The Aristotelian Society 133rd session – 2011/2012 18 June 2012 | 16.15 – 18.00
Agents and Patients Michael Smith
Princeton University
The Woburn Suite Senate House (South Block) University of London Malet Street London WC1E 7HU United Kingdom
** non-citable draft paper **
abstract Can we draw substantive conclusions about the reasons for action agents have from premises about the desires of their idealized counterparts? The answer is that we can. The argument for this conclusion is Rawlsian in spirit, focusing on the choices that that our counterparts have to make simply in virtue of being ideal agents, and inferring from these choices the contents of the desires that they have to have. But whereas Rawls asks us to imagine ourselves choosing principles that will govern the basic structure of a society in which we will live, the argument given here focuses on choices made in much more mundane circumstances in which our basic agential capacities are exercised. What would our ideal counterparts have to choose to do in order to bring their capacities for belief-formation and desire-realization into coherence with each other when they are engaged in certain mundane processes of thought? Somewhat surprisingly, the answer is that they would have to choose not to interfere with the exercise of capacities for belief formation and desire-realization, not just their own but others' as well, and that they would also have to choose to help make sure that there are such capacities to exercise. The upshot is that all agents have reasons to help and not interfere, and, beyond that, that they have reasons to do whatever they desire to do. This conclusion is itself reminiscent of Rawls's own, much as might be expected.
biography Michael Smith is the author of The Moral Problem (1994); Ethics and the A Priori: Selected Essays on Moral Psychology and Meta-Ethics (2004); and the co-author of Mind, Morality and Explanation: Selected Collaborations (2004), a collection of papers written in various combinations by Smith, Frank Jackson and Philip Pettit when they were colleagues at the Research School of Social Sciences, Australian National University. In 2004 Smith moved to Princeton University where is currently McCosh Professor of Philosophy.
!
! AGENTS AND PATIENTS, OR: WHAT WE LEARN ABOUT REASONS FOR ACTION BY REFLECTING ON PROCESS-OF-THOUGHT CASES* Michael Smith Can we draw substantive conclusions about the reasons for action agents have from premises about the desires of their idealized counterparts? Many will agree that we can if we stipulate both that there is a connection between reasons for action and idealized desires, and that the norms governing the idealization are moral norms. But what if the ordinary concept of a reason for action is in play, and the norms governing the idealization are the norms internal to the concept of agency itself? For example, what if we think of an agent as a functional kind defined by the possession and exercise, to some degree or other, of the capacities to have knowledge of the world in which he lives and realize his desires in it, and hence think of the norms governing the idealization as those to which an agent conforms when he fully and robustly possesses and exercises this pair of capacities? My own view is that substantive conclusions about reasons for action follow from premises like these, premises that make no moralized assumptions about the norms internal to the concept of agency (see also Smith 1994). In saying this I am, of course, swimming against a tide. Bernard Williams famously argues that, in the ordinary sense in which people have reasons for action, an agent has a reason to act in a certain way in certain circumstances just in case he would desire that he acts in that way in those circumstances if he were to deliberate correctly, and he further argues that correct deliberation has to be understood in terms of an agent's possession and exercise of the two capacities that I have said characterize an ideal agent (Williams 1981). Williams's account of the ordinary concept of a reason for action is thus much the same as my own, but he thinks this conception entails no substantive conclusions. He thinks it entails instead that what an agent has reason to do is relative to his potentially idiosyncratic and immoral desires, a conclusion that he goes to great lengths to emphasize. There is therefore confusion, or so it might seem, in my supposing that we could move from Williams-style anti-rationalist premises about the nature of reasons for action to a rationalist conclusion about the substance of the reasons that people have. But though there is some truth to this characterization of my argument as moving from Williams-style premises to a rationalist conclusion, I will !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! *
A distant ancestor of this paper was given at Humean Ethics, The Second Warren Quinn Memorial Conference, UCLA, in 1998. I put the paper to one side after that and didn't return to it until I spent a perfect sabbatical year at the Humboldt University in Berlin on a Humboldt Research Award from the Alexander von Humboldt Foundation in 2010-2011. More recent versions of the paper were given as the Inaugural LEAP Lecture, Universitat Pompeu Fabra, Barcelona, in 2011; as a Keynote Address at Reasons of Love, Leuven, 2011; at the Department of Philosophy, Cambridge University, in 2011; at the Department of Philosophy Colloquium, GĂśttingen, in 2011; at the Boston University Ethics Reading Group in 2011; at Thomas Schmidt's Colloquium, Humboldt University, in 2011; at Tag der HandlungsgrĂźnde, Saarland University, in 2011; as the Claire Miller Lecture at the Chapel Hill Colloquium in 2011; at a meeting of Paper Tigers, Princeton University, in 2011; at the Summer Workshop on Rationalism at Melbourne University in 2012; at the Department of Philosophy Colloquium, University of California at Davis, in 2012; at the Department of Philosophy Colloquium, Fordham University, in 2012; at the Murphy Institute, Tulane University, in 2012; as one of three Hourani Lectures given at the University at Buffalo in 2012; at Practical Reason and Metaethics, a conference held at the University of Nebraska at Lincoln in 2012; and at a graduate seminar I gave on the topic of reasons and rationality at Princeton University during the Spring Semester of 2012. Thanks to all those who gave me such excellent comments on these occasions. If nothing else, I hope they will agree that the paper has been much improved by their input. Special thanks to Joseph Biehl, David Braun, David Copp, Chris Cowie, Simon Cullen, Christoph Fehige, Elizabeth Harman, Barbara Herman, Richard Holton, Katie Keene, Errol Lord, Barry Maguire, Julia Markovits, Geoff SayreMcCord, Thomas Schmidt, Noel Swanson, Jay Wallace, Andrew Williams, and Daniel Wodak.
!
1
!
! argue that it involves no confusion. As I see things, theorists like Williams should abandon their anti-rationalism for reasons internal to their own understanding of correct deliberation. Though the argument I will go on to give is rather different from those given by other rationalists for similar conclusions, it will be helpful to note at the outset a respect in which it is similar to an argument that Rawls gives in A Theory of Justice (Rawls 1971). Rawls invites us to imagine ourselves choosing principles to govern the basic structure of a society in which we will eventually live, but he tells us that we must make this choice from a position in which we are ignorant about who we will eventually turn out to be in that society, and ignorant of what our potentially idiosyncratic desires are when we make that choice. Since anti-rationalists think that all choices express our potentially idiosyncratic desires, they are committed to denying the possibility of our making the choice that Rawls describes. They should therefore find Rawls's suggestion that we can make the imagined choice utterly baffling. One attraction of Rawls's argument, as I see things, thus lies in the way it focuses our attention on what's really at issue in the debate between rationalists and antirationalists. If a choice of the kind he describes is so much as possible, as both he and I think it is, then that choice must itself be the expression of a desire we have simply in virtue of being capable of rational choice (we have, after all, abstracted away from our potentially idiosyncratic desires, so all that's left to ground that choice is what's true of us simply in virtue of being rational choosers), and that desire must in turn be one that isn't especially concerned with ourselves (as we have also imagined ourselves ignorant of who we are). The coherence of Rawls's imagined choice thus augurs in favour of rationalism and against anti-rationalism. The argument I go on to give is a lot like Rawls's in that it forces us to focus on choices made in similar circumstances. Moreover, as we will see, it has a very similar upshot to Rawls's argument. But it is unlike Rawls's argument in focusing on choices made in much more mundane circumstances. I have said that theorists like Williams should abandon their anti-rationalism for reasons internal to their own conception of correct deliberation. So what is the problem with that conception? The problem is that the conception purports to be one that honours requirements of coherence, but that it fails spectacularly to do so. The psychology of an agent who fully and robustly exercises maximal capacities to have knowledge of the world in which he lives—and from here-on I will assume that the capacity to have such knowledge reduces to the capacity to access evidence about how the world is and then form beliefs based on that evidence—and desire realization is supposed to be one whose psychology realizes the virtues of coherence in both the theoretical and practical domains. The evidence available to the agent and his beliefs are supposed to cohere well with each other, and the non-instrumental desires, beliefs about means, and instrumental desires he has are also supposed to cohere well with each. But, as we will see, exercises of these two capacities seem not to cohere all that well with each other. The problem is thus that the conception of correct deliberation, while purporting to give pride of place to considerations of coherence, builds in an intolerable amount of incoherence. In order to see that this is so, it suffices that we remember that the capacity for desire realization is one that an idealized agent is supposed to have no matter what the content of his desires turns out to be. His desires can be as idiosyncratic as you like. So imagine an agent who desires that he now believes that p, but imagine further that p isn't the case and that the evidence this is so is available to him. This agent is in a synchronic bind. If he fully and robustly exercises his capacity for desire realization, then he cannot fully and robustly exercise his capacity for belief formation, and vice
!
2
!
! versa. Even if he just so happens to find himself in a world in which he has no such desire, so there is no actual conflict between the exercises of the two capacities in that world, the coherence of their deliverances will at best be a happy accident. The agent still cannot be such that he fully and robustly exercises the two capacities, as he is in a synchronic bind in the possible world in which he does have that desire, but p isn't the case and the evidence that that is so is available to him. The mere fact that there exist possible worlds where the deliverances of the two capacities diverge is thus sufficient to make trouble for the full and robust exercise of the two capacities that Williams tells us underwrite correct deliberation. If we were to stick with Williams's conception of an ideal psychology, then it follows that we would have to give separate scores for the extent to which an agent's psychology is ideal. The situation would be much the same as in Olympic diving where separate scores are given for the difficulty of a dive and its execution. There would be one score for the extent to which an agent possesses and exercises the capacity for accessing evidence and forming beliefs in the light of that evidence, and a separate score for the extent to which he possesses and exercises the capacity for desire-realization. The fact that we would have to give two separate scores is, if you like, a vivid marker of the fact that the two capacities do not cohere all that well with each other. The question is whether we can happily live with this separate scores conception of an ideal psychology, and the answer turns on whether there is some alternative way in which we can bring the two capacities more into coherence with each other. If there is, then the mere fact that the norms of idealization give pride of place to norms of coherence will demand that we revise our conception of an ideal psychology accordingly. There are at least three ways in which we could revise our conception of an ideal psychology so as to ensure that that psychology is more robustly coherent. The first would be for the coherence of an agent's psychology to be wholly based on the extent to which he possesses and exercises the capacity to realize his desires. The possession and exercise of the capacity to know his world might still be required for his having an ideal psychology from time to time, but only insofar as it contributes to desire realization. The second would be for the coherence of an agent's psychology to be wholly based on the extent to which he possesses and exercises the capacity to access evidence and form beliefs on their basis. The possession and exercise of the capacity to realize desires might still be required for his having an ideal psychology from time to time, but only insofar as it contributes to knowing his world. Both of these first two ways of revising our conception of an ideal psychology are seriously revisionary. The first tells us that an ideal psychology is prone to all sorts of dysfunction in the formation of beliefs. The second tells us that an ideal psychology is prone to all sorts of failures of instrumental rationality. For this reason, I take it that neither way of revising our conception is attractive. The separate scores conception is preferable to each of them because it at least admits that dysfunction in the formation of beliefs and instrumental failure are as such departures from the ideal. There is, however, a third way in which we could revise our conception of an idealized psychology, a way that also admits that dysfunction in the formation of beliefs and instrumental failures are departures from the ideal, but is far less revisionary. It is therefore more attractive than the other two revisionary proposals and the separate scores conception. According to the third way of revising our conception of an ideal psychology, having certain coherence-making desires is partially constitutive of what it is to have such a psychology. Focus once again on the agent who desires that he now believes that p. If in order to have an ideal psychology that agent has to have a dominant or overriding
!
3
!
! desire that he does not now interfere with his current exercise of his belief-forming capacities, then there is no conflict of the kind described earlier between the full and robust exercise of his belief-forming capacities and his desire-realization capacities. This is because the only way in which he could possess and exercise the latter in worlds in which he is otherwise ideal, but desires to believe that p, would be by leaving himself free to exercise the former. The third way of revising our conception of an ideal psychology is preferable to the other two because it doesn't require us to suppose that dysfunction in the formation of beliefs, or instrumental failure, is an integral part of the ideal. The coherence of the two capacities with each other is instead grounded in the fact that an ideal agent, one who fully and robustly possesses and exercises the capacities to believe for reasons and realize his desires, has a dominant synchronic coherence-making desire. An agent's having a dominant desire that he does not now interfere with his current exercise of his belief-forming capacities ensures that his exercise of his desire-realization capacities chimes with his exercise of his belief-forming capacities. It is difficult to overstate the importance of this conclusion. I said earlier that Williams emphasizes that an ideal agent's desires can be as idiosyncratic as you like. But if the conclusion argued for thus far is correct, Williams is fundamentally mistaken. Though an ideal agent may well have desires that are idiosyncratic, he will have a reason to act on his idiosyncratic desires only if their satisfaction is consistent with his current exercise of the capacity to form beliefs for reasons. This is because the ideal agent has to have, simply in virtue of being ideal, a dominant or overriding desire that he does not interfere with his current exercise of his capacity to believe for reasons. Note that the argument for this conclusion has been, as advertised, thoroughly Rawlsian in spirit. We discover that an ideal agent has to have such a desire by reflecting on a choice that he can make, indeed must make insofar as he is ideal, a choice that no idiosyncratic desire he has could possibly explain. Subject to a qualification that will be spelled out presently, the upshot is that all agents have a reason not to interfere with their current exercise of their belief-formation capacities, no matter what their idiosyncratic desires happen to be, and that they have this reason in virtue of the synchronic coherence-making desire that their idealized counterparts have to have simply in virtue of being ideal. The main claim of the paper has therefore been established. Premises about the desires of agents' idealized counterparts do indeed entail substantive conclusions about their reasons for action. Once we appreciate that ideal agents must have one coherence-making desire, and hence that all agents share a common substantive reason for action, an obvious question to ask is whether they have to have any other such desires, and hence whether there are any more substantive reasons for action that all agents have in common. The answer is 'yes' to both questions. One remarkable fact about agents is that they exist over time. Their actions take time to perform, and so too does the formation of many of their desires and beliefs. This means that there are two kinds of coherence that an agent can achieve: synchronic coherence, which is what we've been talking about thus far, and diachronic coherence. As we will see, another desire an ideal agent has to have, simply in virtue of being ideal, is a diachronic coherence-making desire. Imagine an agent who is in the business of exercising his capacity to believe for reasons by engaging in a process of thought. Right now he is trying to figure whether p is true. If it turns out to be true, then he will later try to figure out whether p supports q. If it does, he will go on to draw the conclusion that q. However let's also imagine that he now has the desire to believe that not q later. If this agent fully and robustly possesses and exercises the capacity to believe for reasons and realize his desires now, then even if he does have the desire that he does not now interfere with his current exercise of his !
4
!
! belief-forming capacities, he would still be in a similar situation, diachronically, to the synchronic bind we've been talking about thus far. On the assumption that an ideal agent is one who fully and robustly possesses and exercises the capacity to believe for reasons and realize his desires not just synchronically, but also diachronically, we must therefore suppose that the ideal agent has another desire that dominates his potentially idiosyncratic desires as well. He must desire that he does not now interfere with his later exercise of his belief-forming capacities. A diachronic coherence-making desire must therefore also be partially constitutive of what it is to be an ideal agent. Further desires are also required. Imagine the same case, but instead of supposing that the agent has a desire now to believe that not q later, suppose that he is feeling slightly distracted. Perhaps feeling distracted is a condition he has from time to time, one that comes in waves when he has low blood sugar, something that varies predictably during the day. His feelings of distraction make it hard for him to concentrate, but not so hard as to interfere with his exercise of his capacity to figure out whether p is true. He also has one sugar pill that would get rid of his feelings of distraction, so making it easier for him to think now, but he also knows that, even if he were to get rid of his feelings of distraction now, the feelings of distraction will return later when, if all goes well, he will be trying to figure out whether p supports q, and, if they do, he will then feel so distracted that his capacity to figure out whether p supports q will be severely diminished. The question is whether to take the sugar pill now or hold on to it and take it later. It seems to me that if the agent we are imagining is ideal, in the sense that he fully and robustly possesses and exercises the capacity to form beliefs and realize his desires not just synchronically but also diachronically, then we already know the answer. He will hold onto the sugar pill and take it later. But nothing said so far guarantees this result. Since the feelings of distraction will diminish his future capacity to believe for reasons, it needn't be the case that his future exercise of this capacity is being interfered with by his not holding on to the sugar pill now so that he can take it later. This suggests to me that the ideal agent must therefore have yet another dominant desire as well, one with a rather different content. Insofar as he is ideal, he must desire that he now does what he can to help his later self have the capacity to believe for reasons to exercise. His possession of this additional dominant desire, together with the fact that he fully and robustly exercises his capacity to realize his desires now, explains why, as a diachronically ideal agent, he would hold onto the sugar pill and take it later. So far the focus has been entirely on what's required for synchronic and diachronic coherence in the exercise of the capacity for belief-formation. But nearly everything that has been said on this front applies, mutatis mutandis, to an ideal agent's desires concerning his possession and exercise of his desire-realization capacities as well. This is because agents who fully and robustly possess the capacity to realize their desires, not just synchronically but also diachronically, must be invulnerable not just to their having desires concerning what to believe, but also to their having desires concerning which desires to realize. Imagine, for example, an agent who now desires that he later realizes the desire that p, and who therefore lays traps for his future self so as to ensure that no other desire he might happen to have later is realized instead. Agents who fully and robustly possess and exercise the capacity to realize their desires not just synchronically, but also diachronically, do not fall into such traps (compare Nagel 1970). An ideal agent must therefore have versions of the desires already described concerning his exercise of his belief-formation capacities concerning his exercise of his desirerealization capacities too. He must desire that he does not now interfere with his later exercise of his capacity to realize his desires, and he must also desire that he now does
!
5
!
! what he can to help his later self have the capacity to realize desires to exercise. But of course, these desires are subject to a crucial proviso. The desires with which the ideal agent desires not to interfere must themselves be desires whose realization doesn't require that he interferes with the exercise of his capacities for belief-formation or desire-realization. Since this proviso is such a mouthful, I will take it as read in everything that follows. Let me sum up the story so far. I have argued that an ideal agent fully and robustly possesses and exercises two capacities both synchronically and diachronically—the capacity to access evidence and form beliefs on the basis of that evidence and the capacity to realize his desires—and I have further argued that if his exercises of these two capacities are to cohere with each other, then he must have dominant desires that he does not now interfere with his exercise of his capacity to believe for reasons or realize his desires either now or later, and that he now does what he can to help his later self have belief-formation and desire-realization capacities to exercise. These desires must all be dominant in the sense that they must dominate any potentially idiosyncratic desires that he might happen to have as well. But how strong are these dominant desires supposed to be vis-à-vis each other? Imagine, for example, an ideal agent who is suffering from a disease that will eventually cause him to lose all of his mental powers absent treatment, and further imagine that the disease is especially susceptible to mood, so the treatment regime includes a pill that will cause sufferers to believe that they will get better. The ideal agent's dominant desire that he now does what he can to help his later self have beliefformation and desire-realization capacities to exercise will tell in favour of his taking the pill, whereas his dominant desire that he does not now interfere with his current exercise of his capacity to believe for reasons will tell in favour of his not taking it. So how would such a conflict be resolved in an ideal agent? Which of his dominant desires, each of which must dominate his idiosyncratic desires, should dominate which? Though I take it to be obvious that in this case the desire to help his later self have belief-formation capacities to exercise should dominate, in order to give this answer on principled grounds we would have to consider a whole range of conflict cases so that we can see how it follows from a comprehensive account of the relative strengths of the different desires an ideal agent has to have vis-à-vis each other. Given that such an account may well be highly circumstance specific, the provision of such an account is clearly beyond the scope of the present paper. However I mention it just to make it clear that even an ideal agent may find himself in circumstances in which the exercise of his capacity to believe for reasons comes at the cost of his exercise of his capacity to realize his desires, and vice versa, and hence that there are principled limits to the robust exercise of the two capacities. Does this mean that, at the end of the day, the conception of the ideal agent argued for here is no better than the divided reports conception? No it does not mean that. According to the conception of an ideal agent argued for here, the conflicts the ideal agent experiences are themselves all conflicts between desires that an agent has to have simply in virtue of being ideal, and such conflicts are resolved in a principled way, that is, by reference to the relative strengths that these desires have to have vis-à-vis each other simply in virtue of their being desires of an ideal agent. What these residual conflicts show is not that the conception of an ideal agent argued for here is no better than the divided reports conception, but rather that, though the conception argued for here secures more coherence in the psychology of an ideal agent than the divided reports conception, there are principled limits on how much coherence there can be.
!
6
!
! Have we now succeeded in fully specifying the desires that an ideal agent has to have simply in virtue of being ideal? It seems to me that we have not. What's especially striking about all of the desires we have described so far is that the agent himself occurs in their content twice over. He occurs in both the agent-place ("...the agent desires that he does not now interfere with...", "...the agent desires that he now does what he can to help...") and in the patient-place ("...interfere with his exercise of his capacities now or later...", "...help his later self have the capacity..."). However he only occurs in the agent place in the present ("...the agent desires that he does not now interfere with...", "...the agent desires that he now does what he can to help..."), whereas he occurs in the patient place in both the present and the future ("...interfere with his exercise of his capacities now or later...", "...help his later self have the capacity..."). Is distinguishing in this way between the present and future, and between the agent and others, defensible? It is clearly not just defensible, but essential, that the agent in the present occurs in the agent-place in all of these desires, but in the patient place in both the present and the future. The role of these desires is, after all, to secure both the synchronic and diachronic coherence of the agent's exercise of his capacities for belief-formation and desire-realization, and this is something that these desires can do only if they are desires that effect what the agent himself does at the very moment of action, which is always in the present, but where the actions he performs can have effects both in the present and in the future. But is it also defensible that only the agent himself occurs in the patient-place? After all, just as it is a remarkable fact about agents that they exist in time, so it is also a remarkable fact about agents that they exist in a world in which there are other agents. So just as times other than the present have to occur in the patient place of the desires that ideal agents have to have, simply in virtue of their being ideal, shouldn't we suppose that agents other than the agent himself have to occur in the patient place as well? There are various ways in which we could approach this question. One would be to argue on purely theoretical grounds that we cannot make structural distinctions between times and agents when we state rational principles, or the contents of desires that ideal agents have to have simply in virtue of being ideal, as to do so is to make an arbitrary distinction. Rational concern must either be restricted to the agent himself in the present, or extended not just to other times but also to other agents. As I understand it, if we argue in this way then we would be giving a strengthened version of the argument by dilemma that Derek Parfit gives in Reasons and Persons in the chapter on the Appeal to Full Relativity (Parfit 1984). The strengthening of the present argument comes from the fact that, since we have already seen that the ideal agent has to have desires that concern himself at other times, so the only alternative left is to suppose that the ideal agent must have desires that concern other agents as well (see also Pettit and Smith 1997). Another way we could approach this question would be to ask whether it is so much as possible for there to be a community of agents all of who fully and robustly exercise their capacities for belief-formation and desire-realization. If the answer is that this is possible, then we could infer that other agents must occur in the patient place of the desires that ideal agents have to have, simply in virtue of being ideal, from the fact that, in such a community, the agents would all have to be able to rely on each other not to interfere with their exercise of their capacities for belief-formation and desirerealization, and would all have to be able to rely on each other for help in having belief-formation and desire-realization capacities to exercise. Absent such reliance, which would be possible only if ideal agents have desires that concern others as well as themselves at other times, the full and robust exercise of their capacities for belief-
!
7
!
! formation and desire-realization would not be possible, but would be at best a happy accident (for more on this see Smith 2011). And yet another way in which we could approach this question, the way in which I plan to approach it here, would be to ask a much flat-footed question in the spirit of those we have been asking thus far. We could look at process-of-thought cases that are as similar as possible to those we have been discussing, but we could imagine away the fact that the being on whom the agent will have the effect later is the agent himself. We could then ask whether this has any effect on our assessment of what's required for the agent initiating that process of thought to be ideal. As a first step in this direction, imagine a subject who is totally engrossed in a problem. To make matters vivid, imagine that his attention is initially focused on trying to figure out whether p is true, so much so that he is totally oblivious not just to his surroundings, but also to the other aspects of his mental life. If p turns out to be true, then the subject's attention later will be similarly focused on trying to figure out whether p supports q, and, if it does, he will attend to the consequences and go on to draw the conclusion that q. Imagine that everything works out, and that the subject draws the conclusion that q. There is a stream of consciousness throughout in which the subject at different moments attends to the different parts of the problem at hand. Now focus on the subject at the moment at which he is trying to figure out whether p is true. What must be true of him if he is fully and robustly to possess and exercise the capacity for belief-formation? Alternatively put, what is required for him to play his part in securing the maximal synchronic and diachronic coherence of the stream of consciousness? Since this process-of-thought case is much the same as the process-of-thought cases described earlier, the answer has to be much the same as before. But since nothing said in describing this process of thought entails that the subject who goes on draw the conclusion that q is the same subject as the subject who is figuring out whether p is true, it turns out that we cannot insist that the desire that the subject who is trying to figure out whether p is true has to have, simply in virtue of being ideal, is one in which he himself figures in the patient place. To the extent that we think that the answer is much the same as before, then, it seems to follow that we were never really committed to thinking that the desires an agent has to have, simply in virtue of being ideal, are desires in which he himself figures essentially in the patient place. Our idealized counterparts have to have desires concerning the exercise of belief-formation and desire-realization capacities of not just themselves, but of other agents as well. So, at any rate, this process-of-thought case suggests. Note that the argument for this conclusion is once again Rawlsian in spirit. We discover that the ideal agent's desires aren't all just about himself by reflecting on choices that he can make—he can and must choose to play his part in securing the diachronic coherence of the stream of consciousness that begins with his own present conscious thoughts if he is to be ideal—but choices that he couldn't make if the desires he has to have, simply in virtue of being ideal, are desires that only concern the exercise of his own belief-formation and desire-realization capacities. To see that this is indeed the case, let's go through some variations on this last process-of-thought case, variations that make it explicit that there is non-identity of the subject over time. These variations on the case will bring out not just that there is non-identity by the lights of the various standard views of what's required for the identity of a subject over time, but that none of the variations makes any difference at all to our assessment of what's required of the subject who is trying to figure whether p is the case if he is to be ideal.
!
8
!
! To begin, imagine that the Cartesian soul view is the correct account of the identity of a subject over time. Could the process of thought be just as described even though different Cartesian souls are attending to the different parts of the argument? It seems to me that it most certainly could be—and, indeed, it seems further that, for all we know for certain, this is the way things actually are. There is the Cartesian soul who tries to figure out whether p is true, another one who tries to figure out whether p supports q, and yet another one who goes on to draw the conclusion that q. But even if there are different Cartesian souls, so long as the stream of consciousness is exactly the same, with each part of the stream having the same effects and dependencies on the other parts, our judgement about what's required for its robust diachronic coherence remains the same. If the Cartesian soul who is trying to figure out whether p is true is fully and robustly to possess and exercise belief-formation capacities, capacities sufficient to secure not just his own diachronic coherence, but the diachronic coherence of the stream of consciousness as well, then it seems that he must desire not to interfere with not just his own exercise of his capacity for belief-formation, but with the later Cartesian souls' exercises of their capacities for belief-formation as well. His desire cannot be essentially about himself to the exclusion of others. Now let's suppose that the bodily criterion is correct, and let's fix on that part of the body that seems like the best candidate to be the part that secures the identity of a subject over time. Could the process of thought be just as described even though there is no single brain throughout? It seems that it could. The process of thought could be just as described even if, immediately after figuring out that p is the case, the subject suffers a catastrophic accident that so damages his body that, in order to give him the best chance of survival, the doctors on the scene immediately bisect his brain and transplant the two halves into two debrained bodies, and even if both transplants miraculously turn out to be successful. To be consistent with the process-of-thought case as described, we would have to suppose that the subject remains conscious throughout, but is so engrossed in the problem that he is trying to solve that he doesn't realize that the accident and subsequent operations even happen. There would therefore be two streams of consciousness from the moment of bisection onwards, each continuous with the earlier stream. In this variation on the case, no single brain underlies the process(es) of thought as described, so there is no identity of the subject over time. Even so, our judgement about what's required for the diachronic coherence of the process(es) of thought, and the stream(s) of consciousness, remains the same. If he is fully and robustly to possess and exercise belief-formation capacities, as he can, the subject with an intact brain who is trying to figure out whether p is the case must desire not to interfere with the exercises of the capacity to believe for reasons of the later subjects', each of whom has half of his brain. Indeed, the verdict in this case seems especially secure. For on what possible grounds could the agent who is trying to figure out whether p is the case rationally discriminate between his effects on the products of the later surgery even if he knew about the surgery in advance? Could the process of thought be just as described even though there is no psychological continuity and connectedness between the subjects at different stages? It seems that it could. The process of thought could be just as described even though, immediately after having figured out that p is the case, the bulk of the underlying psychology of the subject—all of the psychology except for the part that underwrites the process of thought itself—changes radically. Perhaps the underlying psychology is just like mine up until the subject figures out that p is the case, and just like my wife's thereafter. Or perhaps it is just like a child's thereafter, or just like that of someone suffering from dementia. Of course, there must still be connections between the episodes of thought that comprise the process of thought. The subject who begins attending to whether p supports q must do so because he seems to remember having
!
9
!
! established that p in the past, and that in turn must be so because the earlier subject did establish that p. But, at least relative to the psychology as a whole, these connections might be far too few to secure the identity over time of a single subject over time, given the criterion of psychological continuity and connectedness. But even if this were the case, our judgement about what's required of the subject at the beginning of the stream of consciousness remains unchanged. If the bundle of psychologically continuous and connected psychological states that includes the attempt to figure out whether p is the case is to include the full and robust possession and exercise of the capacity to form beliefs, then it seems that it must also include the desire not to interfere with the exercise of the capacities for belief-formation that are included in the bundle of psychologically continuous and connected psychological states that includes the attempt to figure out whether p supports q and the drawing of the conclusion that q, even if those bundles turn out not to be sufficiently continuous and connected to each other to make it the case that there is a single subject over time. This last case is especially instructive, as it suggests that the desires that an ideal agent has to have, simply in virtue of being ideal, are desires concerning the belief-formation capacities of beings with whom he not only need not be identical, but with whom he needn't have much in common either. Such beings needn't be psychologically similar to him, and nor need they have capacities for belief-formation that are anywhere near as sophisticated as his are either. That's the lesson we learn from the fact that, in spelling out the last variation on the case, the psychology surrounding the process of thought could be like (say) mine initially, but like a child's, or that of someone suffering from dementia, later on. Identity, psychological similarity, and the possession of sophisticated belief-formation capacities, these are all excluded as candidate properties that characterize those beings that figure in the patient-place of the desires that agents must have simply in virtue of being ideal. Other features of the last case, however, though helpful in making the crucial point of non-identity vivid, are plainly irrelevant when it comes to characterizing the beings that figure in the patient-place of these desires. For example, though in the last process-of-thought case we have imagined a stream of consciousness, it plainly isn't necessary that the agent who figures in the patient-place of an ideal agent's desires be conceived of as being conscious, so the desires in question cannot be restricted to conscious agents. After all, if an agent is to possess a robust capacity to believe for reasons diachronically, then he must desire that he does not now interfere with his later exercise of his capacity to form beliefs whether he is conscious or not. And again, though it helps make the crucial point of non-identity vivid, it isn't necessary that the beings with whom he might interfere are engaged in the formation of beliefs whose contents are inferentially connected with the contents of beliefs that he is currently entertaining with a view to making an inference. For, once again, if an ideal agent is to possess a robust capacity to believe for reasons diachronically, then he must desire not to interfere with his later exercise of his capacity to form beliefs whether or not their contents are inferentially connected with the beliefs he is currently entertaining. So what feature must beings possess if they are to figure in the patient place of the desires that an ideal agent has to have simply in virtue of being ideal? When we put our discussion of this last process-of-thought case together with the lessons learned from our discussion of the earlier cases, and when we generalize from those cases to cases in which the exercise of desire-realization capacities is at issue as well, it emerges that the only feature beings who figure in the patient place need to have is that they are such that their exercise of their belief-formation and desire-realization capacities depends on what the agent now does. Dependency, together with a minimal capacity to form beliefs and realize desires, are thus the only features a being needs to have in !
10
!
! order to figure in the patient-place of the desires that the ideal agent has to have simply in virtue of being ideal. The upshot is that we can characterize the psychology of the ideal agent as follows. The ideal agent must fully and robustly possess and exercise the capacity to access evidence and form beliefs on the basis of that evidence, and the capacity to realize his desires, both synchronically and diachronically. However in order for his exercise of these two capacities to cohere with each other to the greatest extent possible, he must also have certain coherence-making desires. In particular, he must have a dominant desire that he does not now interfere with the exercise of the capacity to believe for reasons or realize desires of anyone at any time whose exercise of their capacity to believe for reasons or realize their desires is dependent on what he now does, and he must also have a dominant desire to do now what he can to help anyone at any time whose possession of belief-formation and desire-realization capacities is dependent on what he now does to have belief-formation and desire-realization capacities to exercise. For short, let's call these the desires to help and not interfere. If the agent is ideal then the desires to help and not interfere must themselves be dominant in the sense that they must dominate any potentially idiosyncratic desires that the ideal agent might happen to have, as only so will the robust exercise of his two capacities be guaranteed. But since these desires may conflict with each other, the ideal agent will resolve such conflicts by having each of these desires with an appropriate strength. I haven't said how strong these desires must be vis-Ă -vis each other, so there remains some unfinished business. But unfinished business notwithstanding, it must surely be agreed that this represents considerable progress on the question asked at outset. Can we draw substantive conclusions about the reasons for action agents have from premises about the desires of their idealized counterparts? The answer turns out to be that we can. Since the reasons for action agents have are a function of the desires of their idealized counterparts, and since every agent's idealized counterpart has to have the same dominant desires to help and not interfere alongside whatever other idiosyncratic desires he has, it turns out that every agent has dominant reasons to help and not interfere, and it also turns out that, beyond that, every agent has a reason to satisfy whatever idiosyncratic desires he happens to have. The argument for this conclusion has been Rawlsian in spirit. It focuses on the choices that an agent has to be capable of making simply in virtue of being ideal, and it infers the contents of the desires that ideal agents have to have from the substantive choices that they make. But whereas Rawls asks us to imagine the choice of principles that will govern the basic structure of a society in which the chooser will live, the argument given here focuses on choices made in more mundane circumstances in which our basic agential capacities are exercised. The focus throughout has been on choices made in process-of-thought cases. The question we have asked is what an ideal agent would have to choose to do in order to bring his capacities for belief-formation and desirerealization into coherence with each other when he is engaged in certain processes of thought, and the answer has turned out to be that, quite generally, he would have to choose to help and not interfere. But though the argument for this conclusion has focused on choices made in much more mundane circumstances than Rawls asks us to imagine, the conclusion reached turns out to be remarkably similar to Rawls's own. The reasons we have to help and not interfere are, after all, more than somewhat reminiscent of Rawls's own second and first principles of justice respectively. Nor should this be surprising. If Rawls's argument and mine both simply limn the features of ideal choice, you would expect nothing less than a measure of convergence in their conclusions. !
11
!
! REFERENCES Nagel, Thomas 1970: The Possibility of Altruism (Princeton: Princeton University Press). Parfit, Derek 1984: Reasons and Persons (Oxford: Clarendon Press). Pettit, Philip and Michael Smith 1997: "Parfit's P" in Jonathan Dancy (ed.) Reading Parfit (Oxford: Blackwell) pp.71-95. Rawls, John 1971: A Theory of Justice (Cambridge: Harvard University Press). Smith, Michael 1994: The Moral Problem (Oxford: Wiley-Blackwell) _____ 2011: "Deontological Moral Obligations and Non-Welfarist Agent-Relative Values" in Ratio, XXIV, pp.351-363. Williams, Bernard 1981: "Internal and External Reasons" in his Moral Luck (Cambridge: Cambridge University Press).
!
12