1 I am grateful to Michael Bratman, Mark Green, Patricia Greenspan, Agnieszka Jaworska, Nadeem Hussain, David Hills, Elijah Millgram, Jennifer Rosner, Tamar Shapiro, Fred Schueler, Ken Taylor, Manuel Vargas, David Widerker, and audiences at the Northwest Philosophy Conference, at Ben Gurion University, at Haifa University, and at the Hebrew University for helpful discussion; I am also grateful to Michael Bratman, Agnieszka Jaworska, Nadeem Hussain, Elijah Millgram, Tamar Shapiro, and Ken Taylor for comments on earlier drafts.

2 Millgram, Elijah Practical Induction (Cambridge, MA: Harvard University Press 1997)Google Scholar

3 I borrow the term ‘practical identity’ from Christine Korsgaard, but I do not necessarily follow Korsgaard's use in my understanding of this term. Rather, I understand ‘practical identity’ to refer to the part of one's identity that consists of one's fundamental action-guiding mental attitudes. I use the term in a way that is to a large degree synonymous with Bernard Williams’ notion of one's ‘motivational set.’ See Korsgaard, Christine The Sources of Normativity (Cambridge: Cambridge University Press 1996), 101;CrossRefGoogle Scholar Williams, Bernard ‘Internal and External Reasons,’ in Moral Luck (Cambridge: Cambridge University Press 1981), 102,CrossRefGoogle Scholar 105.

4 When describing the instrumentalist view he says: ‘Only desires count, and so desires have to count simply in virtue of being desires’ (11).

5 Williams; also Brandt, Richard A Theory of the Good and the Right (Oxford: Clarendon Press 1979).Google Scholar Harry Frankfurt is sometimes interpreted as claiming that only desires that are endorsed by our second order desires are reason-giving; Michael Bratman suggests that only intrinsic desires that are endorsed by higher order policies have ‘subjective normative authority’ for the agent. See Bratman, M. ‘Two Problems About Human Agency,’ Proceeding of the Aristotelian Society (2001), 321.Google Scholar His addition to this list, however, should be qualified for two reasons: first because he talks about subjective normativity and not about normativity tout court; and second because it is not clear that he considers himself a Humean-instrumentalist even in a broad understanding of these terms.

6 Note that the notion of instrumentalism discussed by Millgram (and my suggested corrected notion) avoids a recent criticism made against instrumentalism under its more common definitions. Christine Korsgaard has claimed that the idea that instrumental reasoning would give an agent a reason to bring about the means to the satisfaction of one's desires without there being a reason for satisfying the desire itself is untenable. Korsgaard writes: ‘I think the argument shows that the instrumental principle cannot stand alone. Unless something attaches normativity to our ends, there can be no requirement to take the means to them’ (Korsgaard, C. ‘The Normativity of Practical Reason,’ in Ethics and Practical Reason, Cullity, Garrett and Gaut, Berys eds. [New York: Oxford University Press 1997], 251Google Scholar). Accepting the idea that intrinsic desires (or some intrinsic desires) are themselves reason-giving is one way of providing a foundation for the normativity of end-means reasoning.

7 I further explain why Millgram's argument applies to sophisticated instrumentalism in footnote 28.

8 It is usually and implicitly assumed, both by instrumentalists and by non-instrumentalists, that such intentional and direct control of intrinsic desires is impossible. Gilbert Harman is a notable exception. See Harman, G. ‘The Toxin Puzzle,’ in Rational Commitment and Social Justice, Coleman, J. and Morris, C. eds. (Cambridge: Cambridge University Press 1998).Google Scholar

9 Elster, John ‘Sour Grapes,’ in Utilitarianism and Beyond, Williams, B. and Sen, A. eds. (Cambridge: Cambridge University Press 1982), 224Google Scholar

10 In my approach to more complicated cases I follow Michael Bratman. According to Bratman an intentional action is an action done while having a certain present directed intention, such that the present directed intention is appropriately related to the action in a way that is partly determined by the agent's desires and beliefs. Even though a present directed intention is required in order to intentionally ? it need not be an intention to ?. See Bratman, Michael Intention, Plans, and Practical Reason (Cambridge, MA: Harvard University Press 1987), 5–6.Google Scholar

11 In my understanding of the notion of ‘direct actions’ I follow Arthur Danto. His views concerning direct actions are presented both in ‘What We Can Do’ and in ‘Basic Actions,’ though he uses the term ‘basic actions’ to refer to what I am here calling direct actions. See Danto, A. ‘What We Can Do,’ The Journal of Philosophy 60 (1963), 435-6;CrossRefGoogle Scholar and Danto, A. ‘Basic Actions,’ American Philosophical Quarterly 2 (1963) 141–8.Google Scholar

12 According to Millgram our feelings of pleasure are our best guide for the desirability of the objects of our desires (115).

13 As we will see, both Millgram and I will allow into our discussion an example of a change of desire that is not direct, but that example does preserve the important feature of direct changes of desires discussed above; it is a change of desires that is not mediated by any changes in facts about desirability.

14 The story is supposed to bypass the question concerning the mechanisms that would allow us to desire at will and show us that even if we assume that there is no physical limitation on desiring at will (we assume that there is a pill that can do the Job) we would have to admit that desiring at will is impossible. That would show that the impossibility is principled or conceptual. I have explained above why the acquisition of a desire ‘at pill’ does not constitute — in the sense relevant for Millgram's argument — an indirect acquisition of desire.

15 ‘A desire is something that makes certain inferential demands; for example, it is something that, ceteris paribus, I take it that I should act on, something whose object I try to obtain’ (18). This of course does not mean that Millgram believes that our desires ground our reasons for action; they are better thought of as indicators of the fact that such reasons exist.

16 Sydney Shoemaker, in response to Millgram's argument, claimed that if one could change one's taste for broccoli at will then there would be no reason why one couldn't change at will one's desire to eat broccoli. The impossibility of desiring at will, concluded Shoemaker, is an empirical and not a principled impossibility. While criticizing Millgram's argument, Shoemaker implicitly accepts Millgram's main objection to instrumentalism. Desires, the broccoli example suggests, do not ground our practical reasoning. There are practical reasons for having the desires we have, reasons that do not have their origin in our intrinsic desires, and only changing these reasons would enable a rational change in our desires. In this case only a change in our taste for broccoli would make the formation of a desire for broccoli rational.

A desire for broccoli because it tastes good, I suspect, is a misleading example. I may desire to eat broccoli today partly because broccoli tastes good to me (though that certainly would not be reason enough, since many things taste good to me) but that means that my desire to eat broccoli today is a desire I have partly for instrumental reasons. The important question is whether desires that I have for non-instrumental reasons (or those elements of my desires that do not derive from instrumental reasons) can be rationally changed at will. Only if we can reject Millgram's argument concerning the principled impossibility of changing our intrinsic desires at will can we defend instrumentalism from Millgram's objection. Shoemaker, S. ‘Desiring at Will (and at Pill): A Reply to Millgram,’ in Preferences, Wessels, U. and Fehige, C. eds. (Berlin: De Gruyter 1997).Google Scholar

17 Millgram does not say that of his argument, but it is obviously structured as a reductio.

18 Other things being equal.

19 According to Millgram a desire is at least partly identified by the inferences it grounds. If an agent thinks that it would be crazy to act on a certain mental attitude, then that mental attitude will not be the basis for any inference to reasons for action and so it cannot be properly called a desire.

20 Of course this claim should be qualified by a ceteris paribus clause: if someone does not have the money, or thinks that over-consumption is harmful to the environment and has detrimental effects on people in Third World countries, then he may not find it rational to act on his desire to acquire expensive gadgets.

21 And even, I should add, if were it not for the fact that they desired those items, they would consider them unworthy of being desired/brought about.

22 Since Millgram was trying to show that the pill couldn't possibly create real desires my last remark may seem confusing. The point is this. Millgram tries to convince us, in various ways, that whatever the pill creates it would be irrational to act on that thing. My strategy was to resist these kinds of arguments and to say that until we know what attitude the pill creates we cannot know whether that attitude is reason-giving or not. Now imagine that we have an independent way of knowing what kind of attitude was created — independent from information about whether that attitude is reason-giving or not. If we learn that what was created by a certain pill is a real desire then it is reason-giving. If, on the other hand, we learn that what was created by a certain pill is not a real desire then it may not be reason-giving. But all that tells us is that it wasn't a good pill. It did not do what its maker promised us it would do. Even if we learn that no actual pill can create real desires, all we learn is that pills are not good at creating desires. But that does not mean that we do not have our own mechanisms for creating desires, and it certainly does not mean that such mechanisms are conceptually impossible.

23 Frankfurt, for example, may be taken to require that only desires that are really our own be taken into account in deliberation. Quite similarly, Korsgaard (at least where non-moral issues are concerned) requires that only desires that are endorsed by the agent from the perspective of one of her practical identities be taken into account in deliberation. And Bratman requires that only those desires consistent with our current plans be taken into account in deliberation. These are all ways of adding some sort of screening mechanism to the naive instrumentalist view. Instead of accepting all of our desires as reason givers, these theorists say that some should be excluded. However, these screening mechanisms will not screen off all the desires that are acquired at will, since desires that are rationally acquired at will are not particularly likely to be the ones that we do not identify with/do not endorse from the perspective of our practical identities/find inconsistent with our plans. Hence all these theories need not be in conflict with the view that desires can be acquired at will. And even if it turns out that on some occasion a desire that is acquired at will should not count as a reason for action (under one of these views) that would not show that desiring at will is impossible. It would only show that under these views some cases of acquisition of desires at will would yield desires that should not be taken as reasons for action.

24 And even more agree that this is the case in non-moral contexts.

25 It is important to note that my reply to Millgram's argument here does not rest on the assumption that all desiring at will is (or would be) rational. Irrational changes of desires may certainly be part of our mental life even if desiring at will is possible. But this fact does not prevent us from grounding a partial explanation of the range of desiring at will in the claim that our rational capacities limit the extent to which we put to use our ability to change our desires. Similarly the fact that we do not act on every desire we happen to have is partly explained by the limits that our rational capacities put on our decision making — and this explanation holds despite the fact that we sometimes behave irrationally.

26 The word ‘function’ may suggest that I believe that such deliberation would always have only one rational outcome. This is not my view. The idea is only that the rational decision to desire at will would not be less a function of our motivational set than would be the rational decision to act. In fact, desiring at will is best thought of as simply another type of action that is open to a rational agent.

27 Remember that Millgram does not think that desires by themselves are reason-giving; what desires do is transmit a reason for action whose source is external to the desire. However, once we have a desire, and precisely because desires indicate the existence of a source of reason, we can take ourselves to have a reason to achieve what we desire.

28 Since on sophisticated instrumentalist views there are conditions under which a desire is not a real desire and is not reason-giving, one might want to view these conditions as the justifying grounds that Millgram thinks a desire must have in Order to be reason-giving. Therefore, one might think that Millgram would not direct his argument against sophisticated instrumentalism. I think that this is a misunderstanding of sophisticated instrumentalist views. Sophisticated instrumentalism does not provide our desires with the kind of justificatory grounding that Millgram thinks desires must have and that desires do have on non-instrumentalist views. To see why, we should explain what Millgram means when he requires that desires will be normatively grounded in some independent facts. On Millgram's view a desire is just an indicator for the existence of other more basic reasons for action. When Millgram says that a desire gives us reasons for action he means that the desire transfers the normative force of some other facts — it plays a role similar to the one played by an interim conclusion in a long derivation. By itself, it does not contribute anything to the truth of the final conclusion. Sophisticated instrumentalism sees desires in a very different light. On sophisticated instrumentalist views desires make an independent contribution to the reasons we have for action. Even if on sophisticated instrumentalist views there are other factors that determine our reasons for action, these other factors cannot determine them by themselves. Take, for example, Brandf s view that only desires that we will not lose if we gain full information are real desires and therefore really reason-giving. Consider the following scenario: not knowing what the weather is like, I desire to go to the beach. If I knew what the weather is like I would still want to go to the beach. Hence, on Brandt's view, my desire to go to the beach is reason-giving. The conditions that ensure that my desire is reason-giving are facts about the weather, in particular the fact that the weather is nice. But my knowledge of that fact would certainly not be sufficient to justify my going to the beach. I might know that it is a beautiful day and desire to stay at home and drink guava juice and hence have no reason whatsoever to go to the beach. On sophisticated instrumentalist views the conditions without which my desire would not be reason-giving are not sufficient to justify my action. To have a reason for action I additionally need a desire, and thus the desire is a necessary and independent source of reason for action. This is why sophisticated instrumentalism is a kind of instrumentalist view and not a Variation on the views that make our desires mere indicators for some other facts that in turn are the real ground for our reasons for action. And this is also why Millgram would reject sophisticated instrumentalism just as he rejects naïve instrumentalism. On Millgram's view it does not make sense that a desire will be an independent contributor to our reasons for action, a contributor that does not in turn have deeper sources.

29 If one's desires depended for their reason-giving force on evaluative beliefs then if one could believe at will then one could influence a change in one's desires. And that would be the case even if instrumentalism would be false. But such a change in desires would not be a direct change of desires it would be a change that is mediated by a change in one's evaluative beliefs.

30 Yonatan Shemmer, Desiring at Will — Reasons, Motivation and Motivational Change. Doctoral dissertation, Stanford 2002.

31 In a sense instrumentalists have been making similar Claims for a while. It is commonly maintained that agents may discover new means to old goals and even that they may discover that the goals they thought they had were not their real goals. But accepting these suggestions introduces only a limited conception of change in one's practical identity. An instrumentalist who accepts these suggestions may still believe that one's real goals are either static or the outcome of contingent social influences — and thus not rationally changeable. If we accept, on the other hand, the link suggested by Millgram between instrumentalism and desiring at will, we open the way for a truly dynamic understanding of instrumentalism. (See Williams, 105; Brandt, 113.)