Bertrand Russell once observed that puzzles in logic play a role like that of experiments in natural science, meaning that they provide us with ways to test our logical theories. One can make a similar claim for certain puzzles or paradoxes of rationality: they may be useful in determining whether a theory of rationality falls into contradiction and falsehood. This paper lays out and then attempts to solve two puzzles which beset goal-satisfaction theories of rational choice.

The first puzzle has to do with the nature of goals. Many theorists are inclined to assert that reason by itself provides no motive to the will and plays, so far as rational action is concerned, a merely instrumental role in serving the goals given by our desires and aversions. It seems that a rational agent is one who behaves so as to accomplish her higher goals, and that these are just goals which the agent prefers to satisfy above other goals.FN1 Suppose then that we wish to discover whether someone is behaving rationally on a particular occasion. We must try to determine what her goals are in that instance. What other way is there to proceed than to look and see both what she says and what she actually does? If she claims to value one thing while pursuing another, surely her true intentions and goals at the critical moment are revealed by her real choices, rather than by her words. Yet this seems to carry with it the strange consequence that an irrational action is impossible, since one's highest goals are just those that are actually pursued, and the pursuit of one's highest goals is characteristic of rational behavior. Consider the story of Ulysses, who had himself bound in order to hear the Sirens in safety. What makes his subsequent willingness to stay and listen irrational? If we were to take seriously his preference-ordering at the mast, perhaps we would have to agree that giving up everything to hear a little more of the Sirens' song was indeed a rational course of action.

The difficulty is reminiscent of, but is not the same as, the classical Greek problem of akrasia, which is that one who truly knows what his goals are would never do otherwise than pursue them. The worry about akrasia is that a weak-willed person doesn't know enough to avoid an excess of food, drink or sex, while at the same time "knowing better" than to indulge himself. By contrast, our own puzzle lies in determining what a person's ends are in the first place. It will presumably be granted that the goals of some people change more often than do those of other people. The drug addict's may change quite dramatically from one hour to the next. But the fact that goals may vary over time is not by itself inconsistent with a person's being rational at any given moment. To sum up, then, whatever an individual pursues is his highest goal, yet he sometimes fails to pursue it. An agent is always rational, yet sometimes irrational: this is the first puzzle.

The second puzzle is concerned with the rational suppression of reason in the pursuit of our goals. It is hinted at in a well-known line by Thomas Gray: "Where ignorance is bliss, 'tis folly to be wise". More precisely, we are tempted to think that rationality requires one to use one's intellectual resources to the fullest in accomplishing one's goals. In general, it seems irrational to practise self-deception, to reason badly, or to make one's epistemic position weaker in any way. A perfectly rational being would not trick or bind itself in such a fashion. It seems irrational to limit the use of one's reason when one is determining which of two alternative actions ought to be chosen. Yet at times it also seems rational to do just this. For example, to deceive oneself about the chance of success may help one to succeed in some ventures. Another, rather trivial, instance of self-deception is provided by a person who deliberately sets his clock fast in order to avoid being late for appointments. We want to say both that it is rational and that it is irrational of him to do so. According to the second puzzle, then, on some occasions it is both rational and irrational to suppress the use of reason or the results of its use.

With regard to this second puzzle, we should also note that to behave instinctively and without thinking may be the best way to cope with certain emergencies--or even with activities as common as hosting a dinner, driving a car, or conducting a love affair. Discursive thought tends to require time and concentration, and a judicious examination of conflicting claims. It is therefore not always compatible with appropriately prompt or spontaneous action. This circumstance must be of interest to students of philosophy, for whom thinking too much is an occupational hazard. It also interests and lends comfort to certain people who are unfriendly to logic, and who see reason as an inferior faculty. They may feel encouraged to conclude that one can be too rational for one's own good, that "the heart has its reasons, of which reason knows nothing", and so on.

To understand better this puzzle concerning the rational limitation of reason, let us begin by distinguishing two uses of the word 'rational'. In one use, a person is said to behave rationally when she reasons in a logically or epistemically justified way, such as when she draws a valid inference or makes a justifiable assertion; while in another use of the word one is said to behave rationally whenever one deliberately acts so as to satisfy one's goals to the best of one's abilities. Behavior which is rational in the first way need not be rational in the second. For example, so far as the first, logical or theoretical form of rationality is concerned, it might be rational to acquire a well founded belief which makes everyone who knows about it unhappy; while on the other hand, it could be pragmatically rational to deliberately acquire a perfectly groundless belief which somehow improved one's life. In both cases we can speak of a rationally acquired belief, but something different is intended in the two instances. The relationship between pragmatic and theoretical rationality is very interesting, and we might hope to discover that on a more profound level one is a special case of the other, or that these prima facie different forms of rational justification can both be subsumed under some more general account of rationality.

For now, let us simply take note of a certain way in which rational choice is related to reasoning. It is interesting to observe that logical inference seeks to maintain consistency and validity, while irrational choice reveals inconsistency of a sort. When an agent deliberately chooses to avoid realizing his own highest goal, the situation is as if he were to issue the injunction 'Let this goal X be', at the same time saying 'I will do incompatible Y'. It seems that in so speaking or behaving he has committed himself to endorsing in some way the contradictory claims 'X will be' and 'X will not be'. That is why we call his irrational acts 'inconsistent', in an extended use of the word 'inconsistent'. Reason seems to demand that there should be a harmonious account of one's intentions, while irrational action always reveals some disharmony. So, rational thought not only has the ability to facilitate rational choices; it can even be said to prescribe them in a way, since it distinguishes a set of causal principles which tell us what must be done if our goals are to be realized. Reasoning allows us to grasp the truth of various propositions having the form 'If X is your goal, then the choice W is most likely to bring X about'. To say this to someone who has the intention of bringing about X is, under the right circumstances, to recommend or prescribe the choice of W. Because reason tells us how to bring about our goals, we sometimes say that it recommends or prescribes certain actions, and that it forbids inconsistent, or counterproductive actions.

Looking back at our second puzzle we can see how distinguishing pragmatic from theoretical or logical rationality helps us avoid one equivocation which might cause confusion. Clearly, when creatures with a limited ability to understand are choosing the most efficient ways to pursue their projects, they must take into account the time and effort that discursive thinking requires. Understanding exactly what one is doing while one is doing it has advantages, but these may have to be sacrificed, either because the knowledge itself is a distraction or the means to achieving it are inconvenient. It will not be rational to sit proving theorems, for example, when time is a-wasting. Thus, one may exercise pragmatic rationality by passing up a chance to exercise theoretic rationality. Following the prescriptions of reason may require an agent to suppress another particular use of reason, but there is no contradiction in pointing this out. One reasons whenever one inquires, but evidently there is a time to stop inquiring into a question: for instance, when there are more important inquiries to attend to. There will be a time to cease inquiring in order to act. Even so, if his actions are to count as pragmatically or theoretically rational, an agent who behaves instinctively or spontaneously must have some understanding of why he is acting the way he is; otherwise his successes are merely a matter of luck. He must be justified in thinking that further calculation would be unproductive. One might suggest that the person who reasons at inopportune times is too rational. But if she understood everything about her situation she would presumably adjust her behaviour for the better. Therefore, rather than saying that she is too rational it would be better to say that she is not yet pursuing her goals in a rational way.

More interesting problems are presented by instances of self-deception. On occasion it may be most efficient to dispose one's own beliefs in a manner contrary to the best evidence in order to gain a certain objective. In this case too it will be rational to avoid pursuing or acting upon a particular use of reason. We can imagine a sports team which, prior to a championship game, tries to convince itself that it is the underdog, in order to spur itself to greater efforts. Or, think of a situation in which one wants to persuade other people that one will succeed at something, and this is best accomplished by persuading oneself, contrary to all the evidence. Notice that if such a deception is not deliberate then it is irrational, even if it serves to bring about a desired state of affairs. How is this to be explained? It must be because the deliberate manipulation of one's own states of mind demonstrates sophisticated self-knowledge, whereas mere delusion demonstrates only ignorance. We agree that there is something accidental about the successes of ordinary self-delusion and mere wishful thinking. That is because success by itself does not reveal rational choice, but only success which results from choosing to the best of one's abilities. The deliberate self-deceiver acts so as to accomplish some goal, and since his deception is motivated by a grasp of probable causes his actions with respect to this goal are said to be in accordance with reason after all. Still, like spontaneous acts, self-deception brings with it some disadvantages, and to that extent it is incompatible with realizing other goals.

Applying our reasoning to an individual case, we can see that the procrastinator who deliberately sets his clock ahead realizes one goal but not another. Afflicted by the bad habit of putting things off, he nevertheless acts in such a way as to realize the aim of keeping his appointments, and this is rational. It is better to procrastinate and keep one's appointments than to procrastinate and miss them. However, it would be better still not to procrastinate at all; for this would allow him to organize his time more effectively, so that he did not have to rely upon makeshift methods like self-deception. To the extent that he fails to eliminate his vice and reach this higher goal of effectively using his time, he acts irrationally. That is why we are tempted to say both that he is rational and that he is irrational. Altogether, then, we have several individuals to consider: the person who misses his appointments, the one who tricks himself into keeping them, and the one who keeps them without tricking himself. Paradox disappears when we specify which goals are under discussion.

Summing up: in connection with the rational suppression of reason we have discussed two main cases. In one case, a person intentionally foregoes some exercise of thought in order to act in a timely or spontaneous way. In the other case, a person intentionally practises self-deception in order to reach some end in view. Both cases entail certain disadvantages to the agent and require a local sacrifice of theoretical rationality. Hence, there is an appearance of paradox; but rightly considered there is no real contradiction, and no objection to a goal-satisfaction account of rational choice. The agent prevents himself from reasoning on one occasion, but has chosen so as to reach a particular objective.

These considerations might seem to dissolve any puzzle in rationally limiting the use of reason or the results of its use. What we have is just a necessary and temporary suspension of thought as dictated by other thoughts. However, some cases are more difficult to explain, as we see from the following illustration. Suppose an eccentric researcher offers you a large sum of money on the condition that you agree to take his new pill. This pill contains a drug which will strongly dispose you to behave for a short time in a very irrational manner. In fact, we can imagine that if you do not exhibit this disposition then the deal is off and the payment is withheld. Let us further grant that your circumstances make it advisable for you to accept this unusual offer. What shall we say about your actions while you are under the pill's influence? It seems likely that some of them will be irrational. Once again we have a situation in which we are tempted to say both that a certain agent behaves irrationally and that the same agent is somehow rational to do so. Setting out from a more elaborate example in his book Reasons and Persons, Derek Parfit draws the conclusion that if there is some disposition which it would be rational for you to cause yourself to have, and irrational for you to cause yourself to lose, it may still be irrational for you to act upon this same disposition.

Parfit's conclusion seems right. One can behave irrationally as the result of a rationally adopted disposition. However, this is not to admit that it is appropriate to be both rational and irrational in the same way. A person need not be mistaken in disposing himself to make mistakes at a future time. To deliberately thwart one's own overriding goal would indeed be irrational, but in the present case the disposition in question is directly subordinated to achieving an overriding goal. The resulting particular irrational acts, if they serve one's purposes at all, do so accidentally. They are not themselves rational, since in choosing to perform the particular acts one is not choosing to the best of one's ability. On the other hand, when one deliberately performs a locally undesirable act to further some higher goal, no greater paradox exists than when one rests in order to continue working later on. Thus, there need be no contradiction, even where causing irrationality is necessary to reach some higher goal.

Deceiving oneself for the purposes of realizing some goal can be a little like taking a step backward in order to jump forward. By contrast, the experimental pill case is more like being obliged, as a condition of moving forward, to dispose oneself to stagger around a bit. Let us now consider one further situation. There are certain effects such that in order for a person to bring them about, it is necessary for that person to aim instead at other, and even other conflicting effects. This is like finding it necessary, in order to go forward, to so dispose oneself that one no longer tries to go forward, but instead tries to go in some other direction. Sometimes a person who wishes to further her self-interest seems to fall into such a predicament: to systematically pursue one's own well-being may be to cut oneself off from attaining it. Someone who has an effect of this sort as her goal seems obliged to act in an irrational way, deliberately aiming at ends incompatible with her overall objective.

There may be circumstances in which a person must completely abandon any attempt to produce a particular effect if his actions are to help bring it about. And if there are effects of the kind described, then one really would be irrational to try and achieve them. It would indeed be futile to try and bring about a goal if any attempt at all to bring it about, however indirect, would at best achieve nothing and at worst only move the desired object further out of reach. In fact, this is not usually the situation where self-interest is concerned, as can be seen from the fact that nearly all of us profess some concern for our own welfare and think that various prudent steps can help to further it.

We must be wary here of the verbs 'to try' and 'to intend'. Often when we say that someone is trying to do a thing, or has the intention of doing it we mean to indicate that he or she has that thing explicitly in mind. This need not be the case, though, for as we know, someone may put a thing out of mind and still be acting for the sake of bringing it about: he may be trying to bring something about even while he is not thinking of that thing. For instance, when listening to certain pieces of music, one must direct one's attention entirely toward the work itself in order to experience the greatest pleasure from it. To be contemplating anything else, such as one's pleasure, is a distraction. Nevertheless one can be said to desire the pleasure and even to be listening for the sake of it. In fact, we can also say of someone who is asleep that she desires things, and has intentions, and that she is sleeping for the sake of something: her health, perhaps.

Let us say that with respect to an individual a goal is any state of affairs which the individual intends to try and bring about. How could a person rationally aim at a goal which seems to her incompatible with the overriding effect she wishes to obtain? One way is this: suppose that in order to bring about higher goals of type H a person trains herself to aim at lower goals of type L, putting thoughts of H's out of her mind. In general, pursuing L's might be the best way to bring about H's, but perhaps on occasion the two effects exclude each other. On such occasions, it may yet be irrational to stop pursuing L's. The policy

"Change your disposition when such an occasion seems to arise"

may not be appropriate, because one cannot readily identify the relevant occasions with certainty or cannot conveniently alter one's habits. Perhaps the readiness to change one's conscious aim would hinder the overall pursuit of H's. If this is true, the agent may intend to bring about an H, but continue to keep that intention out of her mind while she intends and actively works toward bringing about an L which she thinks is counterproductive.

From the above reasoning we gather that a person may in a way be aiming at a particular goal while putting any thought of it out of his mind, or while he is not conscious of trying to bring that goal about. These considerations return us finally to our first puzzle, and suggest an answer to it. We grant that rationally speaking, some actions ought to be treated as lapses, and products of weakness. Some goals are repudiated with regret and shame in an hour more reasonable than the hour of temptation. Such goals are rejected because their pursuit hinders the pursuit of higher goals. But does one still intend to pursue a goal at the moment when it is neglected, or does it become an object of intent only at other times? The answer seems to be, that goals may still be intended, even if the agent is unconscious of them or not paying attention to them. To discover a person's preferences and intentions it may not be enough to ask him, but it may not be enough to see how he is behaving at this moment, either. When determining what a person really ought rationally to choose we might try treating him as a time worm and conduct an imaginary plebiscite of the temporal slices of this worm. Or we could try to decide which of his possible choices would produce a maximum of pleasure over his whole life. But alternatively, we could try to learn his true intentions at the present moment, including the ones which he is ignoring or is unconscious of; and when he has considered these advise him to act on the one which he has discovered to be his overall intention. This approach would allow for the possibility of changing goals while saving a place for external criticism.

1. I don't explicitly discuss Expected Utility Models of rational choice here because they seem to me best understood as attempts to clarify and extend a Goal-Satisfaction account, rather than as competing theories. back