I. INTRODUCTION
How should we regulate socially productive activities that impose some risk of harm to others? The standard answer to that question in the modern administrative state has been some form of cost/benefit calculus: permit those activities in which the expected social benefits of the risky conduct exceed its expected social costs.
There is a vast and growing literature criticizing this practice on normative, conceptual, and administrative grounds. I focus here on the central objection registered in one form or another by nonconsequentialist critics: that its aggregative procedure, which sums costs and benefits across individuals, fails to respect the separateness of people.Footnote 1 The question I wish to explore here is: What is the alternative? In particular, have critics of aggregation offered an analytically coherent substantive decision rule for regulating risky conduct that does not itself boil down to some form of aggregation?Footnote 2
The short answer is: I do not think so. This is such a large claim and trenches on such an enormous literature that I could not possibly do it justice here. Instead, I want to focus on the proposed decision rule that in one form or another is at the core of most nonconsequentialist proposals for treating harm to others: conduct is impermissible if (1) it results in (serious) harm to others (2) in a fashion that violates the actor's duty of care to those others.Footnote 3 For these purposes, I define “harm” broadly to include all unconsented-to injuries to the interests of others. While the harms we typically worry about in regulating risk involve physical injuries to one's person or property, this broader definition would include psychological and other intangible injuries that cannot be fully redressed with monetary damages.Footnote 4
The first requirement—(serious) harm to the interests of others—is generally taken to exclude the type of harm that motivated John Stuart Mill to articulate the harm principle in the first place: private conduct that does not compromise the interests of others (other than their nosy preferences about how their fellow citizens live their lives). But within the remaining universe of conduct that is (potentially) harmful to others, the requirement of harm to others cannot help distinguish good risk imposition from bad. The contrary view, I suggest, results from confusing ex ante and ex post perspectives on what is identical conduct.
In contrast, the second requirement—that the conduct in question must violate a “duty of care” toward others—is capable of differentiating among different forms of risky conduct. But to do so, it has to specify what sorts of conduct that duty requires, permits, or prohibits. Most nonconsequentialist accounts have left the answer to that question to another day. Where the duty has been given some content, it is not clearly distinct from some version of a cost/benefit calculus.
At the end of the day, then, I believe the “harm plus fault” test has provided either no answer to the question of what dangers it is permissible to expose others to or an answer that reduces to aggregation manqué. It has, however, answered a different question: When are you required to compensate the victims of your risky conduct for their losses? Much of the confusion in this literature, I believe, is traceable to the failure to distinguish between these two questions.
While I am limiting consideration here to the “harm plus fault” standard, I do not think the limitation rules out a lot, as most of the alternatives offered to aggregation are variants on this standard. But a full consideration of the problem would have to consider a number of other criteria that nonconsequentialists have treated as morally relevant to the permissibility of harmful conduct, including whether the harm results from an act or an omission; whether it is intended, in the strong sense; whether it is foreseeable even if not intended; whether the actor has specific or only statistical foresight of the possible harms and possible victims; whether harm is “certain” to result or merely probable; and whether (in neo-Kantian terms) the conduct reflects a regard for others as ends in themselves and not merely as means.
Rather than offering an exhaustive account of the nonconsequential literature on harm to others, I want to put on the table what seem to me the central, recurring problems in the literature that (I believe) have led it down a dead-end path in devising an alternative to aggregation to handle the vast majority of harm-producing conduct. It may be possible to come up with some other alternative that works, although, for reasons touched on briefly at the end of this article, I am skeptical. But I hope what follows helps to clarify what that would entail.
A few preliminary points of clarification. First, I am using the terms “cost/benefit calculus” (hereinafter CBC) and “aggregation” interchangeably to refer to all procedures that evaluate conduct based on its consequences and that value those consequences by aggregating interpersonally expected gains and losses. I do not mean to single out any particular method of aggregation (e.g., conventional methods of cost/benefit analysis) nor to limit “costs” and “benefits” to ones conventionally counted in welfare analyses. In addition, for these purposes, I am treating decision rules that take distributive fairness into account in weighting outcomes as a form of aggregation. I realize this is a somewhat unconventional usage of the term aggregation. The reason for adopting it is that insofar as nonconsequentialists’ core objection is to procedures that fail to respect the separateness of persons, distributionally weighted outcomes are not clearly an improvement on unweighted ones. Distributional weights do treat individuals (or at least classes of individuals) as separate, in the sense that they respond to a particular class's distinct claims to fair treatment. But unless the weights are large enough to function as deontological side constraints, they do not preclude the sorts of interpersonal trade-offs that lead nonconsequentialists to reject CBC.
Second, for these purposes, I mean “viable” in a very undemanding sense: have nonconsequentialists supplied criteria that, as an operational matter, are capable of differentiating among different forms of risky conduct? I do not reach the further question of whether such criteria, if they exist, dominate CBC on normative or practical grounds.
Third, I am interested here solely in the question of what (risky) conduct the state should permit or encourage to go forward and with what level of safety precautions, not the question of who should bear the costs of any harm that results from the conduct (whether permitted or not).Footnote 5 While these questions are related, they are nonetheless distinct, and the failure of many deontologists to distinguish clearly between them is (as I discuss below) one of the central problems in this literature. To the extent deontologists have distinguished clearly between the two questions and addressed themselves only to latter—correcting wrongs—the literature is orthogonal to my main concern here, which is how we ought to decide what sorts of risky conduct are wrongful to begin with.
Finally, I focus only on conduct in which the (risk of) harm to others is an undesired but unavoidable consequence of pursuing other, socially useful ends. Thus I set to the side conduct that would typically fall within the purview of intentional torts or criminal law: conduct that is intended to inflict harm on others or that is pursued so recklessly as to be judged criminally negligent. I am limiting inquiry to the former category for two reasons. The tort system and administrative/regulatory system that regulate risky conduct are almost exclusively concerned with such conduct. Any theory of permissible (risks of) harm to others that cannot tell us what to do in this core case is of limited value. In addition, conduct that falls in the latter category (intentional torts or criminal negligence) is conduct that pretty much everyone—consequentialists and nonconsequentialists alike—agrees the world would be better off without. Insofar as there are disagreements among those groups, they concern how best to respond to its unwanted occurrence (e.g., retributive/compensatory versus deterrence views of punishment). In contrast, we cannot and would not want to eliminate all socially useful conduct that poses some irreducible risk of harm to others. The chief task of any theory of permissible harms to others is therefore to tell us where to draw the line.
II. EIGHT PROBLEMS WITH THE DEONTOLOGICAL LITERATURE ON REGULATING HARMFUL CONDUCT
A. The Dangers of an Immanent Critique
The last twenty-five years have seen the rise of a new (or arguably renewed) scholarly approach to evaluating the common law that might be termed the “immanent critique.”Footnote 6 Rather than starting with an external metric of value—optimizing welfare, correcting wrongs, protecting specified rights, giving parties what they have (or would have) chosen for themselves—and assessing the goodness or badness of an existing legal regime in terms of it, proponents argue that we should start with existing law and work outward. That is to say, we should start by giving the best interpretation we can of existing practice and identifying the norms that are immanent in that practice, and then see whether those norms yield principles we would wish to judge the practice by.
During roughly the same period, moral philosophy has witnessed a similar development, with the rise of a kind of common-law method of moral reasoning about our duties to others. In place of the norms of legal practice, our intuitions about the “right” answers to hypothetical dilemmas function as the social facts that unconsciously embody moral wisdom, a proper understanding of which will lead us, by process of generalization and reconciliation, to a set of moral principles for handling harm to others.Footnote 7
There is nothing inherently wrong with this approach. You can get to the same place whether you work from the inside out or the outside in (or, in Jules Coleman's terms, “bottom up” or “top down”).Footnote 8 Proponents of the immanent critique offer a number of plausible arguments for starting from the inside, including the impossibility of evaluating moral principles divorced from their applications and the fact that widely shared moral intuitions are entitled to some presumption of moral correctness. Thus, starting with what is, understood in the most sympathetic light, may be the most reliable and expeditious route to what ought to be.
But there are obvious dangers in the immanent critique to which both the legal and the philosophical literature on harm to others (to my mind) fall prey. First, it is easy to misunderstand what you are looking at when you view it only from the inside. In seeking to give the “best” explanation of tort law as a freestanding institution, nonconsequentialists in law have committed themselves to the view that it can best be understood by staying within its own borders. But tort law is just one arm of the state's regulatory apparatus for controlling harmful conduct. The decision to analyze tort law in isolation from the rest of that regulatory apparatus leads to a number of misconceptions about tort law itself, some of which I take up below.Footnote 9
A similar problem, I believe, has limited the explanatory power of the nonlegal philosophical literature on harm to others. To smoke out our moral intuitions about what sorts of harms to others we may or may not causally bring about, the literature focuses on an odd subset of hypotheticals (basically, trolley problems and other one-off rescue cases). Those hypotheticals typically share a number of unusual features, including the presence of identifiable victims, consequences that are deemed to be certain to occur and will occur “up close and personal,” alternative courses of action that fall out differently on the act/omission distinction, and costs of rescue that appear manageable only because the case is considered in isolation from a more general practice of rescue. I believe that, consciously or not, those features are driving most nonconsequentialists’ intuitive responses to the hypotheticals. When those features drop out, as they do in run-of-the-mill risk regulation cases, nonconsequentialists are generally left with no clear intuitions at all. That the deontological moral principles gleaned from rescue cases are useless in garden-variety cases of risk imposition not only limits dramatically the domain of human conduct to which such principles are relevant; it also suggests that even within that domain, nonconsequentialists may be misdiagnosing what facts our intuitions are snagging on.Footnote 10
Second, it is easy to lose sight of the fact that what is is entitled at most to a presumption that it is right. It might be wrong. At some point, the presumption of rightness has to be defended in light of articulated norms. In my view, most of the deontological literature on tort law never gets to that last step. As a result, it often confuses convention for some external notion of morality and institutional designs adopted for pragmatic reasons for ones that reflect a deeper normative commitment. (Again, I take up examples below.)
The same limitation is evident in the philosophical literature on harm to others. Within the narrow set of rescue hypotheticals that dominate the literature, nonconsequentialists have identified with ever-increasing subtlety which intuitions about permissible conduct are robust and which are not.Footnote 11 What they have not done, in my view, is to establish that those intuitions are moral rather than emotional or psychological, and hence that we should look for general rules that explain them in the moral rather than the affective realm.Footnote 12
B. Conflating Prohibition and Compensation
The state faces two decisions in regulating harmful conduct: (1) what risks to allow some citizens to impose on others; and (2) who should bear the costs of risky conduct? These are very different questions. In economic terms, the first is an allocative question: What conduct is optimal from a societal point of view? The second is a distributive one: How do we distribute the costs that result from permitting or prohibiting a given form of conduct? To put the difference in operational terms, the first asks, would we want to prohibit (risky) act X from ever occurring, if we could? The second asks, assuming act X has occurred, with or without legal permission, and it has resulted in actual harm to one or more persons, should the victims be compensated, either by the actor or by the state? Of course, how the state resolves the distributive question will indirectly affect the allocative decision by affecting the incentives actors have to engage in act X in the first place. But these are nonetheless separable questions, both analytically and operationally.
The distinction between the permissibility of conduct and the distribution of costs that arise from prohibiting or allowing it has been familiar to legal audiences at least since Guido Calabresi and A. Douglas Melamed's famous 1972 article spelling out the four possible resolutions of the prohibition/compensation question.Footnote 13 In the philosophical literature on harm to others, however, the two questions are frequently conflated. Within the legal academy, the confusion can be traced in large part to the fact that the literature comes at the problem through the lens of the tort system, which (in the case of torts that are not considered “strict liability”) conflates the two questions by design. That is to say, defendants who are found to have acted negligently—meaning to have violated the standard of due care we require of actors in their position—are also required to compensate the victims for the injuries they cause.
But this is not a natural feature of the world. It reflects a prior policy determination to link the two in the design of the negligence portion of the tort system. We can (and in other contexts do) decouple judgments about whether conduct is wrong, in the sense that we would have prohibited it if we could have, from judgments about whether, for sundry policy reasons (corrective justice, individual welfare, cost internalization, risk spreading, etc.), the victim should be compensated, and if so, whether the compensation should come from the person who caused the harm.
The failure to keep these questions separate is, I believe, the first casualty of the immanent critique of law. This point needs to be underscored. In the past, debates about the propriety of analyzing tort law in isolation from the rest of our risk-regulation apparatus have centered on whether tort law should be assumed to (or required to) cohere with the principles underlying the rest of the regulatory state. Why should it?, nonconsequentialists have argued. Why should it not follow its own logic of corrective justice?Footnote 14 The problems with the interpretation of tort law as a freestanding institution that I press here are different. I am not worried that such a practice has encouraged people to ignore the differences (presumptively indefensible) between tort law and the regulatory state. I am worried that it has encouraged them to ignore the similarities and thereby misdescribe the tort system itself. Our tort system is not simply engaged in ex post corrective justice via compensation; it is engaged in ex ante risk regulation as well, via the standards of due care it generates to determine liability for negligence in the first place.Footnote 15
Sometimes nonconsequentialists explicitly undertake to answer both the prohibition and the compensation question but, by conflating the two, answer neither.Footnote 16 More often, what nonconsequentialists really seem to care about is defending the duty to compensate (the corrective justice aspect of the system) but—perhaps because they are not focusing on the prohibition question—they just assume it is resolved pari passu with the compensation question.
Others have acknowledged the distinction between the prohibition and compensation questions but have either postponed or explicitly disavowed the obligation to supply a nonconsequentialist solution to the prohibition question. That tack avoids confusion but at the cost of acknowledging the limited policy relevance of the nonconsequentialist literature on torts.Footnote 17 To the extent the literature is concerned solely with the obligation to compensate others for harm one has done to them and has nothing to say about prohibition, it is simply orthogonal to the question on the table: whether there is a coherent alternative to aggregation for regulating risky conduct.
C. Treating Imposition of Risk and Imposition of Harm as Different Forms of Conduct
Closely related to the foregoing problem, a substantial strain in the nonconsequentialist literature approaches the problem of harmful conduct by splitting it into two different categories: conduct that has already harmed another—or is absolutely certain toFootnote 18—and conduct that imposes a risk of future harm.Footnote 19 But imposition of risk and imposition of harm are not distinct forms of conduct. They are identical conduct viewed from an ex ante and ex post perspective, respectively.
Confusion on this point has (in my view) produced many of the most serious confusions in the nonconsequentialist literature, including (1) the belief that actual harm to others is both a necessary condition for culpability and a sufficient condition to create at least a presumption of culpability; and (2) the belief that we can coherently judge completed harms under nonconsequentialist criteria while ceding risky conduct to some form of aggregation. Finally, confusion on this point has spurred an entire literature trying to solve a nonexistent problem: how risky conduct that has not yet resulted in harm can be deemed wrongful. I take up each of these in turn.
D. Completed Harm Is Necessary for Culpability
Those rights theorists who start with the intuition that the proper way to think about the wrong of harm is from an expected utility framework have no difficulty recognizing that the problem of risk is the problem of harm, viewed from an ex ante rather than ex post perspective.Footnote 20 They are the exception, however. Most rights theorists start with the presupposition that until we have actually harmed others, we have done nothing wrong, leading to the following dilemma: “On what grounds can a rights-based political theory justify prohibiting risky actions” that have not yet ripened into material harm?Footnote 21
Within the legal academy, Ernest Weinrib is the strongest defender of the view that actual injury is necessary for culpability.Footnote 22 But he is hardly alone in that view. Consider this from John Goldberg and Benjamin Zipursky: “The wronging of one person by another is the very essence of the enterprise, and until such an event happens, there is no occasion to inquire whether an actor can or should be held to have acted wrongfully by violating a moral or legal obligation of conduct.”Footnote 23 Or this, from Judith Thomson:
Suppose I play Russian roulette on you. (Gun with six chambers, one bullet.) And suppose that nothing happens: the bullet was not under the firing pin when I fired. Suppose I did this without your knowledge, so that you were caused no fear. Did I infringe a right of yours? It does not seem obvious that I did.Footnote 24
The same concern is expressed by numerous others and drives their search for alternative principles to explain the wrong of risk.Footnote 25
The apparently widespread conviction that we cannot condemn conduct until we know its consequences is surely one of the more perplexing features of the nonconsequentialist literature on harm to others. The view that “results should count a lot” is obviously tangled up with the epistemological difficulties of guessing exactly what risks a given course of conduct poses ex ante. But nonconsequentialists’ attachment to the ex post perspective clearly goes beyond anything that can be explained by such concerns.
Within the legal literature, I think attachment to the ex post perspective is explained in large part by the fact that nonconsequentialists have generally approached the problem of harm to others through the lens of the private-law tort system. Here, then, is the second casualty of the immanent critique: only by viewing the tort system in isolation rather than as an integral part of the state's regulatory apparatus would one be led to conclude that no conduct can be judged impermissible until it has resulted in harm to others.Footnote 26
The United States, like most Western countries, divides the task of regulating harmful conduct between the private-law tort system (which adjudicates only those cases in which there is an identifiable plaintiff with standing to sue) and the public regulatory system (which can intervene at any point, with or without identifiable victims). It is true that the tort system is generally called upon to judge conduct only after such conduct has resulted in harm to identifiable others. But this feature does not reflect deep moral principle; it reflects administrative expediency. The only parties with standing to bring a tort suit are those who have been harmed by the conduct or who are put at a heightened risk of harm.
In theory, one can seek an (ex ante) injunction to prohibit conduct that will put one at heightened risk of serious harm. But to get an injunction, a plaintiff must show that she is likely to be seriously harmed and that compensatory damages will not provide an adequate remedy.Footnote 27 In addition, she would have to know about the risk in time to get into court before the act in question has been completed. Only rarely are all these criteria met. As a result, it is rare for a plaintiff to seek, let alone get, an injunction.Footnote 28 Thus almost all tort suits come up in an ex post posture, in which the remedy the plaintiff seeks is not to prevent the harm (which by hypothesis it is too late to do) but to be compensated for it. Hence actual harm is a necessary element of most tort suits. But that is because of the remedy sought, not the scope of the right.
While the tort system operates largely ex post, much of the rest of the enormous state apparatus for regulating harmful conduct does not. Criminal codes routinely punish conduct that puts others at extreme risk of injury (drunk driving, criminally negligent violation of safety standards, etc.) without requiring that the conduct results in injury first. In the civil sphere, environmental regulations, health and safety regulations (building codes, product standards, workplace regulations, licensing requirements for high-risk jobs), securities regulation, and so on, all operate ex ante. Sometimes ex ante enforcement takes the form of requiring preclearance to demonstrate compliance with regulatory standards (e.g., you cannot market a new drug without FDA approval). Sometimes it takes the form of penalties for violating those standards whether or not actual harm ensues.Footnote 29 Outside the United States, these and other regulatory regimes swamp the tort system in social and economic importance. Even in the United States, which is alone among industrialized democratic countries in its attachment to the “adversarial legalism” of the tort system, regulatory regimes are an increasingly important part of harm regulation and have subsumed many functions that used to be performed by the tort system.Footnote 30 They not only operate ex ante; they are designed to operate ex ante in order to prevent harm from occurring in the first place.
Whatever its source, the belief that actual harm is necessary for conduct to be judged culpable has a number of curious implications. First, it requires us to suspend moral judgment of an act until it is too late to act otherwise, thereby relegating rights-holders to (at best) ex post compensation for the involuntary loss of their rights. That result is surely ironic, given the conception of rights with which most deontologists start. The classic liberal view of a rights violation as an unconsented-to boundary crossing implies that an ex post remedy in the form of cash compensation for the injury is always a second-best solution. The first-best is to prevent the boundary crossing to begin with by formulating required standards of conduct, through injunctive relief, and so on. How can it be that the just state is prohibited from taking the only sort of action that would fully prevent injustice?
Second, as others have noted, a moral theory that cannot judge whether actions are wrongful until they are completed is not a theory of action at all. If we conclude ex ante that it is permissible for someone to drive down the street at twenty-five miles an hour because on balance the activity is socially useful even if it poses an irreducible risk of harm to others; but conclude ex post, after someone is hurt, that it was wrong after all, for reasons not fully captured by a utilitarian calculus, we have created a paradox. That paradox is economically encapsulated in the common practice of describing the wrongfulness of conduct by reference to its actual consequences. Thus Thomson describes the job of moral theory to be to answer the question “what harms we may or may not cause in what circumstance.”Footnote 31 Tony Honoré describes the “conduct” for which one may incur tort liability as including “actions, omissions, and causing untoward consequences [by] conduct of a potentially dangerous sort.”Footnote 32 Goldberg and Zipursky characterize tort law as prohibiting “[t]he doing of realized wrongs,” not “the doing of unrealized. . . .wrongs,” and imposing a “a duty to not injure, rather than a duty to not engage in injurious conduct.”Footnote 33
The paradox created by insisting that we suspend final judgment of conduct until its consequences are known is, of course, the paradox of moral luck: We decide what course of action is morally required of us ex ante, on the basis of expected outcomes, but conclude we made the wrong choice ex post, when things turn out badly through no fault of our own. These time-inconsistent judgments create a paradox, however, only if they purport to answer the same question. Often, they do not. Consider, for example, a variant on Bernard Williams's famous hypothetical of Gauguin, viewed from the perspective of what Gauguin owes to his family.Footnote 34 When Gauguin states ex ante that, all things considered, it is right for him to choose his art over his family, he is proposing a theory of rational (moral) action. When he states ex post, after he has failed as an artist, that he regrets his earlier choice, he is expressing his emotional response (regret, remorse, etc.) to a choice that turned out badly. Our propensity to feel remorse for bad consequences of (all things considered) prudent choices may be unfortunate, in the sense that it guarantees much unhappiness in life to no good purpose. But it is not paradoxical. It simply reflects the fact that our rational and emotional lives operate on different planes and respond to different stimuli.
The same cautionary point holds here. If what nonconsequentialists really mean by a “duty not to injure” is that we have duty to compensate for any injury that results from our conduct—if, that is, they are really proposing a standard for compensation and not conduct—then the duty they propose does not create a paradox. It simply amounts to a version of strict liability (you broke it; you fix it) for conduct that is judged culpable on other, unstated grounds. It may be harder than advocates think to come up with an unmoralized notion of “broke” or to defend a strict liability compensation scheme on moralistic grounds, but the proposed regime is not paradoxical.
If, on the other hand, the “duty not to injure” is intended to prescribe a standard of conduct with respect to the action that resulted in harm to others—you have a duty to act in a fashion that will not cause X consequences, by which we mean that we will prohibit any actions that will turn out to have caused X—it is a paradox and has to be resolved in practice by choosing one temporal perspective or the other. Given that we go through life in only one (temporal) direction, if that duty is meant to provide a standard for conduct, it must be based on information available to the actor ex ante. That information, by necessity, must be about the expected, not actual, consequences of the act under consideration.
Of course, even hardcore adherents to the view that harm is necessary for culpability agree that something is wrong with pointing a loaded gun at someone's head and firing what turns out to be (through no virtue of the shooter's) a blank or brandishing a loaded gun carelessly in a crowded downtown area, whether or not it accidentally goes off. The question is: What exactly is wrong, and does it sound in deontological principles? Nonconsequentialists have split on this question (see Sections II.E and II.F below).
E. The Wrong of Risk Can Be Explained by Deontological Principles
Several possibilities have been floated for explaining the wrong of risk in deontological terms.
1. Harm Includes Expected Harm
The most straightforward possibility is to redefine harm to include having one's prospects statistically worsened. Thus David McCarthy supplants the traditional liberal harm theory with what he calls “risk liability theory,” which holds that X is liable to Y “if X performs an action that she knows or ought to know will impose a risk of harm on Y,” provided there is no excuse for it (“excuses aside”).Footnote 35 Similarly, Joel Feinberg argues that culpability for risky conduct should be judged as a function of the probability and magnitude of expected harm.Footnote 36
The shift to an ex ante (expected harm) perspective provides an answer to the question posed by Thomson: If harm is necessary for culpability, how can the mere act of imposing a risk of future harm be wrong under the harm principle? But it does so at the cost of highlighting a second and more serious problem. Judged ex ante, almost all conduct poses a risk of harm to others. If that is sufficient to make conduct culpable, then almost all action is culpable. If it is not sufficient, then something else is doing all the work in differentiating permissible from impermissible conduct. I take up this problem in Section II.H below.
2. Risk Creation Is a Completed Harm
Others argue that imposing risk on others is a completed harm. Richard Epstein gestures in that direction in arguing that in creating dangerous situations, we create a “store of energy” that is released into a harm-causing force—a view that invites us to view risk creation as just the first step in a continuous, multipart act of harming others, like shooting an arrow from a bow aimed right at someone's heart.Footnote 37 Others go further and argue that imposition of risk is not the first step toward completed harm, but the last. In the case of risks that have run their course and result in no injury, the completed harm imposed by the risk itself could be psychological trauma from a near-miss. In the case of ongoing risks (e.g., exposure to toxic substances that impose a risk of later disease), it could be long-term anxiety about whether the harm will materialize or out-of-pocket monitoring costs to detect it as soon as it does materialize.Footnote 38
This solution does not seem to have much to recommend it beyond expediency. The mere imposition of risk may well constitute a completed harm whether or not the principal threatened harm comes to pass. Knowing that a gun is pointed at your head is surely an injury in itself, with lasting psychological consequences even if the gun never goes off. But our widespread—indeed, universal—intuition that there is something wrong with imposing some forms of risk on others does not seem to depend on the presence of such completed harm nor to track its severity. The victim of an attempted murder need not know the attempt has been made for us to conclude that the attempt was wrongful. We need either to shed those intuitions or to come up with a better explanation for them.
3. Risk Creation Violates a Different Right from the Right to Be Free from Harm
Others concede that risk imposition falls outside the reach of the antiharm principle but argue that it violates other interests of the would-be victims that sound broadly in deontological principles. Several people identify that supplemental right as a right not to be subjected to conduct that expresses contempt for (or otherwise impermissibly devalues) its potential victims. Thus Jean Hampton argues that the reason attempted murder is wrong, even if the attempt fails, is because of what the attempt “conveys about the intended victim's worth.”Footnote 39 Rahul Kumar describes the wrong as “[t]he denial of the value of one's humanity” by “culpably failing to comply with [one's] legitimate expectations.”Footnote 40 The same principle would presumably apply to Russian roulette, criminal negligence, and so on.
Stephen Perry identifies that supplemental right as a “second order interest recursively derived from [one's] first order interest in not being physically injured.” So “I can have a right that you not try to physically injure me, even though a violation of that right which did not cause me physical injury would not itself be a harm.”Footnote 41 Others locate it in an independent, countervailing “right to self-defense” that permits one to take preventive action to forestall injury even when no wrong has yet been committed.Footnote 42
All of these efforts to handle risk under supplementary deontological principles seem similarly unsatisfying. Hampton's supplementary right to be free of “expressive harms,” like the argument that the imposition of risk is an ex post harm, is wildly underinclusive. The contempt for others expressed by certain acts may seem to give a reason to condemn the most egregious forms of risky behavior. But it does not explain why we might want to prohibit or otherwise regulate garden-variety behavior that imposes an excessive risk of loss on others but hardly rises to the level of malum in se conduct (e.g., driving 45 m.p.h. through a crowded city street with a posted 30 m.p.h. speed limit).
Moreover, even as to the most egregious forms of behavior, the right to be free from “expressive harms,” like the recharacterization of risk as an ex post harm, seems to miss the point of what is really wrong about such behavior. Surely most people would think that the central wrong entailed in someone's pointing a loaded gun at your head and attempting to fire it is not the contempt the would-be killer thereby expresses for your worth as a person, or the transitory anxiety he causes you, but the fact that he actually could have killed you and failed to do so through no ‘fault’ of his own. It is hard to see how deontologists can avoid the conclusion that what is wrong with risky conduct, finally, is the risk of ex post harm it poses. At least, it is hard to see how, if (as I believe) most would conclude that what is wrong with actually killing someone, finally, is not the contempt it expresses for the victim or the transitory anxiety it imposes on its about-to-be-killed victim but the killing itself.
Indeed, it is hard to see how ostensibly freestanding principles like “a right to self-defense” or a “right to be free from expressive harms” can be other than parasitic on the judgment that the conduct in question poses ex ante too great a risk of ex post harm to be tolerated. Why else is the expression of contempt in this particular form—brandishing a loaded gun with the intent to shoot—prohibited, when its expression in most other forms is not? Why else do we have a supervening right of self-defense that allows us to prevent certain forms of conduct from occurring even in the absence of any demonstrated harm?
If what is wrong with risky conduct, finally, is the risk of ex post harm it poses, then such “supplementary” principles do not supplement the antiharm principle as conventionally interpreted (that is, to require ex post harm). They quietly eviscerate it by disguising as an unrelated right what is in fact an ex ante judgment on the permissibility of imposing certain risks. Perry's strategy—characterizing the right to be free from risk of harm as a second-order right derived recursively from the first-order right to be free from harm—forthrightly acknowledges that parasitic relationship, raising a different puzzle: What, operationally, turns on bifurcating these interests into primary and secondary ones?
Finally, it is hard to escape the sense that all of these efforts to handle risk under the antiharm principle, notwithstanding the absence of (ex post) harm, like the decision to kick it out and resolve it under other principles, have an ad hoc, jerry-rigged quality to them. What is the metatheory that tells us when we trigger a right to self-defense or a right to be treated with respect and how we are to trade off those supplementary rights against the core (antiharm) principle that the would-be injurer has a right to do what she wishes with herself and her property unless and until she harms another? In most cases, that question is unanswered, again leaving us in doubt what the real criteria for judging the permissibility of conduct are and whether in the end they amount to something distinct from some form of CBC.
F. Risk Should Be Ceded to Welfarism, Keeping (Completed) Harm in the Deontological Fold
Many nonconsequentialists, however, reluctantly conclude that the problem of risk cannot be resolved by nonconsequentialist principles. Consider this lament from Robert Nozick: “It is difficult to imagine a principled way in which the natural-rights tradition can draw the line to fix which probabilities imposed unacceptably great risks upon others. This means that it is difficult to see how, in these cases, the natural-rights tradition draws the boundaries it focuses upon.”Footnote 43
Instead of contorting rights theory to handle risk, they cede the problem of risk regulation to some form of CBC.
Often the concession is implicit, with the move to aggregation couched in the language of rights. One common move is to impose on all risky actors a duty to take reasonable precautions, which duty is cashed out in terms that seem indistinguishable from CBC (I return to this issue below). Others resort to what Nozick at one point disparagingly refers to as a “utilitarianism of rights.”Footnote 44 Nozick himself adopts this approach in deriving the just minimal state from mutually risk-imposing proto-states, concluding that faced with warring threats to rights, we should adopt the solution that minimizes the “total (weighted) amount of the violation of rights in the society,” with weighting apparently to be in accordance with the social importance of the rights to their holders.Footnote 45
Charles Fried concedes as much with respect to his announced standard that “it is wrong to expose the person or property of another to undue risk of harm, but what risk is undue is a function of the good to be attained and the likelihood and magnitude of harm.”Footnote 46 While acknowledging that the standard seems indistinguishable from CBC, Fried gamely tries to keep it in the deontological fold by enshrining its aggregative solution as itself a categorical right: once “the weighing has been done, then it is also absolutely wrong to go against the conclusion of that process.”Footnote 47 (In other words, in return for being willing to rename their welfarist policy recommendation a “right,” welfarists get to specify the content of that right. This is a version of rights theory one imagines any welfarist can live with.)
One can find other implicitly welfarist solutions to the problem of risk, similarly couched in the language of rights, throughout the deontological literature.Footnote 48
At other times, nonconsequentialists are more forthright about ceding risk regulation to CBC. Joel Feinberg, for example, argues that the culpability of an actor should turn on three factors: the probability of expected harm, the magnitude of expected harm, and the independent value of the risky act to the actor himself and to society at large.Footnote 49 Dennis McKerlie suggests that it should be determined based on the social utility of the conduct and whether it is an “ordinary and important part of people's lives”—the latter, one would think, merely a different way of putting the former.Footnote 50 Jules Coleman suggests that while the dyadic, corrective justice model adopted by tort law may make sense within its domain for reasons of institutional competence, both “epistemic and normative,” outside that domain, risk regulation perhaps should be governed by other (apparently welfarist) principles.Footnote 51
But as emphasized above, the problem of risk is the problem of harm to others, judged from an ex ante rather than ex post perspective. Once we recognize that we are talking about the same conduct judged from different temporal perspectives, the problem with the proposed division of labor becomes obvious. It requires the state to judge the permissibility of the same conduct twice and under different standards.
Take the hypothetical case of a new vaccine with the potential to prevent AIDS but carrying some low risk of death from adverse reactions to the vaccine. Suppose the arm of the administrative/regulatory state charged with regulating drug safety (the FDA) reviews the vaccine under some version of CBC and approves it for distribution. The vaccine is widely distributed, resulting in one death and an estimated thousands of lives saved by the end of the first year. The surviving family of the one victim files a tort claim, arguing that the drug company violated its duty of care in selling a defective product.
The court must determine as a threshold matter whether the drug company did in fact violate a duty of care in distributing the vaccine. If—as it routinely does—the court adopts the same standard of due care as the government, either in the course of reviewing the negligence question de novo or in concluding that regulatory safety standards, where they exist, define the level of care required, then the two stages of review present no problem.Footnote 52 But then the court is simply importing CBC into the common law, not adopting a different standard.
But suppose the court decides to impose a different and tougher standard—say, a “duty not to injure”—which duty the court concludes the drug company violated as of the moment that someone died from the vaccine. Now what? Once again, if what the court really means by a “duty not to injure” is not that the company should not have marketed the drug but that it has a duty to compensate victims for any injuries that result, there is no contradiction. The two stages of review are simply answering different questions. The FDA decides the question of prohibition under a CBC and leaves it to the courts to decide the question of compensation under whatever standard they wish.
But if the “duty not to injure” in fact means what it purports to mean—that the company should not have distributed the vaccine in the first place, given that its distribution (as it turned out) resulted in injury—the company (along with the rest of us) has a real problem. Recognizing that continued distribution is likely to result in further deaths, should the company continue to distribute the vaccine or should it not? Do we want it to or do we not?Footnote 53 Assuming someone had the standing to request the court to enjoin future distribution of the vaccine, should the court grant the injunction?
G. Harm Is Sufficient to Create at Least a Presumption of Wrongful Conduct
Few scholars have advocated what one might call an “absolute liability” position: that the mere fact of harm (at least when it was foreseeable ex ante that it could result) is sufficient to condemn the conduct that caused it.Footnote 54 Many, however, have treated the fact that harm resulted from a given act as sufficient to create a strong presumption that the act was wrongful, acknowledging that in most cases an additional factor—generally, some version of fault—may be required to establish culpability conclusively.
As noted above, the presence of harm is a necessary precondition to establish a right to compensation, since without harm of some sort, there is no loss to be compensated for. But it cannot help determine whether the conduct that caused it is impermissible (by which I mean that we would have prohibited it ex ante if we could). From an ex ante perspective—the only perspective from which we can judge the permissibility of conduct without creating the paradox of moral luck—virtually all conduct poses some risk of harm to others, but (by definition) none of those risks has yet eventuated in actual harm. If “harm to others” is defined to include imposing a risk of harm on others, then virtually all conduct is harmful ex ante. If it is defined to include only completed harms, then no conduct is harmful ex ante. Either way, the requirement of “harm” cannot help to differentiate conduct we wish to prohibit from conduct we wish to permit. Something else must be doing all the work.
Once again, I believe that rights theorists reach the contrary (erroneous) conclusion because they approach the problem through the lens of the tort system—the third casualty of the immanent critique. The reason for this is simple. Whatever risks an act may pose ex ante, few of those risks ripen into ex post harm. As a consequence, a rule that treats harm to others as prima facie culpable, whatever its other shortcomings, at least seems not to force us to condemn all action. Goldberg and Zipursky appear to offer exactly this justification for the ex post harm requirement in the following passage, suggesting that tort law's tolerance of moral luck may be at root not a tolerance of luck at all, but instead a defense of liberty. The passage is worth quoting at length:
[A] system that recognized as actionable not only duties not to cause physical injury, but also duties not to cause various increments of risk of physical injury would potentially create undue de facto burdens on citizens’ freedom of action. . . . It may be that, from certain moral perspectives, identical acts of careless driving, one of which ripens into a tort, the other of which does not, are equally deserving of sanction. Hence from those perspectives, there is an element of arbitrariness as to when tort liability attaches. But viewed from the standpoint of liberal political theory, the insistence on realization is not arbitrary. Rather, it harnesses chance to create a kind of buffer zone for free action: Unless and until injurious conduct actually causes an ultimate harm, it is not subject to sanction through a privately commenced lawsuit. In this regard, the employment of duties of non-injury in tort is akin to the rule against prior restraint of speech: It serves as a prophylactic by permitting a certain amount of undesirable conduct in order to ensure that liberty is preserved. Experientially, this appears intuitive: Most people take advantage of the buffer zone created in part by the requirement of ultimate harm at one time or another, for example, by occasionally driving unreasonably.Footnote 55
The authors back away from the more radical implications of this justification in the paragraph that follows, suggesting that perhaps all they are policing against is potential abuse of process by private plaintiffs, along with the administrative burden from the multiplicity of lawsuits that would ensue if we made all risky behavior subject to private causes of action by potentially harmed parties. But taking them at their word, the quoted passage (inadvertently) puts its finger on the moral luck paradox buried in the ex post view. Suspending judgment on the permissibility of conduct until we know its consequences does not, contra the authors, “permit[] a certain amount of undesirable conduct.” It permits it all, relegating the victims to ex post compensation under a (roughly) strict liability standard.Footnote 56
As the authors note, this is precisely the effect of our presumption against prior restraints on speech in the context of the First Amendment. We generally refuse to prohibit speech ex ante; potential victims must wait until the speech in question has been shown to have wronged them (because it is slanderous, libelous, etc.). But by that point, generally the only recourse the injured party has is a suit for compensation. We live with the resulting higher level of harm in the context of the First Amendment because we attach a uniquely high value to freedom of speech. We could follow the same procedure in the context of risky conduct, permitting it all and relegating victims to ex post compensation. Thus, for example, we refuse to judge whether a scaffolding has been put up safely until we see whether it collapses and kills someone.Footnote 57 I seriously doubt this is the result that the authors want. But that is the result they will get—a fact that is easy to miss when looking at the problem from the ex post perspective of the tort system. It is easy to miss because courts, judging conduct ex post, are not called upon to decide whether they would have prohibited the conduct if they could have. That horse is out of the barn. The only decision they are called upon to make is whether compensation is required. And since they do not have to answer the prohibition question, it is easy to think that in answering the compensation question, they have somehow answered the prohibition question as well.
H. “Fault” plus Harm Makes Conduct Wrongful
As noted above, the fact that it resulted in harm is generally taken to create only a presumption that the conduct in question was culpable. Rights theorists acknowledge that in most cases, something more is required to establish culpability conclusively. That “something more” most often reduces to a requirement that the conduct in question be “wrongful.”Footnote 58
Sometimes the requirement of wrongfulness is unconsciously smuggled into the definition of harm-in-fact, like the “benevolent definition of a sou as a small coin to be given to the poor.”Footnote 59 Thus Thomson, after starting with an apparently very broad, unmoralized notion of harm as “unwanted outcomes,” silently accommodates her intuition that many unwanted outcomes in the world are not (and should not be) actionable, by refining the definition of “unwanted outcomes” to state, roughly, that “to cause a person [an unwanted] outcome . . . is to infringe a right of his.”Footnote 60 Goldberg and Zipursky, likewise starting with a very broad notion of our duty of care under negligence law—“to take reasonable care not to cause [a physical] injury through affirmative conduct”—restate it as a duty not to cause “physical injuries [through one's] misfeasance.”Footnote 61
More often, the requirement of wrongfulness is explicitly layered on top of the requirement of harm-in-fact. Thus, to be culpable, conduct must not only cause harm; the harm must also be “wrongful.” The requirement of wrongfulness is phrased in a variety of ways: The conduct must “violate [a] right”Footnote 62 or a “legitimate” or “protected” interest,Footnote 63 fail to give people what they are “due,”Footnote 64 violate a “duty of care,”Footnote 65 constitute a “wrong”Footnote 66 or “improper treatment of the victim,”Footnote 67 be “not excused,”Footnote 68 be unreasonable,Footnote 69 be a harm with respect to which the actor is “at fault”Footnote 70 or “acted faultily,”Footnote 71 or be inflicted “negligently,” “unjustifiably,” or “impermissibly.”Footnote 72 For legal purposes, some of these verbal differences may imply a difference in the types and levels of proof required. But for current purposes, the differences are irrelevant. Each formulation, like Thomson's silently moralized definition of harm itself, boils down to an assertion that culpable harms are all those harms-in-fact that we have a right to be free from.Footnote 73 As a result, we are left with a scheme that essentially says: ex post harm to others’ interests gives rise to liability, except when it does not.Footnote 74
Some rights theorists acknowledge that “fault” is just a placeholder for a standard of permissible conduct yet to be articulated. In the past, Jules Coleman and others have defended that silence by arguing that it is simply not the job of corrective justice theory to answer the question: What duties do we owe others? That job, Coleman suggests, belongs to some other branch of moral philosophy. Its job is, rather, to answer the question: What are the consequences of failing to meet those duties, whatever they might be?Footnote 75 Fair enough. But that is just to acknowledge that the principles of corrective justice speak only to the compensation question and have nothing to say about required levels of conduct. Until nonconsequentialists articulate their position on required levels of conduct, it is impossible to say whether nonconsequentialist principles can supply a coherent alternative to some form of aggregation for distinguishing permissible from impermissible risk imposition.
Where rights theorists try to flesh out a standard for permissible conduct, the criteria they suggest often seem vanishingly close to CBC. Consider, for example, this formulation of our duty of care:
The force the interests of others imposes on our duty to moderate our behavior varies with the circumstances. Sometimes, the likelihood or magnitude of harm to others is so great that the duty we have to others is not to harm them as a result of the actions we choose to undertake. At other times, the balance of interests indicates that we need to take reasonable precautions to guard against harm to others, and no more. Understood in this way, the problem is familiar and not in the least unique to tort law. It is a matter of ordinary morality that the content of our duties to others varies as a consequence of a range of familiar factors.Footnote 76
Or this one, which the author acknowledges reduces to “a calculus of costs and benefits . . . to the actor and other persons”:Footnote 77
By definition, an accident diminishes the quality (and perhaps quantity) of a victim's life and other goods, which produces suffering for the victim. We may assume that human suffering is undesirable and so should be avoided, ex ante, or remedied, ex post. [Fn: Unless the cost of such avoidance or remedy is, by some fair measure, excessive.] Moreover, because accidents consume human and other social resources, society suffers harm to its aggregate stock of goods by accidents. [Fn: Unless the conduct that caused the accident generated more goods than it consumed.]Footnote 78
Or this:
A rough account of why [a] drunk driver has wronged [a] pedestrian [he has put at risk] ought to appeal to the failure to comply with the pedestrian's legitimate expectations of the driver that she operate her vehicle in a manner conducive to keeping the risk at which others are put as a result of her activity within certain acceptable limits.Footnote 79
Either way, rights theorists do not offer a coherent alternative to aggregation for regulating risks—in the first case, because they cede the problem to others (possibly welfarists themselves), and in the second, because they appear implicitly to adopt the welfarist solution as their own.Footnote 80
III. CAN WE RESOLVE THE QUESTION OF WHAT RISKY CONDUCT TO PROHIBIT AND WHAT TO ALLOW UNDER OTHER NONCONSEQUENTIALIST PRINCIPLES?
As stated at the outset, I am limiting consideration here to whether the criteria of “harm” plus “fault” can yield a coherent decision rule for determining what risky conduct to prohibit that does not itself just reduce to some version of CBC. I do not think the limitation rules out a lot, as most of the nonconsequentialist literature on regulating risky conduct in fact revolves around these two criteria. But as I state at the outset, a full treatment of the problem would have to consider other criteria deemed relevant in distinguishing permissible from impermissible harmful conduct, including whether harm will result from an act or an omission; whether the harm is intended, in the strong sense; whether the harm is foreseeable even if not intended; whether the actor has specific or only statistical foresight of the possible harms and possible victims; and whether harm is “certain” to result or is merely possible or probable.
A full consideration of the relevance of each of these criteria to risk regulation is far beyond the scope of this essay. But I want to suggest the possibility that whatever moral traction each of these criteria may have in trolley problems and the other one-off, individual-choice scenarios that dominate the philosophical literature on harm to others (again, I offer no view on that question here), they are irrelevant to large-scale risk regulation.Footnote 81 Here are some of the reasons to think this may be so.
The state is a collective entity, not a real person; unlike private parties, it has an affirmative duty to aid its citizens; and in dispatching that obligation in the regulatory context, it typically articulates rules to govern private conduct rather than acting directly itself. All of these factors may make moot many, if not all, of the agent-centered prerogatives that are taken to support the act/omission distinction.Footnote 82
In addition, the acts I am concerned with here—socially useful conduct that carries some risk of harm to others—are for the most part identically situated ex ante with respect to intentionality, foreseeability of harm, certainty of harm, and identifiability of the type of harm and the likely victim(s). Harm is never intentional, in the strong sense of desired; it is always an unwanted byproduct of conduct pursued for other reasons. Typically, harm is foreseeable as a probabilistic matter; in some cases the frequency and extent of harm is foreseeable to something approaching a statistical certainty. But exactly when and to whom it will occur is not foreseeable, and since no individual instance of harm is certain to occur, it is theoretically possible (although, in many cases, exceedingly improbable) that no harm will result.
Because most socially useful but risky conduct falls out the same way with respect to each of these factors, the factors are incapable of differentiating good risks from bad. If conduct is wrongful whenever it is statistically certain to result in death or serious injury to someone if repeated enough times, then the state would be obliged to prohibit most of the activities we take for granted in our daily lives (administering mandatory vaccinations, constructing roads, bridges, and buildings, performing routine operations, manufacturing automobiles, etc.) On the other hand, if conduct is wrongful only when it is absolutely (not merely statistically) certain to result in harm or when the victims are identifiable ex ante, then virtually no risky conduct of the sort at issue here—socially useful activity that produces unintended harms—is wrongful. In either case, the criteria cannot help us distinguish acceptable from unacceptable risks.
Contractualist approaches to the problem of harm to others are a more complicated case.Footnote 83 I do think they can provide a morally meaningful, nonconsequentialist justification for picking one scheme of risk regulation over another and for choosing one that deviates from standard, unweighted aggregation—for example, one that is sensitive to distributive concerns, measured either ex ante or ex post, or gives greater weight to certain outcomes than individuals themselves might give in choosing for themselves, or makes no attempt to quantify the incommensurate values being weighed in the balance. But at the end of the day, I believe all of the plausible schemes from which to choose will necessarily be aggregative in one form or another.
IV. WHY SHOULD WE CARE?
Millions of decisions are made every day that pose uncertain risks of potentially serious harm to unidentified others. Every time we get in a car, market a new drug, decide whether or not to recall a product, decide what level of precautions to take against earthquakes and other natural disasters or against potential terrorist attacks, we are taking our own and others’ lives into our hands. How we ought to go about making those decisions is of enormous social importance.
There are many reasons to be skeptical of CBC as commonly deployed, and it may be that at the end of the day we would be better off just letting the political process muddle along with some procedural safeguards in place rather than signing on to the false scientism of CBC. But nonconsequentialists reject CBC because they reject on principle the interpersonal aggregation of costs and benefits—the notion that costs to me can be offset by benefits to you. Here, I think it is fair to say that you cannot beat a bad candidate with no candidate, and in my view nonconsequentialists do not offer a viable candidate. That failure, I believe, is masked by the odd perspective from which both the legal and philosophical literature come at the problem of harm to others.
For different reasons, both focus on cases in which consequences are fixed. The philosophical literature focuses on an oddball set of rescue hypotheticals in which the consequences are stipulated to be known with certainty ex ante.Footnote 84 The legal-philosophical literature comes at the problem from the ex post perspective of the tort system. By the time virtually all tort cases land in the court system, the consequences are fixed with certainty; the only decision left for the courts is whether to compel compensation or not. The compensation decision is important, but it suppresses the problem of interpersonal trade-offs between presumptively innocent parties, because what is at stake in the compensation decision is a zero-sum income transfer in which a (presumptive) wrongdoer is simply making whole the victim of his wrong.
In contrast, viewing the problem of potentially harmful conduct from an ex ante perspective pushes to the fore the problem of scarcity, and with it the inevitability of interpersonal trade-offs. Because virtually everything we do (or, acting as the state, permit others to do) carries some irreducible risk of serious harm to others, virtually everything we do (or permit others to do) entails interpersonal trade-offs. That reality is easy to overlook in our individual decision-making capacities, because the one-off nature of individual choices means that most low-probability risks we run will never ripen into harm. This is not the case in the public-policy realm, where we are typically choosing rules that will govern millions or hundreds of millions of events over the long term. As a result, even very-low-probability harms are overwhelmingly likely to occur at some point. If we parole enough prisoners, one of them will turn out to be Willie Horton. If we develop a vaccine for AIDS that has a one out of a million chance of triggering a fatal reaction and administer it to enough people, someone is going to die from it.
We can often reduce those risks by greater precautions. But at a certain point such precautions become prohibitively costly—either in dollars spent relative to the reduction in risk achieved or in new risks the precautions themselves create.Footnote 85 And generally speaking, no amount of precaution will eliminate all risk. If the world's most conscientious child welfare agency has to supervise thousands of children from dysfunctional families, over the long run some child under its watch is going to die from abuse or neglect.
The question is, how should we understand and respond to those bad outcomes when they inevitably come to pass? If their mere occurrence is enough to condemn the actions that produced them, then there is virtually nothing we may as a society do. If it is not enough, then we are acknowledging that interpersonal trade-offs have to be made and that in those trade-offs the numbers will inevitably count.
Deliberately or not, nonconsequentialists’ conviction that we can somehow avoid such trade-offs has reinforced the average citizen's response to bad outcomes: You (government officials) had a duty not to harm anyone, or a duty to keep us safe, and you failed to do so. (As one commentator dryly remarks, “Seldom do we hear a company that was responsible for a deadly accident justify the loss of lives by saying that it was the result of a decision which, in terms of its [expected?] effects, produced far more good than harm.”)Footnote 86 And politicians and policy-makers know how the average citizen will respond. The result is to drive politicians and other public employees to channel enormous resources into preventing the high-visibility bad outcomes for which they know they will be held responsible.
Three years ago, negotiations to ease California's budget crisis and prison overcrowding by early release of nonviolent offenders fell apart over exactly this problem. As one Republican legislator acknowledged with refreshing candor: “If we let someone out early, and that man commits a crime, the Assembly members are worried that that will come back to haunt them like the old famous Willie Horton ads.”Footnote 87 Similar concerns explain why child welfare agencies responded to the rash of high-visibility deaths of children in foster care in the 1990s by directing almost all of their resources to children they think might be at some risk of death, ignoring the thousands of others that are facing serious but (they believe) nonlethal threats of abuse and neglect.
The same desire to avoid ex post blame in the public and private sector drives a substantial portion of our health care expenditures, investments in “homeland security,” and countless other major public-policy decisions. It may be that our propensity to think that bad outcomes from our choices imply bad choices is immune to reason, and the only way policy-makers can curb its influence is to manipulate the psychological salience of bad outcomes. But surely, before we reach that question, we should answer the question of whether it ought to be curbed—whether, that is, nonconsequentialists’ intuition that we can somehow avoid inflicting seriously bad outcomes on some individuals as the price of benefiting many more is in fact plausible.
The full-blown consequentialist will say: of course it is not. It is tragic when reasonable actions have bad consequences, but it is nobody's fault. In a world of scarcity (in the broad sense), whatever we do has potential costs to someone. The best we can do is to act in a way that minimizes aggregate costs relative to aggregate benefits, however we calculate or weight them, and, if the costs to individual victims are serious enough, to remediate them ex post on welfarist grounds.
If there is a viable alternative, it needs to be put on the table so that it can become part of the public debate over risk regulation and be assessed relative to cost-benefit analysis and other versions of aggregation. If there is not, then nonconsequentialists have a moral obligation to acknowledge that, rather than lending the imprimatur of morality to what amounts to nothing more than hindsight bias. We may not be able to change the widespread intuition that bad consequences imply bad conduct; it may be an irreducible part of what it means to be human and may need to be accommodated in some fashion in public policy simply in virtue of that fact. But if the intuition is wrong, philosophers ought to be the ones saying so most clearly and doing what they can to counteract the pernicious public-policy consequences of all of us at least half-believing otherwise.
