis. Strength: moderate that the mechanisms the reports diagnose in proponents operate among warners and advocates too: motivated reading of ambiguity, emotional investment in public positions, asymmetric scepticism. This is well documented within the corpus and consistent with Guidotti’s point (LL2-06, p. 148). The practical implication is symmetry: every diagnostic question in section 15 should also be asked of the critics.

14. Counter-evidence, complications and critiques#

1. Selection on the outcome. - Both volumes chose cases because harm occurred (LL1-00, pp. 11–13; LL2-00, pp. 9–10), and the business chapter’s “virtually all” is close to built in (LL2-25, p. 607; hindsight LL2-25). - The mindsets diagnosed here (confidence, benefit enthusiasm, trust in models) are probably also common among proponents of technologies that turned out fine. The corpus has no base rate (critiques §4, Marchant 2003; Mazur’s design, unused). - What follows: the patterns can say how confident actors fail, not how often confidence is misplaced.

2. Hindsight and outcome bias in calling sincere belief “hubris”. LL1’s own test is “the spirit of the times” (LL1-00, p. 11), and it is not always applied. The historian the asbestos chapter cites disputes its reading of pre-1930 warnings (Bartrip 1998; notes LL1-05). Guidotti defends the 1948 beryllium limit as reasonable consensus (LL2-06, pp. 147–148). Benzene’s leukaemia risk at permitted levels was not shown before 1977 (LL1-04, p. 40). The LL1 editors call DES’s next-generation effects “a complete surprise” (LL1-16, p. 170) while the case chapter says warnings were ignored. Minamata’s pre-1956 foreseeability rests on occupational, not food-chain, literature (notes LL2-05).

3. Paradigm defence was sometimes right, and precautionary conviction sometimes did harm. - Paradigm-based scepticism was largely borne out for mobile phones and food irradiation (hindsight LL2-21, LL2-02). - Precautionary conviction caused harm: - swine flu: 107 Guillain-Barré cases and 6 deaths (LL2-02, p. 28); - Fukushima: measurable harm came from evacuation (2,351 disaster-related deaths) and thyroid overdiagnosis rather than radiation (hindsight LL2-18); - the MMR alarm (hindsight LL2-02). - A lens built only from the reports’ false negatives would misread these confident warners as vindicated heretics.

4. Over-attribution of motive. The documented bad-faith cases are strong, but several broader attributions did not survive hindsight: BSE’s “covertly subordinated” (Phillips), PCE’s “purposefully creating doubt”, DES’s “economic interests predominated”, a tobacco-style “spinning machine” on mobile phones, and LL2’s “for the most part irresponsible corporations” (hindsight LL1-15, LL2-04, LL1-08, LL2-21, LL2-00). Guidotti’s critique of discounting interested science wholesale is partly vindicated by beryllium’s own history (hindsight LL2-06).

5. Firms and professions are not monolithic. Downstream users moved first in several cases (LL1-15, p. 160; LL2-27, p. 647); a refiner declined MTBE on risk grounds; the beryllium producer co-drafted a tenfold-stricter limit (hindsight LL2-25, LL2-06); industry inhalation studies confirmed nanotube effects (LL2-22, p. 535); insurers declined asbestos cover in 1918 (LL1-05, p. 54). My inference: position in the value chain and liability exposure predict behaviour better than “industry” does.

6. States as well as firms. Public authorities held reassuring mindsets or promoted hazards in many cases: MAFF and the Department of Health (LL1-15), DOE and the AEC (LL2-06), Japan’s ministries and prefecture (LL2-05), DFO (LL2-17), USDA (LL1-12), Japan’s nuclear regulators (LL2-18); hindsight adds Flint and Camp Lejeune (hindsight LL2-00). LL2’s corporate-villain frame understates this.

7. Cognition versus interest cannot always be separated. LL1’s Preface puts political will ahead of information (LL1-00, p. 4), Grandjean concedes the same (LL2-26, fn 2, p. 624), and LL2-25 offers two causal stories with no test between them (notes LL2-25). For a lens this matters less than it seems: most safeguards (independent baselines, pre-set triggers, open dissent, separating promotion from protection) work whether the problem is self-deception or strategy.

8. External critiques. Critics mostly attack the precautionary principle rather than the case evidence; for this theme the strongest objections are case selection, hindsight and the missing denominator, and the weakest are those against documented suppression, producer control of research and manufactured doubt (critiques §§1, 4, 9.1). The “innovation principle” that firms launched through the European Risk Forum in 2013 is itself a framing move, casting precaution as the risk (critiques §7).

15. Technology-neutral diagnostic questions#

Each question is tied to the evidence above and should be asked of proponents and critics (Pattern K).

1. Which properties are being celebrated (persistence, stability, potency, inertness, invisibility, scale), and could the same properties cause lasting harm? Pattern A; LL2-11, p. 241; LL1-07, p. 83; LL1-03, p. 31.
2. What model of harm underlies the confidence (endpoint, dose metric, reference population, timescale), and what would we expect to see if it were wrong? Pattern B; LL1-03, p. 33; LL2-06, pp. 133–134; LL1-13, p. 141.
3. When someone says “safe” or “no evidence of harm”, how hard has anyone looked, with what power, follow-up and detection limits, and who decided where to look? Pattern C; LL1-15, pp. 163–164; LL2-16, p. 373; LL1-05, pp. 55, 60; LL2-26, pp. 631–635.
4. Have categorical public reassurances been given that would make later, graded precaution look like an admission of error? Is residual risk stated openly? LL1-15, pp. 161–162; hindsight LL1-15.
5. How are the public, frontline observers and critics described (“hysteria”, “misinformation”, “amateur”, “anecdotal”), and is concern being treated as a communications problem? Pattern D; LL1-15, p. 159; LL1-06, p. 64; LL2-05, p. 105; LL2-06, pp. 133, 136; LL2-17, p. 414.
6. Who counts as an expert: who sits on assessment bodies, which evidence streams are admissible, and do differently constituted bodies reach different verdicts on the same evidence without explaining why? Pattern E; LL2-04, pp. 84–85; LL2-10, pp. 220–223; LL1-15, pp. 162, 165.
7. Is there a documented gap between what developers say privately (internal research, communications with investors or regulators) and publicly, or secrecy about efforts to shape rules, research agendas or evidence? Pattern F; LL2-07, pp. 153–156; LL2-08, pp. 183–186; LL2-06, pp. 136–137; LL2-25, pp. 615, 617.
8. Is scepticism symmetrical, or are warners asked for mechanism-level or universal proof while safety claims rest on assertion or proxies? LL2-05, p. 112; LL2-09, p. 211; LL2-03, p. 50; LL2-26, pp. 631–633.
9. What would admitting a problem cost this organisation (liability, reputation, identity, past statements), how does that cost grow as evidence accumulates, and is there a route for it to change course? Pattern G; LL2-06, pp. 137, 149–150; hindsight LL1-06; LL2-05, p. 105.
10. Who framed the question put to advisers, and could the framing (“sole cause”, “all cases”, “prove obvious harm”, “authorised use only”) guarantee an inconclusive answer? Pattern H; LL2-16, p. 379; LL2-05, p. 99; LL2-03, p. 68; LL1-14, pp. 150, 154.
11. Is precautionary or safety vocabulary (“deliberately conservative”, “controlled use”, “closed systems”, numerical “residual risk”) describing practice that has not changed? LL1-02, pp. 20, 24; LL1-05, p. 57; LL1-16, pp. 174–175; LL2-18, pp. 447–448.
12. Are the promoters also the regulators, funders of risk research or owners of the data, and do bodies without that conflict show the same urge to reassure? Pattern I; LL1-15 and hindsight LL1-15; LL2-22, pp. 546–547; LL2-03, p. 56; LL2-19, p. 468.
13. Where are early warnings coming from (workers, users, affected communities, overruled insiders), what happens to them, and are warners protected for good faith rather than for being proved right? Pattern J; LL2-09, pp. 204–205; LL2-05, pp. 100–101, 105; LL1-15, p. 162; LL2-24, pp. 584–585.
14. If everyone involved is sincere, what would still produce harm: weak feedback from harm to decision-maker, long lags, costs borne by others, commitment to earlier positions? Pattern F; LL2-25, pp. 608, 613–615; LL2-28, p. 678; LL1-08, p. 88; LL2-02, p. 28.
15. Are critics and warners held to the same scrutiny of priors, stakes, investment in public positions and selective use of evidence? Pattern K; LL2-06, p. 148; notes LL2-21, LL2-19; hindsight LL2-21, LL2-02.