(p.167) Prerogatives Without Restrictions
Abstract and Keywords
Shelly Kagan and others have criticized the hybrid theory on the grounds that an agent‐centred prerogative, if unaccompanied by agent‐centred restrictions, might allow agents to harm others in pursuit of their personal projects. Scheffler considers in detail an example of Kagan's that illustrates this criticism and purports to show a conflict between the hybrid theory and common‐sense morality. Scheffler responds that the rationale supporting an agent‐centred prerogative can support either a pure‐cost version of the prerogative or a no‐harm version, and that the latter version avoids Kagan's criticism.
Consequentialists hold that the right act in any situation is the one that will produce the best overall outcome, as judged from an impersonal point of view. By contrast, many non‐consequentialists maintain that, given any impersonal principle for ranking overall outcomes from best to worst, there will be some circumstances in which one is forbidden to produce the best available outcome so characterized, and still others in which one is permitted but not required to do so. They accept, in other words, both what I call agent‐centred restrictions and what I call an agent‐centred prerogative.1
In The Rejection of Consequentialism,2 I argued that an agent‐centred prerogative is easier to explain and defend than are agent‐centred restrictions. For, I suggested, whereas it is possible to identify an underlying principled rationale for such a prerogative, a comparable rationale for agent‐centred restrictions is surprisingly elusive. In consequence, I argued, ‘hybrid theories’, which depart from consequentialism to the extent of incorporating an (p.168) agent‐centred prerogative, but which do not include agent‐centred restrictions, deserve our attention.
In the ensuing discussion, a number of writers have argued that hybrid theories have radically counterintuitive implications and cannot be accepted. However, these writers have disagreed among themselves about what conclusion to draw from this. Some have seen the criticism of such theories as helping to vindicate a version of common‐sense morality, which is taken to include both prerogatives and restrictions, while others have treated the criticism as casting doubt on the advisability of any departure from consequentialism. The former, in effect, see the alleged defects of hybrid views as helping to supply the missing rationale for agent‐centred restrictions. The latter regard the indefensibility of the restrictions as a fixed point, and treat the supposed unacceptability of hybrid views as serving to eliminate another potential rival to consequentialism. In this essay I will explore the force and significance of one of the objections that generates these divergent responses.
In The Rejection of Consequentialism I made the following suggestion about the structure of an agent‐centred prerogative:
(p.169) Suppose, in other words, that each agent were allowed to give M times more weight to his own interests than to the interests of anyone else. This would mean that an agent was permitted to perform his preferred act (call it P), provided that there was no alternative A open to him, such that (1) A would produce a better overall outcome than P, as judged from an impersonal standpoint which gives equal weight to everyone's interests, and (2) the total net loss to others of his doing P rather than A was more than M times as great as the net loss to him of doing A rather than P. As this implies, the agent would always be permitted to perform the act that would have the best overall outcome from an impersonal standpoint if he wished to do so. But he would be required to perform the act that would have optimal results in impersonal terms just in case each alternative would involve a total net loss to others more than M times as great as the net loss to him of performing the optimal act instead of that alternative.
. . . a plausible agent‐centred prerogative would allow each agent to assign a certain proportionately greater weight to his own interests than to the interests of other people. It would then allow the agent to promote the non‐optimal outcome of his choosing, provided only that the degree of its inferiority to each of the superior outcomes he could instead promote in no case exceeded, by more than the specified proportion, the degree of sacrifice necessary for him to promote the superior outcome. If all of the non‐optimal outcomes available to the agent were ruled out on these grounds, then and only then would he be required to promote the best overall outcome.3
This proposal was intended as a very rough gesture at specifying the shape of an acceptable agent‐centred prerogative. Obviously, the suggestion of mathematical precision is wholly artificial, and the proposal is in various respects insufficiently fine‐grained. On the other hand, common‐sense morality seems implicitly committed both to an agent‐centred prerogative of some form, and to at least some rough weightings along more or less the lines I suggested. And it is difficult to see how an acceptable non‐consequentialist view could altogether dispense with such weightings. Thus my proposal was meant to provide a working basis, admittedly oversimplified but with roots in ordinary moral thought, for the discussion of agent‐relative permissions.
Now a number of critics have argued that, when taken by itself, a prerogative of the kind I have described ‘will not only permit agents to allow harm, it will also permit agents to do harm in pursuit of their nonoptimal projects . . . For the prerogative is only sensitive to the size of the loss to others, and not to whether the loss is caused by the (p.170) agent's act.’4 This problem will not arise if the prerogative is accompanied and constrained by an agent‐centred restriction against harming, as it appears to be in commonsense morality. For such a restriction will forbid harming people even in order to produce an optimal outcome overall, let alone to secure some advantage for oneself. But, critics charge, a theory that includes a prerogative unaccompanied by agent‐centred restrictions will permit far too much. Or, more precisely, the only way it can avoid permitting too much is by permitting too little. Thus, for example, Shelly Kagan claims that, on such a view,
Kagan's argument is that a hybrid theory must treat cases of murdering one's uncle in order to gain $10,000 for oneself in just the same way that it treats the case of keeping $10,000 that one already has rather than spending it to prevent a stranger's death. For in each case one must choose a cost of $10,000 for oneself or death for someone else, and, Kagan suggests, there are no differences between the two choices that would lead a hybrid theory to say different things about them. So if the theory allows a person to give enough extra weight to his own interests that he is not required to pay $10,000 to save a stranger's life, it will also be committed to allowing him to kill his uncle for a $10,000 gain. And, on the other hand, if it does not allow him to give enough extra weight (p.171) to his own interests that he can permissibly kill his uncle for a $10,000 gain, it must also require him to pay $10,000 to save a stranger's life. Thus, it seems, the theory either permits too much or requires too much.
It will apparently be permissible to kill my rich uncle in order to inherit $10,000. Lest it be suggested that a plausible M will avoid these results, bear in mind that most of us believe that we would not be required to pay $10,000 in order to save the life of some stranger; any M large enough to save such results will obviously work in the former case as well.5
In the next section, I will explore Kagan's example in greater detail. I will argue that his discussion of that example is misleading in at least two important respects. I will also argue that seeing why this is so enables us to identify considerations which, at the very least, support a significantly reduced estimate of the extent to which hybrid views are committed to tolerating harms inflicted in pursuit of one's non‐optimal ends, and which may even militate in favour of an absolute hybrid prohibition against such harms. In Section III I will discuss a different kind of argument in support of an absolute hybrid prohibition against non‐optimal harming.6 If the arguments of these two sections are correct, hybrid views need not have the intolerable implications that Kagan's example is meant to suggest they have. Nevertheless, such views will undoubtedly conflict with some of our first‐order moral convictions, and I will conclude, in Section IV, by considering the implications of this fact.
The agent‐centred prerogative as I described it says that an agent may do what he prefers to do, P, provided there is no impersonally superior alternative A, such that the total net loss to others of his doing P rather than A is more than M times as great as the net loss of him of doing A rather than P. From this it seems to follow that if it is permissible for a person to allow an n‐sized harm to befall someone else in order to avoid a q‐sized cost to himself, then it must, everything else equal, also be permissible for the person to inflict an n‐sized harm directly in order to (p.172) avoid a q‐sized cost to himself. This ‘symmetry condition’ is the troubling feature of hybrid views to which Kagan's example is meant to call attention. Care must be taken in interpreting the implications of this condition, however. For the cost to me of preventing an n‐sized harm from befalling someone else will typically be much greater than the cost to me of failing to inflict such a harm directly myself. To use an example familiar from the literature, the cost to me of preventing deaths by starvation in faraway lands will typically be much greater than the cost to me of not mailing poisoned food packages to people in those lands. It will involve a greater drain on my resources, and a greater narrowing of the morally acceptable options available to me. Thus the symmetry condition will typically not imply that if it is permissible for me to allow death by starvation in faraway lands, it must also be permissible for me to mail poisoned food packages. More generally, what this example suggests is that the symmetry condition will not license as much harmful behaviour as it may originally have seemed to. That it will nevertheless license an intolerable amount of such behaviour is what Kagan's example is meant to show. But that example, or what Kagan says about it, seems to me misleading in a number of respects, of which two seem particularly important.
Before discussing these, we may first note that the scenarios we are invited to compare in Kagan's example are actually asymmetrical in two obvious ways. First, we are asked to compare a killing in one case with a death of some unspecified kind in the other. Second, we are asked to compare the demise of one's uncle with the death of a stranger. A more symmetrical version of Kagan's example might have read:
(p.173) If this revised example has less apparent force than Kagan's original, that is presumably because it is less than obvious that one would not be required to pay $10,000 to prevent the murder of one's uncle if one were in a position to do so. Although this suggests that the asymmetries in Kagan's example contribute to its impact, I will ignore them in the remainder of my discussion, on the assumption that Kagan could have produced cases that were symmetrical in the two relevant respects, and which served his purposes about as well as the pair he actually used.
it will apparently be permissible to kill my rich uncle in order to inherit $10,000. Lest it be suggested that a plausible M will avoid this result, bear in mind that most of us believe we would not be required to pay $10,000 to prevent someone from murdering our uncle; any M large enough to save such results will obviously work in the former case as well.
So let us turn to the first of the two major respects in which Kagan's example, and his discussion of it, are misleading. Kagan tacitly assumes that because in his example not killing and preventing death both involve $10,000 disadvantages, the relevant costs to the agent of the two courses of action are equal. But this simple identification of cost and financial disadvantage is unwarranted. When considering the cost to an agent of a given course of action, the relevant issue is the extent to which the agent's interests will be damaged if he pursues that course of action. And although financial disadvantage represents one important dimension of damage, it is not the only one. Thus, in considering Kagan's example, we may begin by reminding ourselves of the general empirical observation that the psychological costs of surrendering a benefit one already has tend to be greater than the psychological costs of doing without a comparable benefit that one has not yet acquired. This by itself provides some reason for thinking that the cost of preventing death, in Kagan's example, would be greater than the cost of not killing. This conclusion is substantially reinforced if one reflects on the putatively identical benefits that are forgone in the two cases Kagan compares. For, leaving aside psychopaths and the stick‐figure characters who sometimes inhabit philosophical examples, it is wildly implausible to suggest that the net benefit to a typical agent of killing a relative so as to inherit $10,000 is likely to be just as great as the net benefit to him of retaining $10,000 he already possesses rather than using it to (p.174) prevent a stranger's death. It is true that in each case he ends up with $10,000 he would otherwise have done without. But there the similarity ends. In the first case, the $10,000 is obtainable only by killing. If he gets it, he gets it as part of a package that may, depending on the circumstances, include everything from fear, horror, shame, humiliation, disgust, self‐loathing, and punitive atacks of conscience, to the risk of social ostracism, alienated love, imprisonment, and economic ruin, to profound distortions of personality and of the capacity to lead a fulfilling life. In the second case, the $10,000 is, as Kagan presumably wishes us to suppose, a sum that the agent has obtained legally and that he is legally entitled to keep. Depending again on the specific circumstances, which Kagan does not describe, retaining this money might, if the agent is morally sensitive, cause him some more or less serious pangs of guilt, but it is very unlikely to involve any of the severe penalties associated with the acquisition of the first $10,000.
In view of these considerations, it seems fair to say that the costs to the agent of doing without the two sums of money are likely to be very different. In forgoing the first $10,000 he is, given the severity of the penalties attached to it, arguably better off than he would be with the money. In giving up the second $10,000, he avoids no such penalties; he simply suffers an economic loss plus, perhaps, the associated psychological costs of surrendering a benefit that is already in his possession. Since the net cost to the agent of doing without the second $10,000 thus appears to be significantly greater than the net cost to him of doing without the first $10,000, it would seem that an agent‐centred prerogative could, contrary to what Kagan suggests, allow him to keep the second $10,000 without permitting him to kill his uncle so as to inherit the first $10,000.
It may be objected that all that has been established is that, given certain common but not invariant patterns of emotion and motivation, an agent‐centred prerogative could permit an agent to fail to prevent a death without (p.175) being committed to letting him kill. In other words, given that the agent is a normally motivated person in normal circumstances, and given that for such people murdering relatives for profit involves substantial emotional and psychological penalties that failing to spend one's money saving the lives of strangers does not, an agent‐centred prerogative could treat those two courses of action differently. But the possibility of differential treatment depends on the presence of the ‘normal’ pattern of emotion and motivation, and on the asymmetrical attitudes toward killing and letting die that are characteristic of that pattern. Where there was a symmetry of attitude, the agent‐centred prerogative would no longer have any basis for differential treatment. Such a symmetry of attitude could take a variety of forms. Consider two extreme cases. In the first case, both the emotional cost to the agent of killing his uncle so as to inherit $10,000 and the emotional cost of letting the stranger die rather than paying $10,000 to save him would be so great that it would be in the agent's own interest to avoid doing either. In the second case, neither killing nor letting die would involve any emotional cost to the agent, so that each course of action would simply represent to him a net advantage of $10,000. In both of these cases, the objection might run, an agent‐centred prerogative would have to treat killing and letting die symmetrically; in neither case could it allow the agent to keep the $10,000 he already had unless it also permitted him to kill for his inheritance.
The first of these cases does not pose a clear difficulty for a hybrid theory. For such a theory allows a person to give more weight to his own interests than to the interests of other people. But if, ex hypothesi, killing and letting die would both be very costly for the agent, then his interests and the interests of other people would coincide; it would be better from both points of view if he neither killed his uncle nor let the stranger die. So a hybrid view could satisfy the symmetry condition in this case by saying that the agent should save the stranger and leave his uncle alone, thus doing what would be best for himself and for (p.176) everyone else. (More generally, whenever the net costs to an agent of preventing death and avoiding killing are equalized, not by increasing the cost of the latter but by decreasing the cost of the former, a hybrid view can, plausibly enough, satisfy the symmetry condition by requiring prevention rather than by permitting killing.)
The second case, however, seems much more troublesome for a hybrid view. In this case, ex hypothesi, neither killing nor letting die would be costly for the agent, so that his interests and the interests of others would conflict. In order to prevent the death of a stranger, he would have to pay $10,000 without receiving any compensating benefit for himself. Since the agent‐centred prerogative allows him to give more weight to his own interests than to the interests of others, a hybrid view would, it seems, permit him to keep his $10,000. Or if not, there must be some sum large enough that the agent would be allowed to keep it for himself rather than spend it to save the stranger's life. But then it seems that the prerogative must also, by the symmetry condition, permit him to kill his uncle to inherit the same sum of money, which is outrageous. In this instance, the hybrid view seems to reward psychopathy. The person who can kill for profit without any qualms is permitted to do so; only those who have scruples are prohibited from killing.
This objection might be augmented in the following way. An agent‐centred prerogative can permit normally motivated people to fail to save lives without permitting them to kill because normally motivated people have asymmetrical attitudes toward killing and letting die, as a result of which there are differential emotional costs for them associated with these two ways of behaving. But normally motivated people have these asymmetrical attitudes only because they have internalized common‐sense morality, which says that it is intrinsically worse to kill than to let someone die. Hybrid views, by contrast, do not say this. So if hybrid views ever became generally accepted, agents would have symmetrical attitudes toward killing and letting die in cases where the costs to them apart from (p.177) emotional costs were equal, and as a result the emotional costs in such cases would be equal too. And that means that if hybrid views were generally accepted, then such views would no longer have any basis for permitting normally motivated people to fail to save lives without also permitting them to kill. Thus, to put the augmented objection in its strongest form, as long as the moral view that normally motivated people have internalized is common‐sense morality, then hybrid views will by and large allow only psychopaths to kill for personal advantage. But if there were ever a time when normally motivated people had internalized hybrid views, then those views could no longer allow even normally motivated people to let others die without at the same time permitting them to kill.
This objection may be challenged at a number of points. For one thing, it is not clear that the asymmetry of our ordinary attitudes toward killing and letting die really is dependent on our internalization of common‐sense morality. It seems at least as plausible to maintain that the reverse is true: that common‐sense morality appeals to us in part because it endorses an asymmetrical attitude toward doing and letting happen that we find natural independently of moral considerations. In addition, the objection depends on the idea that the penalties associated with killing one's uncle for profit are all psychological, so that they could be avoided by someone with a non‐standard psychology. But the penalties I enumerated were not all emotional or psychological, and it is not clear that a non‐standard psychology would be sufficient to avoid them. More generally, the fact that an agent had no qualms about such a killing would not suffice to show that it was really in his interest to kill, that it really advanced his good.
These points suggest that the differential costs to the agent of not killing and of preventing death may be more deeply rooted than the objection would have it. But there is also another difficulty with the objection. At the beginning of this section I said that Kagan's example, or his (p.178) presentation of it, was misleading in two respects. I have so far mentioned only one of these: namely, the identification in that example of the cost to the agent with his financial disadvantage. This, I have suggested, helps to create the misleading impression that avoiding killing will quite often be just as costly for agents as preventing death. However, we should now take note of the second thing that is misleading about Kagan's example. This has to do, not with what the example suggests about the relative costs to the agent of killing and letting die, but with what it suggests about the costs to everyone else. In particular, the example creates the impression that the overall consequences from an impersonal point of view of killing for personal gain will typically be no worse than the overall consequences of allowing death. For, in the example, the costs to everyone but the agent of killing and letting die are supposed ex hypothesi to be equal (despite the potentially complicating and therefore unhelpful fact, earlier noted, that we are asked to compare an avoidable killing in one case with a preventable death of some unspecified kind in the other). In reality, however, we cringe at the very idea of a world in which individuals routinely form, develop, and pursue personal projects that require them, even within limits, to kill or injure others. By contrast, the consequences of a practice according to which people typically do not expend great portions of their own resources to save the lives of strangers, while certainly far from optimal, are not likely to be nearly as catastrophic. The difference may amount only to the advantage that the present miserable condition of the world enjoys in comparison with a Hobbesian war of all against all, but that advantage, as limited, intermittent, and insecure as it is, is not one that many sane people would be prepared to forgo. Thus there is, in general, an asymmetry between the consequences of people killing the innocent for profit, and the consequences of their failing to expend large sums of money in order to save innocent lives. Failure to reflect this point constitutes the second respect in which Kagan's example is misleading. When it is taken into account, we (p.179) can see that hybrid views do, contrary to the claim of the ‘augmented objection’, have a basis for treating killing and letting die differently, even when the costs to the agent of avoiding each one are the same. For the cost to the agent is not the only factor that determines what an agent‐centred prerogative will allow. The costs to everyone else must also be considered. And when it comes to cases of killing and letting die such as those we have been discussing, these costs are not the same.
It will be observed, however, that the asymmetry I have identified derives from the presumably catastrophic consequences of a general practice of killing to advance one's own good. It has not been suggested, nor is it true, that there is any single act to which these consequences would all attach. But the agent‐centred prerogative as I have described it is sensitive precisely to the consequences of individual actions. It remains unclear, therefore, to what extent and via what mechanism the catastrophic consequences I have spoken of could qualify as relevant from the perspective of the prerogative, and result in a prohibition of harmful activity designed to advance the individual good.
Similar questions arise, of course, with regard to consequentialist claims about the undesirable effects of a general practice of harming others for the overall good. In that case, such questions have provided one motivation for a move to ‘two‐level’ views, in which consequential assessment is applied to competing rules, policies, or motives rather than to individual acts. It might therefore be suggested that if the disastrous consequences of a general practice of non‐optimal harming cannot be fully taken account of in a hybrid theory's act‐by‐act assessment of individual harms, then this constitutes an argument for a ‘rule‐hybrid’ prohibition against all such harming.7 Of (p.180) course, overall consequences do not carry as much weight in a hybrid theory as they do in a consequentialist theory. In a consequentialist theory, the overall consequences are all that matter. In a hybrid theory, they must be balanced against the agent's interests. Thus, whereas two‐level consequentialist restrictions are defended by appeal to consequential considerations alone, a rule‐hybrid prohibition against non‐optimal harming would have to be defended by appeal to what might be called ‘adjusted consequential considerations’: that is, to the net effect of taking into account both the relevant considerations of overall consequences and the relevant agent‐centred considerations. The argument would have to be that the impersonal advantages of an absolute prohibition against non‐optimal (p.181) harming are great enough to outweigh the costs to agents of having such conduct ruled out.
This does not seem like an implausible thing to claim, but the argument would have to be spelled out in greater detail in order to be persuasive. In particular, a fuller explanation would have to be provided of how, at the level of rules, impersonal and agent‐centred considerations were to be balanced against one another. And even if the argument were presented more fully, it would undoubtedly remain controversial. Some might argue instead that what the appeal to adjusted consequences supports is a rule of thumb that prohibits non‐optimal harming in general, but which fails to apply to individual harms whose adjusted consequences are, by themselves, clearly insufficient to render those harms impermissible. Others might argue that what such an appeal demonstrates is that beyond a certain threshold, non‐optimal harms will have sufficiently bad consequences that they will be disallowed by the prerogative as originally described on the basis of the usual kind of act‐by‐act assessment.
It is evident from this that there is room for considerable disagreement about the way in which the relevant consequential considerations are most appropriately to be accommodated within a hybrid theory. But however this disagreement may be resolved, the importance of those considerations for such a theory seems undeniable. Whether or not they provide support for an absolute hybrid prohibition against all non‐optimal harming, they clearly serve in one way or another to attenuate the degree of a hybrid view's toleration of such harms.
Of course, however, if one doubts whether adjusted consequential considerations do support an absolute prohibition against non‐optimal harming, and if one regards anything less than an absolute prohibition as insufficient, then one is bound to conclude that hybrid theories remain excessively tolerant of this type of behaviour, even if they are less tolerant than Kagan implies. The basic problem, one is likely to feel, is that the purely quantitative structure of the agent‐centred prerogative renders it incapable of (p.182) assigning sufficient importance to the qualitative distinction between harming and failing to prevent harm. In the next section, I will discuss a different kind of argument for an absolute hybrid prohibition against non‐optimal harming: or, as I shall put it, for replacing the ‘pure‐cost’ version of the agent‐centred prerogative with a ‘no‐harm’ version. If this argument is persuasive, then the prerogative need not have a purely quantitative structure.
We may begin by recalling the claim that it is difficult to identify an underlying principled rationale for agent‐centred restrictions. Briefly, the difficulty arises because the restrictions may not ordinarily be violated even to produce optimal outcomes overall. This means that there are circumstances in which they may not be violated, even though a violation would serve to minimize total overall violations of the very same restriction, or other events at least as objectionable, and would have no other morally relevant consequences. And it is very hard to explain how it can be rational to forbid the performance of a morally objectionable action that would have the effect of minimizing the total number of comparably objectionable actions that were performed, and would have no other relevant effects.8
Thus the difficulty in providing a satisfactory rationale for agent‐centred restrictions derives from the fact that the restrictions prohibit acts of certain kinds even when they would produce optimal outcomes overall. At the same time, the objection to hybrid views that we have been considering arises from their apparent failure to prohibit acts of certain kinds even when they would produce non‐optimal outcomes. A modified hybrid view, which included (p.183) a no‐harm version of the prerogative instead of a purecost version, would avoid both difficulties. Like the original hybrid view, it would avoid the first difficulty because it would always permit the production of the best available outcome overall. And by expressly prohibiting all non‐optimal harms, it would avoid the second difficulty as well.9 Thus it would not give agents unqualified permission to devote proportionately greater weight to their own interests than to the interests of other people. Rather, it would only permit agents to do this provided they did not harm others in pursuit of their non‐optimal ends.10
In The Rejection of Consequentialism I argued that it was possible to identify a rationale for an agent‐centred prerogative, but I also argued that the rationale in question underdetermines the specific form taken by the prerogative.11 It is therefore tempting to reason as follows. If the pure‐cost version of the prerogative is subject to an objection that the no‐harm version avoids, and if the rationale (p.184) for an agent‐centred prerogative underdetermines the choice between the two versions, then, other things equal, the fact that only the no‐harm version avoids the objection constitutes a reason for accepting that version in preference to the pure‐cost version. Reasoning in this way, we would conclude that the force of Kagan's example is not to cast doubt on hybrid theories in general, but rather to provide support for those hybrid theories that include a no‐harm prerogative as against those that include a pure‐cost prerogative.
Before this simple solution can be accepted, however, there is an objection to it that must be answered. The objection may be put as follows.12 The rationale for a prerogative is a rationale for allowing people to devote some proportionately greater weight to their own interests than to the interests of other people. This rationale in itself provides no basis for distinguishing between a case in which one directly harms another in pursuit of one's own non‐optimal ends, and a case in which one merely fails to prevent a harm of comparable magnitude because of one's devotion to the pursuit of those same ends. That is, it provides no basis for distinguishing between such cases provided the size of the loss in aggregate well‐being and the size of the advantage to the agent each remains the same in the two cases. The modified hybrid view, however, treats at least some such pairs of cases differentially. There are times when it prohibits an agent from harming an innocent person in order to secure some advantage for himself, but permits him to secure the same advantage in some other way, even at the cost of failing to prevent a harm just as great as the first one from befalling the same innocent person. The modified hybrid view is thus committed to the claim that it is (sometimes) worse to inflict a harm oneself in order to obtain some personal (p.185) advantage than it is to allow a comparable harm to occur in order to obtain the very same advantage. And it is hard to see how there could be considerations that supported this claim without implying that harming is in general worse than allowing harm. But this is just to say that any considerations capable of supporting prohibitions against harming in pursuit of one's non‐optimal ends would also support agent‐centred restrictions that prohibit harming even when it is necessary to prevent still greater harm and so to achieve a better outcome overall. And, contrapositively, any doubts there may be about the rationale for agent‐centred restrictions must also be doubts about the rationale for prohibitions against non‐optimal harming. Thus the modified hybrid view is not a stable theoretical option, for one cannot consistently accept a no‐harm version of the agent‐centred prerogative without also accepting agent‐centred restrictions, and one cannot reject the restrictions without also rejecting the no‐harm version of the prerogative.
At one time I was inclined to regard this objection as decisive. As a result, I was led away from the no‐harm version of the prerogative I had originally formulated,13 to the pure‐cost version sketched in my book. (Thus, purely as a matter of autobiography, what I am now calling the ‘modified’ hybrid theory was really the original version, and what I am now calling the ‘original’ hybrid theory—that is, the one sketched in my book—was in fact a later modification.) However, the objection no longer seems to me conclusive, for the following reasons.
The objection derives its force from the suggestion that there could not, in principle, be a reason for prohibiting non‐optimal harms more strenuously than non‐optimal failures to prevent harm that did not constitute an equally good reason for holding that harming is in general more objectionable than failing to prevent harm. However, the discussion in the previous section has already given us (p.186) some reason to reject this suggestion. For, as that discussion made clear, it would be perfectly consistent for someone to hold that adjusted consequential considerations support a hybrid restriction on non‐optimal harming, while denying that harming is in general worse than failing to prevent harm, or that it is wrong to inflict a harm when doing so will produce optimal results overall. Such a position may or may not be correct. Even if one does not regard it as correct, however, the fact that it is a consistent position, and one that is not obviously unreasonable, suffices to cast doubt on the suggestion that there could not, in principle, be a reason for prohibiting non‐optimal harms more strenuously than non‐optimal failures to prevent harm that was not also a reason for agent‐centred restrictions. Right or wrong, the position offers us one example of something that would count as such a reason.
Here is a second example. There are, it might be said, certain quasi‐practical advantages that prohibitions against harming have in comparison with requirements to prevent harm. The former are more easily taught, internalized, and obeyed than are the latter. Moreover, prohibitions against harming help to demarcate a conception of moral responsibility that coheres smoothly with our natural conceptions of agency and causal responsibility. These considerations, it might be claimed, make it reasonable that a hybrid view should prohibit non‐optimal harms more strenuously than non‐optimal failures to prevent harm. For, given that some but not all non‐optimal conduct is, for independent reasons, to be permitted, such considerations provide a plausible rationale for what is, in effect, one way of drawing the line between permissible and impermissible types of non‐optimal behaviour. (Adjusted consequential considerations might be said to provide an additional rationale for the same way of drawing that line.) By themselves, however, these quasi‐practical considerations provide no support for prohibitions against harms that will produce optimal results overall, for they do not themselves give us any reason to think that such harms are, on balance, morally undesirable.
(p.187) This argument seems to me to have considerable force. Again, however, whether or not one finds the argument persuasive, it is in any case consistent and not obviously unreasonable. And the mere availability of such an argument provides additional evidence against the claim that there could not, in principle, be a reason for prohibiting non‐optimal harms more strenuously than non‐optimal failures to prevent harm that did not constitute an equally good reason for agent‐centred restrictions. It suggests that there is plenty of conceptual space available in which such a reason might be located.
In short, the two examples I have given may convince some people that there are considerations which support a no‐harm version of the prerogative without also supporting agent‐centred restrictions. But it is sufficient for my purposes if those examples are taken only to establish that there could be such considerations. For that weaker conclusion suffices to rebut arguments which deny the possibility of such considerations, and which seek thereby to challenge the modified hybrid view's status as a consistent theoretical option.
Where does this leave us? At the beginning of this section I acknowledged a temptation to argue that if the pure‐cost version of the prerogative is subject to an objection that the no‐harm version avoids, and if the rationale for an agent‐centred prerogative underdetermines the choice between the two versions, then, other things equal, the fact that the no‐harm version avoids the objection constitutes a reason for incorporating that version into a hybrid theory in preference to the pure‐cost version. The claim that acceptance of a no‐harm prerogative would commit one to agent‐centred restrictions posed a threat to this line of thought, for it cast doubt on the idea that a hybrid theory thus modified would be a stable theoretical position. Although that claim has now been rebutted, the line of thought it threatened has not been fully vindicated, because the reason for accepting, a no‐harm prerogative that is suggested by this line of thought is inadequate. The putative reason amounts to saying that we should accept (p.188) a no‐harm prerogative because there is a rationale for a prerogative of some form, and because a pure‐cost version has intolerably counterintuitive implications. But although, as I have argued, there may in principle be considerations that provide support for a no‐harm version of the prerogative without supporting agent‐centred restrictions as well, this is not a good example of such a consideration, since the appeal of agent‐centred restrictions on the intuitive level is also considerable. Moreover, the very enterprise in which I have been engaged consists in looking past the intuitive appeal of various moral doctrines, in an attempt to discover whether deeper rationales for them can be identified. I have argued, for example, that agent‐centred restrictions seem paradoxical despite their intuitive appeal, and that their underlying rationale remains obscure. Again, therefore, I can hardly maintain that the intuitive appeal of a no‐harm prerogative suffices to establish its acceptability. So the line of thought sketched above has not been fully vindicated. On the other hand, neither has it been entirely discredited. A better reason is required for accepting a no‐harm prerogative than the one it advances. However, we have now had occasion, in the course of discussing the objection that threatened that line of thought, to identify two possible candidates for such a reason, one or both of which will be found persuasive by some. And as we have seen, even people who find both of those reasons unpersuasive should agree that we have been given no principled grounds for thinking that a satisfactory reason cannot be found. Nor do prohibitions against non‐optimal harming seem paradoxical in the way that agent‐centred restrictions do. Thus if one believes that no satisfactory reason for accepting a no‐harm prerogative has yet been identified, what can nevertheless be said to have been established is that there is a rationale for an agent‐centred prerogative of some kind, and that such a prerogative need not take a pure‐cost form.
Put this way, the point is parallel to a point about egoism that I made in The Rejection of Consequentialism. There I noted that the choice between a pure‐cost version of the (p.189) prerogative and an egoist version allowing each agent always to advance his own interests was underdetermined by the rationale for the prerogative that I had identified. What I nevertheless claimed to have shown was that there was a rationale for a prerogative of some kind, and that such a prerogative need not take an egoist form. The question of why the prerogative should not take an egoist form—the question what is wrong with egoism—was said to be beyond the scope of my inquiry, and was therefore left open.14 In a similar spirit, what can be said in the present context, by those who believe that no adequate reason for accepting a no‐harm prerogative has yet been identified, is that there is a rationale for an agent‐centred prerogative, and that such a prerogative need not take a pure‐cost form. It can take a no‐harm form instead. The question of why, ultimately, it should take a no‐harm form remains open, on this view. Even on this view, however, we need not feel that if we believe there to be a rationale for an agent‐centred prerogative, we are then faced with a choice between a pure‐cost version on the one hand, and a version that commits us to agent‐centred restrictions on the other.
If the arguments of the preceding two sections are correct, hybrid views need not have the disturbing implications that Kagan's example is meant to suggest they have. This does not mean, however, that hybrid views are wholly congenial to moral common sense. On the contrary, there has never been much doubt that they were counterintuitive to one degree or another. After all, one of their defining features is that they dispense with agent‐centred restrictions despite the intuitive appeal of such restrictions. The interesting question is not whether this is so but rather what it shows.
(p.190) Those sympathetic to consequentialism might argue that if no plausible rationale for agent‐centred restrictions can be found, and if hybrid theories have implications that conflict with first‐order moral convictions in which we have great confidence, then that leaves consequentialism as the preferred option. This is a dangerous argument for a consequentialist to offer, however, since consequentialism itself has implications that conflict with first‐order moral convictions in which we have great confidence.
Those sympathetic to common‐sense morality might argue instead that if it is indeed possible to identify a rationale for an agent‐centred prerogative, and if a moral theory that includes a prerogative but no restrictions is unacceptable, then perhaps the rationale for such restrictions is that they are essential to the design of an acceptable prerogative. This initially surprising thought, that agent‐centred restrictions may depend for their rationale on the agent‐centred prerogative, is not altogether unappealing. If the argument that leads to this conclusion is to be made persuasive, however, the idea that a prerogative without restrictions is ‘unacceptable’, or that restrictions are ‘essential’ to the design of an ‘acceptable’ prerogative, must be given additional content. All that has so far been supposed is that a prerogative unaccompanied by restrictions would have some implications that seem counter‐intuitive and which could be avoided by the inclusion of restrictions. But there was never any doubt about the intuitive appeal of agent‐centred restrictions. The question was whether, in addition to their intuitive appeal, they could be seen to have a plausible underlying rationale, and that question is not answered by simply calling attention once again to their intuitive appeal. This is of course the same point as the one made earlier about the no‐harm version of the prerogative. The fact that there is a rationale for an agent‐centred prerogative of some kind, and that such a prerogative would be seriously counterintuitive if it did not take a no‐harm form, or if it were unaccompanied by agent‐centred restrictions, does not by itself amount to a rationale either for the no‐harm version of the prerogative (p.191) or for the restrictions. Rather, it provides us with an even greater incentive to identify such rationales if we can. Moreover, in the case of the restrictions there is the additional problem that they seem paradoxical despite their intuitive appeal, and that it has proved surprisingly difficult to provide them with a rationale capable of dispelling this air of paradox. If it is true that a prerogative without restrictions would have seriously counterintuitive implications, that fact certainly does not solve this problem or make it go away. On the contrary, it makes it appear even more urgent, because it emphasizes the centrality and importance to us of the seemingly paradoxical moral idea that agent‐centred restrictions embody.
What the two opposing arguments just considered have in common is the assumption that, if hybrid views can be shown to conflict with some of our intuitive moral convictions, then the primary effect is to eliminate such views as candidate accounts of the content of morality. This presupposes that the main interest of hybrid views is as ‘accounts’ of this kind. In my view, however, the most important role of hybrid theories is in helping to frame two challenges: one to consequentialism, and one to common‐sense morality. The challenge to consequentialism is to articulate a more compelling reason why we should accept the hegemony of agent‐neutral considerations in ethics, in view of the fact that it is possible to provide a plausible rationale for one major agent‐relative principle, and in view of the fact that the most obvious objection to departing from consequentialism to the extent of accepting such a principle, namely the counterintuitiveness of some of the results, is one that consequentialism itself appears unable consistently to advance. The challenge to common‐sense morality is different. Since the common‐sense view includes both an agent‐centred prerogative and agent‐centred restrictions, and since the former appears to be better motivated than the latter, the challenge is to identify a plausible rationale for the restrictions, or else to show why the absence of such a rationale is not troubling.
The effectiveness of hybrid views in helping to frame (p.192) the challenges I have described does not depend on their intuitive appeal. Indeed, the less comfortable we are with hybrid views when they are conceived of as embodying an alternative moral outlook in their own right, the more pressing it becomes to find some way of meeting the challenges they present. For those of us who are sympathetic to common‐sense morality the challenge is, in effect, to undertake an exercise in moral self‐understanding: to excavate the sources of the common‐sense principles that we find congenial, in the hope of discovering a more satisfactory way of explaining them, and our allegiance to them, than has hitherto been identified. The less comfortable we are with hybrid views as an alternative to common‐sense morality, the more important it becomes for us to meet this challenge, and thus to convince ourselves that the principles we accept can be explained to our own satisfaction.15
A version of this piece was published in Philosophical Perspectives 6 (1992): 377–97.
(1) I interpret such non‐consequentialist views as holding that each individual has a single general prerogative to devote disproportionate attention to his or her own interests, within certain limits, but that there are many different restrictions that apply to each agent, with each restriction specifying some type of prohibited act. That is why I typically speak of an agent‐centred prerogative (singular), but of agent‐centred restrictions (plural). Of course, there are contexts in which it is appropriate to use the plural for the former as well; since each agent has a prerogative, what groups of agents have are prerogatives. Hence the title of this paper.
(2) Abbreviated in subsequent notes as RC.
(3) RC, p. 20.
(4) Shelly Kagan, ‘Does Consequentialism Demand Too Much?’, Philosophy and Public Affairs 13 (1984): 239–54, at p. 251. The same argument is made by Frances Kamm in ‘Supererogation and Obligation’, Journal of Philosophy 82 (1985): 118–38; and by David Schmidtz in his review of RC in Noûs 24 (1990): 622–7. Kagan's formulation is endorsed by Larry A. Alexander in ‘Scheffler on the Independence of Agent‐Centred Prerogatives from Agent‐Centred Restrictions’, Journal of Philosophy 84 (1987): 277–83. Kagan develops the same point in a more general context in his book The Limits of Morality (Oxford: Clarendon Press, 1989), pp. 19–24.
(5) ‘Does Consequentialism Demand Too Much?’, p. 251.
(6) I am treating a hybrid prohibition against harming others in pursuit of one's non‐optimal ends as equivalent to a hybrid prohibition against non‐optimal harming. My reasons for doing so are explained in n. 7 below.
(7) Note that a hybrid prohibition against all non‐optimal harming would forbid such harms even when they would produce better overall results than some other acts that the agent could permissibly perform. For example, suppose that Edward has three options. Option A is what he would most like to do, and although it would have the worst overall results of the three options, it would not involve Edward's harming anyone, nor would it require the assignment of so much extra weight to Edward's own interests as to render it impermissible from the standpoint of a hybrid theory. Option B would have better overall results than option A, but it would require Edward to inflict a harm, and, since it is not what he would most like to do, it would also represent a sacrifice for him relative to A. Option C would have the best overall results of any act available to Edward, and it would not require him to harm anyone, but it would require him to sacrifice his own interests to an extraordinary degree. A hybrid theory whose agent‐centred prerogative included a prohibition against non‐optimal harming would permit Edward to take either Option A or Option C, but it would prohibit Option B. Nor would these verdicts be different if C also required Edward to inflict a harm.
Since it may not seem natural to think of B as a case of harming someone in pursuit of one's own non‐optimal ends, or as a way of advancing one's own good, it may seem that B is not the type of act to which the consequential considerations I have been discussing above apply. And since a general hybrid prohibition against non‐optimal harming would rule out acts like B, it may seem that such a prohibition is too broad to be considered as a candidate response to those considerations. It may seem that a restriction against harming in pursuit of one's non‐optimal ends should be considered instead. However, for Edward to take Option B, when an act with better overall results was available, would be for him to inflict a harm which served to secure a relative advantage for himself and his purposes at the expense of a worse outcome overall. That is sufficient, in my view, for B to count as a harm inflicted in pursuit of Edward's non‐optimal ends, and to fall within the scope of the consequential considerations in question. Thus, when I speak of harms inflicted to advance one's own good, or in pursuit of one's non‐optimal ends, these expressions (and others like them) should be understood broadly to include harms like B. Given this usage, a hybrid prohibition against harming others in pursuit of one's non‐optimal ends is equivalent to a hybrid restriction against non‐optimal harming.
(9) Here, and throughout most of this paper, I make the simplifying assumption that the second objection is concerned exclusively with the non‐optimal harms that an unmodified hybrid theory appears to allow. Since this is in fact the way Kagan formulates the objection, this assumption is a reasonable one to make in the context of a reply to his particular formulation. However, the same sort of objection might equally well focus on a hybrid theory's apparent failure to prohibit non‐optimal acts of other morally abhorrent types, e.g. lies, theft, etc. Although, for the most part, it is only non‐optimal harms that I explicitly discuss in this paper, I believe that my arguments apply, mutatis mutandis, to these other types of act as well. If this is correct, then the term ‘no‐harm prerogative’ is something of a misnomer, for such a prerogative, fully characterized, would prohibit more than just harming.
(10) A modified hybrid theory with a no‐harm prerogative would differ from the system Kagan describes in his book as ‘neo‐moderate’ (The Limits of Morality, pp. 186–94). For the latter includes, instead of a prohibition against all non‐optimal harming, a prohibition against harming ‘in those cases where this leads to a decline in overall good’ (p. 188); and, instead of a permission to inflict optimal harms only, it includes a permission to inflict any harm ‘that leads to better consequences overall’ (p. 192). Kagan refers to these neo‐moderate features as a ‘zero threshold constraint’ against harming and a ‘zero threshold option’ to harm respectively. The way in which the modified hybrid view differs from the neo‐moderate system, however, is not by imposing any absolute non‐zero thresholds, but rather by maintaining that harms must be optimal—relative to the options available—in order to be permissible.
(12) In ch. 4 of his Ph.D. dissertation (‘The Limits of Morality’, Princeton University, 1982), Kagan directs an objection of this kind against ‘neo‐moderate’ views. However, in his recent book of the same title, which is based on his dissertation, Kagan has eliminated this objection from his discussion of the neo‐moderate position (The Limits of Morality, ch. 5).
(13) In my Ph.D. dissertation, ‘Agents and Outcomes’ (Princeton University, 1977).
(14) RC, p. 70
(15) Earlier versions of this paper were presented to a number of audiences, and many people offered valuable criticisms and suggestions. I am especially grateful to G. A. Cohen, Frances Kamm, and Jeremy Waldron, each of whom provided very helpful written comments on one draft or another.