PRACTICAL REASON AND DESIRE
Ralph Wedgwood
According to the Humean Theory of Practical Reason (HTPR), every rational action must serve some desire that the agent has (or at least would have if she rationally performed that action) -- where this desire is, as I shall put it, a 'contingent desire', a desire that the agent is not in any way rationally required to have.(1) Humeans typically argue for the HTPR on the basis of the following two premisses: first, the Humean Theory of Motivation (HTM), the view that all motivation of action has such contingent desire 'at its source'; and second, the Internalism Requirement (IR), the view that 'reasons must be capable of motivating'.(2)
This argument for the HTPR is familiar, but it has rarely been stated with sufficient precision. In this paper, I shall give a precise statement of this argument and of the HTPR. I shall then rely on this statement to show two things. First, the HTPR is false: it is incompatible with some extremely plausible assumptions about weakness of will or akrasia. Second, this argument for the HTPR is invalid: even if the HTM and the IR are true, they do not support the HTPR.
I. The HTM and the IR
I propose to interpret the HTPR in the light of the argument that Humeans use to support it -- the argument that is based on the HTM and the IR. In this section, I shall explain how I interpret the HTM and the IR. My interpretation is designed, not only to make it prima facie plausible that the HTM and the IR entail the HTPR, but also to make the HTM and the IR at least prima facie plausible themselves; but I will not be able to offer a full defence of the HTM and the IR here.
The HTM is a claim about the motivation of action. But what is motivation? Suppose that a person A intentionally does something -- for example, f-ing. Then A's 'motivation' for f-ing consists of the set of A's mental states that motivated A to f. A set of mental states 'motivates' A to f if and only if we can truly give a certain sort of explanation of A's intentionally f-ing, by saying that A intentionally f-ed precisely because A was in those mental states. More specifically, if those mental states were A's motivation for f-ing, then, given that A was in those mental states, and given certain dispositions that are entirely intelligible in agents like A, it is entirely intelligible and unsurprising that A intentionally f-ed.
For example, suppose that you cross the street, because you believe that there is a florist's shop on the other side of the street, and you want to visit a florist's shop. Given that you had this belief and this desire (and given certain dispositions that are entirely intelligible in agents like you), it is entirely intelligible and unsurprising that you intentionally crossed the street. So, your motivation for crossing the street consists of this belief-desire pair. In this case, it is plausible that neither your belief that there is a florist's shop on the other side of the street, nor your desire to visit a florist's shop, is enough by itself to make it intelligible or unsurprising that you intentionally crossed the street. So neither this belief nor this desire is enough by itself to motivate you to cross the street. Your motivation in this case consists of this belief-desire pair.
As I shall understand it, the HTM claims that there is a certain form that the motivation of action must always take. The first component of the HTM is the claim that necessarily, every intentional action is motivated, at least in part, by some desire that is served by that action (or at least would be served by that action, if the beliefs that motivated the action were true -- but henceforth I shall omit this qualification). It does not matter here exactly how we interpret the notion of 'desire'. My arguments will be equally effective on any reasonable interpretation of this notion. But to fix ideas, let us suppose that according to this first component of the HTM, an intention to f entails a desire to f; so, whenever one acts with the intention to f, one's action is motivated (at least in part) by a desire to f.
Some philosophers take this first component to be all there is to the HTM.(3) But this first component is compatible with the existence of cases in which the desire that motivates an action is itself motivated purely by beliefs, without any input from antecedent desires. But any such case is incompatible with the idea, which seems central to the HTM, that all motivation of action has desire 'at its source'. As Bernard Williams [17, p. 109] puts it, if a process of 'deliberation' gives rise to a 'new motivation', there must have been some antecedent 'motivation for the agent to deliberate from, to reach this new motivation'.(4) So I propose to interpret the HTM as involving a second component. This second component is what Jay Wallace [14, p. 370] has called the 'desire-out, desire-in' principle, according to which 'processes of thought which give rise to a desire (as "output") can always be traced back to a further desire (as "input")'. Let us say that the 'motivational history' of a desire includes not only the mental states that motivated it, but also the further mental states (if any) that motivated those motivating states, the yet further mental states that motivated those further motivating states, and so on. Then, according to the 'desire-out, desire-in' principle, it is necessary that whenever a desire is motivated at all, its 'motivational history' includes some further, antecedent desire.
The relation between a desire and any antecedent desires in its motivational history seems to be both asymmetric and transitive. So, assuming that there is not an infinite regress of motivating desires, the HTM implies that the motivational history of every action must include at least one unmotivated desire -- that is, a desire that is not motivated by any antecedent mental states at all. In this sense, the HTM implies that all motivation of action must flow, at least ultimately and in part, from unmotivated desires.
These first two components of the HTM are compatible with the idea that in some cases, these 'unmotivated desires', from which all motivation of action flows (at least ultimately and in part), are themselves desires that one is rationally required to have -- that is, desires that it would be irrational for one not to have. But this also seems contrary to the spirit of the HTM.(5) I propose then that the HTM should involve a third component: the thesis that, necessarily, the motivational history of every action must include at least one unmotivated contingent desire -- an unmotivated desire that one is not rationally required to have. In short, according to the HTM, all motivation of action must flow, ultimately and in part, from unmotivated contingent desires.
What about the Internalism Requirement (IR)? There are two fundamentally different kinds of 'internalism' discussed in the literature on reasons for action. One kind is 'belief internalism', according to which there is a necessary connection between one's believing that it is rational for one to f (or that there is a reason for one to f) and one's being motivated to f, at least in certain conditions. But such 'belief internalism' will not support an argument from the HTM to the HTPR. To support such an argument, we need a thesis that asserts a necessary connection between its being the case that it is rational for one to f (or that there is a reason for one to f) and one's being motivated to f, at least in certain conditions.
What such connection might there be, between rational action and motivation? The fundamental connection, I suggest, is this: whenever one performs a rational action, that action is rational precisely in virtue of the way in which it was motivated. More specifically, whenever one performs a rational action, the action is rational precisely because it was motivated by means of a process of rational 'deliberation' or 'practical reasoning'. (We should interpret the notion of 'practical reasoning' broadly, so that it includes certain unconscious mental processes, as well as conscious reasoning.) But what is it for a process of practical reasoning to count as 'rational'?
As the term is most commonly used, a process of practical reasoning can still be perfectly 'rational' even if -- through bad luck -- the agent is mistaken or ignorant about some relevant fact. When the term is used in this way, whether or not a process of reasoning is rational depends purely on the quality of the reasoning itself. It does not depend on whether the inputs to that reasoning include false beliefs, or omit relevant true beliefs. However, some philosophers assume that 'rational' practical reasoning, in addition to being 'procedurally rational' in this way, must also not be affected by any such mistaken beliefs, or even by ignorance of anything that a reasonably well-informed agent would know.(6) I shall distinguish between these two senses of 'rational' by using the phrase 'subjectively rational' for the first sense, and 'objectively rational' for the second sense.(7) We may also say that any action motivated by means of 'subjectively rational' reasoning is 'subjectively rational', and any action motivated by means of 'objectively rational' reasoning is 'objectively rational'. Finally, we may say that a belief is 'subjectively rational' just in case no process of reasoning would fail to count as 'subjectively rational' merely because it is based, at least in part, on that belief.
To take an example from Williams [17, p. 102], suppose that you want a gin and tonic, and believe that the stuff in front of you is gin. Suppose that this belief-desire pair motivates you to mix the stuff with tonic and drink it. This was presumably a process of subjectively rational practical reasoning (at least so long as this belief was subjectively rational); so your action of mixing the stuff with tonic and drinking it was also subjectively rational. But suppose that in fact the stuff was not gin but petrol. Then the inputs to this process of practical reasoning included a false belief -- the belief that the stuff was gin -- and omitted a relevant true belief -- the belief that the stuff was petrol. So this reasoning, and your action of mixing the stuff with tonic and drinking it, which was motivated by means of this reasoning, were not 'objectively rational'.
The fundamental connection between rational action and motivation, then, is this: whether or not an action that one performed was rational depends on one's motivation for that action. This connection concerns when it is true to make an ex post statement of the form 'A's f-ing was rational' -- where this ex post statement entails that A actually did f. The IR, as I shall understand it, is a claim about when it is true to make an ex ante statement of the form 'It is rational for A to f' -- where these ex ante statements do not entail that A actually does f.(8)
Let us start with 'subjective rationality'. According to the IR, it is (ex ante) subjectively rational for one to f just in case there is a possible process of subjectively rational practical reasoning by means of which one's present state of mind might motivate one to f. Williams [18, p. 35] gives a similar formulation: according to Williams, A has a reason to f only if there is a 'sound deliberative route' from A's present state of mind to a 'conclusion to f'.
A similar principle applies to 'objective rationality'. Consider an idealisation of one's present state of mind, which includes certain true beliefs that one does not hold, and omits certain false beliefs that one does hold. According to the IR, it is (ex ante) objectively rational for one to f just in case there is a possible process of objectively rational practical reasoning by means of which this idealised state of mind might motivate one to f.(9) For both senses of 'rational', we may say that it is (ex ante) rational for one to f just in case there is a possible process of rational practical reasoning by means of which the 'relevant state of mind' might motivate one to f. For subjective rationality, the 'relevant state of mind' is one's actual present state of mind; for objective rationality, it is an appropriate idealisation of one's present state of mind.
Some philosophers interpret the IR slightly differently. For example, Michael Smith [13, p. 182] interprets it as the claim that it is rational for A to f in circumstances C just in case A would want herself to f in C if she were fully rational. But to say that it is rational for A to f is just to say that it is rationally permissible for A to f, not that A is rationally required to f: even if it is rational for A to f, it might also be rational for A not to f but to do something else instead. So, even if it is rational for one to f, this does not entail that one would want oneself to f if one were fully rational; at most, it entails that wanting oneself to f is compatible with being fully rational. Thus, my version of the IR is preferable to Smith's counterfactual version.(10)
In this way, the IR imposes a motivational constraint on when it is true to make the ex ante statement 'It is rational for A to f'. This will entail a corresponding constraint on statements of the form 'There is a reason for A to f', since it seems intuitively clear that if there is a reason for A to f, then it is rational for A to f -- or at least would be rational for A to f, absent any stronger countervailing reasons. But for simplicity's sake I shall focus on rational action, rather than on reasons for action, for the rest of this discussion.
II. The argument for the HTPR
According to the HTM, every intentional action is motivated by some desire that is served by that action, and this desire is either itself an unmotivated contingent desire, or else is motivated (at least ultimately and in part) by such contingent desires. So -- it may seem -- every intentional action must be motivated (at least ultimately and in part) by a contingent desire that is served by that action. Since the HTM applies to rational actions just as much as to any other actions, it would follow that whenever one performs a rational action, that action must be motivated (ultimately and in part) by a contingent desire that is served by that action; no process of rational practical reasoning can motivate one to f unless that process involves a contingent desire that is served by one's f-ing.
In effect, this point is already a version of the HTPR: according to this point, the ex post statement 'A's f-ing was rational' implies that A's f-ing served, and was motivated by, one of A's contingent desires. Given this point, the IR will imply a parallel point with respect to ex ante statements, such as 'It is rational for A to f'. According to the IR, it is rational for A to f just in case there is a possible rational process by means of which the relevant state of mind might motivate A to f. But according to this point, no rational process can motivate A to f unless that process involves a contingent desire that would be served by A's f-ing. So it would follow that if it is rational for A to f, then there must be a possible rational process that starts out from the relevant state of mind, and involves a contingent desire that would be served by A's f-ing. As I shall interpret it, this is exactly what the HTPR says about ex ante statements of the form 'It is rational for A to f': if it is rational for A to f, then there is a possible rational process that starts from the relevant state of mind and involves a contingent desire that would be served by f-ing.
On this interpretation, the HTPR need not claim that every action that serves one of the agent's desires is rational. Humeans could impose many further restrictions on when actions are rational. For example, Humeans influenced by decision theory, such as James Dreier [4], might propose that the desires served by rational action must take the form of a 'coherent' set of preferences -- that is, a set of preferences that meets the 'axioms' that a set of preferences must meet if it is to be representable by means of a 'utility function'. Other Humeans, such as David Gauthier [5, pp. 26-33], might insist that the desires served by rational action should be desires that one is prepared to avow as well as to act on, or desires that are 'stable under experience and reflection'. Finally, some Humeans, such as Williams [17, p. 105], may even allow that the desires served by a rational action need not have existed before one started deliberating about whether to perform the action: the desires in question may have arisen during the course of deliberation. Nonetheless, all Humeans agree that every rational action must serve at least one contingent desire -- that is, a desire that it would not be irrational for one simply to lack altogether.
This argument from the HTM seems to support several different versions of the HTPR. Ex post versions of the HTPR say that whenever one acts rationally, one's action must have been motivated (ultimately and in part) by some contingent desire that was served by that action. Ex ante versions say that whenever it is rational for one to f, there is a possible process of rational practical reasoning that starts from the relevant (actual or idealised) state of mind and involves a contingent desire that would be served by one's f-ing. There are also two different kinds of both ex post and ex ante versions of the HTPR -- one kind that is concerned with objective rationality, and another kind that is concerned with subjective rationality.
Since I am interpreting the HTPR in the light of the argument that is based on the HTM, I shall concentrate on ex post versions of the HTPR. It is these versions of the HTPR that are most directly supported by the HTM. The HTM seems -- at least at first glance -- to imply that every action must be motivated by a contingent desire that that action serves. If that is true, then it clearly must be true of rational actions as well as all other actions; so the ex post versions of the HTPR must be true -- every rational action that anyone ever performs was motivated by a contingent desire that the action serves. This implies the point that no process of rational practical reasoning can motivate one to f unless this process somewhere involves a contingent desire that one's f-ing serves; and that point, together with the IR, implies the ex ante versions of the HTPR. In this way, the ex ante versions of the HTPR are less directly supported by the HTM than the ex post versions. Moreover, it is hard to see why anyone would accept the ex ante versions of the HTPR unless they also accepted the ex post versions. Why would anyone think that it can only be rational for you to f if there is a possible process of rational deliberation that starts from the relevant state of mind and involves a contingent desire that your f-ing would serve, unless they also thought that if you did rationally f, your motivation for f-ing would have to involve such a contingent desire?
Since we are interpreting the HTPR in the light of the argument that is based on the HTM, there is also reason to concentrate on the (ex post) version of the HTPR that concerns subjective rationality, rather than the version that concerns objective rationality. The HTM implies that every action must be motivated by a contingent desire -- regardless of whether or not the action is motivated by false beliefs, or whether the agent is adequately well informed. So, if the HTM supports any version of the HTPR, it must support the (ex post) version that concerns subjective rationality. If the argument does not support this version, it surely will not support the version that concerns objective rationality -- that is, the version that is restricted to those cases in which the agent is adequately well informed and is not motivated by any false beliefs.
In the next section, I shall show that the ex post version of the HTPR that concerns subjective rationality is false: it is quite possible that a person's action of f-ing was subjectively rational, even if the person had no contingent desire that her f-ing served. I shall also give some reasons for scepticism about the ex post version that concerns objective rationality. For the reasons that I have just given, this is enough, I believe, to undermine the HTPR in general.
III. Why the HTPR is false
My argument against the HTPR will be based on the following three assumptions about akrasia. First, akrasia involves acting against one's better judgment -- that is, willingly acting in a certain way, even though one holds a belief that amounts to a certain sort of 'condemnation' of acting in that way. To fix ideas, let us assume that akrasia involves willingly failing to f even though one believes that (all things considered) one had definitely better f. (My argument would work equally well if we interpreted 'acting against one's better judgment' in other ways -- for example, as 'doing something that one believes oneself to have overriding reasons not to do' or as 'doing something that one believes to be irrational' or whatever.)
Second, such akrasia is necessarily irrational.(11) Indeed, such akrasia seems to involve a sort of internal incoherence: one's beliefs 'condemn' a certain course of action, while the very same course of action is actively 'embraced' by one's will. Moreover, situations can clearly arise in which the only rational way to avoid akrasia is by forming an intention. Suppose that you hold a belief, which it would not be rational for you to abandon, to the effect that you had definitely better f (where f-ing is something that you will willingly fail to do unless you form an intention to f -- henceforth I shall omit this qualification). In this situation, it would be irrational for you to abandon this belief; but if you do not abandon this belief, and also do not form an intention to f, you will willingly fail to f -- which would be akratic and so also irrational. In any situation of this kind, you are rationally required to form an intention to f.
Third, it is possible for there to be agents who are generally disposed to comply with rational requirements of this kind. That is, whenever these agents hold a belief, which it would not be rational for them to abandon, of the form 'I had definitely better f', they are disposed to respond to this situation by forming an intention to f.(12) Indeed, it is arguably irrational for anyone who holds beliefs of this sort not to have this disposition. (If this is so, then a sort of 'belief internalism' would be true of such beliefs: if one is rational and holds a belief of this sort, one must also be disposed to form the corresponding intention.) For this reason, I shall refer to agents who have this disposition as 'rational agents'. However, for present purposes, this may be regarded merely as a label. All that I need here is the assumption that it is metaphysically possible for an agent to have this disposition.
As I am understanding it, this is a disposition to form an intention to f simply in response to one's having a belief (which it would be irrational for one to abandon) of the form 'I had definitely better f'. The stimulus to which this disposition responds is simply the fact that one has a belief of this sort: nothing more is needed to activate this disposition, besides one's having such a belief. It is hard to see why such a disposition should be metaphysically impossible. As I shall argue in the next section, this disposition is quite compatible with the HTM. Anyway, why should anyone -- and especially a Humean -- think that there are a priori or metaphysical constraints on which such causal connections are possible?(13)
This disposition also seems an eminently intelligible disposition to have. It is simply a disposition to comply with rational requirements of a certain kind. It is quite intelligible why this disposition is activated purely by the fact that one has a belief of this sort. As I have argued, any belief of this sort is enough, all by itself, to make it irrational for one not to form the intention to f. So, in forming the intention to f purely in response to the fact that one holds such a belief, one is forming this intention in response to a fact that is enough, all by itself, to make one's forming this intention rationally required.
Suppose that you are a 'rational agent' of this kind, and that you hold a belief (which it would be irrational for you to abandon) that you had definitely better f. Then you must be disposed to form an intention to f, purely in response to the fact that you hold this belief; and it must also be possible for this belief to activate this disposition, with the result that you form the intention to f. Suppose that in normal circumstances, if you intend to f you will actually f, and that circumstances are normal in this way, so that you actually f. If that happened, we could give a true explanation of your f-ing, by saying that you f-ed precisely because you believed in this way that you had definitely better f.(14) Obviously, we could say more to explain why you held this belief; but otherwise, this would be a perfectly adequate explanation of your f-ing. As I have argued, this disposition, to respond to this belief that you had definitely better f by forming an intention to f, is an entirely intelligible disposition to have. So, given your belief that you had definitely better f, and given this entirely intelligible disposition (and given that circumstances are normal, so that if you intend to f you will actually f), it is entirely intelligible and unsurprising that you f-ed. That is, this belief was your motivation for f-ing.
In this case, it seems that your f-ing was subjectively rational. Since it would be irrational for you to abandon your belief that you had definitely better f, this belief is presumably itself subjectively rational. So this activation of your disposition to respond to such beliefs by forming the corresponding intention presumably also counts as subjectively rational practical reasoning. Since this belief motivated you to f by means of such subjectively rational practical reasoning, your f-ing was also subjectively rational.
However, it seems quite possible that in this case you had no contingent desire that was served by your f-ing. Perhaps the HTM is correct to claim that since you intended to f, you did indeed have a desire that was served by your f-ing -- namely, the desire to f. But in this case, this was not a contingent desire. In this case, it would have been irrational for you not to believe that you had definitely better f; and as I have argued, it would also be irrational for you to believe that you had definitely better f without intending to f and so also having a desire to f. So in this case it would also have been irrational for you not to have such a desire to f; this is a desire that you were rationally required to have.
Is there any other reason for thinking that in this case you must have had some contingent desire that was served by your f-ing?(15) Let us suppose that you came to hold your belief 'I had definitely better f' in the following way. You have an adviser, whom I shall call 'Diotima', of extraordinary benevolence and wisdom. Diotima told you that you had definitely better f; you have such strong evidence of her reliability that it would have been irrational for you not to believe her; and you believed her for this reason. However, suppose that Diotima did not have time to explain to you why you had definitely better f; you just took her word for it. So you did not arrive at the belief that you had definitely better f as a direct result of any deliberation that started out from your contingent desires.(16) Thus, neither the fact that you intended to f nor the fact that you had a belief of this sort clearly entails that you had any contingent desire that your f-ing served.
At this point, the Humean might attempt to defend the HTPR by offering an account of the content of the belief that the akratic agent 'acts against' (which for the sake of argument I am assuming to be the belief 'I had definitely better f'). But no account of the content of this belief will imply that it is impossible for one to hold such a belief unless one has a contingent desire that would be served by one's f-ing. We can see this by considering Williams' account of the content of this belief.
According to Williams [17, p. 109], the content of the belief that the akratic agent acts against can only be 'the proposition, or something that entails the proposition, that if he deliberated rationally he would be motivated to act accordingly'. Even if this is the correct account of the content of this belief (which is far from obviously true)(17) this would not help the HTPR. It is quite compatible with the case that I have described that your belief that you had definitely better f is false. (Diotima's wisdom is extraordinary, but she is fallible.) So, even if the content of this belief implies the proposition 'If you deliberated rationally, you would have a contingent desire that would be served by your f-ing', that proposition could be false. So you still need not have had any contingent desire that would be served by your f-ing. And even though Diotima's statement was false, you were not in a position to know that. So you were still rationally required to believe her; and your f-ing was still subjectively rational.
In fact, Williams' account of the content of this belief does not even entail that your f-ing cannot have been objectively rational if you had no contingent desire that would be served by your f-ing. Suppose that your action of f-ing was objectively rational: you were not ignorant of anything that a reasonably well-informed agent would know, and your belief 'I had definitely better f', and any other beliefs that motivated your action, were true. Then, given Williams' interpretation of this belief, it is also true that if you deliberated rationally, you would have a contingent desire that would be served by your f-ing. But all this is compatible with your never actually having 'deliberated rationally' in the relevant way. (After all, Diotima advised you to f, so further deliberation was simply unnecessary.) As Williams himself admits [17, p. 105], such 'rational deliberation' may involve exercising one's 'imagination' in a way that 'can create … new desires' -- presumably including new desires whose satisfaction would not serve any of one's pre-existing desires.(18) So, even if it is true that if you had 'deliberated rationally', you would have had a contingent desire that would be served by your f-ing, it could still be the case that you never actually had any such contingent desire. But your f-ing was still objectively rational. So this account of the content of the relevant belief does not support the HTPR with respect to either subjective or objective rationality.
The only account of the content of this belief that clearly supports a version of the HTPR is a crudely subjectivist account. For example, suppose that the content of this belief is simply: 'One of my contingent desires will be served by my f-ing'. Clearly, this belief cannot be true unless the believer has a contingent desire that will be served by his f-ing. So, on this crude subjectivist account, if your f-ing was objectively rational because it was motivated by this belief, you must have had a contingent desire that your f-ing served. But your f-ing could still have been subjectively rational even if you had no such desire. It is quite possible for one to hold the belief 'One of my contingent desires will be served by my f-ing' (which it would be subjectively irrational for one to abandon), even if that belief is in fact false. If it is akratic to act against this belief, then -- as we have seen -- it is possible for this belief to motivate you to f. If this belief did motivate you to f, then your f-ing was subjectively rational. So even on this extreme subjectivist account, your f-ing could have been subjectively rational even if you had no contingent desire that your f-ing served.
In conclusion, the only way that we have found to defend the HTPR is to resort to an extreme subjectivist account of the beliefs that it is akratic to act against; and even if we do resort to this account, we can only maintain the HTPR with respect to objective rationality (not subjective rationality). It is hard to see what reason there could be to accept this extreme subjectivist account. As I shall argue in the next section, the HTM gives us absolutely no reason to accept any version of the HTPR -- let alone an extreme subjectivist version of this sort.
IV. How the HTM could be true even if the HTPR is false
In the previous section, I argued that it is possible to perform an action -- indeed, a rational action -- even though one has no contingent desire that is served by that action. This might seem flatly incompatible with the HTM. But in fact this is not incompatible with the HTM. According to the HTM, whenever one acts, one's action is motivated (at least in part) by some desire that would be served by that action; and that desire is either an unmotivated contingent desire, or else is motivated (at least ultimately and in part) by antecedent contingent desires. This leaves open the following possibility: the action is motivated by a desire that is served by the action, but that desire is not a contingent desire, but rather a desire that one is rationally required to have; and although this rationally required desire is ultimately motivated by antecedent contingent desires, those antecedent desires are not themselves desires that are served by the action.
Humeans may think that this possibility can be ruled out: how could a rationally required desire to f be ultimately motivated by contingent desires that are not served by one's f-ing? As I shall explain, there is no problem at all in making sense of this possibility. For this reason, the argument from the HTM to the HTPR is invalid.
The argument from the HTM to the HTPR has been criticised before, most notably by Christine Korsgaard [8]. Korsgaard's main point in that discussion, however, was not to criticise this argument, but to criticise any attempt to argue for the HTPR on the basis of the IR alone. The requirement that reasons 'must be capable of motivating rational persons' does not show that all reasons depend on contingent desires, since it may be that a disposition to be motivated by certain reasons is necessarily present in all rational persons [8, pp. 325-29]. Korsgaard does criticise the argument that I am concerned with -- the argument from the HTM to the HTPR -- but not on the grounds that it is invalid. In effect, she criticises this argument on the grounds that it is circular: we could have no reason to accept the HTM, she claims, unless we already had a reason to accept the HTPR [8, e.g. p. 314].
In fact, however, the argument is invalid. It is quite possible for a desire to f to be both rationally required and at the same time motivated (ultimately and in part) by certain contingent desires, even though none of those contingent desires is served by one's f-ing.
As we saw in the previous section, one is rationally required to intend to f, and so also to desire to f, whenever one holds a belief, which it would be irrational for one to abandon, of the form 'I had definitely better f'. Moreover, given the disposition, which all 'rational agents' have, to respond to such beliefs by forming an intention (and so also a desire) to f, this belief is sufficient to motivate this desire. Suppose that this belief, all by itself, did motivate this desire. Then this desire was motivated (at least ultimately and in part) by contingent desires only if this belief was also motivated (at least ultimately and in part) by contingent desires. Given that, in rational agents, these beliefs are sufficient to motivate desires, the HTM clearly implies that, in rational agents, these beliefs must themselves be motivated, ultimately and in part, by contingent desires. For suppose that the motivational history of these beliefs did not contain any such desires. Then a rational agent could arrive at the belief that she had definitely better f, and this belief could motivate a new desire (the desire to f), even though the process of thought that produces this new desire (as 'output') could not itself be traced back to any antecedent contingent desire (as 'input'). So the HTM must insist that, in all rational agents, these beliefs must themselves be motivated, at least ultimately and in part, by contingent desires.
How could it be that, in all rational agents, these beliefs must be (at least ultimately and in part) motivated by -- or in other words, based or grounded on -- such contingent desires? At least part of the answer will be an epistemological story about these beliefs, according to which it is rational to hold these beliefs only if they are based, at least ultimately and in part, on such desires.
There are many possible stories that would imply that the relevant beliefs must be based in some such way on desires. Such stories need not claim that all unmotivated contingent desires play this role as part of the ultimate rational basis for these beliefs; it may be only certain special desires that play this role. It may also be that these beliefs are not directly based on these desires, but are instead based on a complex process of deliberation that is itself based, in part, on these desires. Although the details may vary, the basic idea behind all such stories is that certain unmotivated contingent desires provide (weak, defeasible) support for beliefs about what one had better do, in something like the same way that experiences provide support for ordinary empirical beliefs. These epistemological stories also need not entail that it cannot be rational for you to believe that you had definitely better f solely because Diotima tells you so. At most, they entail that it is not rational for you to trust Diotima in this way unless her advice has tended, in the past, to be supported in the relevant way by your desires.
One example of an epistemological story according to which certain desires play a crucial evidential role in the generation of such beliefs is Michael Smith's account [13, pp. 159-61, 175-77] of how we form rational evaluative beliefs.(19) However, Smith does not see that this epistemological story is the key to reconciling the HTM, as I understand it, with the falsity of the HTPR. Indeed, he does not confront the HTM, as I understand it, at all: the distinction between motivated and unmotivated desires plays no role in his argument.(20)
Suppose that some such epistemological story is true. That is, suppose that whenever rational agents hold beliefs of the relevant kind, these beliefs are based, at least indirectly and in part, on unmotivated contingent desires. Presumably, in some cases, these contingent desires could provide such strong support to the belief 'All things considered, I had definitely better f' that one is rationally required to hold that belief. As I have argued, if one is rationally required to hold that belief, one is also rationally required to desire to f.
However, as I have noted, the support that one's contingent desires provide to such beliefs need not be direct. Instead, these desires may support such beliefs holistically: one's entire set of desires, of the appropriate sort, may form an input to a complex process of deliberation that itself grounds the beliefs in question. So, in some of these cases, this belief could be supported by one's contingent desires even if none of these desires would be served by one's f-ing. Admittedly, these contingent desires are at least part of what made it rational for one to hold this belief, and this belief is what made it rational for one to f; so the fact that one's f-ing was rational is not independent of one's contingent desires. But it is independent of whether one had any contingent desires that were served by one's f-ing. So, these cases are incompatible with the HTPR, even though they are perfectly compatible with the HTM.
Moreover, this point holds even if one's belief 'I had definitely better f' is true. According to this epistemological story, one must rely, at least ultimately and in part, on one's desires in coming rationally to hold any beliefs of the form 'I had definitely better f'. But it does not follow that the truth of these beliefs entails anything about one's desires at all. For all that has been shown so far, it is quite compatible with this epistemological story that in some cases you had definitely better f even though you have no contingent desires that would be served by f-ing -- indeed even though you might continue not to have any such contingent desires, however long you rationally deliberated about the question.(21) This shows that the HTPR could be false, with respect to both subjective and objective rationality, even if the HTM is true.
In conclusion, the HTPR could be false even if the HTM is true -- that is, even if contingent desires are among the unmotivated motivators of every action. The most that follows from the HTM is a broadly sentimentalist epistemology for the beliefs that it is akratic to act against (which I have been assuming, for the sake of argument, to be beliefs of the form 'All things considered, I had definitely better f'). This epistemological story falls far short of the HTPR. According to this story, practical reason does indeed grow from the material provided by our unmotivated contingent desires; it cannot grow -- as the Kantians would have us believe -- in the empty space of pure a priori reason. But practical reason need not always serve desire. As we have seen, practical reason can be guided by a certain sort of rational evaluative belief. Through the attempt to form such beliefs in a rational way, and to act in the light of these beliefs, practical reason may rise high indeed above the lowly desires from which it grows. (22)
Merton College, Oxford
NOTES
1. In calling it the 'Humean' theory, I do not mean to imply that it is Hume's theory -- only that it has been held by philosophers who considered themselves influenced by Hume.
2. This summary of the argument for the HTPR is due to Nagel [11, p. 27]. For a more elaborate version of this argument for the HTPR, see Williams [17].
3. See e.g. the definition of the HTM given by Smith [13, p. 93]. However, Smith later adds the claim that 'beliefs and desires are distinct existences', in the sense that there are no 'besires' -- no beliefs that are, at the same time, desires [13, pp. 118-25].
4. In this passage, Williams seems to be using the term 'motivation' roughly as I use the term 'desire'. In my usage, A's 'motivation' for -ing consists of the set of A's mental states that motivated A to ; this set of mental states will typically contain beliefs as well as desires.
5. In effect, Williams [18, pp. 36-7] accepts this point in responding to Korsgaard [8].
6. Smith seems to use the term 'rational' in this way [13, pp. 156f.].
7. For the terms 'objectively rational' and 'subjectively rational', see Brandt [2, pp. 72f].
8. Compare how Goldman [6, p. 21] distinguishes between ex post and ex ante uses of 'justified belief': 'The ex post use occurs when there exists a belief, and we say of that belief that it is (or isn't) justified. The ex ante use occurs when we … ignore … whether such a belief exists. Here we say of the person … that p is (or isn't) suitable for him to believe'.
9. Objection: What about cases in which the reason why it is 'objectively rational' for one to is precisely that one is not ideally well informed about the relevant facts? Reply: The relevant 'idealised state of mind' need not be completely idealised: it need not contain all relevant true beliefs. Thus, this idealised state of mind may include such true beliefs as 'I am ignorant of certain relevant facts'. (Not every partial idealisation of one's actual state of mind, however, can be used in defining what it is 'objectively rational' for one to do; misleadingly partial idealisations will have to be ruled out somehow. But I cannot go into this question here.) At all events, some appeal to idealisation must feature in any defensible version of the IR that is true of 'objective rationality' (and if the IR is not true of objective rationality, then there is no way of arguing from the HTM to any conclusion about ex ante objective rationality at all).
10. Such counterfactual versions of the IR may also commit the 'conditional fallacy'; see Johnson [7].
11. Hard-line Humeans will object to this second assumption. See Blackburn [1, p. 65]: 'Doing what you know to be bad is bad. We might describe it as irrational, and since Plato many philosophers have done so. But that is not, as it stands, a very interesting thing to say, for it is not at all obvious what further or different specific charge it makes.' But I would argue that irrationality is a specific charge. First, the basic rules or principles of rationality are rules that we can follow directly, without having to follow any other rules in order to do so. Second, as I have argued elsewhere [15], we must have at least some mastery of these rules if we are even to be capable of having the mental states that are governed by those rules.
12. To say that these agents have this disposition is not to say that this disposition is activated in every particular case. Many factors, such as 'depression' or 'listlessness', might prevent this disposition from being activated in particular cases; but that is compatible with these agents' being generally disposed to comply with rational requirements of this kind.
13. The shrewdest defenders of the HTM accept this point. For example, Lenman [10] defends the HTM as a claim, not about which causal connections are possible, but about which such connections are sufficiently intelligible to count as cases of motivation.
14. Essentially the same account of how normative beliefs can explain action is given by Broome [3, pp. 141-42].
15. Some philosophers might suggest that your disposition to respond to your believing 'I had definitely better ' by forming an intention to is itself a desire -- in effect, the desire to do whatever you had definitely better do. As I mentioned, however, this is arguably a disposition that you are rationally required to have, in which case it would not be a contingent desire. Anyway, this purely 'dispositional desire' seems quite different from more ordinary desires: its content would have to be a general proposition ('I will do everything that I had definitely better do'), which you need never have consciously entertained at all.
16. It may be objected that cases of this sort, in which one believes that one had definitely better because one takes someone else's word for it, are parasitic on cases in which such beliefs are based on one's contingent desires. In §IV below, I shall suggest an epistemology for such beliefs that explains how even in non-parasitic cases, one can rationally believe that one had definitely better , even if one has no contingent desire that would be served by one's -ing.
17. I have given my own account of the content of these beliefs elsewhere [16].
18. Williams' idea seems to be this: in deliberation we may imagine 'what it would be like' if a certain state of affairs came about; we may thereby acquire 'a more concrete sense of what would be involved', and this can create a desire for that state of affairs. There is no reason to think that this can only happen if that state of affairs serves some pre-existing contingent desire.
19. Another example is the epistemological story that Korsgaard [9] ascribes to Aristotle.
20. For Smith, this epistemological story is simply a consequence of his analysis of the content of the relevant beliefs. He supports his analysis by arguing that it explains how (a weak version of) 'belief internalism' can be true of these beliefs, even though all motivation involves desires, and beliefs and desires are 'distinct existences' [13, pp. 177-80]. I do not believe that any such analysis of the content of these beliefs is necessary in order to solve this problem. I have given my own explanation of how 'belief internalism' can be true elsewhere [16].
21. Compare Scanlon [12, pp. 368-70]. There is nothing paradoxical about this. There are many beliefs that themselves entail nothing about one's existing mental states, even though one must of course rely on one's existing mental states in coming to hold any belief at all.
22. Earlier versions of this paper were presented to audiences at Duke University and at the 1999 CSPA conference in Norman, Oklahoma. I thank the members of those audiences, and especially my commentator in Oklahoma, Mark van Roojen, and also James John, Martin Stone, Judith Thomson, and a number of anonymous referees, for helpful comments. I am also grateful to the National Humanities Center and the Research Triangle Foundation of North Carolina for a Fellowship at the Humanities Center during 1998-99, when the first draft of this paper was written.
REFERENCES