How Not to Criticize the Precautionary Principle

Abstract

The precautionary principle has its origins in debates about environmental policy, but is increasingly invoked in bioethical contexts. John Harris and S⊘ren Holm argue that the principle should be rejected as incoherent, irrational, and representing a fundamental threat to scientific advance and technological progress. This article argues that while there are problems with standard formulations of the principle, Harris and Holm's rejection of all its forms is mistaken. In particular, they focus on strong versions of the principle and fail to recognize that weaker forms, which may escape their criticisms, are both possible and advocated in the literature.

Keywords:

• ethics
• precautionary principle
• risk
• science
• uncertainty

I. INTRODUCTION

The precautionary principle has its origins in debates about environmental problems and the appropriate policy and regulatory responses to them. In this context the principle has been widely incorporated into international and domestic environmental law and regulation, although it remains controversial, with critics accusing it of excessive and paralyzing conservatism, vagueness, and incoherence. Increasingly, the precautionary principle has also been invoked within bioethics to address issues such as genetically modified foods, genetic therapies, and xenotransplantation. Writing in this journal, John Harris and S⊘ren Holm (2002, p. 357) refer to the “fundamental threat” that this development “poses for scientific advance and technological progress,” and argue that in the bioethical context and elsewhere the precautionary principle should be rejected. ¹

The purpose of this article is to examine and challenge some of the arguments and claims made by Harris and Holm in support of this conclusion. It is not my intention to argue that all of their criticisms are mistaken or that the precautionary principle is without problems. Indeed, I share with Harris and Holm the view that strong forms of the precautionary principle should be rejected. However, Harris and Holm claim that even weak forms of the precautionary principle are invalid. This, I believe, is a mistaken and dangerous conclusion, based on flawed arguments and a misperception of what a weak version of the precautionary principle would assert.

Part of the problem lies with the vagueness which Harris and Holm rightly detect in many formulations of the precautionary principle. This is seen by some supporters as a strength of the precautionary principle: like other widely discussed principles (“sustainability,” for example), the precautionary principle can act as a focus for political action in support of desirable policies, and its ability to be interpreted in different ways allows it to draw together a wider coalition of support. ² The value of such coalition-building presupposes, however, that the policies thus promoted are indeed desirable, and reduces the precautionary principle's status from a decision-making principle to a political slogan. To function as a bioethical principle that can help us to determine which policies ought to be promoted the precautionary principle needs to be specified with some precision. Reasonably precise specification is also necessary in order to assess the accusations of conservatism and incoherence etc. that are levelled against it.

Given the vagueness in many “official” formulations it is therefore necessary for critics to interpret the principle before criticizing it. Their interpretations, however, need to satisfy certain constraints: firstly, if the criticisms are to have any force against the principle that is actually advocated, and not merely against a “straw man,” the interpretations on which they are focused should be informed by the intentions of those who advocate the principle; secondly, if it is to be argued that no form of the precautionary principle is acceptable then the interpretations on which criticisms are focused should represent the range of reasonable interpretations of the precautionary principle. Unfortunately the interpretations of the precautionary principle adopted by Harris and Holm do not fully satisfy either of these requirements.

II INTERPRETING THE PRECAUTIONARY PRINCIPLE

Harris and Holm criticize the precautionary principle on a number of grounds. The core of their critique, however, is what they call the “paradox of precaution” (2002, p. 356). The alleged paradox is that while the precautionary principle instructs us to forego or delay the introduction of potentially risky new practices in order to prevent harm, that delay will itself cause harm in the form, for example, of lost opportunities to prevent disease and death. Thus, in order to prevent this harm the precautionary principle would instruct us to refrain from implementing itself. Whether this is an accurate description depends on how the precautionary principle is interpreted, ³ but even versions of the principle that are not strictly paradoxical may be vulnerable to the related but more general charge of excessive conservatism leading potentially to an increase rather than a reduction in harm (Harris & Holm, 2002, p. 357).

It is important to note, however, that different formulations of the precautionary principle exhibit different degrees of conservatism. Criticisms of the principle as excessively conservative must therefore either focus on one particular formulation, leaving open the possibility that other, weaker formulations may escape the criticism, or else must show that the charge of excessive conservatism applies even to the weakest, least conservative version of the principle.

Like many commentators on the precautionary principle, Harris and Holm take as their starting point the statements of the principle contained in various international conventions and treaties. In particular they quote the formulation contained in Principle 15 of the 1992 Rio Declaration on Environment and Development:

In order to protect the environment, the precautionary approach shall be widely applied by States according to their capabilities. Where there are threats of serious or irreversible damage, lack of scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation. (quoted in Harris & Holm, 2002, p. 357)

Similar statements, Harris and Holm note, are contained in other treaties and conventions, including the Treaty Establishing the European Community (p. 358), but for a somewhat contrasting view they quote the following account of the precautionary principle, issued by participants at the 1998 Wingspread conference:

When an activity raises threats of harm to human health or the environment precautionary measures should be taken even if some cause and effect relationships are not fully established scientifically …. In this context the proponent of an activity, rather than the public, should bear the burden of proof. (quoted in Harris & Holm, 2002, p. 358)

This pair of quotations captures what are often taken to be the key elements of the precautionary principle. Common to both is the view that action should be taken to counter certain kinds of risk in advance of scientific certainty about its existence or seriousness, while the Wingspread statement adds the view that in respect of activities generating the relevant types of risk it falls on the proponents of he activity to prove that it is safe rather than the regulators to prove that it is risky. Harris and Holm observe correctly that the first of these elements (action in advance of scientific certainty) is at the core of most formulations of the precautionary principle, while the second (shifting the burden of proof) is less widely accepted. The account of the precautionary principle that they derive from these quotations is, however, problematic.

The problem with Harris and Holm's formulation of the precautionary principle is that it is both more specific and stronger than many of the widely cited formulations in the literature, including those that they quote. The Rio Declaration states that in the qualifying circumstances (what these are we will consider later) lack of scientific certainty “shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation.” This does not imply that preventive action must be taken on suspicion of risk, but merely rules out one particular ground for not taking such action, namely that the existence of the threat has not been demonstrated with the very high standard of evidence necessary for its acceptance as scientifically established fact. Construed in this way the precautionary principle is a response to those (commercial interests, for example) that would use lack of scientific certainty as grounds for arguing against regulatory interference.

This does not exclude there being other reasons to refrain from preventive action, and indeed one such reason—lack of cost-effectiveness—is specifically referred to in the Declaration. This statement of the precautionary principle is perhaps best viewed as establishing not an obligation but a permission to take action to counter risks in the absence of scientific certainty, so that uncertainty about the threat will not constitute a decisive objection to such actions. In contrast, the Wingspread statement does imply an obligation to take action in appropriate circumstances. However, it is less specific about the nature of the action to be taken, referring only to “precautionary measures” rather than the “preventive action” referred to in the Rio Declaration. ⁴

Harris and Holm's own formulation of the precautionary principle combines the strongest elements of Rio and Wingspread:

PP: When an activity raises threats of serious or irreversible harm to human health or the environment, precautionary measures which effectively prevent the possibility of harm (e.g., moratorium, prohibition, etc.) shall be taken even if the causal link between the activity and the possible harm has not been proven or the causal link is weak and the harm is unlikely to occur. (2002, p. 358, my emphasis)

This combines Wingspread's implication that regulators have an obligation to take action in the face of qualifying risks with Rio's implication that the appropriate action is preventive. (In fact Harris and Holm invoke a stronger notion of preventive action than is implicit in Rio because they omit the “cost-effective” qualifier.) Thus, in respect of the action to be taken, Harris and Holm's definition of the precautionary principle is stronger than either of the statements on which it purports to be based. This, as we shall see, is problematic for their argument.

III WEAK AND STRONG PRECAUTION

Harris and Holm claim to have presented a “fairly weak” formulation of the precautionary principle, and write that “if we can show that this weak formulation is untenable, we will a fortiori have shown that any stronger formulation is untenable” (2002, p. 358). This, however, is misleading. Firstly, their formulation may be untenable for reasons other than being too strong, and this would not have the stated implication. Secondly, we have already seen that Harris and Holm's formulation is, in one respect at least, stronger than those on which it is supposed to be based. Showing that this definition is unacceptably strong would not, therefore, be sufficient to discredit the precautionary principle more generally.

Although I have argued that Harris and Holm's formulation represents a strong rather than a weak version of the precautionary principle there is more to this claim than I have considered so far shown. My claim, above, was that in respect of the action to be taken Harris and Holm's formulation is stronger than either of those from which it is derived. The action to be taken, however, is only one of the respects in which formulations of the precautionary principle can vary in strength.

In general terms we can formulate the precautionary principle as a hypothetical imperative of the following form:

(i) If an activity satisfies trigger conditions T, then take action of type A.

From this we can see that the strength of any specific formulation will depend not only on the nature of the recommended action but also on the trigger conditions. If two formulations recommend the same types of action but vary in their trigger conditions then the formulation with the less demanding trigger conditions will define the stronger principle.

The trigger conditions themselves have several elements including the seriousness of the potential harm—which itself is a function of (at least) its severity for each individual and the number of individuals affected—and the probability, or strength of evidence, that the harm will happen. Thus, a more detailed general formula might be:

(ii) If there is evidence stronger than E that an activity will cause harm more serious than S, then take action of type A.

The trigger conditions for a particular formulation of the precautionary principle will be defined by values of E and S. It may be that the possible combinations are not fully ordered—that is, there may be cases in which we cannot say which of two formulations is stronger, where for example one requires a higher level of evidence to trigger the precautionary principle while the other requires the evidence to be of a more serious threat. ⁵ But, for pairs of formulations that differ in only one of these parameters, or where the parameters differ in the same direction, the stronger formulation will be that which has the lower trigger values.

The concept of strength and weakness, as applied to formulations of the precautionary principle, is, therefore, multi-dimensional: formulations can vary in strength not only in the respect of the action they recommend but in respect of the seriousness of potential harm and level of evidence required to trigger that action. The following sections will make use of this observation to show that Harris and Holm's arguments presuppose an interpretation of the precautionary principle that is strong in all of these dimensions.

IV SERIOUSNESS

In order to make Harris and Holm's formulation of the precautionary principle more precise, and to locate it on a scale of strength, it is necessary to consider what kinds of harms, and what degree of evidence about these harms, would satisfy its trigger conditions. The description of the trigger conditions as involving “serious” or “irreversible” harms is common in the literature but is, as Harris and Holm point out, problematic. They observe that a harm can be irreversible without being serious, and that such a harm would carry little moral weight. Seriousness must therefore be considered a necessary condition for triggering the precautionary principle. Further, Harris and Holm point out that a harm can be irreversible without being irremediable:

If you block your neighbour's driveway so that he has to take a taxi to work on a specific day, the harm that you have done is irreversible (because time is irreversible), but is not irremediable. If you pay for the taxi and compensate your neighbour to his satisfaction, the harm has arguably been remedied. (2002, p. 359)

Harris and Holm take this to show that if the precautionary principle is valid at all, it can only be valid “in cases where there is a risk of harm which is ‘serious and both irreversible and irremediable’” (p. 359). This, however, is mistaken. The argument for regarding the blocking of a neighbor's drive as an irreversible harm surely shows that any harm is irreversible (since all harms take place in time). The adjective therefore serves no narrowing function and should be dropped from the description of the trigger conditions, in favor of “irremediable,” which arguably expresses what those who refer to irreversibility really have in mind.

Moreover, it is not clear that irremediability should be a necessary part of the trigger conditions for the precautionary principle. For one thing even if a harm is remediable there may be doubt about whether remediation will actually take place. ⁶ To impose a harm that is in principle remediable but which in practice one has reason to believe will not be remediated may be just as weighty a matter, morally speaking, as to impose a harm that is irremediable. In cases where remediation does take place this can be thought of as a kind of retrospective consent (assuming that in order to count as remediation the compensation must be accepted as adequate by the victim), but while the expectation of retrospective consent may weaken the presumption against inflicting a harm it does not in all cases render it morally neutral. It is therefore conceivable that precautionary action could be justified by the threat of harms that while serious are not irremediable. In practice, however, it is the question of how serious a harm has to be to trigger precautionary action that is likely to distinguish different degrees of precaution. Harris and Holm appear to believe that whatever level of seriousness is adopted as a trigger (“PP-seriousness”), the precautionary principle will be unacceptably conservative.

If, on the one hand, a serious effect on one person is sufficient for something to be PP-serious, then the PP entails that the inventor of apple pie should have applied the PP, and let the first pie be the last, since there have been people who have choked on apple pie. If, contrariwise, a combined serious effect on health is sufficient for PP-seriousness, then the PP clearly rules out any further procreative acts resulting in pregnancy and child birth. And if, finally, it is sufficient that a harm is serious either at the individual or the group level, then the PP seems to rule out both motherhood and apple pie. (2002, p. 360)

It may seem a little pedantic to criticize examples that are so obviously chosen for comic effect rather than philosophical precision. However, as Harris and Holm do appear to regard them as having serious argumentative force I will list just some of the reasons why the precautionary principle does not (except under extremely strong interpretations) have the implausible consequences that they attribute to it. First, there is (as far as I know) no reason to think that the introduction of apple pie resulted in an increased number of deaths from choking (as opposed to causing different antecedently unidentifiable people to choke on different foodstuffs), so there is no evidence of a net harm that could trigger precautionary action. Second, while Harris and Holm postulate that a risk of even one additional death from eating apple pie would trigger the precautionary principle, such an interpretation would be orders of magnitude stronger than is implicit in the more usual applications of the precautionary principle to events such as nuclear winter, ozone depletion, global warming, zoonotic epidemics, and alterations to the human gene pool with the potential to cause many thousands of deaths and other serious harms.

Third, the risks from both apple pie and motherhood are in the main voluntarily entered into, and for this reason too are beyond the usual range of application of the precautionary principle. Extending the precautionary principle so as to prohibit the voluntary assumption of risks by individuals would result in an unusually strong form of the principle, which could be rejected without any implications for its more usual forms. Fourth, the willingness of people to engage in pie eating and procreation suggests that, on an individual level, the risks are outweighed by the advantages, and again this places the example beyond the range of circumstances to which the precautionary principle is usually applied. Fifth, the examples take no account of the societal benefits of the activities or costs of prohibition, which, as we have seen, the Rio definition allows to be taken into account when deciding on appropriate precautionary action.

In summary, while there is a fair point to be made about the lack of specification of PP-serious harm in standard formulations of the precautionary principle, we have seen that the examples used by Harris and Holm presuppose far weaker trigger conditions and therefore a far stronger precautionary principle than any of its adherents would advocate. The implausibility of recommending precautionary action on the basis of these trigger conditions therefore counts against Harris and Holm's interpretation of the precautionary principle, rather than against the principle itself.

V EVIDENCE

The other main element of the precautionary principle's trigger conditions is the level of evidence for the existence of a risk. Again there is vagueness in the standard formulations. The Rio and Wingspread statements are typical of many in calling for (or at least permitting) action to address risks to be taken in the absence of “scientific certainty,” but, as Harris and Holm note, the Wingspread formulation appears to go much further by advocating that the burden of proof be transferred from the critics to the proponents of a potentially harmful activity.

One obvious way to proceed at this point would be to consider various possible answers to the question: “how much evidence is needed in order to trigger precautionary action?” For reasons that are not entirely clear, however, Harris and Holm choose instead to address the evidential component of the precautionary principle by formulating what they call an epistemic version of the precautionary principle, with a very different structure from the versions so far considered:

E-PP: When an activity raises threats of serious and both irreversible and irremediable harm to human health or the environment, any evidence suggesting a causal link between the activity and the possible harm shall be given (much) greater weight, and the probability of that harm's occurring shall be assumed to be much higher than it would in other circumstances. (2002, p. 360)

This differs from the formulations we have considered so far in treating the strength of evidence of risk not as part of the trigger conditions for precautionary action, but as the focus of the precautionary action itself. Instead of formula (ii) above, we have:

(iii) If there is a possibility that an activity will cause harm more serious than S, then give more weight to the evidence for the risk than the evidence against.

This, Harris and Holm plausibly suggest, is irrational, since we normally believe that the weight given to each piece of evidence should be “a function of its epistemic warrant (i.e., the degree to which we have good reasons for believing this piece of evidence)” (p. 361), and “it is difficult to imagine any justification for an epistemic rule requiring a systematic discounting of evidence pointing in one direction, but not in the other” (p. 362).

Like Harris and Holm's putative examples of PP-serious harm discussed above, however, this epistemic formulation of the precautionary principle has the appearance of having been constructed with a view to easy demolition. I know of no advocate of the precautionary principle who proposes anything resembling this formulation. Douglas Weed, responding to Harris and Holm, argues that (in his specialist area of public health) discounting of evidence only takes place where the varying power of different studies justifies it, and that this can work against as well as in favour of a policy of precaution (Weed, 2004, pp. 319–320). Moreover, the way in which Harris and Holm formulate E-PP suggests that the purpose of the precautionary principle is to guide us in coming to a judgement about the reality of the relevant risks, whereas for many of its supporters the point of the principle is to guide our actions in precisely those cases where we lack the means of reaching a well-grounded judgment about whether the feared harm will materialize. As Stephen Haller puts it, “acting precautiously in response to some model that is predicting global catastrophe is a gamble, not a belief in the truth of that model” (2002, pp. 157–158).

The gap between E-PP and standard formulations of the precautionary principle is also shown by the fact that when Harris and Holm come to assess what they take to be the strongest version of the precautionary principle, the version that includes involves a shifting of the burden of proof, they quietly drop E-PP (despite having said that this is the only interpretation on which the shifting of the burden of proof can be a reasonable corollary to the precautionary principle) and revert to the more conventional interpretation according to which shifting the burden of proof specifies the evidential element of the principle's trigger conditions.

Here too, however, they attack a stronger version of the principle than is often intended by its proponents. The motivation for advocating precautionary action in the absence of scientific certainty is that science sets itself very high standards of evidence. Hypotheses are typically accepted only when the statistical likelihood of error is less than 5%. Such rigorous standards are appropriate in pure science, where the overriding consideration is to avoid the acceptance of falsehoods into the body of scientific knowledge, but would be inappropriate in many practical contexts. I would not, for example, go out without a coat on a day when heavy rain is forecast simply because my confidence in that forecast is 90% rather than 95%. In that case my aim is to keep myself reasonably comfortable, and since the burden of carrying a coat is small I will probably take it with me even if I think the forecast is quite likely to be wrong. In the case of regulating activities with the potential to cause serious harm, the costs of both action and inaction are higher and may need to be carefully weighed against one another. To refrain even from weighing up the risks from the activity against the costs of regulation because that we are less than 95% confident that the threat of harm is real would seem positively irresponsible (Hughes, 2003; Shrader-Frechette, 1991). The idea that regulatory standards of evidence should differ from those of pure science is sometimes likened to the law's requirement for different standards of evidence in civil and criminal cases because of the different significance of false positives and false negatives in each case (Cranor, 1999).

The general claim made by the precautionary principle is that it can be appropriate to address risks for which the evidence falls short of scientific certainty. The strongest versions (in the evidential dimension) would hold that a similar level of evidence for the absence of risk should be required before a potentially risky activity is permitted. Harris and Holm object that since it is “logically impossible to demonstrate the truth of general sentences” such as that “activity A will in no future instance where it is performed cause effect E which is PP-serious … asking for proof of non-harmfulness is … incoherent” (2002, pp. 360–361). This, however, assumes that what is demanded by the precautionary principle is a logically conclusive proof of safety, and once again they present no evidence that any advocate of the precautionary principle makes such a demand. In fact the rhetoric of shifting the burden of proof suggests that what is actually demanded by strong forms of the precautionary principle is “scientific” (e.g., 95%) levels of confidence in the safety of the activity to be undertaken. This is arguably an unreasonable demand, which would lead to irrational and excessively cautious decisions in some or all of the circumstances in which it is made, but it is not incoherent. Weaker versions of the principle may be not only coherent but also reasonable.

VI A PARALYZING DECISION RULE?

Returning explicitly to the more conventional interpretation of the precautionary principle as a decision rule or rule of choice (rather than an epistemic rule), Harris and Holm argue that adopting such a rule would leave us paralyzed. The metaphor of paralysis has been used elsewhere in the literature on the precautionary principle, but with a different meaning. Sunstein (2003), for example, uses it to refer to the idea (similar to Harris and Holm's idea of a “precautionary paradox”) that if applied symmetrically to action and inaction the precautionary principle will yield contradictory prescriptions to refrain from both. Harris and Holm, by contrast, use the concept of paralysis to refer to an aspect of what they see as the precautionary principle's excessive conservatism which in their view would lead to an “infinite regress of precaution,” making it almost impossible to introduce new technologies. Taking the case of genetically manipulated plants, Harris and Holm write that

… there is no doubt that the largest amount of uncertainty about their possible harmfulness to the environment and/or to human beings existed at the time when nobody had yet produced a GM-plant. There were theoretical models showing that harm might occur, but for very obvious reasons no empirical data to back up or dispute these models. The C-PP [i.e. the precautionary principle construed as a rule of choice] would have instructed us not to proceed any further and perhaps even to stop people thinking about going further … and the data to show whether or not the theoretical risks are real risks would never have been produced. (2002, p. 362)

The more general argument that this is intended to illustrate is that a similar regress would prevent the development of almost any new technology; thus, the precautionary principle should be rejected on the grounds of its excessive and paralyzing conservatism. If, however, we examine the assumptions implicit in their example we will see that once again Harris and Holm are attempting to refute the precautionary principle by pointing to problems that arise from one very strong interpretation of it.

Firstly, Harris and Holm's analysis assumes that the evidential threshold that must be reached in order for a risk to be taken seriously is extremely low. On their account precautionary action will be triggered by the existence of theoretical models showing that harm might occur, even if there is no empirical evidence at all to suggest that the models in question are accurate. Later they make the still stronger claim that the precautionary principle asks us to suspend the distinction between logical possibility and our ordinary concept of possibility, and to “act as if the mere fact that [certain harms] are logically possible also means that they are not only possible, but even likely to occur” (2002, p. 363). A version of the precautionary principle that mandated strong precautionary action on the basis of the logical possibility of harm would indeed be paralysing (in both Harris and Holm's and Sunstein's senses), and therefore implausible. To take this as a critique of the precautionary principle in general, however, is to ignore the many discussions of the precautionary principle that explicitly address the need for a reasonable evidential threshold, and also to ignore the possibility that this threshold might vary inversely with the size of the possible harms.

I will give just two examples. Cranor (2001, p. 321) distinguishes two possible formulations, one requiring “credible scientific evidence … appropriately verified” of a threat and the other requiring “a preponderance of evidence,” in order to trigger precautionary action. Both fall short of a requirement for “scientific certainty,” or the criminal law's requirement for “proof beyond reasonable doubt,” but require more than logical possibility or the existence of untested theoretical models. Resnik (2003, p. 336) argues that in order not to be “unscientific” (by which he means irrational), the precautionary principle must address threats that are “credible” or “plausible,” not “feigned, fanciful or disingenuous,” and proposes that plausibility of hypotheses be assessed on the basis of familiar epistemic criteria for theory acceptance including coherence, explanatory power, analogy, precedence, precision and simplicity (p. 337).

Secondly, the “paralysis” argument assumes a very low threshold in the magnitude of possible harm needed to trigger precautionary action. Consider again the GM plants example. The case for carrying out large scale trials before commercial growing, small-scale trials before large-scale trials, and laboratory experiments before field trials is, as Harris and Holm indicate, to acquire evidence relevant, among other things, to the safety of the subsequent step. However, a key reason for seeking this evidence in smaller-scale trials and laboratory experiments is that (excepting the most extreme speculations) the stakes are lower: less is risked in a small scale trial than a larger one in terms of financial cost, damage to wildlife, genetic “pollution,” or any of the other potential harms attributed to GM plants. Thus a moderate form of the precautionary principle might prohibit the large-scale growing of a new GM variety but permit smaller-scale trials and/or laboratory experiments on the basis that the potential harm from the latter falls below the threshold of “serious” harm needed to trigger precautionary action. As indicated above, this threshold might vary depending on the strength of evidence for or against the existence of the threat, and might also be relative to the expected benefits of the new crop variety. Existing accounts of the precautionary principle may be criticisable for not adequately defining where this threshold lies, but we have already noted that many applications of the precautionary principle (global warming, zoonotic epidemic, etc.) are consistent with a very high threshold, and so in the absence of a fully worked-out account we need not assume, as Harris and Holm do, that the threshold must be so low as to effectively halt technological progress.

Thirdly, while the foregoing considerations show that Harris and Holm's paralysis argument presupposes a version of the precautionary principle that is strong in respect of its trigger conditions, it also presupposes a version that is strong in respect of the precautionary action it prescribes. In their analysis of the GM plant example Harris and Holm make the assumption that the only response available to the precautionary principle would be to prohibit the growing of these plants. This is consistent with their own initial formulation of the precautionary principle, which refers to “precautionary measures which effectively prevent the possibility of harm (e.g., moratorium, prohibition, etc.)” (2002, p. 258). In making this assumption, however, Harris and Holm ignore both the possibility and the discussion in the literature of weaker versions of the precautionary principle that are not vulnerable (or at least have not been shown to be vulnerable) to their criticisms. In this case there is clearly conceptual space for a version of the precautionary principle which could, under conditions of uncertainty about the extent of a risk, mandate precautionary action to reduce the probability or magnitude of potential harm, but without prohibiting the technology altogether. The possibility of precautionary measures short of prohibition is also suggested by the reference to cost-effective measures in the Rio Declaration. In the case of GM plants such measures might include a requirement for refuges of non-GM plants (to prevent pests developing resistance to pesticides exuded by GM plants) and buffer zones (to prevent cross-pollination) between GM and non-GM plants. Such measures might not eliminate these risks altogether but might bring them within the thresholds discussed above. Or the risks may remain above the thresholds but the weaker measure be justified by having lower costs than measures that would “effectively prevent” the risks.

We see, then, that Harris and Holm's paralysis argument presupposes a version of the precautionary principle that is strong in respect of both trigger conditions and precautionary action. Weaker versions are possible and a version that was significantly weaker in either respect could avoid the paralysis that they describe.

VII THE PRECAUTIONARY PRINCIPLE AS A MORAL PRINCIPLE

In the last substantive part of their article Harris and Holm consider whether the precautionary principle could be a valid moral principle (as opposed to an epistemic rule or a rule of rational choice). The suggestion that there could be a valid moral rule that was not also a valid rule of rational choice (or at least one that failed to be a valid rule of rational choice in the ways that Harris and Holm maintain the precautionary principle fails) is puzzling, and it is unclear exactly how they view this distinction. Their formulation of the “moral precautionary principle” merely replaces the injunction that “precautionary action … shall be taken” with the statement that “the morally right action is to take precautionary measures” (2002, pp. 359, 363).

Harris and Holm suggest that if the precautionary principle is construed as a moral principle it would have to be a broadly consequentialist principle, though not a maximizing one. They argue that, viewed in this way, it is an implausible moral principle, being unlike any current form of consequentialism but akin to the discredited theory of negative utilitarianism in employing a restricted scale of value which ignores positive benefits and considers only harm. This, however, is at best true only of certain forms of the precautionary principle: those that are strong in the dimension of precautionary action, construing it as action to prevent harm rather than to limit the potential for harm proportionate to the activity's benefits. In fact the comparison between the precautionary principle and negative utilitarianism seems misguided: the asymmetry that undoubtedly exists in many interpretations of the precautionary principle consists not, it seems to me, in taking suffering more seriously than positive welfare (since the harms it seeks to avert can include the loss of positive welfare while among the most important of the benefits that critics accuse it of ignoring is the avoidance of suffering). Rather, it is a matter of taking harms caused by human activity more seriously than “natural” harms that human activity might prevent. There are two possible explanations of this: first, that adherents of the precautionary principle subscribe to an act omission distinction according to which it is a greater wrong to cause a harm by our actions than to allow a similar harm to happen. Second, adherents of the precautionary principle might subscribe to a belief in the beneficence or stability of nature, such that action which changes the natural state of things is more likely to have serious harmful consequences than inaction which leaves things as they are. Both views seem to me mistaken, but in addition the first would imply that if the precautionary principle is a fundamental moral principle it cannot be a consequentialist one, while the second would imply that it is not a fundamental moral principle but at best a lower-level principle intended to assist us in realizing our more fundamental values given certain beliefs about the world.

Harris and Holm do consider the possibility of viewing the precautionary principle as a lower-level principle within some form of indirect consequentialism, i.e., a “rule of thumb” used to make decisions when there is not the time to perform a full calculation of the consequences of our actions, or where the consequences of a mistaken decision are not serious enough justify the costs of undertaking such a calculation. The problem with this, Harris and Holm argue, is that the precautionary principle is usually invoked in relation to decisions that do have serious consequences and where there is not a lack of time for deliberation. There is, however, another possibility that Harris and Holm do not consider: the rule of thumb may be employed not because of the time required for a consequentialist calculation but because of its impossibility in situations where not only the outcomes of different courses of action, but also their probabilities, are unknown. Whether some version of the precautionary principle is the right rule to fill this gap is a matter for debate. However, consequentialists are clearly faced with the need for some decision procedure under such circumstances, and since others commonly proposed such as maximin are themselves problematic it is at least worth considering whether a moderate form of the precautionary principle might be an appropriate guide in certain circumstances.

The test, within a consequentialist approach, would be whether such a principle would, in the long term, produce better consequences than the alternatives. This would be hard to assess with precision but the way to assess it would be to look at the effect of adopting the principle in a range of realistic cases. The approach that Harris and Holm adopt in relation to motherhood and apple pie, and GM plants, is correct in this respect, except that a wider range of examples would need to be considered (including examples in which it appears that precaution has or would have paid off, for example, the phasing out of ozone-destroying CFCs, restrictions on asbestos use, and introduction of health warnings on tobacco), and in order to reach a general conclusion the examples would need to be analysed in relation to weaker as well as stronger interpretations of the precautionary principle.

VIII CONCLUSION

Harris and Holm directly or indirectly highlight a number of problems with the precautionary principle, which those seeking to apply it within bioethics need to take heed of. Many standard formulations lack precision and, in their stronger forms, have implausibly conservative implications. However, Harris and Holm fail to show that all forms of the precautionary principle should be rejected. Although they claim to focus on a weak version of the principle in the hope that arguments against it will apply a fortiori to stronger versions, we have seen that in fact strong versions of the principle feature both in their explicit formulations and in the implicit premises of their arguments.

We have identified several dimensions in which formulations of the precautionary principle can vary in strength. These include the strength of precautionary action recommended, severity of potential harm, number of people potentially harmed, and strength of evidence for the risk of harm. Harris and Holm's account does not give explicit recognition to these parameters and hence does not recognise the possibility of formulating accounts of the precautionary principle that are weaker in each of these dimensions than the versions they attack, and which may not be vulnerable to their criticisms. Nor do they acknowledge that weaker versions of the principle are widely discussed and advocated in the literature on precaution. They might reply that weaker accounts of the precautionary principle are not truly precautionary, but, as we have seen, there is a wide conceptual space between, for example, recommending precautionary action on the basis of risks suggested by untested theoretical models, and the stance often adopted by those with an interest in promoting technological development of refusing to take any such action until the probability of harm is established with scientific levels of certainty.

Harris and Holm frame their discussion as a response to Walter Glannon's (2002) argument against the widespread adoption of technologies to extend the human lifespan. Glannon, it is worth noting, does not refer to the precautionary principle, but Harris and Holm take his argument to entail such a principle, and then present their own construction of the principle, which they argue to be flawed. It is possible (though I have not considered this) that Glannon's argument does presuppose the strong form of the precautionary principle at which Harris and Holm's arguments are directed. But in that case they have shown merely that Glannon's arguments rest on an unacceptably strong form of the precautionary principle, not that the precautionary principle in general must be rejected.

Although the primary aim of this article has been to critique Harris and Holm's argument, it may be useful to consider some more positive conclusions regarding the interpretation and role of the principle that may be drawn out of this discussion.

First, while Harris and Holm's arguments do not apply to all forms of the precautionary principle, they do show that there is reason to reject very strong forms of the principle. In some cases the versions of the principle vulnerable to Harris and Holm's arguments (for example the “motherhood and apple pie” examples, and the suggestion that logical possibility of harm is sufficient to justify strong precautionary action) can be dismissed as straw men, being stronger than anything that appears in the literature or is implied by legal and regulatory formulations. In other cases, however, their arguments target versions of the precautionary principle that, although strong, may be seriously advocated by some of its supporters. For example, versions of the principle that always require very strong evidence of the safety of a practice before allowing it to go ahead (as the Wingspread formulation does under some interpretations, and as are encountered in public debates on issues such as genetically modified crops) are likely to be vulnerable to the “precautionary paradox” argument (since there will be many cases where we do not have this level of evidence for the absence of harmful consequences either in the event of permitting a practice or in the event of prohibiting or restricting it) and to charges of excessive conservatism. This suggests that if there is an acceptable form of the precautionary principle it must be weaker than this in relation to the epistemic aspect of its trigger conditions.

Others who put forward similar arguments to these (e.g. Sunstein, 2003) suggest that any such weakening would render the principle trivial. This, however, is false, since, as we have seen, even a considerably weaker principle would serve as a necessary corrective to the view that restrictions on innovation cannot be justified until evidence for the associated risks reaches the level constitutive of “scientific certainty.”

Second, these arguments assume—correctly in my view—that regulators and legislators should be concerned not only with harms resulting from the introduction of new medical and other technologies, but also with the harms (including opportunity costs) resulting from prohibiting or restricting them. This does not mean, however, that decisions to permit or restrict a technology on grounds of potential harm should always be subject to the same evidential thresholds. In defining the trigger conditions for precautionary action, the evidential threshold should depend on both the severity of the potential harm that the precautionary action is intended to avoid (taking account of the numbers liable to be affected), and the potential harms resulting from the precautionary action itself. Harris and Holm focus on examples where the former are relatively modest (at least in comparison with the global disasters that often form the context in which the precautionary principle is invoked) and the latter are very significant. In these cases the evidential threshold will be high. However, these arguments are also consistent with a strong presumption against a new technology, and a correspondingly low threshold for precaution, in cases where the potential harms are very serious and the costs of precaution low. Even high threshold cases, however, may be regarded as precautionary in comparison with the libertarian approach that rejects restrictions on innovation in the absence of scientific certainty about potential harms.

Notes

1. Harris and Holm have also criticized the precautionary principle in a letter published in Nature (Harris & Holm, 1999), which has been widely cited by opponents of the principle. The letter does not add anything to the arguments in the article discussed here, but the attention it has received makes it all the more necessary to subject their arguments to careful examination.

2. The idea of “A Simply Political Precautionary Principle” is discussed in Gardiner (2006, pp. 39–40).

3. See, in particular, the discussion below of the connection between the precautionary principle and the act omission distinction. If advocates of the principle hold that we have a stronger obligation to avoid causing harm by our actions than to prevent harm from occurring by our omissions then the paradox may be avoided, albeit at the cost of having to defend the act-omission distinction.

4. “Preventive action” is also referred to in Treaty Establishing the European Community (quoted by Harris & Holm, 2002, p. 238).

5. The lack of complete ordering may also apply within the parameter S, for example where one account of S specifies a larger number of people affected while the other specifies a greater harm for each individual. The quasi-mathematical presentation of the trigger conditions should therefore be considered as an indicative guide only.

6. An analogous criticism often made of cost-benefit analysis is that while its application ensures that approved projects will generate enough benefit to compensate for the costs that they impose, it is a separate matter of policy to determine whether those who bear the costs will in fact be compensated.