The Common Structure of Kantianism and Act Consequentialism Christopher Woodard RoME 2009 1. My thesis is that Kantian ethics and Act Consequentialism share a common structure, since both can be well understood as theories of pattern-based reasons. Their common structure consists of the features shared by any such theory. A pattern-based reason is a normative reason for action of the following form: PBR The fact that some action, A, is part of a valuable and eligible pattern of action, P, is a reason to perform A. 1 More commonly we think of reasons for action as act-based, which means that they have the following form: ABR The fact that some action, A, is valuable and eligible is a reason to perform it. 2 Indeed, act-based reasons are the sort of reasons that Act Consequentialism recognises. But we can think of them as a limiting kind of pattern-based reason, in which the pattern P is identical to the action A. Thus the idea of pattern-based reasons is more general than the idea of act-based reasons, and we can properly understand Act Consequentialism as a theory of pattern-based reasons. If pattern-based reasons exist, they are reasons for individual agents to act. It is controversial whether their existence depends on the existence of group agents. 3 But even if it does, they are not supposed to be reasons for group agents to act. If I have a pattern-based reason, it is a reason for me to act. To help keep track of the individual agent whose reasons we are interested in, I will call her the actor. 1 There can be pattern-based reasons against performing actions, too. I omit reference to them here for simplicity. I shall also limit my discussion to teleological reasons, though the idea of pattern-based reasons (and of act-based reasons) is not essentially teleological. Of course, it could be that Kantian ethics is better understood as a non-teleological theory of pattern-based reasons. But I wish to explore the extent to which we can think of it as a teleological theory of such reasons. 2 The value of an action or of a pattern could include any value it has for its own sake, as well as its instrumental value. 3 See Woodard (2008a: 68-80). More precisely, the controversial issue is whether the existence of patternbased reasons other than act-based reasons depends on the existence of group agents. 1
My conjecture is that Kantian ethics too can be well understood as a theory of pattern-based reasons, albeit one that is different in very important ways from Act Consequentialism. According to Kantian ethics, one eligible pattern consists of every rational agent acting as a member of the Kingdom of Ends. What s more, the reasons associated with this pattern defeat any other reasons with which they come into conflict. The result is an ethical view that incorporates exceptionless constraints such as the constraint against lying, or against treating persons as mere means. This is only a conjecture. To make good on the claim that Kantian ethics can be understood as a theory of pattern-based reasons one would need to work out a theory of pattern-based reasons in detail with the features that Kantian ethics has. Even leaving aside issues about what features a theory must have to count as Kantian, I cannot do that. I do not know precisely how to formulate a Kantian account of the value of the pattern consisting of every rational agent acting as a member of the Kingdom of Ends, for example. There are many other details of a projected Kantian theory of pattern-based reasons that I do not know how to give. My aim is more modest. I will explain what I take to be the features of any theory of pattern-based reasons, and try to make it seem plausible that Kantian ethics shares this structure. In doing so, I will focus on two features of Kantian ethics that the structure enables us to explain: the attitude towards other agents that is characteristic of Kantian ethics, and the existence of exceptionless constraints. 2. Any fully-fledged theory of pattern-based reasons must have the following features. First, it must contain an axiology, to explain the value of patterns of action. Second, the theory must contain some account of the eligibility of patterns, by which I mean an account of which possible patterns of actions support reasons. Presumably not every possible pattern supports reasons. Some possible patterns are mere agglomerations, for example. An account of the eligibility of patterns would explain why such patterns do not support reasons, and also which patterns do support them. If it lacked either an axiology or an account of eligibility, the theory could not tell us what reasons there are. Third, a full theory of pattern-based reasons would explain how the reasons it claims exist interact with each other. If it lacked this third feature, the theory could not tell us what we ought to do, all things considered. 4 4 Here I assume that what an agent ought to do is a function of her reasons for action. 2
When I claim that Kantian ethics and Act Consequentialism share a common structure, I mean that they can be understood as theories that would need to have all of these features in order to be fully-fledged. Act Consequentialism as such is not a fully-fledged theory, since it does not contain a specific axiology. But to be a fullyfledged theory, it would need to contain a specific axiology. The structure shared by Act Consequentialism and Kantian ethics is defined by the general features of any theory of pattern-based reasons, which are the idea of pattern-based reasons itself plus the requirement for an axiology, the requirement for an account of eligibility, and the requirement for an account of the interaction of reasons. The idea is that we can generate Act Consequentialism or Kantianism by taking this common structure and completing it with different axiologies, accounts of eligibility, or accounts of the interaction of reasons. For example, if we claim that the only eligible patterns are those that are identical with the agent s options, we get the view that all reasons are act-based. If we add the idea that reasons interact with each other such that the reasons associated with the most valuable options, all things considered, are always decisive for what one ought to do, we get a standard maximising form of Act Consequentialism. If we go further and add that value is a matter of utility alone, we get a form of Act Utilitarianism, which is a fully-fledged theory of pattern-based reasons. Some other familiar views in ethics clearly share this structure. Possibilism, for example, is a theory of pattern-based reasons with a distinctive account of eligibility. According to this account, the eligible patterns are all and only the actor s options together with the best possible responses to those actions that she could make. 5 This implies that the actor has a reason to perform those parts of the best possible sequences or combinations of action she now could perform, whether or not she would in fact perform the other parts later. These reasons are thus based on the parthood relation between her present options and the best sequences, not on any causal relation between these things. It is this account of eligibility that distinguishes possibilist views from actualist ones, such as standard forms of Act Consequentialism. Some forms of Rule Consequentialism are also well understood as theories of pattern-based reasons. 6 In particular, those that specify the value of rules in terms of 5 On possibilism see Zimmerman (1996: Ch. 6). On the connection with pattern-based reasons see Jackson (1987: 106), and Woodard (2009). 6 See Woodard (2008b). 3
compliance with (rather than acceptance of) them can be understood this way, since compliance with a rule or set of rules is a pattern of action. Rule Consequentialist views of this sort claim that there is a reason to perform some action, A, because it is permitted or required by some optimal rule or set of rules, R. This amounts to saying that there is a reason to perform A because it is part of the valuable pattern P that consists in widespread or universal compliance with R. Again, it is chiefly the account of eligibility that distinguishes Rule Consequentialist views of this sort from otherwise similar forms of Act Consequentialism. 7 One complication is that contemporary forms of Rule Consequentialism tend also to include disaster prevention mechanisms. 8 These views probably amount to pluralist theories of pattern-based reasons, which means that different-sized patterns can be eligible in a single case. This allows there to be a rule-based reason to do A in some case but also an act-based reason not to do A. These views also contain a distinctive theory about the interaction of the reasons associated with these different patterns. Roughly, this theory says: rule-based reasons defeat act-based reasons unless the stakes are high enough to constitute a disaster, in which case the act-based reasons defeat the rule-based ones. Other pluralist theories of pattern-based reasons, containing different accounts of the interaction of reasons, or of eligibility, or of value, are also possible. 9 One important class of theories of pattern-based reasons is distinguished by commitment to an account of eligibility that includes some form of Willingness Requirement. According to this requirement, no pattern can be eligible unless (a sufficient number of) the agents required to realise it would be willing to do so. 10 Note than none of the theories we have considered so far takes willingness to be a constraint on the eligibility of patterns. No one claims that an agent has an act-based reason to perform A only if she is willing to perform it, for example. Possibilists and Rule Consequentialists also deny that the agents required to perform the other parts of eligible patterns must be willing to do so. Theories that do accept some version of the Willingness Requirement thereby connect the idea of pattern-based reasons to 7 I claim that Kantian ethics shares a common structure with some forms of Rule Consequentialism. Parfit (2008: 313-318) makes the much stronger claim that Kantian ethics, when suitably interpreted, implies Rule Consequentialism. For critical discussion of Parfit s claim, see Morgan (2009), Otsuka (2009), and Ross (2009). 8 For example, see Hooker (2000: 98-9). 9 See Mulgan (2001), and Woodard (2008a: 105-118). 10 For example, see Regan (1980: 124), and Hurley (1989: 146). Different versions of the requirement use more or less demanding conceptions of willingness, but the underlying idea is that eligibility requires cooperativeness. 4
the idea of cooperation, and we can call all of them cooperative theories. Regan s Cooperative Utilitarianism is one such theory, while McClennen s theory of resolute choice is another. 11 It is striking that so many quite different theories can be understood as theories of pattern-based reasons. This suggests that the idea is quite general. However, all of the theories we have considered so far might be considered variants of consequentialism. What about Kantian ethics? No doubt much that is distinctive about Kantian ethics is due to its axiology. 12 However, much of its distinctive character is surely due also to its accounts of eligibility and of the interaction of reasons; and it is on these features that I will focus in the remainder. The account of eligibility, as I have mentioned, is such that one eligible pattern is that consisting of all rational agents acting as a member of the Kingdom of Ends. 13 Call this the KE pattern. The account of the interaction of reasons associated with different patterns is such that reasons associated with the KE pattern defeat any other reasons with which they conflict. Since Kantian ethics includes no version of the Willingness Requirement, it follows that agents must always act in a way that is consistent with acting as a member of the Kingdom of Ends, even if no other agent is doing so. This explains the Kantian idea that agents strongest reasons do not depend on how others in their environment would react to their actions. In doing so, it also allows us to explain Kantian support for exceptionless constraints. In the following two sections I will discuss each of these claims in more detail. 3. One deep contrast between Act Consequentialism and Kantian ethics concerns the treatment of agents other than the actor s present self. In one respect, Act Consequentialism treats these agents in just the same way that it treats non-agent parts of the actor s environment. Like these other parts of her environment, other agents stand to affect the success of the actor s actions. Since it is oriented exclusively to this success, Act Consequentialism instructs us to consider the response that any practically relevant part of the actor s environment would make to her actions. It takes this same policy with respect to every part of the actor s environment, whether an agent or not. 11 See Regan (1980), McClennen (1990), Woodard (2008a: Ch. 3). 12 On which see Herman (1993: Ch. 10). 13 Kant writes:... a kingdom of ends would actually come into existence through maxims whose rule the categorical imperative prescribes to all rational beings if they were universally followed. Groundwork 4: 438 (Gregor, 1996: 87, emphasis in original). The value of the Kingdom of Ends thus gives each of us a pattern-based reason to play our parts in the pattern that would bring it about. 5
Of course, that is not to say that Act Consequentialist theories do not distinguish in any way between agents and non-agents. They may distinguish between them as bearers of interests or rights, for example. In these or other ways, Act Consequentialist theories will find morally relevant differences between those parts of the actor s environment that are agents and those that are not. But there is, as I said, one important respect in which Act Consequentialism treats them the same. What matters about their possible response to the actor is how they would respond. Contrast this with Kantian ethics. The duty not to lie is completely unaffected by the intentions or subsequent behaviour of the murderer at the door, according to Kant. 14 What matters about his response is not how he would respond, but how he could best respond, namely by acting as a member of the Kingdom of Ends himself. How he could best respond is not dependent on his intentions, but only on his nature as a rational agent. This is in contrast with the attitude Kant appears to take towards non-agents. He seems to take the same attitude towards practically relevant natural features of the actor s environment as is taken by Act Consequentialism. 15 The differing attitudes towards other agents taken by Act Consequentialism and Kantian ethics explain some common criticisms of each of them. Act Consequentialism is said to take an unduly instrumental attitude towards the actor s options, and thereby to undermine her integrity. 16 This might be thought to reflect the way that Act Consequentialism treats other agents. If I am interested in the way you would respond to my actions, I have to tailor my decisions to take account of your intentions. The result is that I regard my own actions in an instrumental way, which supposedly undermines my integrity. By contrast, if I am interested in the best way you could respond to my actions, I can ignore your intentions. This insulates my reasons from changes in your intentions, providing them with a kind of robustness. The downside, as Korsgaard puts it, is that in cases such as that of the murderer at 14 Gregor (1996: 612-13). 15 This is somewhat unclear. At times, Kant seems to say that we should act on the supposition that nature will also respond in the best way possible. That would mean that he treats nature and other agents in the same way, and treats both differently than Act Consequentialism treats them. That may put pressure on my suggestion that his views are properly understood as a theory of pattern-based reasons unless we can interpret them as based on the supposition that the behaviour of nature itself constitutes a pattern of action in the appropriate sense. That may be possible, given Kant s beliefs. At any rate, I shall assume that Kantian ethics treats natural parts of the actor s environment differently, in this respect, than it treats agents. Rawls s remarks on the Kingdom (or, as he calls it, Realm ) of Ends seem to support this interpretation (Rawls, 2000: 311-312). 16 See, for example, Williams (1973: 93-118). 6
the door it seems grotesque simply to say that I have done my part by telling the truth and the bad results are not my responsibility. 17 We can explain this contrast between Act Consequentialism and Kantian ethics if we understand them both as theories of pattern-based reasons. One constraint on the eligibility of patterns, presumably, is that it must be possible for each of their parts to be performed. So, when I treat any agent s action as part of an eligible pattern, I am interested in what he could do. Insofar as I am interested in the best eligible pattern, I am interested in the best he could do. 18 So if my reasons are associated with the best eligible pattern, and this pattern includes this agent s response to my action, what matters is the best way he could respond to my action. How he would respond is irrelevant. Any theory of pattern-based reasons thus distinguishes in the following way between behaviour that is part of an eligible pattern and behaviour that is not. To be part of an eligible pattern, an action must be something the agent concerned could do. If it is part of an eligible pattern, what matters for reasons associated with that pattern is the action s value. If it is not part of an eligible pattern, what matters is whether it would occur. The contrasting attitudes towards other agents that we find in Act Consequentialism and Kantian ethics follow from this rule. Act Consequentialism claims that no other agent s actions are parts of any eligible patterns for me, so if another agent is practically relevant it instructs me to care about how he would respond to my actions. Kantian ethics claims that every other rational agent s actions are parts of an eligible pattern, the KE pattern, with which my strongest reasons are associated. That is why it instructs me to care about how any rational agent could best respond to my actions, and why it claims that my most important reasons are independent of facts about how any rational agent would respond to my actions. Some other views we have considered fall between these two extremes. Possibilists treat the actor s future responses in the way that Kantians treat the responses of all agents, but they treat the responses of all other agents in the way that (actualist) Act Consequentialists treat all responses. Pluralist Rule Consequentialists extend the Kantian stance further than Possibilists, but set it alongside the Act Consequentialist stance in cases where disaster could be prevented only by violating a rule. Other pluralist views may take both stances towards agents responses quite generally, if in 17 Korsgaard (1996: 150). It is notable that Korsgaard explains the Kantian idea in terms of [doing] my part. I am grateful to Guy Fletcher for pointing out the relevance of this passage to me. 18 See Regan (1980: 54-5). 7
each case they recognise reasons associated with some patterns that do, and some patterns that do not, include those responses. 4. Thinking of Kantian ethics as a theory of pattern-based reasons also helps to explain its endorsement of exceptionless agent-relative constraints. Such constraints prohibit certain kinds of action in all cases. To realise the KE pattern it would have to be true, among other things, that no agent ever performs these actions. So each agent s part in realising this pattern includes never performing these actions. According to Kantian ethics this pattern is eligible and supports reasons of such importance that they defeat any reasons that conflict with them. As a result, performing such actions could never be something an agent has most reason to do. Thus no agent ought ever to perform such actions. This gives us the basic explanation of exceptionless Kantian duties. It takes us from the supposition that the KE pattern requires some kind of action, K, to the conclusion that there is an exceptionless duty regarding actions of this kind. Of course, kinds of action can be specified in different ways. If K requires specific actions at specific times, the duty will be perfect. For example, the duty not to lie is perfect because it requires a specific action, namely not lying, on specific occasions, namely each occasion when lying is possible. The KE pattern involves no agent ever lying, and this translates to specific requirements on specific occasions. In contrast, the duty to develop one s talents is imperfect because it does not require any specific action on any specific occasion. It requires instead that, during one s life as a whole, one do things that develop one s talents. Though the KE pattern involves no agent ever failing to develop her talents, this does not translate to specific requirements on specific occasions. We have not yet explained why the KE pattern requires any particular kind of action, however. The conjecture that Kantian ethics can be well understood as a theory of pattern-based reasons involves the idea that this could be explained in the following way. If we could specify the axiology that applies to patterns of action according to Kantian ethics, we could work out what kinds of action this pattern consists in. For example, if the axiology values autonomy, it will be possible to explain why the KE pattern involves no agent ever lying. This will be so for two reasons: first because each instance of lying undermines the listener s autonomy, and so is pro tanto bad, and second because none of the possible responses to failure to lie that could provide a reason to lie (such as murdering the victim whose location is 8
thereby revealed) are consistent with the KE pattern. The result is that the KE pattern will contain no instances of lying. According to my conjecture, a suitable axiology would allow us to explain most or all of the duties that Kantian ethics recognises in a similar way. 19 This sort of explanation dispels the air of paradox that is sometimes thought to surround agent-relative constraints. 20 The appearance of paradox depends on the thought that any plausible explanation of the reason not to perform an action of the prohibited kind will imply the existence of a stronger reason to minimise performance of actions of the very same kind. If there is such a reason it could defeat the reason underlying the constraint. Thus, if my reason not to lie has to do with the badness of undermining the listener s autonomy, I would seem to have a stronger reason to minimise this badness by minimising lying. This is problematic if I could minimise lying by telling a lie myself. It is well-known that there are ways of dissolving this air of paradox. One way is to individuate kinds of action more finely, such that lying by me now is a different kind than lying by you now or lying by me tomorrow, for example. In this or other ways we can try to construct an axiology that permits us to distinguish between the kind of reasons that explain why I should not lie now and the kind of reasons there are to minimise lying overall. Since these reasons are of different kinds, the explanation of one need not imply the existence of the other. 21 This strategy is costly, however, since it may be hard to motivate these axiological distinctions in any way other than by saying that they are necessary to explain the constraints. That may leave them looking ad hoc, and may undermine our confidence in the constraints. If we explain the constraints in terms of pattern-based reasons we can avoid this cost. Once again we can claim that the reason not to lie is of a different kind than the reason to minimise lying. But instead of claiming this because of axiological distinctions that may be hard to defend, we can claim it because the reason not to lie 19 One important issue concerns exceptional cases caused by natural forces rather than other agents. For example, suppose that the eruption of a volcano causes the actor to face a choice between killing one innocent person or letting twenty die. Kantian ethics implies that she should not kill in this case. This seems hard to explain in terms of pattern-based reasons, because the bad consequences of not killing have nothing to do with any agent s response. One possibility is to postulate that nature itself is an agent (see footnote 15, above). A second would be to insist that the relevant description of kinds of action involved in specifying the KE pattern rules out distinguishing between killing when threatened by an agent and killing when threatened by natural forces. If the KE pattern simply requires not killing the innocent, then the corresponding duty will apply even in the case described in the text. A third possibility is to concede that not all Kantian duties can be explained in terms of pattern-based reasons. For related discussion, see Woodard (2008a: 31-36). 20 For example, see Nozick (1974: 30-42) and Scheffler (1994: Ch. 4). 21 See Portmore (2007: 54-5). See also Hurley (1997). 9
is associated with the KE pattern, while the reason to minimise lying is act-based. Since these reasons are of different kinds, we should not expect the explanation of the first to imply that the second kind of reason exists, let alone that it is stronger. That is so even though the badness of my lying now is of the same kind as the badness of your lying now or my lying tomorrow. We can get agent-relative constraints without resort to agent-relative axiology. 22 5. This illustrates a more general contrast between my proposal and the so-called Consequentializing Idea. According to the Consequentializing Idea we can consequentialize any ethical view whatsoever. This involves tailoring an axiology such that, when it is combined with Act Consequentialism, it yields exactly the same verdicts about the deontic status of actions as the original view. We then have a version of Act Consequentialism that is deontically equivalent to the original view. 23 My proposal differs from this idea in several ways. First, it is less general, since I do not claim that every ethical view can be well understood as a theory of pattern-based reasons. Second, the feature that it claims is shared in common is more abstract and more complex than the feature that the Consequentializing Idea claims is shared in common. The Consequentializing Idea claims that Act Consequentialism itself is common to ethical views, in the sense that for any view there is a deontically equivalent version of Act Consequentialism. My proposal is that the general features of any theory of pattern-based reasons are shared in common. These features are more abstract, since the idea of pattern-based reasons is more general than the idea of act-based reasons. They are also more complex, since the structure includes several variables, whereas axiology is the only variable according to the Consequentializing Idea. Arguably, this means that less distortion is involved in representing Kantian ethics, say, as a theory of pattern-based reasons than as a version of Act Consequentialism. Not every distinctive feature of Kantian ethics has to be represented as following from a distinctive claim about value. As we just saw in the discussion of constraints, this may make better sense of the feature in question. 6. If my conjecture is right, it need not follow that we should always think of Kantian ethics or the other views I have mentioned as theories of pattern-based reasons. It 22 As Rule Consequentialists recognise: see Hooker (2000: 109-110). 23 See Portmore (2007). 10
could be that there are several good ways of understanding each of these theories. Some ways could be better for some purposes than for others. Nevertheless, the conjecture would be significant if it is right. It would mean that according to one good way of understanding Kantian ethics, Act Consequentialism, and the other theories I have mentioned, these theories share important features. This would be somewhat surprising, and it would make disputes between these theories seem, in one way, in-house. It would thereby frame the issues at stake in a new way. Kantians could debate eligibility with Act Consequentialists, for example. Given the great differences between these views, these in-house disputes would be very important. But since not every view can be well understood as a theory of pattern-based reasons, part of the character of the views we have discussed would be traceable to the fact that they can be so understood. So there would be important disputes between these views taken as a class and the other views that cannot be well understood as theories of pattern-based reasons. By studying these disputes we might hope to learn more about all of these views, and about the merits of the idea of pattern-based reasons itself. Department of Philosophy University of Nottingham Nottingham NG7 2RD, UK christopher.woodard@nottingham.ac.uk References Gregor, M. J. (trans. and ed.) (1996). Immanuel Kant: Practical Philosophy. Cambridge: Cambridge University Press. Herman, B. (1993). The Practice of Moral Judgment. Cambridge, MA: Harvard University Press. Hooker, B. (2000). Ideal Code, Real World. Oxford: Clarendon Press. Hurley, P. (1997). Agent-Centred Restrictions: Clearing the Air of Paradox. Ethics, 108: 120-146. Hurley, S. L. (1989). Natural Reasons. New York: Oxford University Press. 11
Jackson, F. (1987). Group Morality. In Pettit, P., Sylvan, R., and Norman, J. (eds.) Metaphysics and Morality. Essays in Honour of J. J. C. Smart. Oxford: Basil Blackwell, 1987, pp. 91 110. Korsgaard, C. M. (1996). Creating the Kingdom of Ends. Cambridge: Cambridge University Press. McClennen, E. F. (1990). Rationality and Dynamic Choice. Cambridge: Cambridge University Press. Morgan, S. (2009). Can there be a Kantian Consequentialism? Ratio (new series) XXII, 1: 19-40. Mulgan, T. (2001). The Demands of Consequentialism. Oxford: Clarendon Press. Nozick, R. (1974). Anarchy, State, and Utopia. Oxford: Basil Blackwell. Otsuka, M. (2009). The Kantian Argument for Consequentialism. Ratio (new series) XXII, 1: 41-58. Parfit, D. (2008). On What Matters. Unpublished manuscript, accessed on 12.06.09 at http://users.ox.ac.uk/~ball2568/parfit/parfit_-_on_what_matters.pdf Portmore, D. W. (2007). Consequentializing Moral Theories. Pacific Philosophical Quarterly 88: 39-73. Rawls, J. (2000). Lectures on the History of Moral Philosophy, ed. Herman, B. Cambridge, MA: Harvard University Press. Regan, D. (1980). Utilitarianism and Co-operation. Oxford: Clarendon Press. Ross, J. (2009). Should Kantians be Consequentialists? Ratio (new series) XXII, 1: 126-135. Scheffler, S. (1994). The Rejection of Consequentialism, revised edition. Oxford: Clarendon Press. Williams, B. (1973). A critique of utilitarianism. In Smart, J. J. C. and Williams, B. Utilitarianism: for and against. Cambridge: Cambridge University Press, 1973, pp. 75-150. Woodard, C. (2008a). Reasons, Patterns, and Cooperation. New York: Routledge. (2008b). A New Argument Against Rule Consequentialism. Ethical Theory and Moral Practice 11: 247 61. (2009). What s Wrong With Possibilism. Analysis 69, 2: 219-26. Zimmerman, M. J. (1996). The Concept of Moral Obligation. Cambridge: Cambridge University Press. 12