Late Lessons, Jensen Huang and AI
Markdown version

LL2-25 — Ch25 Why did business not react with precaution to early warnings?#

Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013), Part D “Costs, justice and innovation”. Report pages 607–620 (PDF pages 609–622). Chapter text runs pp. 607–617; references pp. 617–619; p. 620 is blank.

Reading and verification note. I read the whole text extract in order, through the final page marker (PDF 622 / p. 620, which is blank). The chapter is plain two-column text with no figures, tables, boxes or panels. An independent pass over the PDF text layer found no embedded images on any page 607–619. The PDF could not be rendered as images here (the rendering tool was unavailable), so the independent text extraction stands in for the visual check. The two extractions agree. For context I also consulted three other places in the same PDF: the author biographies in Annex 1 (pp. 694, 697), the table of contents (pp. 3–4), and the two places where other chapters refer to this one (Ch 8, p. 193; Ch 27, p. 647). A few factual points were checked on the web, using primary sources where they could be reached (HBR author note; Supran and Oreskes 2017; the Lexchin and O’Donovan 2010 abstract via Europe PMC) and a secondary summary otherwise. Each check is flagged where it appears.

Authors and standpoint#

Authors. Marc Le Menestrel and Julian Rode (p. 607).

Disciplinary position. Business ethics, decision theory, and behavioural/experimental economics. Neither author is a case historian, toxicologist or regulator. The chapter is not a case study. It is a cross‑cutting analytical essay that reuses case material from Late lessons vol. 1 (EEA, 2001) and from other chapters of vol. 2, and interprets that material through management, economics and behavioural‑ethics literature (p. 608). The text calls itself “this paper” (p. 608) and “our article” (p. 616), which suggests it was adapted from a journal‑style manuscript.

Self‑citation. Several supporting references are the authors’ own work: Le Menestrel 2002 (economic rationality and ethics); Aktar and Le Menestrel 2010, an unpublished “Miméo” used for the claim about consumers and voluntary disclosure; Rode et al. 2008; Rode and Le Menestrel 2011; Rauschmayer et al. 2009; and van den Hove, Le Menestrel and de Bettignies 2002, which is the source for the Exxon claim (pp. 608, 609, 610, 614; refs pp. 617–619).

Evident stance. The chapter is explanatory rather than prosecutorial. It explicitly warns that “Blaming business, in particular with hindsight, tends to be a rather typical reaction that may not always be constructive” (p. 616). It concedes that firms have been “the main drivers of innovative activity” with real “benefits they have generated for society” (p. 608), and that many business actors recognise wider responsibilities (WBCSD, 2010; p. 608). It is nonetheless clearly on the side of stronger precaution. It takes as given that the reviewed cases show business neglecting credible warnings. It gives the precautionary principle as guidance to “err on the side of caution” even without “scientifically established evidence” (p. 611). And it ends with normative recommendations to internalise costs, disclose dilemmas and make corporate political activity transparent (pp. 616–617).

Panels and commentaries. None. The chapter has no panels, industry responses or dissenting commentaries.

How the rest of the report frames this chapter (useful context, not part of the chapter): - Ch 8 (vinyl chloride, p. 193) signposts it as the place where “slow or hostile corporate responses to early warnings” are “further examined”. Its footnote 9 adds historical “progressive business” examples (the Quakers; Josiah Wedgwood, spelled “Wedgewood”, who on the passing of the 1833 Factories Act wrote to the UK government “asking for similar laws to be applied to his pottery industry”). - Ch 27 (David Gee, p. 647) summarises the chapter but pushes its emphasis towards externalised costs. Gee’s summary also puts inside quotation marks a sentence that does not appear in this chapter in that form: “in virtually all reviewed cases from both volumes of Late lessons…”. The chapter’s own sentence is “In virtually all reviewed cases it was perceived to be profitable…” (p. 607). Gee then supplies positive corporate examples that this chapter does not use: user companies dropping asbestos in the 1970s, Johnson & Johnson dropping CFC aerosols in 1977, retailers dropping BPA, nanotechnology codes of conduct, and others (p. 647). He observes that responsible behaviour came mostly from “companies selling hazardous products rather than by their manufacturers”. This is a relevant nuance that Ch 25 itself does not draw out.

Section-by-section notes#

Chapter summary (p. 607)#

25.1 Introduction (p. 608)#

25.2.1 Economic rationality (pp. 608–609)#

Reputation (p. 609). - “Conscious consumer” pressure can create incentives (Banerjee et al., 2003; Rode et al., 2008). The logic is still economic: a rational firm “will not sell a product that, for reasons of public concern or lack of consumer trust, may not be profitable or give rise to consumer boycotts”. The end of antimicrobials and growth hormones in European meat production was “driven to some extent by growing public concern” (EEA 2001 Chs 9, 14). - Evenhanded caveat. Such concern works “no matter whether it is driven by the available scientific evidence or… unrelated to evidence, or even overrating the dangers” (Sunstein, 2003). - Usually a weak lever. The public “lacks knowledge about early warnings, underrates the risks… or does not feel affected”, so “reputation does not provide a sufficient economic incentive”. - Firms can manage the channel instead of the risk. They can “influence public opinion in their favour when this appears cheaper than reducing or terminating the potentially harmful practice”, and can withhold private information. Voluntary disclosure “creates the risk that consumers refrain from buying the product”, and consumers “seem to reward transparency and honest disclosure of negative information only under very limited conditions” (Aktar and Le Menestrel, 2010, an unpublished mimeo by one of the authors, whose listed title is “Voluntary Disclosure of Negative Information May Enhance Consumers’ Trust”).

Economic interest in preventing harm (p. 609). - Industries harmed by a hazard can drive precaution: - Newfoundland inshore cod fishers protested and commissioned a report that “gained media attention and eventually lead to an official reappraisal” (EEA 2001 Ch 2). - The Arcachon oyster industry “strongly supported” TBT restrictions (EEA 2001 Ch 13). - Beekeepers “suspected early on” Gaucho’s toxicity and “fought hard for the recognition of the evidence” (vol. 2 Ch 16). - These “seem to be a minority”.

25.2.2 Uncertainty in science and the research environment (pp. 609–611)#

Scientific uncertainty (pp. 609–610). - Uncertainties “today are increasingly characterised by indeterminacy and even ignorance” (Stirling, 2003). Examples: marine ecosystems, mobile phones, GMOs and nanotechnologies (p. 609). - Cross‑reference error. The chapter cites Chs 20, 18 and 21. In the final table of contents these are Ch 21 (mobile phones), Ch 19 (GM crops) and Ch 22 (nanotechnology); final Ch 18 is Chernobyl/Fukushima and Ch 20 is invasive alien species. This suggests drafting against an earlier numbering. The chapter’s other vol. 2 cross‑references (Ch 3 lead, Ch 6 beryllium, Ch 7 tobacco, Ch 16 insecticides and honeybees) are correct. Ch 22’s co‑authors include Andrew Maynard, noted for disclosure. - People struggle to “cognitively deal with and to act upon such strong uncertainty” (Weber, 2006). Cost‑benefit tools need outcome sets and probabilities “which are not always available” (Ashford, 2005). Developing concepts and operational tools for this kind of uncertainty “remains a challenge for social science, and in particular for business research” (p. 609). - Uncertainty is intrinsic to science. “There is no scientific justification of a ‘sufficient level of confidence’ or for the appropriateness of a confidence interval of 99 % or 95 %”. Setting an “acceptable level of risk” is “an ethical or political issue rather than a scientific one” (Crawford‑Brown et al., 2004; Ashford, 2005). The “‘weight of evidence’… is almost always open to subjective judgment and interpretation” (p. 610).

Interpreting and “manufacturing” uncertainty (p. 610). - The structural gap. Ambiguous science meets public “idealistic expectations of science” and a “preference for complete certainty for justifying actions, in particular when the actions involve concrete costs” (Ravetz; van den Hove, 2007; Dana, 2003). “This allows industry lobbyists to oppose or prolong precautionary measures by ‘manufacturing uncertainty’”, using “rhetorical tricks to emphasise the remaining uncertainty and the need for further research” (Rampton and Stauber, 2001; Oreskes and Conway, 2010; Sismondo, 2008). - Examples, as the chapter gives them: - Bayer / Gaucho. Bayer “seems to have repeatedly used selective knowledge and ‘semantic slips’ to blur the evidence” (vol. 2 Ch 16). - Benzene. Manufacturers hired consultants “to introduce irresolvable arguments about dose‑response analysis, which delayed governmental regulation” (EEA 2001 Ch 4). - Monsanto / PCBs, 1960s. A public defence arguing the evidence “was not clear” and that it “would take extensive research, on a worldwide basis” to settle (EEA 2001 Ch 6; Francis, a Sierra magazine article). - Shell / Great Lakes, 1967. A report denying causation, saying chemical fish kills “had not been verified by recent studies” (EEA 2001 Ch 12). - Brush Wellman (spelled “Brush Wellmann”) / beryllium. - 1980s: PR “to reassure customers of the safety of beryllium”, calling reports of disease below the limit “scientifically unsound”. - Late 1990s: “when it was beyond doubt that the established beryllium exposure limit was not effective in protecting workers”, it “initiated more research and convened a conference that propagated the need for further research before any new limit could be set” (vol. 2 Ch 6). - Tobacco. - Minnesota‑trial documents “revealed the extent to which the effects of nicotine were known and intentionally blurred” (Hurt and Robertson, 1998). - New research institutions were funded to “‘keep the debate alive’” on second‑hand smoke (Hong and Bero, 2002; vol. 2 Ch 7). - Exxon / climate. Exxon was “publicly contesting the science” while “presenting itself as ‘a science and technology‑based company’”. Its strategy worked “through efforts in publicly denying the existence of the problem that they had privately identified” (van den Hove et al., 2002, co‑authored by Le Menestrel). - Conclusion. Companies “can contribute effectively to a ‘paralysis by analysis’” (p. 610).

Corporate influence on scientific research (pp. 610–611). - No misconduct needed. Firms “do not even have to manipulate results or engage in other forms of misconduct, which may happen in some cases”, because they can induce “(pro‑industry) design and publication biases” (pp. 610–611). - Vote‑counting. Counts of studies for and against an effect “can easily be altered through the selection of which scientific studies to actually carry out” (Lexchin and O’Donovan, 2010). Because firms “typically have significant financial means”, they can “misrepresent the weight of results in their favour by sponsoring those scientific studies and methods which can be expected to produce favourable results” (p. 611). (The cited Lexchin and O’Donovan paper is a review of conflict‑of‑interest policies in drug regulators; it is an odd source for the study‑selection point.) - Symposia. Industry symposium proceedings, “typically not peer‑reviewed, but still cited as published results” (tobacco; Grüning et al., 2006). - Lead. - Health studies were “conducted and funded exclusively by the Ethyl Corporation and General Motors for over 40 years”. - GM “controlled the publication of results and imposed tight reporting constraints on the regulating US Bureau of Mines”. - Hedged: “it is even said that critical independent scientists had their funding withdrawn and their jobs and lives threatened” (vol. 2 Ch 3; p. 611). - Bayer relied “almost exclusively on their own research” in the 1990s. - Brush Wellman “financed new publications within the beryllium health and safety literature under the names of well‑known academics” (vol. 2 Ch 6). The chapter does not say who wrote them; “ghost‑authorship” would be an inference beyond this wording. - Spillover. These cases “weaken the credibility of privately sponsored research and create a difficulty for companies that have a genuine and honest interest in objective and unbiased research” (p. 611).

25.2.3 Gaps and loopholes in the regulatory framework (pp. 611–613)#

Incomplete information (p. 611). - Part of the problem is inherent ignorance (“currently for nanotechnology or for GMOs”). Part is that regulation “often rel[ies] on information that is generated within the companies whose products or activities have to be assessed”. Even with disclosure rules, “it is frequently up to the companies to reveal such ‘private information’”, and non‑disclosure “can hence delay or distort regulatory action”. - The illustration that “companies may be the ones to first recognise early warning signals” is Bill Joy’s 2000 Wired essay on “genetically modified organisms, robotics and nanotechnology”. Ill‑fitting: this shows an individual’s public, speculative warning, not a firm holding private information about its own products.

Regulation rarely induces full internalisation (pp. 611–612). - Purposes of liability. Beyond ex post justice, liability should give the ex ante incentive to internalise harm, ensure means to compensate, and motivate risk assessment and precaution (Boyd, 1997). Under polluter‑pays this means strict liability. The alternative is “‘assurance bonding’”, requiring firms “to deposit a premium that would cover the costs of potential damage before undertaking the dangerous activity” (Kysar, cited as 2009 but listed as 2010) (pp. 611–612). - Absent liability shifts costs to society. “limited or even complete absence of liability have undermined the polluter pays principle… external costs… are typically borne by society at large” (p. 612): - Fisheries. Restoration paid “by governments of the respective adjacent states”. - MTBE. Most external costs (water treatment, alternative supply, health) fell on society. - UK asbestos. “it is argued that” prices were low because costs were not internalised and “remained with families, health service, insurance carriers and building owners”. Laws from “as early as in the 1930s” were “simply not appropriately implemented”. - Standards and burden of proof. Liability depends on the evidentiary standard (“‘clear and convincing evidence’ vs. ‘more probable than not’ vs. ‘preponderance of evidence’”), the causation test (“‘substantial cause or factor’ vs. ‘contributing factor’”) and who bears “the legal burden of persuasion” (Ashford, 2005). Where the victim must show substantial causation, a firm “can expect fewer costs” (p. 612). Observation (mine): in US civil law “more probable than not” and “preponderance of evidence” are usually the same standard, so the chapter’s three‑point range is really two. - Insolvency. Insolvency undermines internalisation, “especially when harm would only occur in the far future” (Boyd, 1997). “Manville Corporation filed for bankruptcy in 1982 as a means of dealing with asbestos pollution claims… when it was far too late to act with precaution” (p. 612).

Conflicting mandates (p. 612). - USDA, before the EPA, on Great Lakes contamination. As “a supporter of the economic interests of the agro‑industry”, it aligned with pesticide makers and farmers, “demanding proof of causal relationship before ‘massive’ approbations [sic, likely ‘appropriations’] and expenditures… on remedial works” (EEA 2001 Ch 12). The quotation marks are unbalanced in the original. - “US Department of Fisheries and Oceans (DFO)”. Factual slip: this is Canada’s DFO. It “is said to have followed the interests of the fishery sector” and “is accused of having presented biased results, referring to remaining uncertainties, and of arguing against ‘pseudo‑science’ and bad faith of early warnings” (EEA 2001 Ch 2). The hedges are the authors’. - UK MAFF on BSE: “simultaneously to promote the economic interests of farmers and the food industry whilst also protecting public health”. - French Ministry of Agriculture on Gaucho: “contradictory demands of intensive agriculture and beekeepers”. - US DOE on beryllium: “cheap production of nuclear weapons” alongside worker protection.

Corporate influence on regulation (pp. 612–613). - Conflicted experts. Committees “included experts with a conflict of interest, who could shape policy recommendations by interpreting scientific evidence in the interests of the industry” (p. 612). - Tobacco and WHO. The WHO Committee of Experts (2000) found subversion that was “elaborate, well financed, sophisticated, and usually invisible” (cited as “Zelltner”; pp. 612–613). - Benzene. The ACGIH “repeatedly recommended benzene limits higher than those in line with scientific evidence”, with corporate scientists on the TLV Committee (Castleman and Ziem, 1988). - Lead, 1920s. Paid Ethyl consultants simultaneously advised the Bureau of Mines, “providing assurances of ‘complete safety’”. - Pharmaceuticals. “In the pharmaceutical domain, conflicts of interest seem pervasive.” Reviewing three European drug agencies, Lexchin and O’Donovan (2010) are said to “find evidence of widespread potential conflict of interests among scientific experts” (p. 613). Checked (Europe PMC abstract, PMID 19782458): the paper reviews COI policies and procedures at the Irish Medicines Board, the UK MHRA and the EMA. It finds that, although official statements suggest COIs are prohibited, in practice the agencies assume COIs “cannot and need not be eliminated” and can be “managed”, and it advocates prohibition on precautionary grounds. The abstract gives no prevalence figures, so “widespread” may overstate what the paper shows (full text not checked). - Indirect influence via public perception, since “public trust, consumer perceptions or NGOs can have a considerable influence” on regulators (Aerni, 2004; Carter, 2002) (p. 613).

25.2.4 Psychological factors (pp. 613–615)#

Bounded rationality (pp. 613–614). - Premise. “ample evidence from the behavioural sciences” that capacity to recognise and evaluate early warnings “is limited” (p. 613). - How hazards are perceived. They appear as low‑probability events, or are “formulated in terms of a low probability” when evidence is scarce. Vivid events can be overestimated (Kahneman, 2011). But awareness is carried better by “direct experience and the associated emotions, rather than abstract statistical descriptions” (Weber, 2006), and concrete losses outweigh abstract ones (Dana, 2003). - Rare events neglected. Because warned‑of hazards are usually unexperienced, “people tend to neglect the likelihood of rare events” (Hertwig and Erev, 2009). Kahneman is quoted: our mind “is not designed to get it quite right”. This applies “to business decision‑makers and the general public alike” (p. 613). - Reverse effect. “in exceptional cases”, when a low‑probability event does occur, people may “at least temporarily — overrate the probability” (Sunstein, 2003), and “increased public concern may lead to faster regulatory measures”. This “may have been the case for the German decision to phase out nuclear energy after the Fukushima accident in 2011”. The chapter does not call the phase‑out an over‑reaction, and it does not cross‑refer to vol. 2 Ch 18 on Chernobyl and Fukushima. “In most situations, however”, risk perception “seems to impede precautionary corporate action as well as public pressure” (p. 613). - “Pensioner’s party fallacy” (from EEA 2001). Surviving exposed colleagues are taken “as evidence against the existence of harm”, which is survivorship bias; the chapter suggests smoking as an example (p. 613). - Discounting (p. 614). Present costs versus distant, uncertain harms: discounting, including hyperbolic discounting, “works against precautionary measures”. The anti‑paternalist view (free “time preferences”) is acknowledged, then qualified: discounting on “the time preferences only of the present generation may be questionable from an ethical point of view” (O’Neill et al., 2008). The Stern/Weitzman/Spash/TEEB discounting debates are noted. - Other effects (p. 614): “finite‑pool‑of‑worries” (Weber, 2006); “single action bias”; cultural differences (Bian and Keller, 1999). - Candid gap. “even though we have not found psychological studies on the phenomenon”, cases show companies exploiting the tendency to read “‘no evidence of harm’ as ‘evidence of no harm’” (vol. 2 Chs 3, 6; Zeltner) (p. 614).

Bounded ethicality (pp. 614–615). - Concepts. “ethical blindness” (Palazzo et al., 2012), “ethical biases” (Banaji et al., 2003), “bounded ethicality” (Gino et al., 2008). - Self‑serving bias (Babcock and Loewenstein, 1997) leads to “self‑deception that helps them reinterpret or disguise” ethical violations. Such phenomena “can be largely unconscious” and are tied to reducing “cognitive dissonance” between “making profit and acting ethical” (Festinger, 1957) (p. 614). The bias “seems of high relevance when business decision‑makers face uncertain early warnings signals but precautionary measures are not in the economic interest of the company” (p. 614). - Mechanisms. “language euphemisms” and “‘slippery‑slope’ decisions” (Tenbrunsel and Messick, 2004); hiding “from relevant knowledge” (Ehrich and Irwin, 2005); neglecting reasoning “that may reveal them as responsible” (Rode and Le Menestrel, 2011). - Uncertainty as excuse. “Strong uncertainty… may also serve as a welcome ‘excuse’ and justification about why the profitable action may not be so unethical after all” (p. 614). - The public’s blind spot. People overlook harmful conduct by others “when that behaviour is not clear, immediate and direct, and when it has not yet resulted in a bad outcome” (Gino et al., 2008). This helps explain weak action by “consumers, voters or engaged citizens” (p. 614). - Manville, as told by Sells (1994, HBR). - The asbestos disaster “was a management blunder, and the blunder was denial”, with managers “unwilling or unable to believe in the long‑term consequences of these known hazards” (p. 614). - Attribution note. The chapter says Sells “cites one of the presidents of Manville”. HBR’s author note (web‑checked) says Sells ended his career at Manville “as a senior vice president and president of the Fiber Glass Group”, and the article is a first‑person account. The words are therefore probably Sells’s own. The chapter’s phrase may be a clumsy way of referring to Sells himself, but it reads as if he is quoting someone else. - Organisational culture (p. 615). Hedged by the authors (“It is as if”): economic interest, uncertainty and psychology “concur to trap business executives in an organisational culture where the danger is minimised and alternative business solutions unattainable”. Sells: where culture “encourages denial, problems get buried”, and Manville’s “good people” “fostered a culture of self‑deception and denial”. - Reversal: Manville fiberglass (p. 615). - Asbestos and bankruptcy “acted for Manville Corporation as a lesson for the company to stop its culture of denial” and changed its approach to product stewardship. In 1986, “shortly after learning that its fiberglass products could be related to an increase in cancer rate”, its leadership “took precautionary action with regards to its operations and voluntary re‑labeled these products as possibly carcinogenic despite the reluctance of their lawyers”. - It “benefited… thanks to a successful indemnification and marketing strategy, proving that what may be perceived as a conflict of interest could well lead to a successful alignment of business and social values”. - “It took then nearly five years to realise that the excess detected in respiratory cancer… were not sufficiently significant to justify such a warning label” (Sells, 1994; Paine and Gant, HBS cases 394‑114 to 394‑118). - Checks. The featured success story of precaution involves a warning later judged unfounded, and the chapter draws no lesson from that. The “nearly five years” (to about 1991) describes the company’s own reassessment, not a formal reclassification, and formal bodies moved much later. Checked against a secondary summary (Wikipedia, “Glass wool”): IARC reclassified insulation glass wool to Group 3 in October 2001, and the US NTP delisted biosoluble glass wool in June 2011, ending required cancer labels. From memory: IARC had classified glass wool as 2B in 1987–88. The “five years” probably reflects the company’s own reading of the epidemiology.

25.3 Lessons and reflections about business and early warnings (pp. 615–617)#

The three lessons (p. 615–616; full content under “The authors’ own lessons and conclusions” below): - Lesson 1 (p. 615): “early warning signals often entail conflict of values for business actors, who expect to be in their economic interest not to respond with precautionary ‘business actions’”. Because health and environment “are regarded as issues pertaining to society at large or at least to multiple societal actors or groups”, “these conflicts of values are often conflicts of interest between business actors and public interest”. - Lesson 2 (p. 615): “characteristics of the research environment and the regulatory context can provide business actors with opportunities to enter into ‘political actions’ to undermine early warning signals”. - Firms have “a further incentive to suppress, contradict or downplay” warnings. In tobacco, “such actions contributed to discrediting national and international institutions and NGOs, weakening their ability to produce or relay early warnings signals”. - The definition: “‘Political actions’ are not aimed at maximising profits within the political and regulatory contexts but rather aim at influencing these political and regulatory contexts in the pursuit of profits”. The source is cited as “Scherrer and Palazzo, 2010”, which is missing from the reference list; Ch 27 cites Scherer and Palazzo 2011. - Lesson 3 (pp. 615–616): “psychological and cultural factors contribute to neglecting early warning signals”. - Risk perception and time preference are “biased towards underrating uncertain hazards”. People avoid value‑conflict dissonance and show self‑serving bias. The belief that profit maximisation “is the most appropriate way for business to serve society” provides “a powerful justification for dismissing the relevance of these value conflicts” (p. 616). - Reading note: the sentence adds “and to increase self‑perception of responsibility”. Most coherently, believing profit serves society lets managers see themselves as responsible while dismissing conflicts.

Reflections (pp. 616–617): - Regulation necessary, not sufficient. Legal, fiscal and financial mechanisms “still have a large potential”; “assurance bonding should be considered”. But “it seems unrealistic to believe that complete alignment of business interests with interests of society at large will always be feasible”, so open, rational discussion of value conflicts “is an absolute necessity” (p. 616). - Understanding over blame. This comes “notwithstanding the necessity to strengthen the accountability of business actors”. Blame “with hindsight… may not always be constructive” (p. 616). When firms put business interest before precaution, “it is not necessarily because they willingly act against the interests of society or to harm the environment”. Some actors “may well acknowledge the need to sacrifice some business interest, but may consider in good faith that the early warning signals are not strong enough”; others “may be unaware of the full extent of their conflicting interests and of their self‑serving biases” because of cultures “trapping them in a short‑sighted economic approach” (p. 616). - Proposed institutional role. “a crucial first step towards any solution is awareness and acceptance of the dilemmas business actors are facing, and of the various temptations”. The authors “can imagine that public institutions could support progressive business by analysing and publically disclosing the dilemmas and temptations”. That covers profit‑versus‑harm conflicts, temptations to hide from them, to “use gaps and loopholes of regulation or to influence the regulators”, and “to influence the scientific evidence”. “Rather than prescribing specific precautionary business actions, such institutions could then promote more open, transparent, and stakeholder‑inclusive participatory decision frameworks” (Stirling, 2008) (p. 616). The proposal is framed tentatively (“can imagine”, “could”). - Expected effects. Exposition of the dilemmas “will create further incentives for responsible actors to share and communicate their precautionary responses” (an unhedged prediction). “Clear and factual descriptions… if possible devoid of judgemental considerations, may contribute to reducing unconscious denials, force business organisations to openly discuss the factors driving their decision‑making… and promote more transparency, proactive attitudes and innovative responses” (p. 616). - Position against CSR idealism. This would “more realistically complement initiatives based on the idealised principle that being socially responsible is economically profitable, typical of Corporate Social Responsibility (Porter and Kramer, 2011)” (p. 617). Note: the principle that social value creation can pay is indeed Porter and Kramer’s, but their “shared value” article presented itself as distinct from, and a replacement for, CSR, so labelling it “typical of” CSR is loose. - Political actions. “Even though they could be regarded as a natural tendency to justify and protect one’s own interest”, they “have the potential to disrupt an honest debate”. “The fact that some business actors spend sophisticated efforts to hide or keep secret their political actions can be seen as a signal that their behaviour is of bad faith and would not be socially acceptable”. Regulation making them “more transparent can help to sustain a sound balance of power, thereby maintaining our ability to benefit from early warning signals” (p. 617).

References (pp. 617–619)#

64 entries, mostly business ethics, behavioural economics and precaution literature. Case evidence is second‑hand: via EEA (2001) and vol. 2 chapters; peer‑reviewed analyses of tobacco‑industry documents and industry influence (Hurt and Robertson in JAMA; Hong and Bero in BMJ; Grüning et al. in AJPH; Castleman and Ziem); the WHO expert committee report (Zeltner et al., 2000); a Harvard Business Review insider account (Sells) and HBS teaching cases; plus journalism (Francis in Sierra) and advocacy (Rampton and Stauber).

Irregularities: - Scherer and Palazzo cited but not listed. - Year mismatches: Kysar 2009/2010; Francis 1998/1994; Ravetz 2005/2004. - Misspellings: Zelltner, Biana, Tenbrusel and Messik, Raffensberger, O’Donavan. - One garbled entry merges a tobacco‑documents paper (Muggli, Hurt, Blanke) with O’Neill, Holland and Light’s Environmental Values (2008).

None of this changes the argument, but it indicates light editorial checking.

Case timeline (adapted: this is a cross-cutting chapter, not a case study)#

The chapter does not build its own timelines. It uses short vignettes, mostly undated, and relies on other chapters for detail. The table records what this chapter says, with any dates it gives.

Case (chapter’s source) Business or regulator behaviour as characterised Dates given here Page
Fisheries, Newfoundland cod (EEA 2001 Ch 2) Profit from “most efficient fishing methods”. Inshore fishers (harmed) protest and commission a report that leads to official reappraisal. DFO (misnamed “US”) accused of following the sector, citing uncertainty, calling warnings “pseudo‑science”. No corporate liability; governments pay restoration none 608, 609, 612
Benzene (EEA 2001 Ch 4) Consultants hired to downplay evidence, introducing “irresolvable” dose‑response arguments that delayed regulation. ACGIH TLV committee with corporate scientists recommended limits above the evidence none 608, 610, 613
Lead in petrol (vol. 2 Ch 3) Ethyl and GM exclusively funded and conducted health research for 40+ years. GM controlled publication and constrained the Bureau of Mines. Paid consultants advised government and assured “complete safety”. Critics allegedly defunded and threatened. “No evidence of harm” framing 1920s onward; “over 40 years” 608, 611, 613, 614
Asbestos (EEA 2001 Ch 5; Sells 1994) Low price because costs were externalised to families, health service, insurers and building owners. Laws from the 1930s not implemented. Manville bankruptcy “as a means of dealing with” claims. A culture of “denial” laws “as early as the 1930s”; bankruptcy 1982 608, 612, 614–615
Manville fiberglass (Sells 1994; Paine and Gant) Voluntary “possibly carcinogenic” relabelling despite lawyers’ reluctance; commercially successful. Warning later judged not justified 1986; “nearly five years” later 615
Gaucho / honeybees (vol. 2 Ch 16) Bayer used “selective knowledge”, “semantic slips” and almost exclusively its own research. Beekeepers as early warners. French Ministry of Agriculture with a dual mandate 1990s 608, 609, 610, 611, 612
Growth hormones, antimicrobials (EEA 2001 Chs 14, 9) Profitable use; ended in Europe “driven to some extent” by public concern none 608, 609
TBT antifoulants (EEA 2001 Ch 13) Arcachon oyster industry supported restrictions none 609
PCBs (EEA 2001 Ch 6; Francis) Monsanto’s public defence: evidence unclear; extensive worldwide research needed 1960s 610
Great Lakes (EEA 2001 Ch 12) Shell report denying causation. USDA (pre‑EPA) aligned with pesticide makers, demanding proof of causation before remedial spending 1967 (Shell) 610, 612
Beryllium (vol. 2 Ch 6) PR to reassure customers; reports “scientifically unsound”. Company‑financed publications “under the names of well‑known academics”. Once the limit was “beyond doubt” ineffective, more research and a conference calling for more research before any new limit. DOE with a dual mandate. “No evidence of harm” framing 1980s; late 1990s 610, 611, 612, 614
Tobacco (vol. 2 Ch 7; Hurt and Robertson; Hong and Bero; Grüning; Zeltner) Internal knowledge of nicotine effects blurred. Research institutions funded to “keep the debate alive”. Non‑peer‑reviewed symposia. Deliberate subversion of WHO. Discrediting of institutions and NGOs. Pensioner’s‑party fallacy “likely” relevant to smoking; “no evidence of harm” framing Minnesota trial documents (Hurt and Robertson 1998); WHO report 2000 610, 611, 612–613, 614, 615
Climate / Exxon (van den Hove et al. 2002) Publicly contesting the science while presenting as “science and technology‑based”. Public denial of a privately identified problem “early days of the debate” 610
MTBE (EEA 2001 Ch 11) Most external costs (water treatment, supply, health) borne by society none 612
BSE (EEA 2001 Ch 15) MAFF’s dual promote/protect mandate none 612
Pharmaceuticals (Sismondo; Lexchin and O’Donovan) Funding‑driven design and publication bias; pervasive COI in drug agencies none 610–611, 613
Nuclear / Fukushima “In exceptional cases” an occurring event may lead to temporary over‑weighting and faster regulation; the German phase‑out “may have been” such a case 2011 613
Emerging technologies: mobile phones, GMOs, nanotechnology (vol. 2 Chs “20, 18, 21”, i.e. 21, 19, 22) Cited only as examples of indeterminacy and ignorance, and of regulators lacking information; Joy (2000) cited as an insider raising early warnings Joy essay 2000 609, 611

Lags. The chapter does not compute warning‑to‑action lags. The durations it mentions are: - lead: 40+ years of industry‑controlled research (p. 611); - asbestos: protective laws from the 1930s “not appropriately implemented”, with bankruptcy in 1982 “far too late” (p. 612); - beryllium: calls for further research continued even after the limit was “beyond doubt” ineffective in the late 1990s (p. 610); - fiberglass: the reverse, a warning acted on at once and judged unwarranted about five years later (p. 615).

For actual lags and harms the case chapters must be used.

The authors’ own lessons and conclusions#

Lessons the authors derive from their evidence (analytic): 1. Value conflict plus economic incentive. Early warnings create conflicts between short‑term business gain and long‑term health and environmental values. Because those values are public, these become conflicts of interest between business and the public. Firms, following the profit‑maximising paradigm within existing rules, tended not to take precautionary “business actions” (p. 615; built from pp. 608–609). 2. Political actions. The research environment and regulatory context gave firms the opportunity to “suppress, contradict or downplay” warnings, by pressuring science, lobbying and shaping opinion, sometimes discrediting institutions and NGOs (tobacco). These are “political actions” aimed at changing the context of profit‑seeking, distinct from business actions (p. 615; built from pp. 610–613). 3. Psychological and cultural barriers. Biased risk perception and time preference, dissonance avoidance and self‑serving bias lead decision‑makers to dismiss warnings. The profit‑maximisation ideology and organisational cultures of denial reinforce this (pp. 615–616; built from pp. 613–615).

Recommendations and advocacy (normative, not tested in the chapter): - Regulatory, fiscal and financial mechanisms to internalise external costs of uncertain hazards; consider assurance bonding (pp. 611–612, 616). - Accept that “it seems unrealistic to believe that complete alignment of business interests with interests of society at large will always be feasible”, and create room to discuss value conflicts “rationally and openly” (p. 616). - Seek to understand rather than blame with hindsight, while strengthening accountability (p. 616). - Public institutions (proposed tentatively: “we can imagine… could”) to analyse and publicly disclose industry‑specific “dilemmas and temptations”, and to promote participatory, stakeholder‑inclusive decision frameworks rather than prescribing actions (p. 616). - Non‑judgemental exposition of dilemmas, to reduce unconscious denial, push firms to discuss their decision drivers openly, and give responsible firms incentives to share their precautionary responses (p. 616). - Treat the business‑case version of CSR as idealised; the proposed approach “more realistically” complements it (p. 617). - Regulate for transparency of corporate political actions, to preserve a “sound balance of power” and society’s capacity to benefit from early warnings (p. 617). - Implicit research agenda: better concepts and operational tools for business decisions under indeterminacy and ignorance (p. 609).

Unstated but implied. Burden and standard of proof in liability law should shift towards industry (p. 612 lays out the effect but stops short of recommending it). Regulatory mandates to promote an industry and to protect health should be separated (p. 612 documents the problem but offers no explicit remedy).

Mechanisms and dynamics#

1. The decision frame, and the three leaky channels. The chapter’s core model: firms count harm to others only when it returns to them as a business cost, via liability, regulation or reputation (p. 608). Each channel is shown to leak. - Reputation: publics are unaware, underrate risk or feel unaffected. Firms can shape public opinion when that “appears cheaper” than reducing the risk. Voluntary disclosure risks lost sales and seems to be rewarded “only under very limited conditions” (p. 609). - Regulation: regulators depend on firm‑generated information, have conflicting mandates, and include conflicted experts (pp. 611–613). - Liability: it may be absent; standards and burdens favour defendants; long latency and insolvency let firms escape (p. 612).

Note-taker’s synthesis: on this account the precautionary deficit is less a matter of bad character than of how weakly social costs are connected back to the firm’s decision frame. (The chapter’s own treatment of “political actions” as possible bad faith, p. 617, qualifies this.) The ideology that profit‑seeking within rules serves society (p. 616) makes that gap feel legitimate.

2. The incentive not to know, and information asymmetry. Research that might confirm harm is itself a cost to be avoided (p. 609). Firms often hold the first and best information (p. 611), and disclosure is voluntary in practice even under disclosure rules (p. 611). The result is a structural bias toward strategic ignorance and non‑disclosure. In the tobacco case, private knowledge became public through litigation (documents “made public in the Minnesota trial”, p. 610). The Exxon account likewise turns on the gap between private recognition and public denial (p. 610).

3. Competitive and first‑mover pressure. Competition raises pressure to use cheap hazardous inputs and rewards “early introduction of innovative products or methods” with potential monopoly (pp. 608–609). This is the chapter’s only explicit link between innovation dynamics and non‑precaution, and it is asserted rather than shown.

4. Uncertainty as a structural resource. Three features combine. - (a) Genuine indeterminacy and ignorance defeat probabilistic cost‑benefit tools (p. 609). - (b) Evidential thresholds (95%/99%) have “no scientific justification”, “acceptable risk” is “an ethical or political issue rather than a scientific one”, and weight of evidence is “almost always open to subjective judgment” (p. 610). - (c) Publics expect certainty before costly action (p. 610).

Together these make uncertainty exploitable. The chapter’s cases (p. 610) yield a small taxonomy of tactics: - selective citation and “semantic slips”; - methodological disputes that cannot be resolved; - demands for more (worldwide) research, including after inadequacy is “beyond doubt”; - counter‑reports; - labelling adverse findings “scientifically unsound”; - new research bodies to “keep the debate alive”; - a science‑based self‑image alongside public denial of privately recognised problems.

The result is “paralysis by analysis” (p. 610). Note-taker’s gloss: in these cases the demand for “more research” works as a delay tactic, especially when voiced by the party that benefits from delay.

5. Control of the research agenda (bias without misconduct). The strongest version of the claim: firms need not falsify anything. Choosing which studies are done, their design, their publication, and their venue (non‑peer‑reviewed symposia) shifts the apparent count of evidence (pp. 610–611). The extreme case is lead, where two firms were the sole funders and controllers of health research for 40+ years and constrained the regulator’s reporting (p. 611). Company‑financed publications “under the names of well‑known academics” (beryllium) borrow credibility (p. 611). A second‑order dynamic follows: such conduct degrades trust in all privately sponsored research, a collective credibility cost that also falls on honest firms (p. 611).

6. Regulatory structure: dual mandates and capture. - Five cases across the US, Canada (misattributed), the UK and France show agencies charged with both promoting a sector and protecting against its harms (p. 612). Biased behaviour is described for two of them: demanding proof of causation before expenditure (USDA) and, hedged (“is said to”, “is accused of”), invoking uncertainty and dismissing warners as “pseudo‑science” (DFO). For MAFF, the French Ministry of Agriculture and the DOE (whose mandate included “cheap production of nuclear weapons”), the chapter describes only the conflicting mandate, not its effects. - Capture operates through conflicted experts on standard‑setting committees (benzene TLVs; lead consultants advising the Bureau of Mines while paid by Ethyl) and through shaping public opinion, which then shapes regulators (pp. 612–613).

7. Liability design allocates the cost of uncertainty. Evidentiary standard, causation test and who bears the burden of persuasion determine the expected cost of a hazard to a firm (p. 612). Victim‑bears‑burden regimes lower expected costs and therefore lower incentives for precaution. Harms with long latency interact with limited liability and insolvency: bankruptcy converts future liabilities into social costs (Manville 1982, p. 612). Assurance bonding is proposed as an ex ante fix (pp. 611–612, 616).

8. Distribution of costs and benefits. Profits are private and short‑term. Costs fall on society at large (governments restoring fish stocks; public water and health costs for MTBE; asbestos costs borne by families, health services, insurers and building owners) and on future generations through discounting (pp. 612, 614). The chapter notes that when harm falls on another economic actor (fishers, oyster farmers, beekeepers), that actor becomes an early warner and a force for precaution (p. 609). Where precaution comes from depends on who bears the costs.

9. Cognition: why early warnings are cognitively weak signals. - Early warnings concern harms that are unexperienced, probabilistic, abstract and distant. Human judgment underweights exactly these (description‑experience gap, discounting, finite pool of worries, single‑action bias) (pp. 613–614). - This dampens both managerial precaution and public pressure “in most situations”. “In exceptional cases”, when a low‑probability event actually occurs, people may temporarily overrate its probability and regulation may speed up (the German phase‑out after Fukushima “may have been” such a case, p. 613). Note-taker’s extrapolation: this suggests a step‑change dynamic, with long periods of under‑response punctuated by sudden responses. The chapter does not develop this or call the German response excessive. - Survivorship bias (“pensioner’s party fallacy”) and the conflation of “no evidence of harm” with “evidence of no harm” are both natural errors and rhetorical resources for reassurance (pp. 613–614).

10. Motivated ethics and organisational culture. - Self‑serving bias means uncertainty is doubly useful to an interested party: hard to model, and a “welcome ‘excuse’” (p. 614). - Self‑deception is sustained by euphemism, slippery slopes, wilful ignorance and avoiding responsibility‑revealing reasoning (p. 614). It can be unconscious and collective: “good people” build cultures where “problems get buried” and alternatives seem “unattainable” (p. 615). - Publics, for their part, overlook harmful conduct by others when it is indirect and has not yet produced a bad outcome, so external correction is weak too (p. 614). - Lawyers appear as a force against disclosure (Manville’s lawyers were reluctant to relabel, p. 615), a hint that legal‑risk thinking can oppose precautionary communication.

11. Business actions versus political actions. This is the chapter’s most distinctive conceptual contribution (p. 615). - Profit‑seeking within rules (continuing or stopping a product) is a business action. Efforts to change the rules, the evidence base and public opinion in pursuit of profit are political actions. - Note-taker’s inference (not stated in the chapter): the distinction undercuts the “stay within the rules” defence (p. 611). If firms shape the rules and the evidence, rule‑compliance no longer shows social acceptability. - It also names a systemic harm beyond any single product: political actions can degrade society’s early‑warning capacity itself by discrediting institutions and NGOs (tobacco, p. 615). - The authors treat secrecy about political action as a signal of bad faith (p. 617). This lets them combine charitable interpretation of business decisions with a harder line on political conduct.

12. Mental models of the actors. - Managers. Profit maximisation is embedded in tools and training (p. 608). Ethics counts only instrumentally. Profit‑seeking is believed to be “the most appropriate way for business to serve society”, which justifies dismissing value conflicts and, in the chapter’s words, serves “to increase self‑perception of responsibility” (p. 616): managers can feel responsible while setting the conflicts aside. Managers were “unwilling or unable to believe” long‑term consequences (Sells, p. 614). - Firms’ public posture. Confident reassurance (“complete safety”, p. 613), scientific self‑image (“a science and technology‑based company”, p. 610), and a stated preference for “more research”. - Regulators with promotional mandates. They think in terms of proof of causation before spending (p. 612) and treat warners as partisan (“bad faith”, p. 612). - Public. Expects certainty and overlooks indirect harms (pp. 610, 614).

13. Framing and language noted in the text: “complete safety” (p. 613); “scientifically unsound” (p. 610); “keep the debate alive” (p. 610); “pseudo‑science” (p. 612); “semantic slips” (p. 610); “a science and technology‑based company” (p. 610); “paralysis by analysis” (p. 610); “no evidence of harm” vs “evidence of no harm” (p. 614); “language euphemisms” as a self‑deception mechanism (p. 614).

14. Counter‑forces the chapter identifies. Other industries that bear harm (p. 609); public concern, sometimes out of proportion to the evidence (p. 609); vivid events (p. 613); litigation that forces out internal documents (p. 610); crisis‑induced cultural change inside a firm (Manville, p. 615); and, prospectively, liability reform, bonds, disclosure of dilemmas, and transparency of political activity (pp. 616–617).

15. What the chapter does not analyse. - Lock‑in and path dependence appear only as a hint: cultures making “alternative business solutions unattainable” (p. 615). - Substitutes and alternatives are not discussed. - Innovation effects of precaution (positive or negative) get only a passing mention of “innovative responses” and “Innovative solutions such as assurance bonding” (p. 616). - Trade and international competition are absent beyond a citation. - Differences among firms (producer vs user, size, ownership, investors, insurers) are not considered. - Existing EU instruments that address the gaps it identifies go unmentioned, for example REACH’s industry data requirement (“no data, no market”, 2006) and the Environmental Liability Directive (2004/35/EC).

Transferable insights (technology-neutral)#

  1. Precaution is unlikely where social harms do not feed back into the decision‑maker’s own accounts. When a warning conflicts with revenue, a profit‑maximising frame produces precaution only if harm returns via liability, regulation or reputation. - Evidence: pp. 608–609, 615; recurring across fisheries, benzene, lead, asbestos, insecticides, hormones. - Strength: moderate. A consistent pattern and standard economics, but drawn from cases selected as failures, with no comparison set.

  2. Each feedback channel has predictable leaks. Reputation fails when publics are unaware or unaffected, or when opinion is cheaper to shape than the risk is to reduce. Regulation fails through information dependence, dual mandates and conflicted experts. Liability fails through absent rules, victim‑side burdens, strict causation tests, latency and insolvency. - Evidence: pp. 609, 611–613. - Strength: moderate. The liability and regulatory leaks are illustrated by several cases each (fisheries, MTBE, UK asbestos, Manville; USDA, DFO, benzene TLVs, lead consultants) plus established law‑and‑economics reasoning, all second‑hand. The reputation leak is argued mainly from general reasoning and one unpublished study by an author; the chapter’s reputation cases (antimicrobials, hormones) show the channel working.

  3. There is an incentive not to know. Research that might confirm harm is treated as a cost, and firms often hold the first and best information, with disclosure effectively voluntary. - Evidence: pp. 609, 611; tobacco documents and Exxon (p. 610). - Strength: moderate. The logic is compelling and illustrated by prominent cases, but the chapter asserts rather than measures its prevalence.

  4. Competitive and first‑mover pressure works against precaution. The chance of market dominance from early introduction raises the cost of waiting. - Evidence: pp. 608–609. - Strength: asserted. One sentence citing an economic‑theory paper and the Gaucho case study; the chapter shows no case in which the mechanism operated.

  5. Uncertainty is structurally exploitable. Because evidential thresholds and weight‑of‑evidence are judgment calls, and publics expect certainty before costly action, interested parties can “manufacture uncertainty” and turn “more research” into delay (“paralysis by analysis”). - Evidence: p. 610; seven named cases. - Strength: strong for the existence and recurrence of the pattern, with multiple documented cases including disclosed internal documents; later corroborated for Exxon (see limitations). Its frequency across industries is not established here.

  6. Agenda control biases evidence without misconduct. Selecting which studies are funded, their design, publication and venue can shift the apparent balance of evidence. Sole‑funder control of a field is the extreme case. - Evidence: pp. 610–611 (lead 40+ years; Bayer; beryllium publications financed under academics’ names; tobacco symposia). - Strength: moderate–strong. Well supported by the cited funding‑bias literature and cases. Individual case claims carry hedges (“it is even said”).

  7. Distorted sponsored research creates a collective credibility cost. Manipulation by some sponsors lowers trust in all privately funded research and penalises honest actors. - Evidence: p. 611. - Strength: suggestive. A plausible inference with no evidence presented.

  8. Dual promote‑and‑protect mandates bias regulators toward the promoted sector. This shows up as demands for proof of causation before action and dismissal of warners. - Evidence: p. 612 (USDA, DFO, MAFF, French Ministry of Agriculture, US DOE). - Strength: moderate. Dual mandates are shown in five cases in four countries, but biased behaviour is described for only two (USDA; DFO, hedged “is said to”, “is accused of”). All rely on secondary characterisations.

  9. Conflicted experts on advisory and standard‑setting bodies shape “safe” limits. - Evidence: pp. 612–613 (benzene TLVs, lead consultants, pharma agencies, WHO/tobacco). - Strength: moderate. Documented in the historical cases (benzene TLVs via Castleman and Ziem; lead; WHO/tobacco). The pharma claim is weaker than stated: the cited paper reviews COI policies and finds that agencies “manage” rather than prohibit conflicts; its abstract gives no prevalence figures.

  10. Who bears the burden and standard of proof decides who bears the cost of uncertainty. Regimes that require victims to prove substantial causation lower firms’ expected costs and weaken incentives for precaution.

    • Evidence: p. 612.
    • Strength: moderate. Sound legal‑economic reasoning, not empirically tested in the chapter.

  11. Long latency plus limited liability and insolvency turns future harms into social costs, so ex ante financial assurance is proposed.

    • Evidence: pp. 611–612 (Manville 1982), 616.
    • Strength: moderate for the mechanism; asserted for assurance bonding as a remedy (no evidence of feasibility or effect).

  12. Early warnings are cognitively weak signals. They concern unexperienced, probabilistic, delayed harms that people systematically underweight, which “in most situations” impedes both precaution and public pressure. In exceptional cases, once a low‑probability event actually occurs, a temporary over‑weighting can speed regulation.

    • Evidence: pp. 613–614 (Fukushima and the German phase‑out, offered tentatively).
    • Strength: moderate for under‑weighting: robust laboratory psychology, but the chapter does not show these biases operating in the specific business decisions. Suggestive for the post‑event reversal: one tentatively offered example (“may have been”), described by the chapter as exceptional.

  13. Survivorship bias and the confusion of “no evidence of harm” with “evidence of no harm” recur in reassurance.

    • Evidence: pp. 613–614.
    • Strength: suggestive. The authors say they found no psychological studies on the latter; it rests on case reports.

  14. Self‑interest turns ambiguity into an excuse, and good people can build cultures of denial. Self‑deception works through euphemism, slippery slopes and wilful ignorance, and can be unconscious.

    • Evidence: pp. 614–615 (Sells on Manville).
    • Strength: moderate. Well‑established behavioural ethics findings; one strong insider testimony. Generalisation to other firms is inferred.

  15. Cultures can change after a crisis, and early voluntary precaution can pay commercially.

    • Evidence: p. 615 (Manville fiberglass 1986).
    • Strength: suggestive. A single case from an insider account and a business‑school case. The warning was later judged unfounded, and the motive plausibly included post‑asbestos liability fear, so the case fits economic rationality as much as ethical change.

  16. Distinguish business actions from political actions. Efforts to shape the evidence, the rules and public opinion undermine the defence that profit‑seeking “within the rules” is acceptable. They can also weaken society’s capacity to detect and relay warnings at all.

    • Evidence: pp. 611, 615, 617 (tobacco discrediting institutions and NGOs).
    • Strength: moderate as an analytical distinction grounded in the cases. The claim that secrecy signals bad faith (p. 617) is a normative inference.

  17. Harmed third parties are natural early warners and allies of precaution. Economic interests are not one‑sided: actors whose livelihoods are damaged by a hazard often detect and publicise it first.

    • Evidence: p. 609 (cod fishers, oyster farmers, beekeepers).
    • Strength: moderate for the existence of the pattern; the chapter itself calls these cases a minority.

  18. Market pressure is not calibrated to evidence. Public concern can end a practice whether or not it is well founded, and fail to arise where the risk is real but unfelt.

    • Evidence: p. 609 (antimicrobials, hormones; Sunstein).
    • Strength: suggestive–moderate.

  19. Full alignment of private and public interest is unrealistic, so open, non‑judgemental exposition of sector‑specific dilemmas and temptations is needed alongside regulation.

    • Evidence: pp. 616–617.
    • Strength: asserted. A recommendation with no evidence that such disclosure changes behaviour.

  20. Transparency of corporate political activity helps preserve the balance of power needed for warnings to be heard.

    • Evidence: p. 617.
    • Strength: asserted / advocacy. A reasonable inference from Lesson 2 but untested.

Limitations, contestation and bias check#

Nature of the evidence. - This is a narrative literature review, not a systematic or comparative study. No case‑selection protocol, coding or comparison group is described (p. 608). - The case base is the Late lessons corpus, which EEA assembled precisely as examples of warnings neglected. “In virtually all reviewed cases it was perceived to be profitable…” (p. 607) is therefore close to built in by selection. The chapter cannot say how often firms do act on warnings, nor whether its factors distinguish the cases where firms acted from those where they did not. That is selection on the dependent variable. - The chapter does not engage with vol. 2 Ch 2 (Hansen and Tickner, on false alarms), the report’s own treatment of regulatory false positives. - Evidence for individual cases is second‑hand, filtered through EEA chapter authors and a mix of sources: peer‑reviewed tobacco‑documents and industry‑influence studies (JAMA, BMJ, AJPH; Castleman and Ziem), a WHO expert report, an insider HBR account, journalism (Sierra magazine) and advocacy works (Rampton and Stauber). - The authors often hedge appropriately (“seems to have”, “it is even said”, “is said to have”, “is accused of”, pp. 610–612). The digest should preserve those hedges rather than harden them. - Self‑citation is substantial. One empirical claim, that consumers reward disclosure of negative information only under “very limited conditions” (p. 609), rests on an unpublished mimeo by one of the authors.

Accuracy issues found in this read. - “US Department of Fisheries and Oceans” should be Canada’s DFO (p. 612). - The company is Brush Wellman, not “Wellmann”. - Cross‑references to Chs 18, 20 and 21 should read 19, 21 and 22 (p. 609). - The Sells quotation is attributed to “one of the presidents of Manville”. Sells was himself president of Manville’s Fiber Glass Group (HBR author note) and wrote in the first person, so the words are probably his own; the chapter’s wording is at best ambiguous (p. 614). - The Friedman quotation is truncated before “without deception or fraud” (p. 611). - Porter and Kramer (2011) are labelled as typical of CSR although they presented “shared value” as distinct from CSR (p. 617). - Lexchin and O’Donovan (2010), a review of COI policies, is cited for “widespread potential conflict of interests among scientific experts” (p. 613) and for study‑selection bias (p. 611); its abstract supports neither claim directly. - Several reference errors (see the References section above), including one citation missing from the list and one garbled merged entry. - The Joy (2000) example does not illustrate the point it is used for (p. 611).

The one positive case is double‑edged. The Manville fiberglass story (p. 615) is presented as proof that business and social values can align. But (i) the warning was later judged not significant, so this was a false positive the firm acted on. The chapter records this without discussing what it implies for the costs of precaution or for when warnings should trigger action. (ii) The firm had just been bankrupted by asbestos liability, so the relabelling is consistent with economic rationality under heightened liability and reputational exposure, not only with ethical renewal. (iii) The “nearly five years” refers to the company’s own reassessment (about 1991). Formal bodies moved much later (IARC Group 3 in 2001; NTP delisting in 2011, from a secondary summary; IARC 2B in 1987–88, from memory), so the chapter’s account of when the warning was shown to be unfounded rests on the company’s view.

Two causal stories not fully reconciled. The chapter offers both a calculative account (economic rationality, political actions, deliberate manufacture of doubt, public denial of private knowledge) and a non‑calculative one (bounded rationality, unconscious self‑deception, culture). The cases it cites most prominently (tobacco documents, Exxon, lead) point to deliberate strategy. The “understand, don’t blame” reflection (p. 616) fits the non‑calculative account better. The authors partly resolve this by applying charity to business actions (good‑faith disagreement, unconscious bias) and suspicion to political actions (secrecy as a sign of bad faith, p. 617). That is a defensible, useful interpretive key, but it is not stated as such, and the chapter gives no way to tell the two accounts apart in a live case.

Balance, in both directions. - Fair to business. The chapter acknowledges business benefits and innovation (p. 608), progressive actors (pp. 608, 616), good‑faith judgments that signals are too weak (p. 616), the risk of hindsight blame (p. 616), that public concern can be baseless (p. 609), that post‑event over‑reaction occurs (p. 613), and that honest firms bear a credibility cost (p. 611). This is more even‑handed than much of the report. - Less fair or incomplete. - No discussion of the costs of precaution: forgone benefits, substitution risks, or innovation lost to false alarms. - No examination of cases where firms led on precaution. Ch 27 supplies several. - No attention to differences among firms, such as user companies acting earlier than manufacturers (noted by Ch 27, p. 647). - The psychological section’s lab findings apply to regulators, NGOs and early warners as much as to business, but only business and public under‑reaction is developed. - Existing policy instruments that already address the gaps (REACH data duties; the Environmental Liability Directive) are not assessed.

Recommendations are thin on mechanism. - Which “public institutions” would analyse and disclose dilemmas, with what powers, and with what evidence that disclosure reduces denial is not specified (p. 616). - Assurance bonding is endorsed without discussing how to size a bond for hazards the chapter itself describes as marked by “indeterminacy and even ignorance” (pp. 609, 616). There is a tension between the critique of probability‑based cost‑benefit tools and a remedy that needs some quantification. Kysar’s “symmetric humility” argument, in the cited reference title, may address this, but the chapter does not explain it. - The transparency‑of‑political‑action recommendation is sensible but general (p. 617).

Hindsight. The authors warn against hindsight blame (p. 616) while building their evidence entirely from hindsight‑constructed case histories. They do not ask what these warnings looked like at the time, relative to the many warnings that proved unfounded.

Later evidence noted in passing (for the hindsight strand; only the Exxon item was checked against the primary paper, and the fiberglass dates against a secondary summary, Wikipedia’s “Glass wool” entry, not the IARC or NTP documents themselves): - Exxon. Supran and Oreskes (2017, Environmental Research Letters), a content analysis of 187 ExxonMobil communications from 1977 to 2014, found that 83% of peer‑reviewed papers and 80% of internal documents acknowledged climate change as real and human‑caused, against 12% of advertorials, 81% of which expressed doubt. This strongly corroborates the chapter’s p. 610 claim, which rested in 2013 on a 2002 paper co‑authored by Le Menestrel. ExxonMobil has contested this body of research. - Fiberglass. IARC Group 3 (2001) and NTP delisting (2011) come from a secondary summary and should be confirmed against the IARC monographs and NTP Report on Carcinogens; see above. - Not checked here, for follow‑up: post‑2013 EU restrictions on imidacloprid (Gaucho’s active ingredient) and related neonicotinoids (2013 partial; 2018 outdoor ban); uptake of assurance bonds or financial‑security requirements for novel hazards; lobbying‑transparency reforms (e.g. the EU Transparency Register); later syntheses of “manufactured doubt” tactics across industries.

Notable quotes#

  1. “In virtually all reviewed cases it was perceived to be profitable for industries to continue using potentially harmful products or operations.” (p. 607)
  2. “ethical values are reasons to act if and only if they contribute to the expected economic benefits for the business actor” (p. 608)
  3. “Economic rationality is thus remote from a proactive precautionary response to early warning signals.” (p. 608)
  4. “economically rational companies can decide to influence public opinion in their favour when this appears cheaper than reducing or terminating the potentially harmful practice.” (p. 609)
  5. “it is an ethical or political issue rather than a scientific one to determine an ‘acceptable level of risk’” (p. 610)
  6. “by emphasising the lack of scientific certainty companies can contribute effectively to a ‘paralysis by analysis’ that prevents precautionary measures” (p. 610)
  7. “They do not even have to manipulate results or engage in other forms of misconduct… but they can effectively bias research results in their own interest” (pp. 610–611)
  8. “Strong uncertainty… may also serve as a welcome ‘excuse’ and justification about why the profitable action may not be so unethical after all.” (p. 614)
  9. “‘Political actions’ are not aimed at maximising profits within the political and regulatory contexts but rather aim at influencing these political and regulatory contexts in the pursuit of profits.” (p. 615)
  10. “The fact that some business actors spend sophisticated efforts to hide or keep secret their political actions can be seen as a signal that their behaviour is of bad faith” (p. 617)

Open questions#

  1. Discrimination. Which of the named factors (economic stake, uncertainty, regulatory gaps, psychology, culture) actually distinguish cases where firms acted with precaution from cases where they did not? This would need a comparison set that includes positive cases (such as those Ch 27 lists) and false alarms (vol. 2 Ch 2).
  2. Deliberate or unconscious? How can one tell, in real time, a good‑faith judgment that a signal is too weak from motivated reasoning or deliberate doubt‑manufacture? Is the chapter’s implicit test (secrecy about political action signals bad faith) usable in practice?
  3. Producers and users. Does precaution come disproportionately from users and harmed third parties rather than producers, as p. 609 and Ch 27 (p. 647) suggest? What does that imply for where to look for early warners?
  4. Assurance bonds under ignorance. Can assurance bonds or financial‑security requirements be designed for hazards whose probabilities and magnitudes are unknown, and what happened where they were tried after 2013?
  5. Disclosure of dilemmas. Has any public institution tried the proposed disclosure of sector‑specific “dilemmas and temptations”, and did it reduce denial or reward precautionary firms?
  6. Transparency and delay. Have transparency requirements for corporate political activity (lobbying registers, research‑funding disclosure) measurably reduced manufactured‑doubt tactics or shortened warning‑to‑action lags?
  7. Manville fiberglass. What actually happened: the epidemiology, the dates, the commercial outcome, and how the “nearly five years” squares with IARC 2B (1987–88), Group 3 (2001) and NTP delisting (2011)? Does the case show ethical renewal, liability‑driven rationality, or both?
  8. The research‑agenda critique. How should sponsor‑funded research be weighted in weight‑of‑evidence assessments without discarding honest industry science (the collective credibility problem, p. 611)?
  9. Competition. Does competitive or first‑mover pressure (pp. 608–609) show up empirically in the case histories, or is it only asserted in this chapter?
  10. The reversal dynamic. If post‑event over‑reaction (p. 613) is a common counterpart to pre‑event under‑reaction, what governance designs smooth the cycle rather than amplifying either phase?

Audit log#

Independent audit against the full text extract (PDF pp. 609–622), with spot checks in the PDF: a pypdf check confirmed no embedded images on PDF pp. 609–622; the table of contents (PDF pp. 5–6); bios (PDF pp. 696, 699); Ch 8 p. 193; Ch 27 pp. 647 and 667 (reference list). Changes made: