Late Lessons, Jensen Huang and AI
Markdown version

LL1-16 — Ch16 Twelve late lessons#

Report: Late lessons from early warnings: the precautionary principle 1896–2000, EEA Environmental Issue Report No 22 (2001) Chapter: 16, “Twelve late lessons”, credited to the “Editorial team” Report pages: 168–191 (text pp. 168–189; references pp. 189–191) PDF pages: 168–191 (report and PDF page numbers are identical in this file) Read: every page of the text extract in order, through the final page marker (p. 191). Box 16.1 (p. 170) and Table 16.1 (p. 184) checked visually against the PDF; the extract is faithful for both. Front matter (pp. 1, 3–6) and author biographies (pp. 195–199) consulted only to identify the editorial team.

Authors and standpoint#

Who wrote it. The chapter is credited to the “Editorial team” (p. 168). The front matter (p. 1; acknowledgements p. 6) names that team:

The Preface (p. 3) says the case authors’ lessons “were then distilled into twelve ‘late lessons’ by the editorial team, under the guidance of the EEA Scientific Committee.”

Standpoint. The chapter is openly normative and pro-precaution. Its stated purpose is to give “a basis for the practical implementation of the precautionary principle” (p. 169). Six things shape how it reads:

  1. A prior framework shaped the lessons. The lessons were organised using the ESTO project on technological risk and the management of uncertainty (Stirling, 1999). The chapter says this “provided the initial framing of this analysis” and gave “an opportunity to test or elaborate” ESTO points against the case material (p. 168). The chapter describes the process as two-way. The lessons address “the key issues that emerged from the case studies” (p. 168), and the Preface says the case authors’ own lessons were “distilled” into the twelve (p. 3). Note-taker’s inference: the organising categories (risk, uncertainty and ignorance; framing; diversity; participation) closely match the editors’ own research programme (Stirling; Wynne), so the lessons are best read as case material sorted into a pre-existing framework, not as a purely inductive result. No method for the “test” is described.
  2. Several editors were also case authors. Four of the seven editors wrote or co-wrote case chapters: fisheries (MacGarvin), asbestos (Gee), PCBs (Keys) and MTBE (Harremoës) (contents, pp. 7–8). They were synthesising their own evidence among others.
  3. Section 16.3 draws substantially on the editors’ own published work. It cites Stirling (1999), Stirling and Mayer (1999), Wynne (1992, 2001), Wynne et al. (2000, 2001) and allied scholars (Tickner, a peer reviewer of the report (Acknowledgements, p. 6); Ashford; O’Brien; Raffensperger). It also cites a wider policy and academic literature, including the NRC, RCEP, WBGU, Kourilsky and Viney, Krohn and Weyer, Underwood, Copeland et al., Popper and von Hayek. The chapter says the twelve lessons follow a rule of not introducing “material extraneous to the case studies”, but that the final section deliberately goes beyond them (p. 169).
  4. Advocacy backgrounds are declared. They are in the biographies (Gee: trade unions, Friends of the Earth; MacGarvin and Keys: consultancy for NGOs and industry). This does not invalidate the analysis, but it matters when weighing contested claims, such as “persistent obstruction and misinformation by vested interests” (p. 179).
  5. Framing from outside the chapter. The Preface is signed by the EEA Executive Director (Domingo Jiménez Beltrán), not the editorial team. It adds a point that sits in some tension with the information-centred lessons: “the absence of political will to take action to reduce hazards, in the face of conflicting costs and benefits, seems to be an even more important factor in these histories than is the availability of trusted information” (p. 4). The chapter’s own treatment of political will is brief (“short-term economic and political interactions”, p. 168; “lack of political will”, p. 181).
  6. Forward references. Box 16.1 and 16.3.1 point to “the last part of this report” and “the ‘Implications’ part of this report” for more on levels of proof and Type I/II errors (pp. 170, 184). Ch 17, “Conclusions” (pp. 192–194, with Table 17.1), does discuss levels of proof (p. 193), so it is the likely target for that point. But Ch 17 has no Type I/II discussion, and the Preface says implications “will be further explored” in “a separate EEA publication” (p. 3). So the “‘Implications’ part” may refer to material that ended up outside this report. Ch 17 is outside this section and was only skimmed here to check the reference.

Panels or commentaries. None. The 2001 report has no panels, and this chapter carries no dissenting or industry voice. Counter-positions appear only as the editors characterise them: “some views critical of precaution” (p. 185); the US view that EU GMO risk assessment is “irrationally exaggerated” measured against “sound science” (p. 183); and “fishers can be less precautionary” (p. 177).

Section-by-section notes#

These are paraphrased notes. Verbatim quotes are kept to key phrases.

16.1 Introduction (pp. 168–169)#

Method. Summarising all the cases would just replicate them, so the editors return to the four questions put to every case author (p. 168): 1. When was the first credible early warning? 2. What were the main actions or inactions, and when? The key point is the gap between identifying a problem and effective action. 3. What were the resulting costs and benefits? 4. What lessons follow for future decisions?

Main findings. - The gap was long: “many years or decades, and in some cases over a century” (p. 168). - Even after the need for precaution was articulated in the 1970s–80s, “unequivocal precautionary action” was “relatively scarce” (p. 168). - The editors push back on the framing of precaution as an EU versus North America contest: the case histories “suggest another story”, one of “different degrees of acceptance of the need for precaution within different institutions” on both sides of the Atlantic (p. 168). - Core diagnosis: adequate hazard information was often available well before action, but was not brought to decision-makers early enough, or was discounted. In some cases early and even “‘loud and late’” warnings were ignored because of “short-term economic and political interactions”: asbestos, PCBs, Great Lakes, sulphur dioxide (p. 168).

Costs and benefits were the hardest question (p. 168). There were two reasons. Case authors were technical experts, not economists. And there is “no credible way” to reduce pros and cons to “a single figure”, because of incommensurables and uneven distribution across interest groups. The editors add that “there are constructive ways of dealing with these complications”, but a general analysis “lay beyond the scope” of the report. Some treatment of costs and benefits was unavoidable in answering the fourth question (p. 168). This is a significant self-declared limitation. The Preface invites readers to judge whether earlier action would have come “at a lower overall cost to society” (p. 3). This chapter does not demonstrate that systematically.

Framework. The ESTO project on technological risk and uncertainty (Stirling, 1999) “provided the initial framing”. The cases were used to “test or elaborate” ESTO’s points (p. 168).

The twelve lessons as listed (pp. 168–169): 1. Acknowledge and respond to ignorance, as well as uncertainty and risk, in technology appraisal and public policy-making. 2. Provide adequate long-term environmental and health monitoring and research into early warnings. 3. Identify and work to reduce “blind spots” and gaps in scientific knowledge. 4. Identify and reduce interdisciplinary obstacles to learning. 5. Ensure that real world conditions are adequately accounted for in regulatory appraisal. 6. Systematically scrutinise the claimed justifications and benefits alongside the potential risks. 7. Evaluate a range of alternative options for meeting needs alongside the option under appraisal, and promote more robust, diverse and adaptable technologies so as to minimise the costs of surprises and maximise the benefits of innovation. 8. Ensure use of “lay” and local knowledge, as well as relevant specialist expertise, in the appraisal. 9. Take full account of the assumptions and values of different social groups. 10. Maintain the regulatory independence of interested parties while retaining an inclusive approach to information and opinion gathering. 11. Identify and reduce institutional obstacles to learning and action. 12. Avoid “paralysis by analysis” by acting to reduce potential harm when there are reasonable grounds for concern.

Wording drift between the list and the section headings: - No. 10 is awkward in the list. Its heading (p. 178) reads “Maintain regulatory independence from economic and political special interests”. - No. 9’s heading is “Take account of wider social interests and values”. - No. 6’s heading is “Systematically scrutinise and justify the claimed ‘pros’ and ‘cons’”. - Use the list wording when quoting a lesson.

Caveats the editors attach (p. 169): - The distinctions are “illustrative, rather than definitive” and interlinked. - Many lessons concern how information is produced and used within “a more participative and democratic process”. - Appraisal should be scaled to the likely size of the consequences. This is an implicit proportionality principle. - The lessons were meant to add nothing “extraneous to the case studies”. Wider points are confined to 16.3.

16.2.1 Respond to ignorance as well as uncertainty (pp. 169–171)#

Central lesson. Every activity studied had some form of risk assessment, “(formal or informal)”. What was neglected was the “virtual certainty” that some factors lay outside its scope. That is the domain of ignorance and “inevitable surprises” (p. 169). - “Surprise is inevitable”. Research that yields positive surprises (“discoveries”) will also yield negative ones. - Complex, cumulative, synergistic and indirect effects have been poorly handled by regulators. - Precaution means acknowledging the possibility of surprise. It does not mean “blanket opposition to innovation”. It means humility about the science and broader appraisal: more disciplines, knowledge types and constituencies (p. 169).

Illustrations (pp. 169–170): - CFCs: the very possibility of an ozone hole was unappreciated before 1974. CFCs were “relatively inert and benign” under “normal” conditions, and “less conventionally hazardous than the substances they replaced”. But they behaved differently under conditions the appraisal never considered. Note-taker’s inference, not stated in the chapter: the safety property (inertness) was bound up with the harm. - DES: the next-generation effects were “a complete surprise”. - TBT: accumulation in top predators was “simply not envisaged”. - Antimicrobials: the case authors say the later dilution of the Swann Committee’s advice rested on what was “precisely known rather than” what was not. They call for science with “more humility and less hubris” (p. 170).

Lesson. Appraisal should extend to as wide a range of conditions and effects as can reasonably be anticipated. Surprises will still occur, but the cases suggest “it is possible to do rather better than in the past” (p. 170).

Practical proxies for ignorance (pp. 170–171). This is the most operational content in the chapter. - Irreversibility can be weighed even when consequences are unknown, though this is “not quite as simple as it seems” (p. 170). - For halocarbons, PCBs and MTBE, three things known at the outset were warning signs: novelty as artificial chemicals, persistence, and a tendency to disperse and become ubiquitous. It could have been deduced that any problem would take years to “‘go away’” (pp. 170–171). - Where persistence emerged later, “regulators were often slow to react”. Examples are TBT’s greater-than-assumed persistence and the long-known permanence of asbestos dust (p. 171). - Screening rule: using persistence and bioaccumulation to screen out hazards should make future “‘surprises’” smaller and less serious (p. 171). - Irreversibility applies beyond chemicals: antimicrobial resistance is long term, and collapsed fish stocks may never recover. - Scale: a global hazard offers “only one ‘experimental’ model” (p. 171).

Anticipation is possible. The Swann Committee anticipated much of the antimicrobial problem in the 1960s. That sensitivity was later “overwhelmed by scientific over-confidence”. PCB animal tests in 1937 “might have been taken as more of a warning” (p. 171).

Correlated harms and sentinels (p. 171): - A known harmful property may be correlated with less obvious harms. - Acute effects preceded chronic ones, sometimes by decades, for sulphur dioxide, radiation, benzene, asbestos, TBT and PCBs. - The relationship is “asymmetrical”: chronic harm need not be preceded by acute signs (BSE, halocarbons), so this is not a general rule. - Wildlife effects can be “sentinel events” for humans, which argues for integrated ecological and health appraisal.

Institutional versus societal ignorance. This is a key distinction (p. 171). - Institutional ignorance: relevant knowledge exists in society but is “not available to the decision-makers”. It is present in “most of the case studies”. The resulting surprises “though they can be serious, may be quite localised”. The remedy is better communication and social learning. - Societal ignorance: nobody knows. It is “more intractable”, and it is “also exemplified in many of the case studies (including BSE)”, so the two conditions coexist across the cases. The remedies are research plus “greater diversity, adaptability and flexibility” in decisions and technology choices. - Definition note (note-taker’s reading): as defined on p. 171, institutional ignorance covers knowledge that is “not available to the decision-makers”. Warnings that reach decision-makers and are then ignored for “short-term economic and political interactions” (p. 168) look like a separate failure mode. But the chapter’s own usage is looser: on p. 174 it applies “‘institutional’ ignorance” to disciplinary capture, including MTBE, where water-pollution information “was available” but “essentially disregarded”. The boundary between non-delivery and discounting is therefore blurred in the text itself.

Box 16.1, “Risk, uncertainty and ignorance” (p. 170; checked visually): - Opening claim: the precautionary principle is “seen principally as a way to deal with a lack of scientific certainty”, and there is “an urgent need” for a more systematic way of thinking about how uncertainty pervades regulatory appraisal. - Risk: outcomes are known and probabilities can be assigned. Here risk assessment “is a valid technique that can save lives”, though judgements about what counts as “at risk”, and about the balance struck, are still value-laden. - Uncertainty: there is no adequate basis for probabilities, because of novelty, complexity or variability. Conventional risk assessment is “too narrow”. Safety factors and scenario or sensitivity analysis help but do not suffice. - Ignorance: some possible outcomes are themselves unknown, so there is a “continual prospect of surprise”. This needs humility and “an institutional capacity for open reflection” on the quality of available knowledge. Appraisal must spell out the implications of alternative assumptions and how they relate to different social groups. - Robust, transparent and accountable treatment of all three is “one crucial means of regaining public confidence”. - Levels of proof: decisions should be explicit about the level of proof needed to justify action. Options range from “scientifically based suspicion” to “beyond reasonable doubt”, each with different cost and benefit implications for different groups. This is more sophisticated than “simple pronouncements of truth or falsity”. - Minor inconsistency: the Box labels differ slightly from Table 16.1 (p. 184).

16.2.2 Research and monitor for “early warnings” (pp. 171–173)#

Critical-path questions left unasked (p. 172): - BSE: identified in 1986. Research on maternal transmission, on which MAFF’s early position depended, did not start until 1989, and it found maternal transmission did occur. Scrapie-to-cattle experiments began only in 1996. No survey of infectious but symptomless cattle entering the food chain had ever been done. Meanwhile government reassurance cited absence of evidence “when no evidence was actually being sought”. The chapter calls this “a classic example of ‘no evidence of harm’ being misinterpreted as ‘evidence of no harm’”. - Antimicrobials: the resistance concerns of “the 1967 Swann Committee” (date inconsistent elsewhere) were not followed up until the 1990s, “despite longstanding knowledge that widespread antibiotic usage could lead to the rapid development of resistance”. - Asbestos: no systematic health monitoring, despite recommendations for mortality studies between 1898 and the 1920s and feasible methods. - TBT: the case authors stress that baseline studies are “frequently under-rated”. - MTBE and PCBs: research on identified concerns was lacking. - Monitoring alone is not enough. Results must be reported, disseminated and used.

Research is no panacea (pp. 172–173): - Research may turn ignorance into uncertainty and uncertainty into risk, but not necessarily. It can “compound uncertainty”. - A Canadian multi-species fisheries model became less predictable as more data were added. Great Lakes research “amplified the uncertainties” about bird population crashes. - “At face value”, this challenges the idea that more research, or more disciplines, is automatically precautionary (pp. 172–173). - Likewise, the antimicrobials, sulphur dioxide and PCB cases show research opening up “the research domain and … sources of ignorance as increasing complexities are discovered” (p. 173). This reinforces the point above; it is not a counter-example. - Symmetry principle: where research “genuinely reveals” a concern is unfounded, for example by showing a convincing alternative mechanism, it is “in no way precautionary” to keep restricting the agent (p. 173).

Procedural recommendation, from the Swann model (p. 173). Appraisals should state: - the precise research question; - how long the research will take; - who funds it; - how independent the researchers are; - whether hazard reduction should come before or after the research.

16.2.3 Search out and address “blind spots” (pp. 173–174)#

Blind spots in the dominant discipline (p. 173): - Halocarbons: the depletion mechanism was published in Nature in 1974, yet regulators neglected it until “firm empirical evidence” arrived. The 1985 confirmation was “essentially by accident”. An earlier satellite programme had detected depletion, but the results were “considered suspect and set aside”. The chapter’s point is that analytical assumptions can “prejudge the results”. Blind spots sat at the centre of the relevant disciplines, in both theory and data. - Antimicrobials: the Swann Committee’s 1968 evidence anticipated later environmental, welfare and health problems. Its recommendations were “initially influential” but became marginalised over the following decades. Had they been implemented and sustained, they “might at least have mitigated” today’s difficulties. - MTBE: persistence was apparent at the outset but raised too few questions in the formal process. - TBT: degradation rates rested on marine-environment assumptions that were wrong in many areas. - Hormones: regulators missed that young children with low natural oestrogen were the likely at-risk group. - Radiation: risk estimates relied on atomic bomb survivors at atypically high doses and dose rates, and long overlooked the resulting uncertainty.

The “problem already solved” blind spot (pp. 173–174): - Asbestos: from 1906 onwards there were successive claims that past disease reflected superseded conditions and was “not so likely to occur (in future)”. Each small improvement reset a clock that took decades to run. - Substitutes: the first asbestos substitutes shared properties such as fibre size, and eventually proved to carry “essentially similar, if much lower, risks”. - Sulphur dioxide: taller stacks relieved local effects but not cumulative, long-range ones (p. 174).

Lesson (p. 174): search systematically for blind spots “at the heart of” the disciplines historically in charge. Engage other disciplines and knowledge sources to provoke “uncomfortable interactions” that expose false assumptions.

16.2.4 Identify and reduce interdisciplinary obstacles to learning (p. 174)#

The discipline whose effects appear first can come to dominate appraisal, even to hold it “‘captive’”. This produces institutional ignorance. Examples (p. 174): - Asbestos and radiation: standards were shaped by clinicians’ focus on acute effects, and chronic toxicology and epidemiology were neglected. - MTBE: appraised through engine, combustion and air-pollution knowledge. Water persistence and taste and odour were “essentially disregarded, though the information was available”. - Sulphur: a health-built regulatory process struggled to absorb ecological effects. - Hormonal growth promoters: a human-health focus, and wildlife concerns got little attention. - The reverse, antimicrobials and BSE: veterinary framing marginalised human impacts. Swann’s 1968 recommendation of a single advisory process with “overall responsibility for the whole field of use of antibiotics” in humans, animals, food preservation and other purposes was not taken up, in the UK or elsewhere, “for many years”. - BSE: UK veterinary officials judged transmissibility to humans “acceptably slight”. The US had treated a scrapie–CJD link as possible since the 1970s and kept infected animals out of the food chain.

16.2.5 Ensure that real world conditions are fully accounted for (pp. 174–175)#

The problem is well recognised in principle, yet the cases show many incomplete assessments (p. 174).

Assumed performance versus actual practice (pp. 174–175). “In the human sphere it is often assumed that technologies will perform to the specified standards”, and it “may be a long time” before the gap is noticed: - MTBE: US appraisal underestimated storage-tank leakage and hence exposure. Better tank design can be undone by poor installation. - PCBs: containment in “‘closed’” systems proved impossible. The results were Yusho and Yucheng, losses from poor maintenance, and illegal disposal into the food chain. - Halocarbons: optimistic assumptions about engineered containment and the efficiency of decommissioning reduced the effectiveness of control measures. - Benzene: some public exposure routes were rigorously controlled, but “these are not necessarily the routes of greatest exposure”. Petrol exposure had no mitigation practices “or even warnings”. - Radiation: despite an optimal-dose principle from 1949, doses for the same examination still vary 100-fold between hospitals. - Growth promoters: WHO/FAO expert committees assessed only authorised use and single substances, not combinations, and gave little attention to misuse (doses, injection sites, retained implants, shortened withdrawal).

Scope creep and exposed groups (p. 175): - Regulation did not follow asbestos’s spread into consumer products and housing. Users and nearby residents, not just workers, were recognised as at risk only late. - The 2001 WTO Appellate Body found “controlled use” could not be relied on to protect workers in real conditions.

Non-compliance (p. 175): illegal PCB disposal; poor antimicrobial practice; evasion of fisheries rules; unrealistic assumptions about UK slaughterhouse compliance after BSE; acute effects in Puerto Rico and Italy blamed on “the illegal or incompetent use” of growth promoters; halocarbon smuggling “threatening the effectiveness of global controls”.

Natural systems departing from assumptions (p. 175): - PCBs concentrate at high latitudes, and bioaccumulated PCBs are more toxic than commercial mixtures. - Single-stock fisheries models ignore interactions between species. The chapter concedes “enormous, perhaps insoluble, problems” in broadening the stock-modelling approach, but says the single-stock approach “does not reflect actual conditions”. - The stratospheric chemistry of halocarbons was unanticipated. - Sensitive subgroups (“young boys”, p. 175; “young children”, p. 173) respond differently from the “average”.

16.2.6 Systematically scrutinise and justify the claimed “pros” and “cons” (pp. 175–177)#

Even-handed starting point (pp. 175–176). All sides want benefits examined. Some fear risk-talk obscures benefits; others think benefit and efficacy claims go unscrutinised. Either way, appraisal should test benefit claims and the conditions under which they would or would not materialise.

Partial appraisal of fixes (p. 176): - Tall stacks and smokeless fuel solved the 1950s urban respiratory episodes. That success may have diverted attention from long-range acid transport: “a classic case of an end-of-pipe ‘solution’” creating “less visible hazards”. An integrated approach eventually addressed all of it. - MTBE “apparently promised a simple solution” to lead, so its taste, odour and groundwater-persistence problems were “perhaps more readily overlooked than might otherwise have been the case”. - Hormonal growth promoters: health, environment and welfare were “‘not given significant attention’”.

Distribution (p. 176). For sulphur dioxide, the distribution of pros and cons between the UK and Scandinavia “was very different, or at least was perceived to be very different”. For natural-environment effects, “Sweden seemed to be suffering the most”. Only when UK building damage became evident did both see themselves as victims.

Foreseen but not acted on (p. 176). For BSE and Canadian cod, the pros and cons of early action were identified in advance. But uncertainty was significant and costs high, so action was limited, leading to “far greater costs (many of which were anticipated)”. Letting depleted stocks recover pays off even without a collapse.

Efficacy, DES (p. 176). 1953 trial data showed DES was ineffective against miscarriage in some groups and actively harmful. This was not appreciated at the time. Use declined slowly, there was no regulatory action and marketing continued. Bans came only 20–30 years later, after cancers in daughters. Closer early scrutiny of efficacy might have prevented some of them.

Prior justification, radiation (p. 176). The ICRP’s 1950s “justification principle” (the chapter says “International Committee”) responded to dubious uses: ringworm, children’s shoe-fitting, cosmetic hair removal, psychiatric treatment. Yet “even with this criterion”, justification “remains open to question”. Surveys find that doses “have reduced drastically”, but a large proportion of medical X-rays are still “of doubtful clinical use”. Note-taker’s reading: the principle has therefore had only partial success. (The “1950s” dating is flagged for checking under “Factual and dating issues” below.) Sweden’s reduction of antimicrobial use is offered as another wider weighing of pros and cons.

Pricing and liability (pp. 176–177). Correct allocation of costs and benefits is a “pre-condition” for choosing between technologies. Polluter-pays tools (taxes, subsidies, liability) serve equity. Because the prices of asbestos, halocarbons and PCBs excluded their health and environmental costs, they gained “an unjustifiable advantage in the marketplace”, keeping “technically superior substitutes” out for longer than was socially optimal. Internalisation and liability are “controversial” but “essential” for efficiency and equity.

16.2.7 Evaluate alternatives; promote robust, diverse and adaptable solutions (p. 177)#

Lock-in. Framed as “one concern”: once a technology is committed, “institutional and market processes act to reinforce its position, even if markedly inferior to potential alternatives” (p. 177). - MTBE: alternatives existed (bioethanol, better engines, higher-octane fuel), but “little formal scrutiny” occurred at adoption. California is now evaluating all successors. - TBT: the case authors warn against “‘chemical for chemical’” substitution. - X-rays: diagnostic substitutes are “underutilised”. - Second-generation CFC substitutes were “perhaps … unduly tolerated” because they beat the originals, while more benign options were not examined. - Antimicrobials: husbandry without routine antimicrobials exists in several countries but is “not actively promoted”.

This raises “challenging issues” about how regulation relates to private product development. Alternatives should be developed within “eco-efficiency”, “clean production” and closed-loop material flows, to shrink future surprises (p. 177).

16.2.8 Use “lay” and local knowledge as well as specialist expertise (pp. 177–178)#

Who and why. Lay people include workers, users, neighbours and heavily exposed consumers. The claim is not that they know more or care more. Their knowledge is complementary, “sometimes” better grounded in real operating conditions, and independent of narrow professional perspectives. Lay knowledge may also rest on different assumptions about “what is salient, or what degree of control is reasonable to expect or require”, whereas specialists “may simply respond to granted authority without further reflection” (p. 177).

Examples (pp. 177–178): - Asbestos and PCB workers knew of ill health before regulators did. Communities spotted clusters first (Love Canal). - Fishers are sometimes less precautionary than others, but often want to act precautionarily and are blocked by “systems failure”. Canada and others increasingly involve them. - Swedish farmers’ knowledge of alternative husbandry let them cut antimicrobials voluntarily ahead of regulation. Widely held knowledge is needed even to recognise alternatives. - BSE: rules separating specified bovine offal were “widely flaunted” (sic) and inspection failed to report it. Industry workers were “apparently better informed” about operational realities than senior advisers and officials.

Balance. Lay knowledge deserves the same scrutiny as expertise and may be more vulnerable to error. An example is the “‘pensioners’ party fallacy’”, where asbestos workers took healthy retirees at a Christmas party as proof of safety (p. 178). Still, methods should be developed to bring such knowledge in. Doing so can strengthen appraisal, governance and legitimacy.

16.2.9 Take account of wider social interests and values (p. 178)#

16.2.10 Maintain regulatory independence (pp. 178–180)#

Principle (pp. 178–179). Interested parties should argue their claims. Independent, accountable institutions should adjudicate.

Undue influence (p. 179): - Benzene was known as a bone-marrow poison in 1897, asbestos respiratory effects in 1898, and “PCB-induced chloracne” in 1899 (flag), with worker effects by the late 1930s. Yet real progress came only in the 1960s–70s. - The UK BSE regulator was “responsible first to the industry and only second to consumers”. - The 1974 temporary lifting of the US DES growth-promoter ban “followed strong pressure from the farming lobby”, despite available alternatives (flag).

Weak refutations, obstruction and suppression (p. 179): - Poorly substantiated “refutations” sustained over long periods: of the identification of PCBs as widespread pollutants (1960s); and in the UK response to acid deposition (1970s to early 1980s). - Asbestos: “persistent obstruction and misinformation by vested interests”. By 1918, some US and Canadian insurers were already wary of covering asbestos, but did not sustain that stance, at a cost of billions. - Benzene: action was held up by demands for animal-carcinogenicity evidence despite human evidence, and by statistically erroneous low-dose claims. - Great Lakes: suppression of reports and intimidation of publishers. - Antimicrobials: critical research was delayed. - BSE and asbestos: critics were discredited. - Californian sardine fishery, 1930s: agency scientists were dismissed. - “‘Shooting the messenger’” since Galileo “rarely, if ever, promotes societal welfare”.

Self-censorship (p. 179). The 1988 Southwood BSE committee thought a ban on all cattle brains in the food chain might be scientifically justified, but not “politically feasible”. It trimmed its advice to what was “‘realistic’” or “‘achievable’”.

Information dependence (pp. 179–180): - Appraisal “frequently fails” because it depends on information “produced and owned” by the firms being assessed. - Independent information is “necessary, if not sufficient”, and was often missing. - The editors concede that not all cases show that dependence distorted outcomes. But hazard evidence tends to meet “low public profile interpretative jockeying” to justify inaction. - The editors call it “very difficult to dispute” that, had an independent public body controlled and disseminated such information, the policy debate “would have been more open, and more pluralistic”, and diverse interests “more equitably, thoroughly and probably more rationally represented”. - Independent information institutions, with rights, resources and responsibilities, are key to real independence. - Signs of change: EU advisory committees moving from “‘producer’” directorates to the Health and Consumer Directorate; new independent food agencies nationally and at EU level (p. 180).

16.2.11 Identify and reduce institutional obstacles (pp. 180–181)#

Short-termism. Government and business cycles work against medium- and long-term welfare (asbestos, benzene, PCBs) (p. 180).

Transitions (p. 180): - BSE: a 1979 UK commission recommended minimum rendering standards. A new administration withdrew the draft regulations the same year as “an unnecessary burden on industry”. Whether they would have helped is “not clear”, but similar standards featured in that government’s 1996 BSE response. - California sardine fishery: a precautionary programme was reversed by a change of government.

Levels and departments (p. 180): - California, 1930s: the US Federal Bureau of Fisheries opposed the state’s precautionary proposal as a drag on commerce. - BSE: MAFF told the Department of Health about BSE only 17 months after it first knew, and only because DoH assent was needed to keep clinically affected cattle in the food chain. - Governments can exert “more or less subtle influence” even over formally independent agencies. - Agencies can become “part of the issue” through their own past decisions. In BSE, economic concern for farming and fear that inconsistency would damage credibility delayed recognition.

National differences (pp. 180–181): - Sulphur: Norway accepted the causal science in 1976, the UK only in 1985. The editors suggest the UK’s scepticism was plausibly coloured by the uneven distribution of costs and benefits. - Sweden and the EU: EU membership rules obstructed Sweden’s national antimicrobial policy. - Global problems need global institutions (TBT and the IMO). - Same evidence, different decisions: in the mid-1970s the USDA, but not MAFF, excluded scrapie-affected animals from food and feed on the same evidence. National DES decisions also varied widely in timing on the same information (p. 181).

16.2.12 Avoid paralysis by analysis (pp. 181–182)#

The acknowledged tension. The lessons so far say “know more”, so how much is enough? Paralysis arises from “information overload or lack of political will”. The example is the US Supreme Court benzene decision’s “anti-precautionary ‘straightjacket’”, which demanded “layer upon layer” of information (p. 181).

“We know enough” (p. 181): - Swann (1969): the evidence was a sound enough basis for action, and “the cry for more research should not be allowed to hold up our recommendations”. - Other cases, asbestos and BSE, “suggest” that more or better-targeted research at an earlier stage “would have helped minimise future costs”. - The US Ecosystems Principles Advisory Panel on fisheries held that uncertainties are not “acceptable excuses to delay”.

Prospective versus retrospective appraisal (p. 181). This is a key analytical move. - The Great Lakes authors argued that bringing in new disciplines raised uncertainty and could produce paralysis. - Most cases are prospective: they start from an agent and look for effects. Broadening is precautionary there, because it finds more possible effects. - The Great Lakes is retrospective: it starts from observed effects and looks for causes. Broadening can raise doubt about targeting any one agent. - The editors’ resolution: restricting the wrong agent is not precautionary. But precaution applies to uncertainty about agents as much as about effects. It may “entirely legitimately” defend continued action on an agent until the uncertainties are resolved. - That this did not happen in the Great Lakes is “more a reflection of the value judgements in the prevailing legal and socio-political context” than of “intrinsic inconsistencies in the concept”. - “If the culture is not prepared to act even when there is demonstrable evidence of cause and effect”, precaution will fail. Regulatory cultures and sub-cultures vary, for example the differing US approaches across BSE, fisheries, MTBE, benzene and the Great Lakes (pp. 181–182).

Distribution decides what counts as prudence (p. 182). Whether more information is “paralysis” or “prudent and careful evaluation” depends on how the pros and cons “impinge upon” each group. Early action is likely when benefits are large, harms small and both evenly spread. When they are uncertain and unevenly spread across groups or time periods, consensus is harder.

Innovation (p. 182): - “It may be that in some cases” a field or direction will need to be severely curtailed or ended “where society judges the risks to be unacceptable”. - But discouraging one pathway differs enormously from “channelling” innovation elsewhere. - Asbestos, halocarbons, PCBs and antimicrobials show that curtailment “may actually serve to foster and intensify innovation in other areas”, and may give leading countries “a competitive edge”. - Intelligent foresight and precaution can lower social costs, stimulate innovation, encourage “systems-based science” and improve decisions. - No data are given for these claims.

16.3 The wider implications of precaution: introduction (pp. 182–183)#

Self-assessed criteria for the lessons (p. 182): empirically grounded; general enough for “virtually any risk management problem”; balanced and fairly comprehensive; concrete enough for practice, though the precise measures in any case “will necessarily be defined by variable local circumstances”.

Admitted limits (p. 182). One short discussion cannot anticipate “the full range and diversity of detailed practical precautionary measures and procedures”, still less their contextual application. The editors point to “a wide and burgeoning literature” (O’Riordan and Cameron, 1994; Harding and Fisher, 1999; Raffensperger and Tickner, 1999; Stirling, 1999; O’Riordan et al., 2001).

Convergent policy literature (pp. 182–183). The chapter says many lessons have been developed elsewhere, often without the “precaution” label: - US National Research Council, Understanding risk (1996), and the Presidential/Congressional Commission (1997); - UK Royal Commission on Environmental Pollution (1998), on “framing assumptions”; - France, Kourilsky and Viney (1999); - Germany, WBGU (2000); - Sweden’s chemicals policy, which uses persistence and bioaccumulation as “‘proxies’” for unknown impacts (p. 183).

GMOs (p. 183). Specialists (van Dommeln, 1997) find EU GMO risk assessment too reductionist and narrow; the editors call these “cogent critiques”, so they are not neutral here. US counterparts find the same assessment “irrationally exaggerated” by their “‘sound science’” standard. The chapter’s conclusion: framing all questions as resolvable by existing risk knowledge is “radically incomplete”, above all in deciding which questions are salient.

Institutional experiments (p. 183): - Danish and Dutch consensus conferences and scenario workshops; - UK “strategic commissions” (food, human genetics, agricultural genetics); - the BSE Inquiry (Phillips et al., 2000) and the mobile phones inquiry (IEGMP, 2000). These appraisals explored “various of these lessons”, with specific recommendations on issues “such as” institutional conflicts of interest and unrealistic expectations of science as “touchstone, or arbiter, of ultimate truth”.

No single recipe (p. 183). The source text garbles this sentence (“avoid caution against over-reliance”), but the sense is clear: do not over-rely on any one set of precautionary prescriptions. Implementation draws on many methods, many “on the same continuum as orthodox risk management”. Choosing among them is “an essentially political business”, subject to deliberation, review and democratic accountability. Precaution began in environmental policy, but a proper framework “must encompass these wider domains” of decision-making, “even if only indirectly”. The section then introduces three themes: science, innovation and governance.

16.3.1 Precaution and science (pp. 183–186)#

Statistics (pp. 183–184): - Hypotheses cannot be proved, only falsified, so tests reject a null hypothesis at 95% or 99% confidence. - The tradition of minimising false positives (Type I errors) means “not being wrong is more important than being safe” (p. 184). The chapter notes policy is increasingly aware of this. It points to Table 16.1 and to the report’s “‘Implications’ part” for more (p. 184). - Small, costly monitoring programmes and high natural variability make false negatives (Type II errors) likely. Citing Underwood (1999), there is little concern about erring against the environment. - Low-power monitoring wastes money, for example ship-borne marine sampling unable to separate effects from noise (HELCOM, 1996). - Further pitfalls: the wrong question; factors dismissed by erroneous prior assumption; sampling design; rare events; non-linearity. - In epidemiology, small exposure misclassification biases towards missing associations (Copeland et al., 1977). Low power produces false reassurance from “‘negative’” studies. - A bias that generates false negatives, when those are disasters “as in most of these case studies”, is “not sound public policy” and not precautionary (p. 184).

Table 16.1, “Different levels of proof for different purposes: some illustrations” (Source: EEA; p. 184; checked visually):

Verbal description Examples
“Beyond all reasonable doubt” Criminal law; Swedish chemical law, 1973 (for evidence of “safety” from manufacturers)
“Balance of evidence” Intergovernmental Panel on Climate Change, 1995 and 2001
“Reasonable grounds for concern” European Commission communication on the precautionary principle
“Scientific suspicion of risk” Swedish chemical law, 1973, for evidence required for regulators to take precautionary action on potential harm from substances

Analytical note. The Swedish 1973 law appears twice: a very high bar for manufacturers to show safety, and a low bar for regulators to act. One regime can set asymmetric standards of proof for different parties. This is a concrete design for shifting the burden of proof.

Fundamental issues (pp. 184–185): - “Uncertainty” should be differentiated “at least” into risk, uncertainty and ignorance. A wider examination of how agreement on the “facts” is reached would also have to take account of “other distinct dimensions such as complexity, indeterminacy, ambiguity, and the nature of disagreement” (Wynne, 2001; Stirling, 1999). - Precaution has prompted new thinking in the sociology of science and philosophy, including recognition of “new kinds of ignorance underlying the very processes of knowledge procurement” (p. 184). - Krohn and Weyer (1994): full knowledge of an innovation’s consequences comes only by using society and the environment as “the experimental laboratory”. This has unrealised implications for democratic policy (pp. 184–185). - Commitment to precaution responds to science’s innovative powers “outrunning its capacity to anticipate the consequences”. Too often scientists denied this “waning ability to predict”. The democratic demand for circumspection concerns “our ability to know”, not innovation as such (p. 185). - Raising such questions is not anti-scientific. Denying them is closer to it. Citing Hayek, “‘pretence at knowledge’” undermines science’s authority (p. 185). - Dual identity: among peers, science is provisional; externally, it is asked for certainty. The contradiction goes “unacknowledged — but not, it appears, unnoticed” (p. 185).

Publics and institutions (p. 185): - “Public surveys” on GMOs “on both sides of the Atlantic” (Levy and Derby, 2000, a US FDA focus-group report; Wynne et al., 2000, the final report of the EU-funded PABE project, listed with an editor as first author) are said to show that non-experts make “a basically correct distinction between uncertainty and ignorance”. Levy and Derby is a focus-group study by its title; PABE was also focus-group based (general knowledge; verify). Neither is a large representative survey. Their concern is unacknowledged ignorance, and hence the purposes driving an enterprise and “who benefits”. - Quoting Wynne: if consequences can never be fully known, publics want the purposes and interests controlling responses to surprise to be “good ones”. - Official efforts to reassure by researching known uncertainties are “futile”. They misread public concern as a demand for zero risk. The institutional approach “only feeds public mistrust by inadvertently demonstrating its own denial of ignorance and lack of intellectual mastery”. - Against critics who call precaution populist anti-science, the editors claim “ample evidence” that people accept more radical uncertainty than institutional science acknowledges. No evidence is cited at that point; the nearest empirical sources are the two GMO studies above (with Wynne, 1992 and 2001, cited for the framing). Precaution rejects “reductionist, closed and arbitrarily narrow science” in favour of more robust science (Stirling, 1999).

Uncertainty is a public matter (pp. 185–186): - “Perhaps the most fundamental general insight”: uncertainty, like knowledge used to authorise policy, is not a private matter for scientific bodies to settle before policy sees it. - The NRC (1996) and RCEP (1998) both said prior questions need public deliberation: what science should address, which factors are salient, and what makes good science for environmental policy (for example comprehensiveness versus precision). - Facts and values (p. 186): precaution does not yield one “correct” outcome. Citing Popper, policy cannot be derived from facts alone, and fact-heavy policies that ignore values will not win acceptance where opinion is divided (RMNO, 2000). - Stakeholders from the start (p. 186): involvement must “begin at the beginning”, not only at the risk-management stage. Appraisal, management and communication are not sequential. The NRC, the RCEP and the Nice Council (2000) recognise this. This is not a new paralysis; ignoring these complexities makes progress slower and mistakes worse.

16.3.2 Precaution and innovation (pp. 186–187)#

Starting point (p. 186). The German Vorsorgeprinzip of the 1970s treated stimulating innovation, employment and forward planning as integral to precaution. The “overarching principle” is that innovation and regulation should be less separate and antagonistic. Building the lessons into innovation itself could end their adversarial relationship.

Chemicals example. Risk assessment identifies a chemical’s use but does not use that to assess alternatives (p. 186).

Constructive technology assessment (CTA) (Netherlands; pp. 186–187): - Technologies are “more than hardware”; they depend on social goals, relationships and skills. - The greater a system’s “scope, power, complexity or interconnectedness”, the more these social and institutional aspects matter. - CTA brings in innovators, regulators, users and stakeholders early, instead of leaving society only an “(often adverse) downstream reaction”. - Systems tend to “‘lock-in’” early, foreclosing options and raising switching costs. Winners may prevail through “chance and first-leader advantage” rather than merit (pp. 186–187).

Technological options analysis (TOA) (US; Ashford, Tickner; p. 187). It routinely considers “‘off the shelf’” and “‘on the horizon’” alternatives and can be run by agencies or firms. With CTA it forms an “alternatives assessment” culture (O’Brien, 2000) that treats innovation as open-ended and chosen. The in-text Ashford dates (1981, 1994) do not match the reference list (1984, 1991).

Diversity as insurance (p. 187). Robust, diverse, adaptable technologies stimulate innovation and insure against surprises such as mesothelioma and ozone loss. Surprises are smaller with several competing technologies than with “one, global, near monopoly”, as asbestos, halocarbons and PCBs were. Diversity helps with intractable “‘societal ignorance’”.

16.3.3 Precaution and governance (pp. 187–189)#

Scope. Governance means formal and informal management and regulation, and more broadly “the conduct of life or business”. Precaution requires more than new decision rules: it requires learning across policy, industry, science and civil society (p. 187). The section has three parts.

(a) “Evolution, not revolution” (pp. 187–188). Existing tools can be extended: - Multi-criteria mapping (Stirling and Mayer, 1999) handles divergent framings. - Life cycle analysis covers full chains. - Cost-benefit analysis, despite “serious limitations”, usefully insists on weighing pros against cons. - Sensitivity and scenario analysis support humility. - Integrated environmental assessment can synthesise these. The lessons suggest it should add: - openness about framing agendas; - the distinction between risk, uncertainty and ignorance; - “the risk and consequences of being wrong”; - all values; - wider pros and cons; - earlier involvement; - “what-if” and participatory scenarios (p. 188).

(b) Participation and subsidiarity (p. 188): - Regulation has historically centralised and “disenfranchise[d]” many groups. Participatory initiatives are vulnerable to the same centralising pressure. - Without safeguards, a consensus conference can become consultation “driven by the sponsor’s agenda”. Appeals come too late for anything but marginal issues. - Participation should be early, broad and local, within democratically legitimate strategic frameworks. It need not mean “paralysing indiscriminate full public participation in every single decision”. - Examples: New Zealand’s subsidiarity-based environmental law; Australia’s oceans policy, which found local interest groups’ views differ from national NGO or industry positions. - The chapter concedes participatory tools are “in various stages of development, and the challenges are far from trivial” (Brookes, 2001). It sets this against traditional approaches, where “the costs of failure can also be high”: public rejection of irradiated food, the abandoned Brent Spar dumping, and “the response to GMOs”.

(c) Awareness and ethics (pp. 188–189): - Participation needs an engaged public and education from home and school, for critical skills, not “indoctrination”. - Media and communicators favour “‘positive knowledge’” and “‘clear and simple’” messages, excluding ignorance and complexity (p. 189). - Ethics of responsibility: because responsibility has been bounded by scientific knowledge, consequences beyond prediction are treated “by definition” as beyond responsibility, even though surprises are known to come. Precaution requires enlarging responsibility to cover unknowns “predictable in principle even though not in specifics” (p. 189). - Closing conviction: a cultural shift is needed towards civil responsibility and involvement in policy, including science and technology policy. Expert institutions of science, industry and policy must “learn to trust, to challenge and to build” frameworks for civil society to take responsibility. Once accepted, precaution “leads far beyond the simple definition” and is “playing its part in the development of civil society and policy-making during the early 21st century” (p. 189).

16.4 References (pp. 189–191)#

The list is short and mostly consists of risk-governance and STS sources (Stirling; Wynne ×4; O’Riordan; Raffensperger and Tickner; NRC; RCEP; Krohn and Weyer; Hayek; Popper), plus a few statistical and monitoring sources (Underwood; Copeland et al.; HELCOM). There is almost no primary case literature; that sits in the case chapters. The one exception is the BSE Inquiry report (Phillips et al., 2000), cited only in general terms (p. 183).

Minor citation errors suggest hurried assembly but do not affect substance: - Ashford dates: 1981 and 1994 in the text, 1984 and 1991 in the list. - “Omen” for Omenn; “Raffensberger”; “H. Feinberg” for Fineberg. - Wynne (1992) listed as Global Environmental Change “Vol. 6” (the 1992 article is in Vol. 2; general knowledge). - Godet (1992, p. 187) and Brookes (2001, p. 188) are cited but missing from the list. - Year mismatches: National Oceans Office 2000 in the text, 2001 in the list. WBGU is dated 2000 in both text and list, but the list entry describes it as the “Jahresgutachten 1998” (annual report 1998); this may be publication year versus report year rather than an error. - Renn et al. (1996), cited for Danish and Dutch consensus conferences and scenario workshops (p. 183), is listed as a Swiss landfill-siting paper.

Case timeline#

Ch16 is a cross-case synthesis, not a case study. Below are the dates and lags as Ch16 states them. They are second-hand and should be checked against each case chapter’s notes. Flags mark internal inconsistencies or likely errors.

Case Early warning(s) cited in Ch16 Response / action cited Lag or pattern noted in Ch16
Benzene Bone-marrow poison 1897 (p. 179); 19th-century evidence, “not collated” (p. 172) “Significant progress” 1960s–70s (p. 179); US Supreme Court “straightjacket” (undated, p. 181); petrol exposure route never addressed (p. 175) ~60–80 years; erroneous low-dose statistics claims (p. 179)
Asbestos Acute respiratory effects 1898 (p. 179); mortality-study recommendations 1898–1920s (p. 172); “not so likely to occur” claims from 1906 (p. 173); insurers wary by 1918 (p. 179) Progress 1960s–70s (p. 179); similar-fibre substitutes (pp. 173–174); WTO Appellate Body 2001 (p. 175) ~60–70 years; “persistent obstruction and misinformation” (p. 179)
PCBs Chloracne “1899” (p. 179, flag); 1937 animal tests (p. 171); workers’ effects by late 1930s (p. 179); identified as widespread pollutants in 1960s (p. 179) Progress 1960s–70s (p. 179); “closed systems” failed: Yusho, Yucheng, illegal disposal (p. 174) ~30–40 years from 1930s evidence
Ionising radiation (Preface: first injury reports 1896, p. 3); optimal dose principle 1949 (p. 175) ICRP justification principle 1950s (p. 176) Dose variation ×100 persists (p. 175); risk models built on atypical high doses (p. 173)
Halocarbons Mechanism in Nature 1974 (p. 173); satellite depletion data “set aside” (p. 173) “Regulatory neglect until firm empirical evidence” (p. 173); ozone hole confirmed 1985 “essentially by accident” (p. 173); HCFCs “unduly tolerated” (p. 177); smuggling (p. 175) ~11 years to confirmation, per Ch16’s framing (flag: see bias check on aerosol bans and Montreal timing)
DES 1953 trial data showed ineffective and harmful (p. 176) No regulatory action; bans “20–30 years” later after daughters’ cancers (p. 176); US growth-promoter ban temporarily lifted 1974 (p. 179, flag) 20–30 years
Antimicrobials (growth promoters) Swann Committee dated 1960s (p. 171), “1967” (p. 172), “1968” (pp. 173, 174), “1969” (p. 181), inconsistent Concerns not followed up until 1990s (p. 172); Swedish reduction and voluntary controls (pp. 176, 178); EU membership an obstacle to Swedish policy (p. 180) ~25–30 years
Sulphur dioxide / acid rain 1950s urban episodes (p. 176) Tall stacks and smokeless fuel (p. 176); Norway accepted causal science 1976, UK 1985 (p. 180); UK “refutations” 1970s–early 1980s (p. 179) 9-year transnational lag on accepting the same science
MTBE Persistence “apparent at the outset” (p. 173); taste and odour info available (p. 174) Adopted without scrutiny of alternatives (p. 177); California evaluating successors (p. 177) Not dated in Ch16
BSE US treated scrapie–CJD link as possible from 1970s (p. 174); same evidence available to UK mid-1970s (p. 181); 1979 rendering standards recommended, then withdrawn (p. 180); BSE identified 1986 (p. 172) MAFF told DoH 17 months after being alerted (p. 180); Southwood 1988 self-censorship (p. 179); maternal-transmission research 1989 (p. 172); scrapie-to-cattle experiments 1996 (p. 172); no asymptomatic surveys ever (p. 172) “Absence of evidence” reassurance while not looking (p. 172)
Fisheries Californian sardine precautionary programme, 1930s (p. 180); Canadian cod pros and cons “identified prior to the event” (p. 176) Sardine scientists dismissed; state action opposed by federal bureau; programme reversed with change of government (pp. 179–180) Collapse costs “many of which were anticipated” (p. 176)
TBT Persistence “relatively quickly established” (p. 171) Depends on IMO global agreement (pp. 180–181) Hazard reduction did not account for long-term effect early enough (p. 171)
Great Lakes Manifest wildlife effects (retrospective case) (p. 181); Love Canal community awareness (p. 177) Suppression and publisher intimidation (p. 179); precaution not invoked because of legal and socio-political culture (p. 181) Broadened research amplified uncertainty (pp. 172–173)
Hormones as growth promoters Children or young boys as sensitive group (pp. 173, 175); wildlife concerns raised (p. 174) JECFA assessed authorised use only, and single substances (p. 175) Misuse ignored (p. 175)

Overall claim on lag: “many years or decades, and in some cases over a century” (p. 168).

The authors’ own lessons and conclusions#

Lessons the editors present as derived from the case evidence#

These are the twelve lessons (pp. 168–169), argued in 16.2 with case illustrations (pp. 169–182):

  1. Ignorance, not just risk and uncertainty. Surprise is inevitable. Use proxies such as irreversibility, novelty, persistence, bioaccumulation, dispersal and scale. Distinguish institutional from societal ignorance (pp. 169–171).
  2. Monitoring and research for early warnings, including baselines. Absence of evidence is not evidence of absence. Research can also compound uncertainty (pp. 171–173).
  3. Blind spots within dominant disciplines, including the “problem already solved” blind spot (pp. 173–174).
  4. Interdisciplinary capture produces institutional ignorance (p. 174).
  5. Real-world conditions: performance, misuse, non-compliance, changing applications, complex environments and sensitive subgroups (pp. 174–175).
  6. Benefits and efficacy deserve as much scrutiny as risks. Includes justification, internalising costs and polluter pays (pp. 175–177).
  7. Alternatives and diversity: lock-in, substitution and clean production (p. 177).
  8. Lay and local knowledge, subject to scrutiny (pp. 177–178).
  9. Values and social interests: public intuitions “may sometimes prove quite robust” relative to the framing of regulatory science (p. 178).
  10. Independence from interested parties: information dependence; independent information institutions (pp. 178–180).
  11. Institutional obstacles: short-termism, transitions, inter-departmental and inter-level friction, agencies defending past decisions, national scepticism, supranational constraints (pp. 180–181).
  12. Avoid paralysis by analysis: act on reasonable grounds for concern. The prospective/retrospective distinction; thresholds depend on the distribution of costs and benefits (pp. 181–182).

Headline empirical conclusions: - Warning-to-action gaps were long, and unequivocal precautionary action was scarce even after the 1970s–80s (p. 168). - In many cases the information existed but was not delivered or was discounted (p. 168). - The acceptance of precaution varies by institution on both sides of the Atlantic, rather than by continent (p. 168).

Recommendations and advocacy, beyond what the cases show#

These are mostly in 16.3 and flagged by the editors as going beyond the cases (p. 169): - Use explicit, graded and potentially asymmetric levels of proof (Box 16.1, p. 170; Table 16.1, p. 184). Correct the scientific bias against Type II errors (p. 184). - Put prior framing questions (what science should ask, what counts as good science) to public deliberation (p. 186). - Involve stakeholders from the start, not only at risk management (p. 186). Participation should be early, broad and local, with safeguards against sponsor capture (p. 188). - Integrate innovation and regulation through CTA, TOA and alternatives assessment (pp. 186–187). Promote diversity as insurance (p. 187). - Evolve existing tools (MCM, LCA, CBA, scenarios, IEA) rather than replace them. The editors call this “Evolution, not revolution” (pp. 187–188). - Create independent information institutions with rights and resources. Move advisory committees out of producer directorates (pp. 179–180). - Internalise costs via polluter pays and liability (pp. 176–177). - Pursue education, public awareness and more responsible media communication of complexity and ignorance (pp. 188–189). - Enlarge ethical responsibility to cover “unknowns, which are predictable in principle even though not in specifics” (p. 189). Seek a cultural shift towards civil responsibility (p. 189). - Claims about innovation: precaution can channel rather than stop innovation, foster alternatives and confer competitive advantage (p. 182). These are asserted with examples, not evidenced here. - Claims about public understanding: publics correctly distinguish uncertainty from ignorance, and reassurance-by-research is futile (p. 185). These rest on two GMO studies, one listed with an editor as first author (Wynne et al., 2000) and one US FDA focus-group report (Levy and Derby, 2000). - The editors’ own caveats: the lessons are “illustrative, rather than definitive” (p. 169). One should not over-rely on a single set of prescriptions, and choosing responses is “an essentially political business” (p. 183). Precaution does not produce one “correct” outcome (p. 186).

Mechanisms and dynamics#

The analytical account below is supported by this chapter’s text. Page references are to Ch16. The underlying evidence sits in the case chapters.

How warnings arose and how they were handled#

Contestation and suppression of warnings#

Mental models of proponents, experts and regulators#

Industry, regulators and conflicts of interest#

Law, courts and trade#

Economics: costs, incentives and who bears harm#

Lock-in, path dependence, substitutes and alternatives#

Time lags, irreversibility and scale#

Complexity and systemic dynamics#

Institutional behaviour and culture#

Lay knowledge, publics and media#

Framing and language#

Many terms in the chapter function as frames: - “reassurance” (BSE, p. 172) and “safety” (over-confidence in antimicrobial safety, p. 171; manufacturers’ burden to show “safety” under Swedish law, Table 16.1, p. 184); - “controlled use” (asbestos, p. 175); - “closed systems” (PCBs, p. 174); - “realistic” and “achievable” (Southwood, p. 179); - “unnecessary burden on industry” (p. 180); - “sound science” (US framing of GMO assessment, p. 183); - “paralysis by analysis” (p. 181); - “not so likely to occur (in future)” (asbestos, p. 173).

The chapter’s own counter-frames are “humility” versus “hubris” (p. 170), “pretence at knowledge” (p. 185) and “shooting the messenger” (p. 179). Its central reframing: the democratic demand for circumspection “was not about innovation and risk themselves, but about our ability to know” (p. 185).

Innovation effects#

Transferable insights (technology-neutral)#

Strength ratings: - Strong: well supported across several cases in the report and consistent with wider evidence. - Moderate: supported by several examples but with gaps or hindsight risk. - Suggestive: a plausible pattern with thin or single-case support in this chapter. - Asserted: a normative or theoretical claim not evidenced here.

Ch16 summarises. The underlying evidence sits in the case chapters, so ratings reflect how well the pattern is supported across the cases Ch16 cites.

  1. A recurring failure is not missing knowledge but knowledge that exists somewhere and either does not reach the decision point (“institutional ignorance”, p. 171) or reaches it and is discounted (p. 168). Evidence: pp. 168, 171, 179; benzene 1897, asbestos 1898, PCBs 1930s, DES 1953, Swann 1960s, halocarbons 1974. Strong. The chapter says institutional ignorance is exemplified in “most of the case studies” (p. 171). Two caveats: it also finds genuine society-wide ignorance in “many” cases (p. 171), so this is not the only failure mode; and hindsight makes “available” knowledge look clearer than it was.

  2. Uncertainty comes in kinds: risk, uncertainty and ignorance. Tools valid for one (probabilistic risk assessment) are inadequate for the others (“too narrow in scope”, Box 16.1), and ignorance guarantees surprise. Evidence: Box 16.1 (p. 170); pp. 169–170 (CFCs, DES, TBT). Strong as a conceptual framework (well established in decision theory). Moderate as a causal explanation of the historical failures.

  3. When the specific harm cannot be predicted, properties that raise the cost of being wrong can serve as proxies for caution. These are novelty, persistence, mobility or ubiquity, accumulation, irreversibility and global scale with “only one experimental model”. Evidence: pp. 170–171, 183 (Swedish policy). Moderate. It is well illustrated for persistent chemicals and logically general. The trade-off that the same property can be a safety virtue (CFC inertness, p. 170) is not addressed by the chapter; this is the note-taker’s point.

  4. Reassurance is only as good as the search behind it. “No evidence of harm” is often produced by not looking, or by studies too small to detect harm. Evidence: BSE (p. 172); low statistical power and exposure misclassification (p. 184). Strong. BSE is well documented and the statistical logic is uncontroversial.

  5. Absent deliberate monitoring and baselines, the implicit assumption that harms will announce themselves in time fails for slow, diffuse or delayed harms. Evidence: asbestos, benzene and PCBs (p. 172); TBT baselines (p. 172). Moderate-strong. Several cases, and the logic fits long latency. The ozone case cuts both ways (see bias check).

  6. Questions identified early as critical to a reassuring position can go unresearched for years. Appraisals should name the question, timeline, funder, independence, and whether action waits for the answer. Evidence: BSE, 1986 to 1989 and 1996 (p. 172); Swann to 1990s (p. 172); procedure (p. 173). Moderate. Clear in BSE and antimicrobials; thinner elsewhere. The chapter does not claim delay is worse when the answer is inconvenient, though it notes antimicrobial research “that might have been expected to reinforce a critical position was delayed” (p. 179).

  7. More research does not necessarily reduce uncertainty. In complex systems it can reveal more complexity. Broadening an effects-to-causes (retrospective) inquiry can weaken the case against any single cause, while broadening a cause-to-effects (prospective) inquiry widens the net. Evidence: pp. 172–173, 181 (fisheries model, Great Lakes). Moderate. Two cases; the distinction is analytically sound and even-handed.

  8. Whichever discipline first “owns” a problem frames it, and effects outside its lens (chronic versus acute, ecological versus health, human versus veterinary, water versus air) are systematically neglected. Evidence: p. 174 (asbestos, radiation, MTBE, SO2, hormones, antimicrobials, BSE). Moderate-strong. Seven cases illustrate it, in both directions.

  9. Appraisals assume ideal performance, but real use involves leakage, poor installation and maintenance, misuse, deliberate non-compliance, smuggling, sensitive subgroups and expanding applications. “Controlled use” assumptions often fail in practice. Evidence: pp. 174–175; WTO 2001 (p. 175). Strong. Many cases, plus an independent legal finding (specific to asbestos).

  10. Belief that a new practice has fixed a historic problem is itself a recurring blind spot. Incremental improvements reset the clock on long-latency harm, and substitutes that share the problematic property reproduce the risk. Evidence: asbestos claims from 1906; fibre substitutes (pp. 173–174); HCFCs (p. 177). Moderate-strong.

  11. Fixes that remove a visible problem can shift harm to less visible scales or media, and the fix’s success can suppress attention to the displaced harm. Evidence: tall stacks (pp. 174, 176); MTBE as a lead substitute (p. 176). Moderate.

  12. Benefit and efficacy claims deserve the same scrutiny as risk claims, including whether the thing works and whether each use is justified. Prior “justification” requirements are rare and, where they exist, only partly effective. Evidence: DES efficacy (p. 176); ICRP justification (p. 176); even-handed framing (pp. 175–176). Moderate. DES is clear. Radiation shows an institutional model, but the chapter says that even under it justification “remains open to question” and many X-rays are “of doubtful clinical use” (p. 176).

  13. When harms are not priced, hazardous incumbents gain an unwarranted market advantage and substitutes are delayed. Liability and cost-internalisation are the corrective, though controversial. Evidence: pp. 176–177 (asbestos, halocarbons, PCBs). Suggestive. The economics is sound but the “technically superior substitutes” claim is not demonstrated in this chapter.

  14. Early commitment triggers self-reinforcing institutional and market processes (lock-in). Which option wins can be arbitrary, so alternatives must be assessed before commitment. Evidence: pp. 177, 186–187 (MTBE; CTA/TOA theory). Suggestive/asserted. It rests on imported theory with one illustration.

  15. Diversity is insurance against ignorance: surprises are smaller when needs are met by several technologies than by a single near-monopoly. Evidence: pp. 171, 187 (asbestos, halocarbons, PCBs). Suggestive. The logic is compelling but no counterfactual is shown, and diversity has costs not discussed.

  16. People closest to operations and exposures (workers, users, neighbours, harvesters, farmers) often hold earlier or truer knowledge of real practice and emerging harm. That knowledge needs channels and scrutiny, and has its own fallacies. Evidence: pp. 177–178 (asbestos, PCBs, Love Canal, fishers, Swedish farmers, BSE slaughterhouses; pensioners’ party fallacy). Moderate. Several examples, and the chapter is notably balanced.

  17. Public intuitions of unease about situations far outside normal experience can be a rational response to uncertainty, and sometimes anticipated real hazards. Evidence: BSE offal revulsion; antimicrobial misgivings (p. 178). Suggestive. The examples are selected where intuition proved right, with no counter-cases.

  18. Relying on hazard information produced and controlled by the party whose product is assessed is a structural weakness. Independent information capacity is necessary but not sufficient. Evidence: pp. 179–180. Moderate. Several cases, but the editors concede not all show distortion.

  19. Hazard evidence can meet low-visibility interpretative contest (“can be accompanied by”, p. 179), weak but persistent “refutations”, and sometimes suppression or discrediting of messengers. Evidence: p. 179 (asbestos, benzene, PCBs, acid rain, Great Lakes, BSE, sardines). Moderate. Well documented in some case chapters; second-hand here, and attribution of motive varies.

  20. Advisers self-censor towards what seems politically “realistic”, so scientific and political judgements get fused before they reach decision-makers. Evidence: Southwood 1988 (p. 179). Suggestive. One instance. The chapter cites the BSE Inquiry (Phillips et al., 2000) only in general terms (p. 183), not as a source for this episode; check against the BSE chapter notes.

  21. Institutional design drives delay. The drivers are promoter-regulator dual mandates, political transitions, friction between departments and levels, and agencies defending past positions for credibility. Evidence: pp. 179–181 (BSE especially; sardine fishery; DES). Moderate-strong for BSE; the others are single illustrations.

  22. The same evidence yields different decisions in different institutions and cultures. Outcomes are not determined by evidence alone. Evidence: p. 181 (US versus UK scrapie; DES timing), p. 180 (Norway versus UK); p. 168 (institutions, not continents). Moderate.

  23. Whether a demand for more evidence is prudence or obstruction depends on who bears the costs of acting and of not acting. Uneven distribution across groups, countries or time makes consensus on action harder. Evidence: p. 182; p. 176 and p. 180 (UK and Scandinavia). Suggestive-moderate. One strong example plus theory.

  24. Standards and burdens of proof are design choices with distributional consequences. They can be explicit, graded, and set differently for different parties. Evidence: Box 16.1 (p. 170); Table 16.1 (p. 184). Strong as a descriptive point. The normative choice remains contested.

  25. Conventional inference guards against false alarms far more than missed harms. Low-power monitoring and “negative” studies generate false reassurance. Evidence: p. 184 (Underwood; HELCOM; Copeland). Strong as a statistical point. The policy implication, that significance conventions drive regulatory outcomes, is moderate, since regulators need not use those thresholds.

  26. Institutions that present provisional knowledge as certain, and respond to public concern by researching only known uncertainties, erode trust. Public concern centres on unacknowledged ignorance and on purposes and interests (“who benefits”). Evidence: p. 185. Suggestive. It rests on two GMO focus-group studies, one of them co-authored by an editor (Wynne et al., 2000), and on the editors’ own theoretical work.

  27. Responsibility tends to stop at the edge of scientific foresight, leaving surprises that are foreseeable in principle unowned. Evidence: p. 189. Asserted. A normative and philosophical claim.

  28. Engagement must come at the framing stage to matter. Late consultation touches only marginal issues, and participatory forms can be captured by sponsors. Evidence: pp. 186, 188. Asserted/suggestive. Drawn from wider literature, not the cases.

  29. Restricting one pathway can redirect rather than stop innovation. Evidence: p. 182. Asserted/suggestive. Examples are named but no data are given.

  30. Where early action is costly and uncertainty high, systems default to limited action and incur larger later costs that were often foreseen. Evidence: BSE and cod (p. 176). Moderate. Two well-documented cases.

  31. Short political and business horizons can discount long-latency harms (they “can militate against social welfare in the medium and long term”, p. 180). Evidence: p. 180 (asbestos, benzene, PCBs). Suggestive. Stated with examples but not detailed here.

  32. Transboundary and commons harms need institutions at the matching scale. National or supranational rules can block a more cautious jurisdiction, and smuggling or non-compliance undermines global controls. Evidence: pp. 175, 180–181 (TBT and the IMO, Sweden and the EU, halocarbon smuggling). Moderate.

  33. A known harm can be a clue to less obvious ones. Readily seen acute effects often came before chronic harms, sometimes by decades, but the relationship is asymmetric: the absence of acute signals is no reassurance. Harm in non-human populations can act as a “sentinel” for human harm. Evidence: p. 171 (acute before chronic for sulphur dioxide, radiation, benzene, asbestos, TBT, PCBs; chronic without acute warning for BSE and halocarbons). Moderate for the acute-then-chronic pattern (six cases named, and the chapter itself warns it is not “a general rule”). Suggestive for the sentinel role, for which Ch16 names no case.

Limitations, contestation and bias check#

Advocacy versus analysis#

Case selection and hindsight#

Risk-risk trade-offs#

Internal tensions#

Factual and dating issues to verify against case chapters and primary sources#

Points where the chapter is fairer than its reputation#

Counter-arguments not engaged#

Hindsight pointers (for the hindsight strand; from general knowledge, to verify at source)#

Notable quotes#

  1. “What remained neglected, however, was the virtual certainty that there would be factors that remained outside the scope of the risk assessment.” (p. 169)
  2. “If persistence and bioaccumulation are used as screening for eliminating potential hazards, then the size and seriousness of future ‘surprises’ are likely to be smaller.” (p. 171)
  3. “Presumably the assumption was made that if there were harmful effects, evidence would emerge of its own accord and in good time for corrective action.” (p. 172)
  4. “a classic example of ‘no evidence of harm’ being misinterpreted as ‘evidence of no harm’” (p. 172)
  5. “once a technological commitment is made, a host of institutional and market processes act to reinforce its position, even if markedly inferior to potential alternatives” (p. 177)
  6. “Regulatory appraisal frequently fails due to the dependence of risk assessment on information produced and owned by the very actors whose products are being assessed.” (p. 179)
  7. “If the culture is not prepared to act even when there is demonstrable evidence of cause and effect, then appeals to precaution are unlikely to succeed.” (p. 181)
  8. “In effect not being wrong is more important than being safe.” (p. 184)
  9. “its growing innovative powers were increasingly outrunning its capacity to anticipate the consequences” (p. 185)
  10. “Any possible future consequence which lies beyond existing scientific knowledge and predictability is deemed by definition to be beyond responsibility.” (p. 189)

Reserve quotes: - “the contradiction between intrinsic provisionality and pretended certainty often goes unacknowledged — but not, it appears, unnoticed” (p. 185) - “there is an enormous difference between the discouraging of a particular innovatory pathway, and the channelling of innovation into alternative routes” (p. 182) - “‘Shooting the messenger’ has been a typical response to those bearing disturbing news ever since Galileo” (p. 179) - “can potentially degenerate into little more than a form of consultation driven by the sponsor’s agenda” (p. 188) - “may do so for arbitrary reasons which have nothing to do with intrinsic qualities, and everything to do with chance and first-leader advantage” (p. 187)

Open questions#

  1. Selection effects. How would the lessons change if the case set included warnings that proved false, or precautionary actions with net costs? Compare LL2’s false-alarms chapter. Which lessons survive a balanced sample? Candidates are information dependence, real-world conditions, and absence of evidence. Which are weakened? Candidates are “act on reasonable grounds” and trust in public intuition.
  2. Operationalising “ignorance”. The proxies (persistence, novelty, irreversibility, scale) are concrete for chemicals. What are their technology-neutral equivalents, and how do we handle cases where the proxy property is also the safety feature (inertness)?
  3. Ratchet or symmetry? What evidential standard should lift a precautionary restriction under the retrospective scenario (p. 181)? The chapter does not say how uncertainty over an agent gets “resolved”.
  4. Costs of action versus inaction. Given the admitted inability to answer Q3 (p. 168), what can be said, in hindsight and case by case, about whether earlier action would have been net beneficial?
  5. Independent information institutions. Did the post-2001 independent agencies reduce information dependence in practice, or does dependence on applicant-generated data persist within them?
  6. Participation. Did early, “upstream” participation deliver the epistemic gains the chapter claims, or mainly legitimacy? How often did participatory processes suffer the sponsor capture the chapter warns about (p. 188)?
  7. Innovation effects. Is there systematic evidence, beyond anecdote, that curtailing a hazardous option “channels” innovation and confers competitive advantage (p. 182)? Under what conditions does it instead just delay benefits?
  8. Culture versus evidence. The same evidence led to different decisions (p. 181). What institutional features predicted the more precautionary response (for example USDA versus MAFF): mandate, independence, legal standard of proof?
  9. Monitoring design. How should low-power monitoring (p. 184) be designed or pooled so that “negative” results carry information? Did regulators later adopt power requirements or equivalence testing?
  10. Chronology reliability. How many of Ch16’s compressed dates (Swann, PCBs 1899, DES 1974, halocarbon “neglect”) survive checking against the case chapters? Should the project quote Ch16’s chronology only after verification?
  11. Distribution and thresholds. Can the insight that evidence thresholds track who bears costs (p. 182) be turned into a diagnostic? For example: map who bears action costs and who bears harm costs, and predict where demands for more evidence will arise.

Audit log#

Independent audit against the full text extract (pp. 168–191), with front matter (pp. 1, 3–8), Ch 17 (pp. 192–194) and biographies (pp. 195–199) checked in the PDF. Table 16.1 re-verified with an independent layout-mode extraction. No contemporary-technology references found.