Law, Probability and Risk (2006) 5, 19−31 Advance Access publication on August 8, 2006

doi:10.1093/lpr/mgl010

The precautionary principle: a new approach to public decision-making? K ATIE S TEELE† Department of Philosophy, The University of Queensland, Australia [Received on 22 January 2006; revised on 10 July 2006; accepted on 10 July 2006] The precautionary principle recommends preventing possible harm to human health and environment. It has gained support in the international community as a higher-order legal principle that should guide public policy and the formulation of specific laws. But it is also the target of much criticism, with many arguing that the principle is vacuous, inconsistent or based on an excessively conservative attitude towards risk. I argue that the value of the precautionary principle is that it emphasises aspects of good decision-making that go beyond the scope of formal decision theory, and are often neglected in practice. It is best conceived as providing guidelines for formulating a decision problem, as opposed to challenging standard decision rules. To this effect, the principle advocates assessment of acts relative to feasible alternatives, and proper representation of all potential outcomes of acts (including those outcomes that are probabilistic or scientifically uncertain). In terms of determining suitable outcome utilities and burdens of proof, I claim that the precautionary principle appeals to ethical ideals associated with ‘sustainable development’. I finally outline some general implications of the principle for environmental decision-making. Keywords: precautionary principle; risk; uncertainty; decision-making; sustainable development.

1. Introduction The ‘precautionary principle’ (hereafter referred to as the PP) requires that public policy include measures to avoid or diminish morally unacceptable harms that may result from human actions. The harms need not be certain outcomes of an action; it is sufficient that they be scientifically plausible.1 There are different views in the literature about what this general maxim amounts to. Of course, cataloguing social harms is not straightforward, and there is disagreement about the proper balance between particular types of goods and harms, especially in cases where there is much uncertainty. Some versions of the PP make claims in this respect that are unwittingly controversial by the lights of formal decision theory.2 Such interpretations of the PP only serve to fuel criticism that it † Email: [email protected] 1 This is similar to the proposed definition of the PP in COMEST’s (2005, p. 14) report. 2 Resnik (1987, Preface) calls decision theory ‘the collection of mathematical, logical and philosophical theories of decision-making by rational individuals—taken alone, in competition, and in groups’. I use ‘formal decision theory’ more narrowly to refer to formal choice rules that identify what options are admissible from a set of possible options. In Section 6, I draw attention to some debates within decision theory (in the area of individual decision-making). For the most part, however, I associate ‘formal decision theory’ with the dominant ‘normative’ decision model—expected utility theory. According to this model, we should choose the option that maximises expected utility, where the expected utility of an option is calculated by multiplying the utility of each possible outcome by its probability, and then summing these terms. Various axiomatisations of expected utility theory relate this cardinal representation to an ordinal preference ranking satisfying a few key conditions. See c The Author [2006]. Published by Oxford University Press. All rights reserved.


20

K. STEELE

is ‘unprincipled’ or ‘unscientific’.3 The PP literature in large part, however, contains many valuable recommendations with regard to good decision-making. These positive contributions indicate that the value of the PP lies in the focus it gives to decision-modelling issues that arise in the course of applying formal decision theory, chiefly to the public policy domain. The PP makes recommendations about the initial specification of a decision problem, an important element of decision-making that takes place prior to the application of specific decision rules. With respect to formulating a decision problem, the PP demands a comprehensive survey of the space of possibilities, and evaluation of outcomes and thus acts based on sound ethical principles. There are a few major proposals to this effect, which I will discuss in turn: • Acts should be evaluated relative to feasible alternatives, and all potential outcomes of acts (including those outcomes that are probabilistic or scientifically uncertain) should be acknowledged. • Different types of scientific uncertainty should be represented appropriately in a decision model. • When it comes to evaluating act outcomes, harm to individual persons is just one source of disutility. Damage to public goods and harm to future generations are also important. • Burdens of proof with respect to proving that an action is permissible should where feasible reflect the idea of ‘polluter pays’ or the right not to be harmed by others. In terms of ethical stance, I claim that the PP is aligned with the principle of sustainable development. It therefore inherits the (by no means trivial) problems of interpretation and application of the sustainable development ideal. Given the ambitious ideals associated with the PP, some might claim that it is too abstract to be of any use to public policy, even when the principle is interpreted modestly as advice on problem specification (rather than as a new way of managing risk).4 But if the PP is properly conceived as a higher-order legal principle, then ambiguities with respect to its application are to be expected. Such principles are necessarily abstract, because their purpose is to guide and provide justification for a broad range of more concrete laws and policy measures. According to the COMEST (2005) report on the precautionary principle, legal principles of this kind first provide a standard that ‘allows evaluation of the validity of a law’, second they ‘assist in the interpretation of laws’ and third they ‘have the capacity to fill legal gaps’.5 In this vein, the PP is an overarching rationale for public policy. It is a decision procedure intended to guide the development of laws that constrain the activities of public and independent actors in the interests of human and environmental health. 2. Surveying the decision space The PP emphasises the importance of comprehensively modelling a decision situation. This involves considering not just a singular act of interest but rather the full set of feasible alternatives. Tickner Resnik (1987, pp. 88–96) for an outline of one such axiomatisation (Von Neumann and Morgenstern’s) of expected utility theory. 3 Sandin et al. (2002, p. 288) cite a number of criticisms of the PP, including the claims that it leads to increased risk-taking and that it marginalises the role of science. (They then go on to defend an interpretation of the PP that does not have these weaknesses.) 4 Starr (2003, p. 1), for instance, claims that ‘an analytic basis to support [the PP’s] verification and predictability as a principle does not exist.’ 5 Refer to the COMEST (2005) report for an overview of references to the PP in other international agreements.

THE PRECAUTIONARY PRINCIPLE

21

(2003, p. 274) refers to this assessment of alternatives as ‘the heart of the solutions-oriented approach of the PP and central to sound, forward-looking environmental decision-making’. It involves considering different strategies and technologies that might be employed to realise a particular goal, rather than confining attention to some pre-established way of doing things. In terms of policy recommendations, this aspect of the PP might amount to a requirement on public agencies to survey and assess a range of strategies for carrying out a particular social service. Consideration of the set of feasible alternatives moreover affects whether the actions of an independent agent should be restricted given a certain level of risk to human health and environment (Tickner, 2003). For example, if the introduction of novel crops to a particular region is considered the only viable strategy for producing food, more lenience is justified with regard to environmental risks and liability for damage. If a number of low-risk agricultural options are otherwise available, however, the pursuit of risky alternatives should be less tolerated. Given any particular course of action, the PP calls for acknowledgement of all potential outcomes, even those that are scientifically uncertain. In this regard, Resnik (2003), Keeney & von Winterfeldt (2001) and Sandin et al. (2002) all claim that PP appeals to a difference between practical decision-making and epistemological decision-making. In practical decision settings, all potentially significant consequences of an action must be taken into account, not just those that are established as scientific fact. Epistemological decision-making, on the other hand, could be understood as giving credence only to hypotheses that are shown to be highly probable in light of the relevant data/statistical tests.6 To give an example, theoretical speculation might predict a certain chemical to have potentially harmful effects on a particular ecosystem, and if so, this should affect public decisions regarding the use of the chemical, even before the precise causal interactions are established through scientific experiment.7 This basic point is expressed in the Rio Declaration, which is considered by many to be an early statement of the PP in international law.8 It refers to decisions about whether to undertake preventative action, given incomplete scientific knowledge about whether a current practice is environmentally damaging: Scientific uncertainty shall not be used as a reason to postpone action given appropriate cost-benefit analysis. It may seem obvious that we should not ignore potential consequences of an action just because they are not known with certainty. In public decision scenarios, however, uncertain negative consequences can be overlooked, given the interests of the parties involved and the misconception that objectivity requires certainty. The idea that not just scientifically proven but also merely plausible scenarios should be taken into account in decision-making invites discussion as to what kinds of possibilities should count as plausible. Consider the so-called ‘catastrophe principle’ which maintains that if the probability of a catastrophic event (an event with extremely high disutility) is nonzero then we should act now to 6 This distinction is a bit crude. Indeed, Rudner (1998) denies any distinction between practical and epistemological/theoretical decision-making. Moreover, hypotheses with low probability at a given time are also important for theoretical science as new data may raise their plausibility. In support of the practical/theoretical distinction, however, we could say that scientists have an extra obligation to communicate speculative hypotheses that are relevant to public decision-making, even if such ideas would not normally receive much prominence in the course of theoretical work. 7 Adams (2002, p. 308) considers the PP to be particularly relevant ‘where a causal link to effects is unclear’. 8 The Rio Declaration was determined at the 1992 United Nations Conference on Environment and Development (UNCED) (see Adams, 2002, p. 306).

22

K. STEELE

prevent it. Many theorists, including Manson (2002), rightly dismiss this simplistic interpretation of the PP, as it is impossible to prove that any event, catastrophic or otherwise, in any situation, has nonzero probability.9 In effect, the catastrophe principle does not discriminate between different courses of action, because all actions involve some chance of having catastrophic consequences. Furthermore, any remedial action that is performed to avert a catastrophe will itself involve some chance of causing catastrophe. As Marchant (2001) points out, we would not be able to distinguish one strategy from another, because all could be potentially disastrous. If the PP is to make a useful contribution to decision-making, it must advocate discretion in terms of which potential outcomes of an action should be taken into account. Even without assuming the ‘maxi-min’ decision calculus10 endorsed by the catastrophe principle, the incorporation of far-fetched possibilities within a decision framework, in the name of representing scientific uncertainty, does not make for efficient decisionmaking. Criteria must be negotiated for determining what are plausible threats associated with an action. Given a particular type of harm, one technique that is employed is the ‘de minimis’ rule. According to this rule, a minimum probability level is stipulated for a certain type of harm (with an agreed disutility value), and if the actual probability of the harm is calculated to be below this probability level, then the harm is disregarded in the decision process (Sandin et al., 2002, p. 292). Jasanoff’s (1991) criticism of this rule is that it is too simplistic and thus prone to irresponsible application, as it depends on how different harms are identified—the addition of many small and thus neglected risks might end up amounting to a large risk. This problem can be overcome, however, by situating any particular course of action within the broader field of actions that have similar consequences. For example, if only one person was to drive a leaded-petrol car then presumably this would present negligible threat to human and environmental health. But if the entire population was to drive leadedpetrol cars then the dangers would be much more significant, and would justify placing restrictions on each individual case. Decision analysis should be used to determine what level of risk to the community is acceptable for any particular type of activity, and thus what threshold probabilities are relevant to individual cases. The de minimis rule may be useful for standard types of problems that reference considerable statistical data,11 but it does not really help in cases where the threats involved and the manner of detecting them are less conventional. Resnik (2003) provides a more holistic set of criteria that can be used to determine whether particular consequences of an action are plausible. He appeals to criteria that are commonly employed in science to determine the status of an untested hypothesis, or even a general law or theory; criteria such as coherence, explanatory power, analogy, precedence, precision and simplicity. For example, while it is not completely impossible that the sun will not rise in the east tomorrow morning, this sort of event does not cohere well with the rest of our scientific theories. It lacks all the criteria cited above for determining whether a hypothesis is well grounded. Such a scenario should not, therefore, figure in any of our decisions regarding what should be done today!

9 The exceptions are events that are logically impossible. 10 Maxi-min is a strategy for making decisions under uncertainty that attends to the worst possible outcome of an action.

It selects the action that has the best worst-outcome. Resnik (1987, pp. 26–27) gives a formal outline of the maxi-min rule. 11 Examples include toxicology cases in which the task is to determine whether a chemical is present at harmful levels in the environment.

THE PRECAUTIONARY PRINCIPLE

23

3. Representing scientific uncertainty Connected with the task of acknowledging any scientific uncertainty involved in a decision, is the more demanding project of representing uncertainty in a comprehensive manner. It is reasonable to assume that the more informatively scientific uncertainty can be characterised, the better the decision process. The PP, in line with general trends in risk assessment, demands better reporting from scientists (at least those scientists whose work affects policy-making), of the uncertainty that is associated with their findings. Not only does this encourage a standard of reporting that is more informative but it also promotes a self-critical attitude among scientists, in terms of the judgements and practices that are incorporated in their work.12 Lynn (1986) notes that inadequate reporting of uncertainty can mask error or unwarranted bias in scientific findings. Scientists cannot avoid value judgements altogether13 (not least because values like simplicity and explanatory power are part of normal scientific practice), but these sorts of judgements are better made publicly, and with greater awareness. There are some straightforward ways to improve the communication of scientific uncertainty relevant to public decisions. Mayo (1991, pp. 267–273) comments on the appropriate way to report results from classical statistical tests used in, say, risk assessments. (Typical cases are tests to ascertain whether a chemical is present at toxic levels in the environment.) Mayo claims that negative Neyman–Pearson-type statistical results are often misinterpreted. This is because investigators fail to attend to the frequency that a test will yield a type II error, i.e. that it will advise accepting the null hypothesis when the null is in fact false, and an alternative hypothesis is true. (Frequency of type II error is indicated by the power of the test.)14 Mayo refers to a mortality study of Du Pont workers as illustration. The study reported that the workers did not have a significantly higher cancer mortality rate (a negative result) relative to the general populace. However, the power of the test was such that it had a very small chance of rejecting the null hypothesis (zero increase in mortality rate) even if the true increase in mortality rate was substantially greater than zero. In other words, the result was bound to be negative. Even more useful for decision-modelling purposes is to report a confidence interval for the parameter of interest, in this case the increase in mortality rate. Confidence intervals show the range of uncertainty associated with experimental findings, as opposed to a ready-made ‘yes’ or ‘no’ decision regarding whether a hypothesised effect is significant.15 Informative reporting of statistical results is particularly relevant to the PP given the attention it draws to the relevance of value judgements in risk assessment. The level of acceptable risk associated with actions is a social decision that should be made by policy-makers rather than scientists; thus there should be a demand for scientists to communicate statistical findings in a manner that does not make prior judgement on this issue. Some risks (such as exposure to toxins) can be quantified through empirical testing and communicated via established practices of statistical reporting. Various other types of uncertainty, however, are more difficult to characterise. Regan et al. (2002) make a distinction between linguistic 12 For this reason, Levidow (2003, p. 116) speaks favourably about references to the PP in genetic engineering debates; he claims that attention to scientific uncertainty encourages scientists to sharpen their critical approach and engage more fully with questions of methodology. 13 Notably, naive positivists claim that science should not involve value judgements, but I agree with many others including Shrader-Frechette (1991) who suggest that this view is indeed naive. 14 I further discuss the relationship between type I and type II errors in Section 5. 15 I note that the Bayesian framework arguably provides a sounder basis for statistical inference, including the specification of confidence intervals, than does classical statistics. The framework facilitates comparison of (the probabilities of) a range of hypotheses, and makes explicit the subjective aspects of assessing hypotheses in light of new evidence. Howson & Urbach (1989) argue extensively for the advantages of Bayesian over classical statistics.

24

K. STEELE

and epistemic sources of uncertainty, where epistemic sources include traditional categories such as measurement error and model uncertainty, and linguistic sources include vagueness and ambiguity associated with natural language. Proposed ways of reporting and correcting these various types of uncertainty are markedly different. Indeed, it is suggested that not all types of uncertainty can be represented probabilistically16 which presents a further challenge in terms of how alternative types of uncertainty should be handled in decision-making. This is an ongoing problem for those working in formal decision theory. In any case, probabilistic risk estimates should not be based solely on analysis of frequency data if there exists other (perhaps more elusive but nonetheless significant) sources of uncertainty. 4. Considerations of justice With regard to ethical stance, a number of sources emphasise the connection between the PP and the principle of sustainable development. The two principles are linked in both a conceptual and a historical manner: they have a shared ethical vision that features concern for public goods and the rights of future persons, and have both been developed chiefly in response to problems of environmental degradation. O’Riordan & Jordan (1995, p. 200) claim the PP to be ‘the political basis of sustainability’. Paul Stein (2000, p. 7) gives the following analysis: The Precautionary Principle needs to be considered in the broader context of the wider principles and philosophies forming the concept of ecologically sustainable development (ESD) . . . In essence, ESD is development which aims to conserve and effectively manage the environment for the benefit of future generations. The COMEST (2005, p. 20) report on the PP states that the PP is ‘directly related to the principle of sustainable development’. This is to say that both intra- and intergenerational equity is a central concern, and by implication, environmental health. As per the principle of sustainable development, the PP could be said to advocate extension of the liberal ‘no harm’ principle to take account of damage to public goods and harms to future generations, as well as traditional harms to individual persons. These justice issues are receiving increasing attention, given awareness of the serious long-term effects that some groups in particular have on the shared environment. As discussed, the PP emphasises the fact that uncertain as well as certain consequences of proposed courses of action should be taken into account. This is a strong commitment to environmental well-being and intergenerational equity. It advocates placing legal restrictions on the activities of independent actors where there is a significant ‘chance’ of harm to community interests. It is worth pointing out that there are problems with interpreting these lofty ethical ideals. How to suitably respect future interests is not a trivial question. Even for a single individual, it is not easy to work out, let alone maintain, a desired balance between long-term and short-term goals. To what extent should I ensure that my short-term actions do not diminish my future prospects? In the context of social decision-making, these kinds of questions are even more difficult to answer. People have differing attitudes towards the relative importance of their immediate well-being versus their well-being at future times. Moreover, there are differing attitudes towards risk and the kinds 16 Regan et al. (2002, p. 624) make this claim in passing. See Colyvan (in press) for a more detailed argument as to why

not all types of uncertainty can be represented probabilistically.

THE PRECAUTIONARY PRINCIPLE

25

of gambles involving future well-being that are permissible.17 The issue of intergenerational justice is yet a further challenge, because it involves thinking about others who do not yet exist, and who thus cannot voice preferences. In this case, added to the uncertainty about future consequences of an action, is uncertainty about what will actually constitute future well-being. There is a significant literature on sustainable development that engages questions regarding intergenerational justice.18 There remains, however, considerable divergence of views about the topic; ‘sustainable development’ is variously interpreted, as it is sensitive to different conceptions of what is owed to future generations, and how these obligations can be fulfilled. Versions of sustainable development that prescribe the preservation of specific resources or productive capabilities are generally considered to be problematic because they prejudice the question of what will constitute future well-being. Better conceptions of the principle include the (more general) goal of ‘nondeclining well-being’, or the stipulation that the present generation should ensure future generations have the means to satisfy their needs, whatever these needs might be. Weiss (1992) argues that this involves conservation of the options available to future generations (through maintenance of the ‘diversity’ and ‘quality’ of the natural and cultural resource base) and conservation of rights of access to social goods via institutions that promote equality of opportunity. As can be expected, such broad definitions of intergenerational justice invite continual controversy regarding how they should be interpreted and applied. Although difficult to resolve, such debates have ongoing relevance to the PP. 5. Liability and burdens of proof In connection with a commitment to maintaining public goods for present and future generations, the PP apportions significant responsibility to producers, or those who stand to gain from potential environmental damage.19 A typical scenario is whether a producer should be allowed to market a product that potentially harms humans and/or their environment. If the consequences of allowing production to go ahead are probabilistic, then we should use decision-theoretic reasoning to determine whether the action is acceptable, i.e. there should be a requirement for the expected utility of the action to be greater than some benchmark. (Some compensation may need to be provided to losing parties so that the costs and benefits are distributed fairly.) In this way, a suitable decision model can be devised to determine whether an action is acceptable, but there remains a further important issue regarding who holds the responsibility for establishing this result, whether before the action in question is undertaken, or when disputes arise about its effects. This is an extra justice consideration that turns on whether particular types of action should be initially presumed harmful or otherwise safe. If it is thought that the producer (the party who stands to gain/the potential polluter) should be considered ‘guilty until proven innocent’, then they are effectively required to ‘do the research’ that is required to show (via a decision model) that the benefits of their proposed action sufficiently outweigh the costs. 17 Hollander (1991) notes that whether a risk is tolerable depends not only on cultural values but also on the extent to which it is voluntarily accepted. 18 Wolf (2003) gives a good overview of the problem of intergenerational justice and the question of sustainable development. Weiss (1992) also surveys the various proposals regarding what is owed to future generations. Her own proposal places importance on the preservation of biological and cultural diversity, as well as equitable institutions. 19 This is made explicit in the COMEST (2005) report on the PP, in the sections entitled ‘Liability and burden of proof’ (p. 24) and ‘Culpable ignorance and the PP’ (p. 18).

26

K. STEELE

We can rephrase the above in terms of determining the appropriate balance between type I and type II errors when testing whether a practice is safe. The standing assumption or null hypothesis might be that the practice is safe and the alternative hypothesis that it is not safe. Then a type I error would be to assert that the practice is unsafe when it is in fact safe. A type II error would be to assert that the product is safe when it is in fact not safe. Experimental design typically involves minimising the possibility of a type I error, but this is an arbitrary convention and has many implicit value judgements regarding the choice of null hypothesis and the social costs of making a type I versus a type II error (Howson & Urbach, 1989). Particularly when the results have practical significance, it is more appropriate to employ expected utility analysis to determine what is the acceptable balance between type I and type II errors, if a hypothesis test is to serve as the locus of decision (Malinas, 1984). Moreover, as stated above, the PP recommends that the responsibility for proving that there is an acceptable balance between type I and type II errors belongs (to the extent feasible) to the producer, or the party who stands to gain from the transaction. Regulations may be instituted that prohibit producers from engaging in a particular risky activity in the first place. If these regulations are transgressed the producer will be liable through fault. Otherwise, producers may be strictly liable for damages to human health and environment. (The producer need not be at fault; it is sufficient that their activity caused the damage in question.) This rationale is referred to as polluter pays. It is implicit in COMEST’s (2005, p. 20) statement that ‘the costs of human-made risks should not be externalised, neither to the local environment nor to the environment of other societies or nations’. In ‘a White Paper on environmental liability’, the European Commission Directorate-General for the Environment (2000) discusses the viability of requiring producers to compensate for environmental damage. The general rationale of polluter pays is supported in the European Commission White Paper, but due to problems of transparency and insurance cover, the recommendation is that ‘liability is only effective where polluters can be identified, damage is quantifiable and a causal connection can be shown between damage and polluter’. Despite the fact that only a limited form of environmental liability appears feasible, this would still be a major step towards realising the ethical standards of the PP. The issue of who holds the ‘burden of proof’ again arises in the context of environmental liability. Discussion of burdens of proof is clearly not new; it is rather an integral part of legal decisionmaking. The legal system involves standards for the kind and amount of evidence that should be demanded before particular legal remedies are provided. Some cases call for evidence ‘beyond all reasonable doubt’, while others allow a lesser degree of certainty, such as a ‘preponderance of evidence’. There are also stated legal presumptions against which evidence must be stacked. The PP emphasises that this rationale should be extended to public goods, and it generally recommends a shift in burden of proof towards producers.20 A thorough extension of legal practice to the realm of social goods would involve establishing appropriate connections between the three relevant categories identified by Manson (2002)—type of harm, level of certainty/uncertainty permitted regarding occurrence of the harm (and what is the legal presumption) and suitable remedy. 6. Implications for decision rules Given that its origins are in public policy, it is arguably not the role of the PP to adjudicate on debated issues in formal decision theory. (There have been some very misguided suggestions for 20 Cranor (2001) and Weiss (2003) argue that existing legal standards should inform discussion about environmental

liability.

THE PRECAUTIONARY PRINCIPLE

27

decision rules in the PP literature, such as the above-mentioned catastrophe principle.) In any case, expected utility theory and its variants have provisions for handling uncertainty, and can reflect the values of society, whatever these values turn out to be.21 I claim that the PP is better conceived as emphasising important aspects of formulating a decision problem that must be settled prior to the application of specific choice rules. As well as promoting the assessment of options relative to all feasible alternatives and better reporting and handling of scientific uncertainty, the PP endorses care about the future and ethical priorities derived from the principle of sustainable development. I should note, however, that some consider the PP to contribute to that part of formal decision theory concerned with decision-making under considerable uncertainty or ignorance. In such cases, we cannot assign precise probabilities (or utilities) to the possible outcomes of actions; there is thus not enough information available to calculate precise expected utilities. For example, if there is little information available about the characteristics of a particular ecosystem, then it may not be possible to determine a single ranking of conservation management options based on precise expected utility estimates. One plausible probability distribution (representing, say, extinction risk of a particular species) might make a specific management option optimal while another plausible probability distribution makes a different option optimal.22 There are various suggested methods within formal decision theory for deciding what to do in these cases, including (among others) the maxi-min rule, the ‘maxi-max rule’, the ‘mini-max regret rule’ or the ‘principle of insufficient reason’.23 Each of these rules responds to the risk associated with uncertainty (about probabilities or utilities) in a distinct way; e.g. maxi-min reasoning is concerned with the worst possible consequence (or worst possible expected utility if it is a case of partial ignorance) of each action, and selects the action with the most optimal worst-case outcome. This is clearly a pessimistic strategy, and it is particularly unattractive when an alternative action offers the opportunity of much larger gains for only a small decrease in worst-case utility. The other decision rules, however, also have their problems. It seems indefensibly optimistic to base public choices on the best possible outcome of options (as per the maxi-max rule). The mini-max regret rule has the significant weakness that it makes the preference relation between two options depend on what other options are in the set.24 And the principle of insufficient reason also suffers from the devastating criticism that it makes the ranking of options sensitive to the way the state/outcome space is partitioned into apparently equi-probable states/outcomes, as illustrated by Bertrand’s paradox (H´ajek, 2003, pp. 187–188).

21 I note that there are a number of suggested variants to expected utility theory. For example, Machina (1991) discusses various decision rules that relax the Independence axiom of expected utility theory. Such models have been developed largely as a response to observed choice behaviour that violates Independence but is not obviously irrational, such as that exposed by Allais’ (1953) paradox. (The Independence axiom holds that preference between options should not depend on what happens in circumstances where the two yield identical outcomes. This precludes an agent’s preferences being sensitive to the variance as well as the expectation of an option’s outcome distribution. The fact that many reasonable agents’ preferences are affected by this kind of risk consideration provides some reason to relax Independence.) Other decision models accommodate different kinds of risk considerations, such as the ‘regret theories’ of Bell (1982) and Loomes & Sugden (1982). 22 Margules & Pressey (2000) give prominence to this kind of decision-theoretic approach in conservation planning. (They claim that the utility of conservation management options depends on an appropriate measure of biodiversity persistence.) 23 These decision rules are outlined in Resnik’s (1987) introductory decision theory book. Resnik assumes complete ignorance about the probabilities of outcomes, but the rules can be adjusted to apply to cases of partial ignorance as well. 24 Resnik (1987, p. 31) notes this criticism of the mini-max regret rule. In other words, the mini-max regret rule violates the Independence of Irrelevant Alternatives condition, as outlined by Sen (1979, p. 17).

28

K. STEELE

Resnik (2003) gives an interesting account of the PP: he presents it as an alternative approach to decision-making under ignorance, superior to the maxi-min or other established rules. Resnik suggests a way to expand the set of options under consideration in a decision situation in response to maxi-min-type considerations. For example, if I would be inclined not to go bushwalking because I want to avoid the worst-case scenario of getting lost and spending the night very tired and hungry, then perhaps I should expand my option set to include the action ‘go bushwalking with emergency food. (If I get lost with the emergency food I will at least not be so hungry and so the utility of the worst-case outcome is increased.) This is a constructive suggestion with respect to decisionmaking under ignorance, but it is not a genuine alternative to the maxi-min rule. Resnik’s strategy is effectively to augment the option set, and so it is another example of the PP providing guidance about formulation of a decision problem. If it says anything about decision rules, it is in favour of maxi-min, contrary to Resnik’s claim. Relatively recent work on handling imprecise or indeterminate probabilities (and utilities) gives a more credible basis to discussion about decision-making ‘under ignorance’, or partial ignorance, as the case may be. Walley (1991) gives a comprehensive account of the mechanics of imprecise probabilities. Levi (1986), Seidenfeld (1983) and Seidenfeld et al. (1995), among others, investigate choice rules that accommodate indeterminate (imprecise) probabilities/utilities. They argue that whatever secondary choice rule is favoured (be it maxi-min, maxi-max or some other alternative),25 only those options that maximise expected utility for ‘some’ plausible probability and utility distributionpair make for admissible choices. Even granted this primary criterion, there remains scope for disagreement about what is the appropriate secondary choice rule.26 It is not clear that the most suitable decision rule for negotiating ignorance will be distinctly cautionary or risk-averse. But this is not at odds with the ‘precautionary principle’, because, as I have argued, the focus of the PP is rather the set-up or formulation of a decision problem. 7. Special case—the environment I have been discussing the PP in general terms, but many consider it to be specifically directed towards environmental decision-making. This is to be expected, since the principle initially emerged from concerns about treatment of the environment. Given the complexity and dynamics of ecological systems, the scientific issues encapsulated by the PP—tracking uncertainty and long-term consequences—are particularly pertinent to environmental management decisions. Furthermore, I have stressed the link between the PP and the principle of sustainable development, and the latter gives prominence to the significant impact that ‘the environment’ has on the health and happiness of the present and future generations. Many people also consider the natural environment, and specifically biodiversity, to be intrinsically valuable. (Either way, there is likely to be much agreement about conservation priorities.) In any case, there is good reason for the PP to focus on environmental conservation. The extent to which this value should be prioritised above other conflicting human values (perhaps even cultural diversity) is of course a point of contention. Soule (2000) distinguishes between ‘weak’ and ‘strong’ versions of the PP. The ‘strong’ version of the PP is given this label because it privileges the 25 Levi (1986), in particular, argues for a lexical choice rule. 26 There is also scope for more fundamental disagreement. For example, Ben-Haim (2001) offers yet another decision

calculus (‘info-gap theory’) that is intended for cases in which there is ‘extreme’ ignorance about decision parameters.

THE PRECAUTIONARY PRINCIPLE

29

goal of maintaining sustainable ecosystems above all others. Weak versions weigh up environmental concerns with economic and other types of social factors; the assumption is that environmental quality can be legitimately balanced against these other goods. I find the weak/strong categories somewhat misleading, because a principle is not weak or strong in any absolute sense, but only relative to particular concerns—in this case, the natural environment. The strong version is perhaps too heavy-handed with respect to values—clearly human interests are diverse, and not limited to the preservation of biodiversity. Given that economic goods have in the past been given undue priority, however, it is not surprising that many conservationists now stress the absolute importance of environmental protection. Biodiversity is not only considered by many to be intrinsically valuable but also essential (at some critical level) for basic human well-being and thus all other social goods, so it is surely wise to give environmental conservation a high priority. Local damage to the environment is assessed in various ways, but it generally refers to loss of ecosystem integrity and biodiversity. As discussed earlier, where there is highly uncertain but serious environmental damage at stake, production activities may need to be restricted from the outset. Holding producers strictly liable for damage already done is appropriate only in certain settings. The European Commission Directorate-General for the Environment’s (2000) White Paper on environmental liability distinguishes between damage to biodiversity and more tangible environmental contamination (due to, say, an oil spill or discharge of agricultural or factory chemicals). Damage to biodiversity is the more difficult to assess, and it is not easy to determine what kind of legal remedies are ‘practical’, due to the tricky balancing of environmental and other human values (for instance economic values). The European Commission White Paper recommends that costs of restoration be proportional to level of biodiversity damage. Full restoration of ecosystems following damage is the ideal, but this may not always be feasible given the costs involved. In some cases, it may be more sensible for the ‘polluter’ to compensate by protecting some other area with appropriate biodiversity value. Thus, while the PP advocates polluter pays with respect to environmental damage, economic and other considerations must inform how this maxim is fleshed out in practice. 8. Conclusion The precautionary principle has received much recent exposure in public debate, particularly in relation to international environmental management issues such as climate change and the adoption of modern biotechnology products. It has been introduced (explicitly in international contexts) as a higher-order legal principle that is intended to guide the formulation of more specific laws and public policy. In both policy and academic settings, however, both the scope and content of the principle have been widely interpreted, often in a biased manner to promote a particular agenda. The PP has been criticised for being excessively risk-averse or unscientific. If we consider versions of the principle proposing decision-making procedures that are inadvertently at odds with expected utility theory, then such criticisms are valid. But I have argued that better versions of the PP complement rather than compete with established formal choice rules. I claim that the PP is best conceived as providing broad guidelines for formulating or specifying a decision problem, particularly in the public policy context. I have discussed its major contributions in this respect. To begin with, the PP demands thorough surveying of the decision space. This includes considering acts alongside all feasible alternatives, and representing both the certain and uncertain outcomes associated with each act. Second, the PP promotes an ethical outlook consistent with the ideal of sustainable development. This means that intra- and intergenerational equity issues should

30

K. STEELE

significantly influence the evaluation of act outcomes. Environmental conservation is accordingly given much emphasis. Finally, to make these ethical commitments more concrete, the PP advises a shift in responsibility for damage to public goods towards potential polluters. While these are all broad and rather abstract recommendations, this is fitting to a higher-order legal principle. It is the ongoing task of legal practitioners and scholars alike to determine how the precautionary principle should be applied in more specific contexts. Acknowledgement Many thanks to Mark Colyvan, William Grey, Julian Lamont, an anonymous referee and Jay Koehler, an editor of this journal, for very helpful and detailed comments on earlier drafts of this paper. While doing this research, the author has also benefited from a formal affiliation with the Australian Research Council Centre of Excellence for Integrated Legume Research. R EFERENCES A DAMS , M. D. (2002) The precautionary principle and the rhetoric behind it. Journal of Risk Research, 5, 301–316. A LLAIS , M. (1953) Fondements d’une th´eorie positive des choix comportant un risque et critique des postulats et axioms de l’´ecole americaine. Econom´etrie, 40, 257–332. B ELL , D. E. (1982) Regret in decision making under uncertainty. Operations Research, 30, 961–981. B EN -H AIM , Y. (2001) Information-gap decision theory. San Diego, CA: Academic Press. C OLYVAN , M. Is probability the only coherent approach to uncertainty? Risk Analysis (in press). C RANOR , C. F. (2001) Learning from the law to address uncertainty in the precautionary principle. Science and Engineering Ethics, 7, 313–326. E UROPEAN C OMMISSION D IRECTORATE -G ENERAL FOR THE E NVIRONMENT (2000) White Paper on Environmental Liability. Luxembourg: European Commission. ´ , A. (2003) Conditional probability is the very guide of life. Probability Is the Very Guide of Life: H AJEK The Philosophical Uses of Chance (H. E. Kyburg, Jr. & M. Thalos eds). Chicago, IL: Open Court, pp. 183–203. H OLLANDER , R. D. (1991) Expert claims and social decisions: science, politics, and responsibility. Acceptable Evidence: Science and Values in Risk Management (D. G. Mayo & R. D. Hollander eds). New York: Oxford University Press, pp. 160–173. H OWSON , C. & U RBACH , P. (1989) Scientific Reasoning: The Bayesian Approach. La Salle, IL: Open Court. JASANOFF , S. (1991) Acceptable evidence in a pluralistic society. Acceptable Evidence: Science and Values in Risk Management (D. G. Mayo & R. D. Hollander eds). New York: Oxford University Press, pp. 29–47. K EENEY, R. L. & VON W INTERFELDT, D. (2001) Appraising the precautionary principle—a decision analysis perspective. Journal of Risk Research, 4, 191–202. L EVI , I. (1986) Hard Choices: Decision Making under Unresolved Conflict. Cambridge, New York: Cambridge University Press. L EVIDOW, L. (2003) Precautionary risk assessment of Bt maize: what uncertainties. Journal of Invertebrate Pathology, 83, 113–117. L OOMES , G. & S UGDEN , R. (1982) Regret theory: an alternative theory of rational choice under uncertainty. The Economic Journal, 92, 805–824. LYNN , F. M. (1986) The interplay of science and values in assessing and regulating environmental risks. Science, Technology and Human Values, 11, 40–50.

THE PRECAUTIONARY PRINCIPLE

31

M ACHINA , M. J. (1991) Dynamic consistency and non-expected utility. Foundations of Decision Theory (M. Bacharach & S. Hurley eds). Oxford, MA: Basil Blackwell, pp. 39–91. M ALINAS , G. A. (1984) Pesticides and policies. Journal of Applied Philosophy, 1, 123–131. M ANSON , N. A. (2002) Formulating the precautionary principle. Environmental Ethics, 24, 263–274. M ARCHANT, G. (2001) The precautionary principle: an “unprincipled” approach to biotechnology regulation. Journal of Risk Research, 4, 143–157. M ARGULES , C. & P RESSEY, R. (2000) Systematic conservation planning. Nature, 405, 243–253. M AYO , D. G. (1991) Sociological versus metascientific views of risk assessment. Acceptable Evidence: Science and Values in Risk Management (D. G. Mayo & R. D. Hollander eds). New York: Oxford University Press, pp. 249–279. O’R IORDAN , T. & J ORDAN , A. (1995) The precautionary principle in contemporary environmental politics. Environmental Values, 4, 191–212. R EGAN , H., C OLYVAN , M. & B URGMAN , M. A. (2002) A taxonomy and treatment of uncertainty for ecology and conservation biology. Ecological Applications, 12, 618–628. R ESNIK , D. B. (2003) Is the precautionary principle unscientific? Studies in History and Philosophy of Biological and Biomedical Sciences, 34C, 329–344. R ESNIK , M. D. (1987) Choices: An Introduction to Decision Theory. Minneapolis, MN: University of Minnesota Press. RUDNER , R. (1998) The scientist qua scientist makes value judgements. Introductory Readings in the Philosophy of Science (E. D. Klemke, R. Hollinger & D. W. Rudge eds). New York: Prometheus Books, pp. 492–498. S ANDIN , P., P ETERSON , M., H ANSSON , S. O., RUD E´ N , C. & J UTHE , A. (2002) Five charges against the precautionary principle. Journal of Risk Research, 5, 287–299. S EIDENFELD , T. (1983) Decisions with indeterminate probabilities. The Behavioral and Brain Sciences, 6, 259–261. S EIDENFELD , T., S CHERVISH , M. J. & K ADANE , J. B. (1995) A representation of partially ordered preferences. The Annals of Statistics, 23, 2168–2217. S EN , A. K. (1979) Collective Choice and Social Welfare. Amsterdam, The Netherlands: North-Holland. S HRADER -F RECHETTE , K. (1991) Reductionist approaches to risk. Acceptable Evidence: Science and Values in Risk Management (D. G. Mayo & R. D. Hollander eds). New York: Oxford University Press, pp. 218–248. S OULE , E. (2000) Assessing the precautionary principle. Public Affairs Quarterly, 14, 309–328. S TARR , C. (2003) The precautionary principle versus risk analysis. Risk Analysis, 23, 1–3. S TEIN , P. (2000) Are decision-makers too cautious with the precautionary principle? Environmental Planning and Law Journal, 17, 3–23. T ICKNER , J. A. (2003) Precautionary assessment: a framework for integrating science, uncertainty, and preventive public policy. Precaution: Environmental Science and Preventive Public Policy (J. A. Tickner ed.). Washington, DC: Island Press, pp. 265–278. WALLEY, P. (1991) Statistical Reasoning with Imprecise Probabilities. London: Chapman & Hall. W EISS , C. (2003) Expressing scientific uncertainty. Law, Probability and Risk, 2, 25–46. W EISS , E. B. (1992) Intergenerational equity: a legal framework for global environmental change. Environmental Change and International Law (E. B. Weiss ed.). Tokyo, Japan: United Nations University Press, pp. 385–412. W OLF, C. (2003) Intergenerational justice. A Companion to Applied Ethics (R. G. Frey & C. H. Wellman eds). Malden, MA: Blackwell Publishing, pp. 279–294. W ORLD C OMMISSION ON THE E THICS OF S CIENTIFIC K NOWLEDGE AND T ECHNOLOGY (COMEST) (2005) The Precautionary Principle. Paris: UNESCO.