Deontology, Paradox, and Moral Evil Critics of deontology have long noted that its proscriptions seem paradoxical since, in contrast with welfare utilitarianism, they forbid some acts that maximize welfare overall. Recently some philosophers have suggested that deontology harbors a special paradox; that thinking certain actions morally objectionable—for example, rape—it forbids minimizing such actions by doing one.' For example, Samuel Scheffler states the following about a deontological constraint. [I]t is a restriction which it is at least sometimes impermissible to violate in circumstances where a violation would [serve to minimize total overall violations] of the very same restriction,... and would have no other morally relevant consequences.^ The passage suggests that there is something particularly irrational about forbidding a deontological violation that, if done, would have the effect of reducing the total number of such violations. To take another example, if intentionally kilhng the innocent is wrong, it seems irrational 'Samuel Scheffler, "Agent-Centred Restrictions, Rationality, and the Virtues," Mind 94 (1985): 409-19, p. 413 (reprinted in Samuel Scheffler (ed.), Consequentialisni and Its Critics (Oxford: Oxford University Press, 1988), pp. 243-60). See also his The Rejection of Consequentialism (Oxford: Clarendon Press, 1982); and "Deontology and the Agenf A Reply to Jonathan Bennett," Ethics 100 (1989): 67-76. Among recent philosophers who defend deontology, see Philippa Foot, "Morality, Action, and Outcome," in Ted Honderich (ed.). Morality and Objectivity: A Tribute to J.L Mackie (London: Routledge & Kegan Paul, 1985), pp. 23-38; and "Utilitarianism and the Virtues," reprinted in Scheffler (ed.), Consequentialism and Its Critics, pp. 224-42; F.M. Kamm, "Harming Some to Save Others," chap. 7 in Morality, Mortality, Vol. II: Rights, Duties, and Status (Oxford: Oxford University Press, 1996); and "Non-Consequentialism, the Person as an End-inItself and the Significance of Status," Philosophy and Public Affairs 21 (1992): 354-89; Henry S. Richardson, "Beyond Good and Right: Toward a Constructive Ethical Pragmatism," Philosophy and Public Affairs 24 (1995): 108-41; Jorge L.A. Garcia, "AntiConsequentialist Moral Theory," Philosophical Studies 71 (1993): 1-32; Christopher McMahon, "The Paradox of Deontology," Philosophy and Public Affairs 20 (1991): 35077; William H. Shaw, "On the Paradox of Deontology," Journal of Philosophical Research 16 (1991): 393-406; Richard Brook, "Agency and Morality," Journal of Philosophy SS (1991): 190-212. Scheffler, The Rejection of Consequentialism, p. 80. © Copyright 2007 by Social Theory and Practice, Vol. 33, No. 3 (July 2007) 431 432 Richard Brook to forbid killing one innocent person to prevent five murders. I argue, considering work of Scheffler and others, that thinking deontology paradoxical in this sense commits one to believing that minimizing moral evil is a goal distinct from minimizing harm. Although perhaps correct, I explore what I think to be serious difficulties with this view. Some preliminaries: I note that orthodox welfare utilitarianism aims to optimize happiness or pleasure without constraining that aim by conditions of distributive or retributive justice. However, it is explicit for Scheffler, and implicit for others I consider, that what is subject to minimization is undeserved harm. In the text, therefore, "harm" will mean "undeserved harm." Following Christopher McMahon, I call minimizing violations, those that reduce the number of identical violations, "preventative violations."^ I will use "deontology," "common-sense morality," and "ordinary morality" interchangeably. "Moral evil," as used here, characterizes any violation of ordinary moral prohibitions, for example, those against theft, kidnapping, rape, or murder. And although there are debates about what constitutes harm, for this essay I count as harm a person's experienced diminution of welfare. Scheffler, however, expresses deontology's "air of paradox" in two ways. Responding to an essay by Jonathan Bennett,^ Scheffler writes about his earlier book. The Rejection of Consequentialism: I was concemed with the air of paradox surrounding the idea that it is morally impermissible to minimize morally undesirable activity, the idea, more specifically, that because certain kinds of acts are so objectionable, one must not perform one such act even if that means that more acts of the very same kind will be performed or that other equally undesirable events will transpire.' "Fqually undesirable events" denotes equivalent harm naturally or accidentally caused. Scheffler further comments: He [Bennett] rightly notes that deontological restrictions do not apply only to cases in which, for example, killing an innocent person oneself is the only way to prevent more numerous killings committed by other people (call these 'A-type cases'); they also apply to cases in which killing an innocent person oneself is the only way to prevent a greater number of deaths due to natural causes ('B-type cases').^ He claims, with respect to A-type cases, that if we're concemed, pace 'McMahon, "The Paradox of Deontology," p. 350. "•For a conception of harm that doesn't require victims of harm to suffer, see Joel Feinberg, "Harm to Others," in The Moral Limits of the Criminal Law (New York: Oxford University Press, 1984), particularly chap. 1. 'Jonathan Bennett,"Two Departures from Consequentialism," Ethics 100 (1989): 5466. 'Scheffler, "Reply to Bennett," p. 73. 'Ibid., pp. 73-74. Deontology, Paradox, and Moral Evil 433 Bennett, with people making bad things happen, then it seems rational to do one making—a preventative violation—to minimize such makings. In B-type cases, the issue of "makings" vs. "allowings" (of violations) obviously doesn't arise. But nothing here suggests that A-type cases are paradoxical in a different way from B-type cases. However, in a later essay, commenting on work of Philippa Foot, Scheffler expresses ordinary moral constraint's paradoxical character in the following way: "How can the minimization of morally objectionable conduct be itself morally objectionable?"^ No mention is made of harm "caused by equally undesirable events." Scheffler's point against Foot is that even if one grants that conceptions of morally better and worse lack meaning outside the dictates of ordinary morality, a principle of maximization comes into play. For Foot, optimizing welfare overall, aside from expressing the particular virtue of benevolence, is not morally overriding, say, against actions that express the virtue of justice. Scheffler concedes this for the sake of argument but contends that [a]ll we need is the recognition that fewer violations will occur if I act one way rather than another, together with the idea that such violations are morally objectionable, in the ... sense that it is morally preferable that no such violations should occur than that any should.' Scheffler's concession (for the sake of argument) to Foot disallows him from conceiving constraints as problematic simply because they forbid some violations that are optimific. But the fact that he still thinks deontology paradoxical suggests that A-type cases (stealing, for example, to minimize the number of thefts) have a special status. Preventative violations now appear paradoxical not simply because prohibiting them diminishes harm, but because such prohibitions conflict with a general teleological principle that Scheffler calls "maximizing rationality" (MR). The core of this conception of rationality is the idea that if one accepts the desirability of a certain goal being achieved, and if one has a choice between two options, one of which is certain to accomplish the goal better than the other, then it is, ceteris paribus, rational to choose the former over the latter.'" Forbidding preventative violations, Scheffler thinks, violates a plausible application of MR, since if certain acts are morally objectionable, it appears reasonable to do one such act to minimize the number of such acts. Even in The Rejection of Consequentialism, which doesn't mention MR, Scheffler claims that the rationality of preventative violations concerns disvalues other than harm. 'Scheffler, "Agent-Centred Restrictions," p. 250; see Philippa Foot, "Utilitarianism and the Virtues," p. 159. 'Scheffler, "Agent-Centred Restrictions," p. 250 '"ibid., p. 252. 434 Richard Brook It makes no difference, in particular, which feature of a violation is singled out as having a high disvalue: no difference, for example, whether the focus is on the victim of the violation, the agent, or the relationship between them. Echoing this point, Henry Richardson, discussing the apparent ubiquity of the problem, writes: "for paradoxical cases can apparently be invented that work with any characterization of what is bad about an action and argue that surely it is better to have less of it than more." Take a particular kidnapping. I consider the harm to the victim, the badness of the kidnapper, and perhaps the corruption of the relationship between perpetrator and victim to be all of disvalue. A preventative violation minimizes all three. Suppose I could save someone from kidnapping or another from an accidental but equal and unjustified restriction of liberty. If I am required to minimize both harm and moral evil I should save the first. Yet this conflicts with an intuition Scheffler himself suggests we have about the following case. Consider two twins, equally innocent. While we are strongly inclined to say that it would be impermissible to kill the first twin in order to prevent the accidental death of the second twin, even if that were the only way to prevent the second twin's death, we have no comparably strong inclination to say that it would be impermissible to prevent the accidental death of the second twin instead of preventing the murder of the first twin by some other person, if one could only prevent the death or the killing but not both." In this example, we assume that a background ceteris paribus condition is met. The up front distinction is between intentional and accidental threat, either of which we might nullify. In fact, imagine that the sole moral dilemma for anyone would be that particular rescue decision. Thus, the only morally objectionable act committed would be the murder. Although saving the twin threatened by accident results in a world with one objectionable action rather than none, that rescue seems permissible. MR applies here apparently only to preventing a death.'"* One could argue that preventing the murder deprives the prospective killer of the satisfaction of killing, and therefore is right to do. But we could design the case in which the villain mistakenly believed he succeeded. In fact, although I don't pursue it in detail, an argument can be made to prevent the accidental death. For the murder to be evil it must be freely chosen. Under a libertarian conception of freedom, there is a real "Scheffier, The Rejection of Consequentialism, p. 89. '^Richardson, "Beyond Good and Right," p. 118. "Scheffler, The Rejection of Consequentialism, p. 109. "'Thomas Nagel appears to take this view. "Admittedly," he writes, "the wickedness of the murder is in some sense a bad thing; but when it is a matter of which of them there is more reason to prevent [murder or accidental death] the murder does not seem to be a significantly worse event, impersonally considered." The View From Nowhere (New York: Oxford University Press, 1986), p. 178. Deontoiogy, Paradox, and Morai Evil 435 as opposed to merely epistemic probability that the prospective murderer won't kill. Thus a real possibility exists that by preventing the accident, no lives are lost. Of course, epistemic probability might be all we have. The murderer has a track record sufficient for the epistemic probability of the murder being equal to the probability of accidental death. But there is something odd in combining an assumption that the murder will occur if the accident is not prevented and as well thinking the murder a moral evil. The assumption treats the prospective murder much like a natural threat.'^ Two more recent papers explicitly take the view that ordinary morality contains a paradox other than prohibiting some optimific actions. Christopher McMahon believes that both a violation's moral evil and consequent harm are prima facie relevant to rescue decisions. He asserts that "[c]ommon-sense morality regards the disvalue of the violation of a deontological constraint as different in kind from, and greater than, the disvalue of the accidental production of its effects."'* "Murder," he writes, "makes the world a worse place from everyone's point of view."'^ Therefore, in the twin example, apparently anyone would have at least prima facie reason to prevent the murder. Yet McMahon, as a utilitarian, thinks it irrelevant which death we prevent. Given apparently contradictory beliefs, he presents himself with the following project. We need only explain why the fact that a murder makes the world worse from everyone's point of view than an accidental death—and worse in a way that we could prevent does not translate into a stronger reason to prevent it than to prevent an accidental death.'' McMahon's "explanation" involves a theory of rectification; the murder, he thinks, can be "repaired after the fact." Rectification nullifies its special disvalue by denying the violator benefits gained by her action. I don't consider this argument, though I think it has limited application. Of more significance, McMahon gives no reason for thinking that common"Bemard Williams's discussion of a blackmail case illustrates the problem of considering causing moral evil both as a real choice and akin to a natural disaster. A person (Jim), as a guest in a South American town, is honored by a sadistic army captain by being allowed to pick one of twenty terrified natives to kill. If he refuses, the captain will kill all twenty. Although Williams doesn't propose a clear answer about what Jim should do, he suggests that one's intuitive reluctance to support Jim's killing expresses the significance we give to agency; that we should be properly concemed about our own actions in a way that is different from our concern about the free agency of others. Yet Williams also requires the assumption that the captain will in fact kill twenty if Jim doesn't kill one. Bernard Williams, "Utilitarianism and Integrity," in J.J.C. Smart and Bernard Williams, Utilitarianism, For and Against (Cambridge: Cambridge University Press 1973) pp. 96-117. ""McMahon, "The Paradox of Deontology," p. 353. '^Ibid., p. 352. Ibid., p. 353 (emphasis in original). 436 Richard Brook sense morality, even as a prima facie guide to rescue choices, sums (or should sum) two kinds of disvalue, the moral evil of a transgression, and the harm caused. McMahon offers another, perhaps intuitively stronger, example to show a violation's moral evil and harm caused are prima facie additive. He writes: "More than three times as many people died in the epidemic of Spanish influenza following World War I as died in the Holocaust, yet intuitively, the Holocaust is a much worse thing to have happened."^'^ He claims that since the Holocaust was a "worse thing to have happened" there is prima facie reason for anyone, given a choice, to prevent the Holocaust. William Shaw claims that deontology harbors a special paradox only when preventative violations are forbidden. No paradox occurs simply because the constraints of common-sense morality forbid some optimific actions. He writes: " when ... the good is defined so as to include the absence of actions of a sort forbidden by the deontological theory, then a paradox begins to emerge."^" In Shaw's view then, deontology becomes paradoxical only when it forbids minimizing moral evil, and not simply because it prohibits action that would reduce harm however caused. Again the point is that if doing A is evil, why isn't it rational to permit A if, without other moral effect, that minimized the doings of A? Why not, Shaw asks, echoing Robert Nozick, use someone as a means if that minimized the number so used?^' We should remember, however, that the question of minimizing constraint violations only arises because they cause harm. Future attempts at violations, assuming that they will fail and have no welfare consequences, don't demand minimization though for some they are evil as successes. We might even think missed opportunities at murder—you've set up your machine gun at Market and Main, but the victim doesn't appear—as bad as murder itself.^^ So those like Shaw and MacMahon, who evidently think common-sense morality paradoxical in a special sense, must take both the harm and evil of violations to be subject to minimization. Yet this raises a problem. If the disvalues of harm and moral evil are, in fact, additive, the question arises how to sum them for rescue decisions. Perhaps a given murder can only be balanced by ten accidental deaths. We compare the harm of a constraint violation with harm natu"ibid., p. 352 (emphasis in original). ^"Shaw, "On the Paradox of Deontology," p. 394 (my emphasis). ^'ibid., p. 396. See Robert Nozick, Anarchy, State, and Utopia (New York: Basic Books, 1974), p. 32. ^•^Some philosophers believe what I've called "failed attempts at violations" to be, in fact, constraint violations. David McNaughton and Piers Rawling take this view in "Deontology and Agency," The Monist 76 (1993): 81-100, p. 92. Deontology, Paradox, and Moral Evil 437 rally or accidentally caused. If harm and moral evil are in fact commensurable, one might think about rescue choices that a point exists where naturally or accidentally caused harm balances the moral evil plus harm of a violation. Of course, even if these disvalues are additive, no precise point of balance may exist. Perhaps there is a threshold above which a violation (like the Holocaust) is so reprehensible that preventing it trumps a rescue that minimizes much greater naturally caused suffering. But thoughts about where that threshold lies seem, at least in many cases, ineluctably subjective. They need include not only how to rank the immorality of constraint violations, but of more difficulty, I think, how to rank the amount of naturally caused harm that even roughly balances the moral evil of violations. Although perhaps possible, it is difficult to envision how this calculus would work. On the other hand, these parameters may be incommensurable and nonadditive. No common coin permits summing them. This is consistent with noting that harm, for example, the loss of your manuscript, may be caused by theft. But it might be caused by a natural event; the wind blew it in the river. In either case, others might have an equal though defeasible obligation to prevent the loss. Under this view, we begin, in rescue decisions, with the aim to minimize harm. That the harm results from theft simply reflects the circumstances of its origin. In some cases we might say, "it's odd to forbid doing something morally objectionable if that reduces the number of identically objectionable actions." But the description of the latter actions—"that they are morally objectionable" as opposed to the description "causing harm we have reason to prevent"— may have no pride of place. Similarly in rescue choices, the fact that one threat is by human intention and the other by nature or accident may be irrelevant. Unless the demand for minimization applies to moral evil itself, it is not clear that the causal origins of preventable harm should affect rescue choices. On this view, MR would apply in rescue cases but again only, as in the twin case, at the level of minimizing harm. The view, then, pace Scheffler and Richardson, that paradox arises for deontology about "any characterization of an action's disvalue" is mistaken unless moral evil itself is subject to MR. If it isn't, then forbidding preventative violations causes no difficulty for ordinary morality distinct from prohibiting some optimific actions. Deontology forbids intentionally killing one person to save two either from murder or accidental death. A recent argument for thinking that, for deontology, rescue decisions should favor potential victims of violations rather than those threatened by nature or accident stems from the belief that if ordinary moral constraints are universalized in what Philip Pettit calls the "straightforward 438 Richard Brook way," then preventative violations appear to be rational. He writes: As a would-be non-consequentialist thinker, my initial claim must have been that the point is to instantiate P [e.g., not murdering] in my own life, not promote it generally. But I countenance the general claims of the P-pattem when I universalize in the straightforward way: I prescribe general conformity to that pattern, not just conformity in my own case. Thus it now seems that what I must think is that this general conformity is to be promoted, even if that means not myself instantiating the pattern in my own behaviour or psychology or relationships. The "straightforward way" of universalizing is that each agent should think about every agent including herself that she not violate constraints. I, for example, hold about each agent including myself that he shouldn't intentionally kill innocent people. However, it is not clear what follows from this. Suppose that if I don't intentionally kill an innocent person, each of five other people intentionally kills one. I refuse; consequently five are murdered. Unless we beg the question in favor of consequentialism, what follows is simply that five people have violated the constraint and I haven't. It is difficult to see how universalizing in the "straighforward way" commits one to an overriding concern to promote the "general conformity" to constraint adherence. By overriding here I mean a concern that permits violating a constraint simply to minimize identical violations. Obviously, if universalizability entailed moral symmetry between duties not to harm and duties to prevent harm, or entailed that my responsibility for my own actions equals my responsibility for the actions of others, it would be right to murder one to prevent two murders. But that symmetry, even if true, isn't entailed by the requirement that we must think about any common-sense moral constraint that no agent should violate it. To see this, suppose we phrase Pettit's puzzle as follows: if I think it bad that (unjustified) harm, regardless of its cause occurs, shouldn't it be rational to bring about some harm to minimize the total amount of harm? But at that level of generality the genesis of the harm drops out as significant, and we are back to the conflict between deontology and welfare utilitarianism. Pettit's presentation suggests, then, that there is, for him. "Philip Pettit, "Non-Consequentialism and Universalizability," Philosophical Quarterly 50 (2000); 175-90, p. 183 (my emphasis). In a recent essay, Scheffler gives an independent moral argument for the greater moral gravity of doing harm than allowing harm to occur. I note, however, that doings in general are not always, qua doings, subject to constraints, but rather intentional doings, or something comparable like using someone as a mere means. Rightly or wrongly, merely foreseen harm by one's action, say, collateral damage in war, is considered often subject only to constraints of proportionality. Yet such harming does exemplify what Scheffler calls the "primary manifestation" of one's agency. Perhaps Scheffler believes only an agent's intentional doings are subject to what he calls the "norms of responsibility." But we would want to know the reason for that. See Samuel Scheffler, "Doing and Allowing," Ethics 114 (2004): 215-39. Deontology, Paradox, and Moral Evil 439 something morally distinctive about cases in which I might minimize a kind of wrongdoing by committing an instance of that wrong. This echoes Scheffler's contention that if some actions are so objectionable that no one should do them (as he says, "as a first order strategy"), it seems irrational to forbid a minimizing violation (as a "second order strategy"). This does single out A-type cases—in which harm and moral evil are joined—as of special concern. Perhaps a case can be made that we are obliged to minimize moral evil and not just naturally caused harm. But the mere requirement of universalizability is inadequate to justify that claim. A given agent A prefers that for all P (including A) P not kill an innocent person. It doesn't follow that S overridingly prefers a world in which there is as little killing of the innocent as possible. I might think, for example, that each person including myself should eat more fruits and vegetables for her own good. It doesn't follow simply from that thought that I should believe as many people as possible should eat more fruits and. vegetables for their own good, certainly not that I should eat less to insure that a greater number eat more for their own good. For deontology, universalizing a course of action is a necessary condition for its permissibility, and that condition can certainly be satisfied without thinking preventative violations are permissible. A person, in fact, can successfully universalize the following: Each agent, including herself, should not violate a constraint simply to minimize constraint violations. Conclusion Although this paper isn't a defense of deontology, I have argued the following: If deontology contains a special paradox, one distinct from prohibiting some optimific actions, then preventing moral evil would be a proper goal of action in addition to preventing harm. That may well be true, and if so, would be an important result. In fact, such a result would undermine deontology itself, since general commitments to constraints would entail permission to violate them simply to minimize the number of identical violations. However, if, to get a total measure of an act's badness, we must sum the moral evil and harm of constraint violations, then there should be some account of how to do that. Why again shouldn't the badness of one murder equal the badness of ten or fifteen accidental deaths? Why not save one triplet from murder rather than her two sisters threatened by an avalanche? We wish to know, even roughly, how much moral evil trumps prospective harm in rescue decisions.^'* ^""Scheffler briefly mentions, without endorsing, a possible consequentialism that would minimize murders rather than deaths. The Rejection of Consequentialism, p. 108. 440 Richard Brook True, MacMahon's example of the Holocaust vs. the influenza epidemic, unlike Scheffler's case of the twins, suggests that we think we have some obligation to prevent evil although as a consequence a great deal more naturally caused harm results. But the general problem remains. We need some credible argument to show that in making rescue choices, the moral evil of a transgression adds weight to the harm caused. Absent that argument, deontology appears to harbor no paradox beyond forbidding some actions that would reduce harm overall. The etiology of that harm again would not be significant. This doesn't mean deontology is out of the woods. But the woods, in this case deontology's conflict with welfare utilitarianism, are old-growth timber. Richard Brook Department of Philosophy Bloomsburg University of Pennsylvania dchardb@ptd.net