Jump to ContentJump to Main Navigation
The Rejection of ConsequentialismA Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions$

Samuel Scheffler

Print publication date: 1994

Print ISBN-13: 9780198235118

Published to British Academy Scholarship Online: November 2003

DOI: 10.1093/0198235119.001.0001

Show Summary Details

(p.152) Deontology and the Agent: A Reply to Jonathan Bennett

(p.152) Deontology and the Agent: A Reply to Jonathan Bennett

Source:
The Rejection of Consequentialism
Author(s):

Samuel Scheffler

Publisher:
Oxford University Press
DOI:10.1093/0198235119.005.0002

Abstract and Keywords

Scheffler argues, in response to criticisms by Jonathan Bennett, that there is a non‐trivial sense of ‘agent‐centredness’ that applies to both an agent‐centred prerogative and agent‐centred restrictions. In further articulating this notion of agent‐centredness, Scheffler anticipates and responds to four objections that Bennett might make to his argument. He also discusses the relation between agent‐centredness and the doing/allowing distinction.

Jonathan Bennett objects to my use of the term ‘agent‐centred’. He thinks that there is no interesting notion of agent‐centredness that applies equally well to what I call an ‘agent‐centred prerogative’ and to what I call ‘agent‐centred restrictions’. Thus, applying the term ‘agent‐centred’ to both the prerogative and the restrictions is misleading, for it wrongly suggests that they have some interesting property in common.

Bennett says that there is a trivial sense in which the prerogative and the restrictions both can be said to be agent‐centred: that is, they both speak to the question of how agents ought to conduct themselves. However, Bennett rightly indicates that this notion of agent‐centredness will not serve my purposes. After all, in this sense even consequentialism is agent‐centred, for it too speaks to the question of how agents ought to conduct themselves.

Bennett believes that there is no non‐trivial sense in which the restrictions can be said to be agent‐centred, but he thinks that a more substantive notion of agent‐centredness may be available in the case of the prerogative. For the prerogative, he says, is a permission ‘to be guided by what is best for oneself rather than by what is best’, and this, he thinks, provides ‘a good, solid notion of agent‐centredness’.1 Having said this, however, he quickly (p.153) goes on to add that ‘even here there is a complication’ (p. 59). This complication consists in an argument, attributed to Frances Howard, which it will be illuminating to consider. In brief, this argument maintains that the prerogative as I characterize it is really a permission (sometimes) to do what one wants to do even if that will not produce the best results overall; that as such it has the same ‘practical import’ as a permission not always to promote optimal outcomes; that it could therefore be ‘thrown out’ in favour of a permission of the latter type; and that, in consequence, ‘even on the “prerogative” side of Scheffler's line the notion of agent‐centredness does no real work’ (p. 59).

Bennett tries to devise a response to this argument on my behalf so as to rescue me from its damaging conclusion, but the reply he constructs is not one that I myself would offer. Instead, I would make the following observations. In the first place, the two permissions described in the argument are not equivalent as they stand, and they do differ in their ‘practical import’, so that the first of them could not be ‘thrown out’ in favour of the second. They are not equivalent because the claim that one may perform some non‐optimal acts, although entailed by the claim that one may perform some non‐optimal acts that one wants to perform, does not entail it. For from the fact that there are some non‐optimal acts that one is permitted to perform, it obviously does not follow that any of those acts is an act that one wants to perform. Nor do the two permissions have the same ‘practical import’, as is evident from the fact that if I have resolved never to act impermissibly, then one of the permissions assures me that I will, compatibly with that resolution, be able to perform some of the non‐optimal acts that I would like to perform, whereas the other gives me no such assurance and leaves open the possibility that, in view of my resolution, I will have to refrain from performing any of the non‐optimal acts that I would like to perform. It may therefore make a considerable practical difference to me which permission I have. Since the two permissions are thus neither equivalent (p.154) nor possessed of the same practical import, the claim that one could be ‘thrown out’ in favour of the other is false.

More important, however, suppose that the agent‐centred prerogative could be described simply as a permission not always to promote optimal outcomes. Would it follow that the notion of agent‐centredness was doing ‘no real work’? The claim that this would follow is based on the idea that a prerogative describable in this way would ‘clearly’ not involve ‘non‐trivial agent‐centredness’ (p. 59). But if this is meant to suggest that such a prerogative would be agent‐centred only in the trivial sense earlier identified—only, that is, in the sense of being addressed to the question of how agents ought to conduct themselves—then it is false. The force of the adjective ‘agent‐centred’ would here be greater than that; it would be to indicate that the permissibility of what one does is not entirely determined by the impersonal optimality of the results. This means, among other things, that consequentialist theories would not be agent‐centred in this sense, as they would be if ‘agent‐centred’ were taken in the trivial sense of ‘being concerned with the conduct of agents’. This shows both that the present notion of agent‐centredness is not the same as the earlier, trivial notion and that the present notion does ‘real work’. Excluding consequentialism is ‘real work’.

This point is significant because the notion of agent‐centredness just suggested actually comes closer to what I had in mind than any of the other candidates Bennett considers, including the ‘good, solid’ notion of agent‐centredness that he himself thinks may apply to the prerogative. I did try to explain, toward the beginning of The Rejection of Consequentialism, what I meant by the term ‘agent‐centred’. In the course of introducing the prerogative I wrote: ‘Such a prerogative would be a genuinely agent‐centred prerogative, for it would have the function of denying that what an agent is permitted to do in every situation is limited strictly to what would have the best overall outcome impersonally judged’ (p. 14). Thus, to elaborate, what makes an agent‐centred prerogative (p.155) agent‐centred is the fact that it makes the rightness or wrongness of certain acts turn, at least in part, on considerations about the act or the agent which are assigned significance independently of any appeal to overall (actual or expected) value or disvalue as judged from an impersonal standpoint. Similarly, what makes agent‐centred restrictions agent‐centred is the fact that they make the rightness or wrongness of certain acts turn, at least in part, on considerations about the act or the agent which are assigned significance independently of any appeal to overall (actual or expected) value or disvalue as judged from an impersonal standpoint. Note that consequentialism is not agent‐centred in this sense, so there is thus nothing trivial about the notion understood in this way.2

Thus, pace Bennett, there is a notion of agent‐centredness which is not trivial and which does apply equally well both to the prerogative and to the restrictions. Bennett fails to take this notion seriously, in part at least because he appears mistakenly to equate it with the genuinely trivial notion of ‘being concerned with the conduct of agents’, which he rightly dismisses. From various things that he says, however, it is possible to imagine four different objections Bennett might make to my reliance on the notion of agent‐centredness I have articulated.

First, it might be said that after denying that

  1. (a) a permission to do what one wants to do even if that will not produce the best results overall

could be thrown out in favour of
  1. (b) a permission not always to produce the best results overall,

I have now gone ahead and conceded that the agent‐centred prerogative should be understood on the model of b anyway. This concession, however, is misleading and unwise. For my book clearly is concerned with something (p.156) like a, as opposed to other sorts of permissions not to produce optimal results, such as permissions not to optimize when doing so would violate deontological principles. Thus, to use Bennett's terminology, what I have done here and in the passage from my book that I have quoted is to define the agent‐centred prerogative as simply a ‘permissive departure from consequentialism’, even though it is clearly intended to express a permissive version of a ‘morality of personal concerns’, and even though the concept of a permissive departure from consequentialism is not coextensive with the concept of a permissive morality of personal concerns.

It is worth noting that since it relies on the idea that there is an important difference between a and b, this criticism is inconsistent with the suggestion that a could be thrown out in favour of b. Since that suggestion has already been shown to be false, however, the fact that it is inconsistent with the present criticism is in itself no objection to the latter. Where the present criticism does go wrong is in failing to distinguish between my notion of agent‐centredness and what is, in effect, the complete stipulative definition I give of ‘agent‐centred prerogative’. An agent‐centred prerogative, as I frequently say in the book, is a prerogative that gives each agent permission to devote energy and attention to his projects out of proportion to their weight in an impersonal ranking of overall states of affairs. I also emphasize (pp. 41, 69–70) that such a prerogative, so defined, could take very different forms and that my focus in the book is on a certain range of those forms, roughly, those that make the prerogative responsive to considerations about integrity, but do not license unlimited or unconditional attention to one's own projects and interests. For present purposes, the salient point is that what I call an ‘agent‐centred prerogative’ is in fact to be understood more on the model of a than on the model of b. Nor have I said otherwise in this paper. All I have said is that what makes it appropriate to use the term ‘agent‐centred’ as part of the label with which I refer to such a prerogative is the fact that the prerogative makes (p.157) the rightness or wrongness of certain acts rest, at least in part, on considerations about the act or the agent which are assigned significance independently of any appeal to overall value. It is perfectly true that a permissive departure from consequentialism could be agent‐centred in this sense without qualifying for the label ‘agent‐centred prerogative’ as I use it. Bennett's example of a permission not to optimize when doing so would violate deontological rules is a case in point. All this shows, however, is that the adjective ‘agent‐centred’, as it occurs in the term ‘agent‐centred prerogative’, doesn't convey complete information about the kind of prerogative the latter term refers to, and that someone else could legitimately use the term ‘agent‐centred prerogative’ to refer to a different kind of permissive departure from consequentialism. Neither of these points, however, constitutes an objection to my usage, and parallel points hold for many other bits of technical jargon in philosophy (compare ‘difference principle’, say, or ‘ontological argument’).

It must be admitted, however, that there is a slightly different relation between my notion of agent‐centredness and my characterization of the kind of prohibition I refer to as an ‘agent‐centred restriction’. ‘An agent‐centred restriction’, I say in the book, ‘is a restriction which it is at least sometimes impermissible to violate in circumstances where a violation would prevent either more numerous violations, of no less weight from an impersonal point of view, of the very same restriction, or other events at least as objectionable, and would have no other morally relevant consequences’ (p. 80). Now just as an agent‐centred prerogative, as I understand it, could take very different forms, so too there could be many different agent‐centred restrictions, in my sense of the term. And just as I focus exclusively on a certain range of forms that an agent‐centred prerogative could take, so too I focus only on a certain class of agent‐centred restrictions, namely, those that occur in standard deontological theories. However, while there could also be a permissive departure from consequentialism that was agent‐centred but did not qualify (p.158) as an agent‐centred prerogative in my sense, it is not the case that there could be what Bennett calls a ‘mandating’ departure from consequentialism that was agent‐centred but did not qualify as an agent‐centred restriction in my sense. For instance, Bennett's example of a moral prohibition against optimizing when doing so ‘would be less favorable to satisfying one's natural wants’ (p. 56) counts as an agent‐centred restriction in my usage, albeit not one that I focus on, since it is not part of standard deontological theory as I understand it. Although this represents a genuine asymmetry between the prerogative and the restrictions as I have characterized them, it is not, so far as I can see, incompatible with anything that I say or imply about either.

Let me sum up my reply to the first objection. There is, as I have insisted, a non‐trivial notion of agent‐centredness that applies equally well to what I call an ‘agent‐centred prerogative’ and what I call ‘agent‐centred restrictions’. The applicability of this notion does not by itself suffice to make the agent‐centred prerogative responsive to considerations about the moral significance of the individual's desires, projects, and point of view. Nevertheless, the full characterization I give of the prerogative makes it clear that, so characterized, it is indeed responsive to those considerations. Thus, the objection fails because the agent‐centred prerogative as I construe it is properly understood more on the model of a than on the model of b, despite the fact that the term ‘agent‐centred’, although by no means insignificant, does not by itself convey that information.

This reply to the first objection may, however, give rise to a second objection. I have agreed that my focus in the book is restricted to a subset of those departures from consequentialism that could legitimately be called ‘agent‐centred’ in my sense.3 This subset includes permissions (p.159) that allow one to devote disproportionate but not unlimited attention to one's own projects, and prohibitions of the kinds that figure in standard deontological theories. The focus on this subset may lead to the charge that there is no adequate motivation for singling out these particular departures from consequentialism for joint treatment. For, it may be said, all that they have in common is their agent‐centredness and, as I have conceded, they have that much in common with other imaginable departures from consequentialism. To this objection my reply is simple. What the departures I single out have in common is that they are, in my view, the most conspicuous and important agent‐centred components both of common‐sense morality and of standard deontological views. That is the reason for focusing on them as opposed to other imaginable agent‐centred departures from consequentialism, and it seems to me to be reason enough. This is obviously not to deny that an investigation of the other departures might also be interesting, or that different principles for grouping various departures together might be appropriate in other contexts and for other purposes.

This reply in its turn leads naturally to the expression of the third and fourth objections. The third objection is that my language in the book creates the impression that there is more to agent‐centredness than I am now saying there is: that it is a richer or more interesting notion than the one I have articulated here. Accordingly, it may be said, the book gives readers the impression that the prerogative and the restrictions have something more in common than I am now claiming they do. The fourth objection, which complements the third, is that by presenting my enterprise as a comparative investigation of two agent‐centred departures from consequentialism, I falsely suggest that what is most interesting, salient, or significant about each is precisely the property of agent‐centredness (p.160) which it shares with the other, and I obscure or direct attention away from the really significant features of each departure, features that are not the same in the one case as in the other. In the case of the restrictions, most notably, what is salient and problematic about them is not that they are agent‐centred but that they rely on a distinction between doing and allowing, or some related contrast.

In reply to the third objection, I can only say that I certainly did not intend in my book to suggest any notion of agent‐centredness other than the one I have articulated here. As I have noted, my explanation of the term ‘agent‐centred’ in the book, although doubtless too brief, does in fact invoke the same notion of agent‐centredness as the one presented here. Nevertheless, if some readers received a contrary impression, then I must not have expressed myself as clearly as I might have done, and I am indebted to Bennett for furnishing the occasion for this clarification. It should be noted in this connection that where I use the term ‘agent‐centred’ some other writers use the term ‘agent‐relative’, and perhaps that expression would have created less confusion.

As far as the fourth objection goes, it would surely be an exaggeration to say that I present the shared property of agent‐centredness as the most significant feature of both the prerogative and the restrictions, since the main burden of argument in the book is that there is an important difference between the two, namely, that it is possible to identify a plausible principled rationale for the former but surprisingly difficult to do so for the latter. And this difference evidently derives from other differences between them. It is quite true, however, that my discussion of agent‐centred restrictions does not focus primarily on the distinction between doing and allowing or, as Bennett would have it, between positive and negative facts about one's conduct. The fact that standard deontological views rely on some distinction of this kind has of course been a traditional point of departure for much criticism of deontology, and Bennett has himself been among the most effective critics. My own starting point in the book was (p.161) somewhat different, however. I was concerned 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. This idea is, implicitly if not explicitly, a feature of standard deontological views.4 One may, of course, attempt to defend the idea, as I noted in the book, by saying that it is worse to do certain things than to allow comparably bad things to happen. However, since what one allows to happen may be more doings of the very same kinds by others, such a defence cannot rest solely on the claim that there is a morally significant difference between doings and allowings. It requires the additional claim that the significance of that distinction is agent‐relative in character; what matters are the agent's doings, not the total number of doings overall. And it was with the rationale for such agent‐relativity that I was concerned.

Bennett says that the ‘typical deontological position forbids certain makings more strenuously than it frowns on the corresponding allowings; so deontological moralities do offer restrictions that are making‐focused; they do not offer restrictions that are in any significant sense agent‐centred’ (p. 63). Now I have already explained the sense in which such restrictions are agent‐centred, and Bennett's remark inadvertently helps to underline the significance of the fact that they are agent‐centred in that sense. To speak of deontological restrictions as ‘making‐focused’ is clearly to underdescribe them, for two reasons. Both of these reasons can be brought out by attending to a distinction that Bennett himself draws. He rightly notes that deontological restrictions do not apply only to cases in which, for example, killing an innocent person oneself (p.162) 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’). From this he concludes, again rightly, that what makes deontological restrictions agent‐centred cannot be that they apply only in cases where there is a contrast between oneself and other agents, as there is in A‐type cases, since no such contrast is present in B‐type cases, to which the restrictions also apply. However, just as it would be a mistake to forget that deontological restrictions do not apply only to A‐type cases (Bennett's point), so too it would be a mistake to forget that they do apply to A‐type cases, among others. Whereas the first point shows that one imaginable characterization of agent‐centredness is too narrow, the second point reveals one reason why Bennett's characterization of deontological restrictions as ‘making‐focused’ is too broad, an underdescription. For in A‐type cases that term applies equally well to the relevant deontological injunction, which is not to kill, and to the relevant consequentialist injunction, which is to minimize killings. To discriminate between these two diametrically opposed directives, one must go beyond the term ‘making‐focused’ and add the information that deontological restrictions are concerned in such cases with the agent's making rather than with the total number of makings overall—which is of course to say that they are, in the sense I have defined, agent‐centred.

The second reason that the term ‘making‐focused’ underdescribes deontological restrictions is this: whereas in A‐type cases it underdetermines the choice between the opposing directives given by deontology and any standard form of consequentialism, in general it is neutral between deontology and forms of consequentialism which hold that certain kinds of makings are always to be minimized. For example, a consequentialist principle according to which one should always minimize the number of killings in the world, even if that means that more total deaths overall will be allowed to occur, could appropriately be described (p.163) as making‐focused. In order to alter or supplement Bennett's phrase so that it distinguished deontology from such forms of consequentialism, one would again have to use language that called attention to the agent‐relativity of deontological restrictions; the point, again, is that insofar as what deontological principles focus on are ‘makings’ of certain kinds, they forbid an individual to perform one of those makings himself, rather than requiring him to minimize the total number of makings overall.

Note that whereas in A‐type cases the term ‘making‐focused’ underdetermines the choice between the opposing directives given by deontology and any standard form of consequentialism, in B‐type cases the same directive would be given by deontology and a form of consequentialism which held that one ought to minimize killings (as opposed to other kinds of deaths). Thus in A‐type cases the term ‘making‐focused’ underdetermines the choice between the opposing answers given by deontology and even those forms of consequentialism that are not themselves making‐focused; but in general it underdetermines the choice between deontology and making‐focused forms of consequentialism, two diametrically opposed moral orientations which give the same directives in B‐type cases.5

(p.164) The point of all this is not that deontological restrictions are not typically making‐focused, or even that in A‐type cases such restrictions cannot legitimately be represented as relying on a distinction between making and allowing. The point is rather that these features by themselves do not suffice to reveal what is distinctive about deontological restrictions; agent‐centredness is one defining feature of such restrictions, and one cannot characterize them fully without making reference to it.

Contrast this with Bennett's argument ‘that the notion of agent‐centredness has no role in characterizing deontological morality’ because in deontology ‘the relevant contrast is between what I might do and what might come about because I do not prevent it, in which case everything on both sides of the line is defined by how it relates to me’ (p. 62). The argument fails because, as I have defined ‘agent‐centred’, deontological restrictions are indeed agent‐centred despite the fact that there is reference to the agent on both sides of the line Bennett speaks of. Bennett's argument no more shows that deontological restrictions are not agent‐centred than my argument shows that they are not making‐focused. We could, indeed, embrace the ecumenical conclusion that deontological restrictions are simply agent‐centred restrictions on makings of certain kinds, but for the fact that it is notoriously difficult to draw the making/allowing distinction in such a way as to capture all and only those things that deontological restrictions are intuitively supposed to apply to, so that given any non‐circular way of drawing the line there will almost certainly be some ‘allowings’ to which the restrictions are also meant to apply. Nevertheless, we can certainly say this. ‘Agent‐centred’ and ‘making‐focused’ call attention to two different distinctions, both of which have roles to play in the characterization of deontology. Bennett and many others have emphasized the role played by the second of these distinctions; I was concerned in my book with the problems deontology faces in virtue of the role played by the first.

There are a number of other points in Bennett's essay that deserve comment, but in the space available I must (p.165) limit myself to some brief concluding remarks about its final section, where Bennett identifies three instances in my book in which the ‘structure’ implied by my use of the term ‘agent‐centred’ can, he says, be seen to be illusory. The second of these instances, to begin with, is a bit odd, since it is a case in which I deny the existence of a certain connection between the agent‐centred prerogative and agent‐centred restrictions; that is, I deny that someone who accepts the prerogative must, to be consistent, accept the restrictions as well. Bennett's complaint in this instance is just that I take too long to repudiate this ‘crazy’ suggestion and that I do so because I labour under the illusion that the prerogative and the restrictions are both agent‐centred in some significant sense. Well, as I have argued, they are both agent‐centred in some significant sense, so I don't see anything illusory here. However, I am delighted that Bennett thinks the suggestion is crazy, especially since a number of the book's critics have argued that something very close to it is true.6 Bennett's third instance of illusory structure is also a bit odd since there he agrees, if I understand him correctly, that a certain parallel I draw between the prerogative and the restrictions is genuine; he just thinks it's a coincidence. About this we simply disagree; I think the parallel grows directly out of the agent‐centredness of the prerogative and the restrictions and that this is clear once ‘agent‐centred’ is understood in the way I have indicated. That leaves the first ‘illusion’, which concerns what Bennett calls a ‘minor episode’ in my discussion, and it really is very minor. For what it's worth, however, the parallel whose existence is here said by Bennett to be illusory is the perfectly genuine parallel between: (p.166)

  1. (a) Once you fully understand what consequentialism says, you will see that in real‐world circumstances it would not systematically undermine the integrity of all agents, though it might undermine the integrity of some,

and
  1. (b) Once you fully understand what consequentialism says, you will see that in real‐world circumstances it would not systematically require agents to engage in abhorrent activities like killing or harming innocent people, though it might require such acts in some cases.

Notes:

First published in Ethics 100 (October 1989): 67–76. © 1989 by The University of Chicago. All rights reserved.

(1) Jonathan Bennett, ‘Two Departures from Consequentialism’, Ethics 100 (1989): p. 59. Future page references to this article will be given parenthetically in the text.

(2) It should also be pointed out that the permissions and prohibitions imposed by rule‐consequentialism and related ‘two‐level’ views do not count as agent‐centred in this sense.

(3) Are there any departures from consequentialism that could not legitimately be called ‘agent‐centred’ in my sense? Yes. Imagine, e.g., a principle which held that one should always perform that act which, of those available to one, would have the second‐best overall outcome from an impersonal point of view. Such a principle would represent a departure from consequentialism. It would not be agent‐centred, however, for it would leave the rightness or wrongness of actions entirely dependent on the overall value from an impersonal standpoint of their outcomes.

(4) For additional argument on this point, see ‘Agent‐Centred Restrictions, Rationality, and the Virtues’, Mind 94 (1985): 409–19, reprinted in this volume, pp. 133–51 above.

(5) In objecting that Bennett's term ‘making‐focused’ underdescribes deontological restrictions, am I not making a mistake strictly parallel to the mistake earlier noted of assuming that the adjective ‘agent‐centred’, as it functions in the label ‘agent‐centred prerogative’, must convey complete information about the kind of prerogative that label refers to? No. I treat ‘agent‐centred prerogative’ as a piece of technical terminology which refers to a permission of a certain kind, but Bennett does not introduce ‘making‐focused restriction’ as a technical term to refer to deontological restrictions. What he does instead is to draw a contrast between a description of deontological restrictions as ‘making‐focused’, which is said to be accurate and informative, and a description of them as ‘agent‐centred’, which is said to be either false or trivial. Against this specific claim, I have argued that the term ‘agent‐centred’ provides important information about deontological restrictions which is not provided by the term ‘making‐focused’, as witness the fact that the latter term differs from the former by failing to distinguish between deontology and certain forms of consequentialism. It may, however, be added, incidentally, that the fact that ‘making‐focused’, unlike ‘agent‐centred’, does fail to distinguish deontology from some forms of consequentialism, suggests that ‘making‐focused restriction’ would actually be less apt than ‘agent‐centred restriction’ as a piece of technical terminology used to refer to deontological restrictions. This judgment is reinforced by the point, which I mention later in the text, that it is not clear that deontological restrictions do apply exclusively to makings.

(6) See, e.g., L. Alexander, ‘Scheffler on the Independence of Agent‐centred Prerogatives from Agent‐Centred Restrictions’, Journal of Philosophy 84 (1987): 277–83; S. Kagan, ‘Does Consequentialism Demand Too Much?’, Philosophy and Public Affairs 13 (1984): 239–54, and The Limits of Morality (Oxford: Oxford University Press, 1989); and F. Kamm, ‘Supererogation and Obligation’, Journal of Philosophy 82 (1985): 118–38.