It is natural to assume that whether one is rational to believe depends on one's reasons to believe. Such a view is not inevitable. Reliabilists or wide-scopers may hold the view that reasons have very little to do with the rationality of belief. But I am going to begin by assuming that reasons are what determine rationality. Similarly, it is natural to assume that whether one is rational to act is determined by one's reasons to act. This raises an interesting question: Are there interesting differences between reasons to believe and reasons to act – in particular is there an important difference in their structure? With an eye toward this question, I'll be considering Mark Schroeder's arguments that we can give a general account of the structure of reasons that applies both to reasons to believe and reasons to act.Footnote 1 I will argue that the prospects for Schroeder's project are dim. I will also argue that the assumption that rationality (and in particular rational requirement) is determined by reasons alone cannot be sustained.
1. PRACTICALLY RATIONAL BELIEVING?
Most epistemologists agree that for purposes of doing epistemology, the distinction between reasons to believe and reasons to act is inadequate. They think there can be both practical and “epistemic” reasons to believe, the former of which determines the practical rationality of believing, the latter of which determines the epistemic rationality of believing. On this view, these reasons can conflict. Suppose your evidence supports that you will very likely die from your illness, but your evidence also supports that believing you will survive will increase your chances of surviving. This latter kind of reason is sometimes called “the wrong kind of reason.” In such a case, it is said that one has epistemic reasons to believe one will die, but one also has practical reasons to believe one will survive. So one is epistemically rational to believe one will die, but practically rational to believe one will survive.
While this kind of case is suggestive, it certainly does not force us to accept that believing can be practically rational. As others have argued, we may instead distinguish only between the rationality of believing you will die and the rationality of doing what you can to bring it about that you believe you will survive (or perhaps the rationality of desiring or intending to bring it about that you believe you will survive). So rather than there being two senses in which it is rational for you to believe, there is simply a way for you rationally to believe, and a way for you rationally to act.
The original description of the case entails that whatever you believe you are irrational – either epistemically or practically. But it is not clear that you must be irrational in such a case. Suppose you respond to your reasons by believing you will die, while doing everything you can to bring it about that you believe you will survive. You start hanging out with people who are skeptical of medical science, you undergo hypnosis, you take pills that are supposed to induce optimism, etc. Unfortunately, you do not succeed. You continue to believe you will die, as your evidence indicates, and lament that you failed to succeed in bringing it about that you believe you will survive. Surely, the fact that the hypnosis and the optimism pills did not work does not make you irrational. So your belief that you will die is rational, and your actions directed at inducing the belief that you will survive are rational. Indeed there does not seem to be any basis for attributing any kind of irrationality to you. You have met all the demands of rationality. This means that the original description of the case according to which you are practically irrational if you fail to believe you will die is wrong.
1.1 The balance of reasons account of rationality
How much are reasons for belief and reasons for action alike? It is interesting that talk of reasons is much more common in discussions of rational action, than in discussions of rational belief.Footnote 2 By and large, epistemologists talk about justification, reliability, having evidence, etc. So it is perhaps best to begin by looking at how reasons function in accounts of rational action, and then see to what extent that account can be generalized to reasons to believe. Mark Schroeder's project is to show that we can give a general account of the structure of reasons according to which reasons for belief (or what he calls “epistemic reasons”) and reasons for action are special cases. As Schroeder recognizes, such an account faces significant difficulties. I will argue that his suggestion for overcoming these difficulties itself faces serious difficulties.
When we say that it is rational for S to do A, we could mean either that S is rationally permitted to do A or that S is rationally required to do A. These notions are duals in the logical sense. In the theory of reasons for action, they correspond to the distinction between having sufficient reason to do A and having conclusive (or decisive) reason to do A. Standardly “S has sufficient reason to do A” = df S is rationally permitted to A, and “S has conclusive reasons to do A” = df S is rationally required to do A.
In the theory of rational action, the difference between having sufficient reasons and so being rationally permitted and having conclusive reasons and so being rationally required corresponds to a difference in the relative strength of one's reasons:Footnote 3
Act Suff: S has sufficient reasons to do A just in case S has at least as strong reasons to do A as S has not to do A.
Act Con: S has conclusive reasons to do A just in case S has stronger reasons to do A than S has not to do A.
Schroeder calls these accounts “balance of reasons” accounts. The idea is that we explain having sufficient/conclusive reason (being rationally permitted/required) to do A, an all-things-considered notion, in terms of the pro tanto notions of having a reasons to do A and reasons not to do A. The difference between being rationally required to do A and being rationally permitted to A is a matter of whether one's reasons to do A are better than or merely as good as one's reasons not to do A.
2. RATIONAL PERMISSION TO BELIEVE
The balance of reasons approach construes sufficiency as solely a matter of the relative strength of one's reasons. This has two consequences worth noting. The first is that one can be rationally permitted to do A even though one's reasons to do A are very weak. So long as one's reasons to do A are at least as good as one's reasons not to do A, one is rationally permitted to do A. So I can be rationally permitted to raise my arm, even though I do not have much of a reason to do so, simply because I do not have much of a reason not to raise my arm. The second is that one can be rationally permitted to do A even when one has strong reason not to do A. So long as one has an equally strong reason to do A, one is rationally permitted to do A. So even if having the surgery has a high risk of dying, I may be permitted to have it provided that not having the surgery is at least as risky. Both of these features seem plausible for reasons for action. But, as Schroeder notes, they cause trouble for extending this analysis to reasons to believe:Footnote 4
Bel Suff: S has sufficient reason to believe p, i.e., S is rationally permitted to believe p, just in case S has at least as strong reasons to believe p as S has not to believe p.
A natural assumption is that one's reasons to believe consist in one's evidence. It is important to see that, as Schroeder notes, evidence is always evidence for some proposition, i.e., evidence for p or for ~p. There cannot be evidence for not believing p that is not evidence for ~p. Thus if reasons to believe are evidence, whether one has sufficient reason to believe will depend on the relative strength of one's reasons to believe p and one's reasons to believe ~p. This means we can equivalently express Bel Suff as:
Bel Suff': S has sufficient reason to believe p, i.e., S is rationally permitted to believe p, just in case S has at least as strong reasons to believe p as S has to believe ~p.
Now we see a clear difficulty in extending the balance of reasons account to sufficient reasons to believe. As we noted earlier, it can be rational for one to act, even if one has only a weak reason, provided one does not have a stronger reason not to act. Belief is different. I am not rationally permitted to believe that a fair coin toss will land heads. My having weak evidence in support of the coin's landing heads is not compensated for by my having a weak evidence in support of the coins not landing heads. We also noted that one can be rationally permitted to do A even if one has strong reasons not to do A, provided one has equally strong reason to do A. But again, belief is different. I am not rationally permitted to believe p if I have conflicting expert testimony concerning p. My strong testimonial evidence for ~p is not compensated for by my equally strong testimonial evidence for p. Being rationally permitted (having sufficient reasons) to believe p requires both that one have sufficiently strong reasons to believe p and that one's reasons to believe p are (sufficiently) stronger than one's reasons to believe ~p.
Schroeder finds the “sufficiently stronger” condition to be too imprecise. What counts as sufficiently strong and sufficiently stronger? I agree it would be nice to know the answer to this question. But this may prove difficult. ‘Rational’ is a comparative adjective and on the standard view, the truth-value of an utterance containing ‘rational’ is context sensitive.Footnote 5 So there is reason to think that context will play a significant role in determining what counts as sufficiently strong reasons to believe p, and what counts as sufficiently stronger reasons to believe p than to believe ~p. There may not be a uniquely correct answer to this question.
3. REASONS TO SUSPEND
Schroeder recognizes the problem of extending the balance of reasons account to reasons to believe and proposes a solution. He argues that there can be non-evidential reasons for suspending judgment. If there are, then whether one is rationally permitted to believe is not simply a matter of the balance of evidence for p and for ~p. Independent non-evidential reasons to suspend will also play a role. Schroeder argues that recognizing this kind of reason is the key to responding to the problems faced by the balance of reasons account. On the assumption that there are non-evidential reasons for suspending, we can return to our original formulation of sufficiency:
Bel Suff: S has sufficient reason to believe p, i.e., S is rationally permitted to believe p, just in case S has at least as strong reasons to believe as S has to not believe p.
And given that reasons not to believe p will consist of both evidential reasons to believe ~p and non-evidential reasons to suspend on p, we can formulate the account thus:
Bel Suff”: S has sufficient reason to believe p just in case S has at least as strong evidential reason to believe p as S has evidential reason to believe ~p, and S has at least as much evidential reason to believe p as S has non-evidential reason to suspend on p.
Before we examine whether reasons to suspend can help with the difficulties encountered by Bel Suff”, we first need to take up the issue of whether there are such things and what they look like. Schroeder gives several examples. The most promising, in my view, derive from the alleged phenomenon of pragmatic encroachment.
Here is a case from Fantl & McGrath (Reference Fantl and McGrath2009): Suppose I am on a hike and I encounter a large frozen lake. I could save a little time by crossing it rather than walking around it. I ask the park ranger whether the ice is thick enough to hold me. He says that it is. Now consider two sub-cases. In the first, the lake is shallow and falling through the ice will result only in my pants getting soaked. In the second, the lake is deep and falling through will result in my drowning. Proponents of pragmatic encroachment argue that in the first sub-case, I can rationally believe the ice will hold me, but in the second sub-case I cannot. Schroeder's explanation for why I cannot rationally believe the ice will hold me in the second case is that the risk of drowning gives me a reason not to believe, i.e., to suspend on whether the ice will hold me. And because the risk is not evidence for or against my falling through the ice, it constitutes a non-evidential reason to suspend on whether the ice will hold me. The issues here are complex, and this is not the place to discuss them in any detail. I will simply point out that this description of the case is not forced. One can rationally believe p without being certain that p. So we can say that although I am rational to believe the ice will hold me, because of the massively asymmetrical stakes, my expectation for crossing is negative. After all, one has little to gain by crossing and much to lose. Thus even though it is rational for me to believe the ice will not break, it can still be rational for me to refuse to cross.Footnote 6
Schroeder (Reference Schroeder2015a) argues that there are also examples of non-evidential reasons to believe p stemming from a rational expectation of forthcoming evidence relevant to whether p:
Suppose that both Sophia and Zoe have significantly better evidence for p than for ~p, but their situations differ in the following way: although Sophia is in a position to be confident that no further evidence which might bear on the matter is forthcoming, Zoe is waiting on the results of an experiment that has the potential to provide more conclusive evidence than any of her evidence collected so far. Zoe's expectation of further evidence is not evidence against either conclusion, but it does seem to raise the bar for how conclusive her existing evidence must be, in order to make it rational for her to believe. Here's a natural explanation of why: it's because the expectation of further evidence is an epistemic reason not to believe.
Perhaps there is a pull toward saying that in this case Sophie could be rational to believe p while Zoe is not. But I think this can be explained without appealing to non-evidential reasons to withhold. For these kinds of cases, the crucial question is this: Does knowing about the experiment give Zoe any reason to change her expectations for getting evidence for or against p? There is always forthcoming evidence that is relevant to whether most of the things we believe are true. Every morning, as I prepare to leave my house, I know there is forthcoming evidence concerning whether I live in Tucson. But knowing that does not give me a reason to suspend on whether I live in Tucson. So what is different about the case of Zoe?
By the description of the case, Zoe has significantly better evidence for p than ~p. Let's say her evidence supports p to 0.9 and ~p to 0.1. If all she knew were that she would be getting random new evidence that bears on p, her probability for its supporting p should be 0.9 and her probability for its supporting ~p should be 0.1. Let's call these her default probabilities. (Here I'm assuming she has no special reason to think the evidence will be misleading.) The relevant question is whether expecting this new evidence gives her any reason to change her default probabilities concerning what that evidence will support. If not, then the fact that she is expecting new evidence does not constitute a reason to suspend (or believe or disbelieve). Again, when I leave my house, I will be getting new evidence that bears on whether I am in Tucson. But that fact does not give me a reason to suspend on whether I am in Tucson.
In the case at hand, Zoe learns that an experiment will produce new evidence that bears on p. How should that information affect her default probabilities? The mere fact that an experiment is being conducted should not change them at all. After all, we could view leaving my house as an experiment concerning whether I am in Tucson. I think Schroeder's case makes us think Zoe has a reason to suspend because we make the natural assumption that Zoe knows the experiment is being done by knowledgeable researchers who think it is in some sense an open question whether p. We do not generally expend resources to investigate questions when we are confident what the answer is. On that assumption, it is clear that Zoe's default probabilities concerning what nature the forthcoming evidence will support should change. Her probability that the evidence will support ~p should go up (from 0.1) and her probability that the evidence will support p should go down (from 0.9). In that case, what she knows about the experiment is evidence for her that the experimental evidence will support ~p over p. And evidence that there is evidence in support of ~p is itself evidence for ~p.Footnote 7 Thus we can given an evidential explanation for why Zoe has a reason to suspend that Sophia does not have. Zoe, unlike Sophia, has further evidence for ~p. There is no need to appeal to a non-evidential reason to suspend.
The last example of a reason to suspend, according to Schroeder (Reference Schroeder2012), derives from the costs of forming beliefs.
To withhold (suspend) is to not make up your mind, to have formed no belief. Consequently, any disadvantage of forming beliefs – of making up your mind – is potentially a reason to withhold. For example, forming beliefs takes cognitive resources – we are finite creatures, and logical space is infinite. This could be what explains why it is rationally permissible not to believe all of the disjunctions, one of whose disjuncts is something that you are rationally required to believe. For even though the disjunction is as well-supported by evidence as the disjunct, there are further reasons to withhold with respect to it – namely, that you get all of the benefit by believing the disjunct, and believing the disjunction merely takes up further cognitive resources. This sounds like a plausible story, but telling it requires appealing to reasons to withhold.
This last case raises issues about conclusive reasons for believing, i.e., being rationally required to believe. Before we examine it, I want to consider the plausibility of the balance of reasons account for sufficiency even if we were to grant that Schroeder's examples are indeed cases of non-evidential reasons to suspend. Recall the untoward results that arise for sufficiency on the assumption that all reasons to believe are evidential: one is rational to believe p even when one has only very weak reasons to believe p, provided that one does not have stronger reasons to believe ~p. In addition, one is rationally permitted to believe p when one has strong reasons to believe ~p provided one has equally strong reasons to believe p.
How would non-evidential reasons to suspend on p help? The first question to ask is whether whenever one has equally strong reason to believe p and to believe ~p, one has a reason to suspend on p? (It is important to remember that reasons to suspend are independent of evidential considerations.) There would seem to be cases where one has no reasons to suspend of the sort Schroeder mentions, i.e., there could be no practical costs for believing p, no further evidence forthcoming, and no “pragmatic encroachment” effect. So there is nothing to prevent the balance of reasons account from yielding the result that one is rationally permitted to believe p, when one has equally strong reasons to believe ~p.
But suppose there were always reasons to suspend. Presumably, reasons to suspend come in degrees of strength. Suppose one has equally strong reasons to believe p and to believe ~p, and one has a reason to suspend that is no stronger. This does not change the fact that one is not rationally permitted to believe p. If anything, the reason to withhold counts against rationally believing p. But the balance of reasons account will still allow that one is rational to believe p. For one's reasons to believe p will be at least as strong as one's reason to believe ~p and at least as strong as one's reason to suspend on p. It is hard to give a concrete example here because I do not know how to compare the relative strengths of evidential and non-evidential reasons. But suppose one has arbitrarily strong reason (short of certainty) to believe p and equally strong reason to believe ~p. This could come about from opposing testimony from arbitrarily reliable experts. One would not be rationally permitted to believe p. And given that one's reason to believe p is arbitrarily strong, supposing there is a reason to suspend cannot block the result that one will be rationally permitted to believe p on the balance of reasons account. By stipulation, one's reason to believe p will be at least as strong as one's reason to suspend on p.
The balance of reasons account could avoid this result if (implausibly) reasons to suspend always trump reasons to believe. But this would present a dilemma. Either there are always reasons to suspend or there are not. If the former, then on the assumption that reasons to suspend always trump reasons to believe, it would never be rational to believe p. If the latter, then there will be cases where one has equally strong reasons to believe p and to believe ~p and yet one is rationally permitted to believe p.
4. RATIONAL REQUIREMENT TO BELIEVE
So far we have seen that it is far from clear that reasons to suspend exist, and even if they do, it looks very doubtful that they can rescue the balance of reasons account for sufficient reason to believe. It is easy to see that slight modifications in the cases we have been discussing would raise the same doubts about a balance of reasons account for conclusive reasons to believe. If we change them only by making one's reasons to believe in the above cases minimally stronger, we get the result that one is rationally required to believe p. But if one has weak reasons to believe ~p, having only slightly stronger reasons to believe p does not even permit one to believe p, let alone require one to believe p. Similarly, if one has strong reasons to believe ~p, having slightly stronger reasons to believe p does not even permit one to believe p, let alone require one to believe p. And as we saw in the case of sufficient reasons, the appeal to reasons to suspend does not help.
So what makes one rationally required to believe? Consider Schroeder's remaining example of an alleged reason to suspend. It illustrates why giving a balance of reasons account of the difference between rationally permitted belief and rationally required belief cannot work. For Schroeder, the explanation for why one can be rationally required to believe p without being rationally required to believe p v q is that once one believes p, there is no further inferential benefit to believing p v q. Because believing takes up cognitive resources, one has a reason to suspend on p v q. And having this reason to suspend explains why believing p v q although permitted, is not rationally required. (Here I ignore the pressing issue of how to weigh evidential reasons to believe against non-evidential reasons to suspend.)
But suppose that while considering one's evidence for p, one considers whether p v q? Now it is no longer an option whether to believe p. Once one attends to whether p v q, one is required to believe p v q (assuming one is still required to believe p). Moreover, the alleged non-evidential reason to suspend remains. One is still using limited cognitive resources for no additional inferential benefit. So the non-evidential reason to withhold cannot explain why one is not required to believe p v q.
What does explain why one is not required to believe what one is rationally permitted to believe? The balance of reasons approach views requirement as unique permissibility. When one is permitted, but not required to believe p, one is also permitted to suspend on p and perhaps even to believe ~p. This follows from the fact that one's reasons for adopting these attitudes are equal in strength. But on the standard view, one's evidence can support only one doxastic attitude toward p – either belief, disbelief, or suspension. It is very hard to see how one's evidence could support believing p while also supporting suspending on p, let alone how one's evidence could support believing p while also supporting believing ~p.
But if only one attitude is permitted, isn't that attitude is required? The analogous point is true for action. If only one act is permitted, then that act is required. The reason for this is that you have to do something. You can only not do A by doing something else. Belief is different. Schroeder notes that believing p and believing ~p are not the only two doxastic attitudes one may take toward p. One may suspend on p. And suspending is a way of not believing p without believing something else. Still suspending is a doxastic attitude one may take toward p. If only one of these three attitudes toward p is permitted, doesn't it follow that the attitude is required? No, because there is a fourth possibility. One may take no doxastic attitude whatsoever toward a proposition – one may neither believe p, nor believe ~p, nor suspend on p.Footnote 8 In fact, that is surely our position with respect to most propositions. We are not required to have any attitude toward them simply because we have never considered them. Here it is important to see the difference between suspending on a proposition and not taking any attitude. One suspends on p only if one has considered whether p. For example, if you do not know me well, before I mentioned it, you had no attitude toward whether I have a sister. If you are rational, you have now suspended on whether I have a sister. So you went from not having any attitude toward the proposition that I have a sister, to suspending judgment on whether I have a sister.
What was rational for you to believe concerning whether I have a sister prior to my raising the issue? Your evidence uniquely supports suspending judgment. So you were uniquely permitted to suspend. But given that you had never considered the question, you were not irrational for not having taken that attitude. That is to say, you were not required to suspend. If we were required to have attitudes toward propositions we had never considered, we would all be massively irrational. Thus, before I raised the question, there was a unique rationally permitted attitude for you to have toward the proposition that I have a sister. But you were not required to have that attitude.Footnote 9
The same holds for when the uniquely permissible attitude is believing p. No matter how strongly my evidence supports believing p, I am not irrational for not believing p, if I have not considered whether p. Suppose p is an obvious consequence of my evidence? I am clearly rationally permitted to believe p. Am I irrational for failing to believe p? Surely I am not. There are indefinitely many propositions that are obvious logical consequences of my evidence. Again, if we were required to believe all the obvious logical consequence of our evidence, we would all be massively irrational.
So is one ever required to believe p (or to believe ~p, or to suspend)? I think so, and the reason is suggested by our discussion so far. One is required to believe p if one is rationally permitted to believe p on the basis of one's evidence, and one is considering whether p. As we saw, this is what transforms believing p v q from being rationally permitted to being rationally required. Despite the fact that inferring p v q from p is a pointless use of limited cognitive resources, if one rationally believes p and considers whether p v q (while remaining rational in believing p), one is rationally required to believe p v q.Footnote 10 If this is correct, whether one is rationally required to believe p does not supervene on one's reasons for p.
Consider another example:
Suppose I read a weather report that says there is a massive cold front moving through the upper Midwest. I am rationally permitted to believe that it is cold in Chicago. Am I required to so believe? Suppose I were in a rush and so distracted that I never considered whether it is cold in Chicago. Surely I am not irrational for failing to so believe. But suppose I am reading the report because my wife is going to Chicago and I'm wondering whether she will be cold. Then it seems that I am irrational for failing to believe it is cold in Chicago. Now it might be that when I first read the report, I get distracted and so never consider whether it will be cold in Chicago. Instead I am thinking of the awful toothache I have and whether I should go to the dentist. Or I may see an old friend walking by. Or perhaps I am just thinking about where I should make a dinner reservation. In that case, I never consider whether it is cold in Chicago. So I'm not irrational for failing to believe it.
We can even see how attention works in perceptual cases. Suppose I am sitting in my favorite coffee shop working on a paper. I'm staring off into space and it appears to me that Juan is in the coffee shop. On standard views, I have sufficient reason to believe Juan is in the shop. But suppose that although Juan's image is in my visual field, I'm not paying attention to what's in front of me. I'm deep in thought about the nature of reasons to believe. I am certainly rationally permitted to believe (on the basis of my evidence) that Juan is in the coffee shop. But surely I am not irrational if, owing to my being distracted by philosophical reflection, I don't believe Juan is in the shop. But if I actually pay attention to what is in front of me, then I am rationally required to believe he is in the coffee shop.Footnote 11
These cases further support the idea that one is required to adopt a doxastic attitude toward p just in case one is permitted to believe p and one is attending to whether p. Admittedly it is hard to be precise about how attention figures in making beliefs rationally required. But there seem to be clear cases where it does. Moreover this notion of attention is vague. This introduces another aspect of vagueness beyond the vagueness in the notion of evidential support. There will be borderline cases of rationally required believing that have nothing to do with vagueness in the notion of evidential support.
The important point here is the permitted/required distinction looks very different for belief than it does for action. In the case of action, the distinction hinges on the relative strength of one's reasons to do A and not to do A – between having sufficient reasons and having conclusive reasons. In the case of belief, the distinction between requirement and permission is not a distinction between sufficient and conclusive reasons. One is rationally permitted to believe p just in case one's reasons permit one to believe p. Whether one is required to believe is not a further fact about one's reasons, i.e., there is no such thing as having conclusive reasons to believe (where having conclusive reasons entails being required to believe). Being rationally required to believe is not determined by the reasons you have to believe. Rather it involves facts about one's psychology.
Unfortunately, this makes matters somewhat messy. Suppose I have sufficient reason to believe p and I attend to my evidence for p. Reasoning from that evidence to p takes some time, however small. If attending to my reasons makes me required to believe p, am I irrational for not believing p in the period between when I begin attending to my reasons and when I infer p? Surely that would be unduly harsh. Since none of us can reason instantaneously, our mental lives would be replete with minute intervals of irrationality. The absurdity of this indicates that when one has sufficient reason to believe p, rationality will tolerate an interval between when one first attends to one's reasons and when one infers p. How long of an interval will it tolerate? This is a difficult question. It will no doubt depend on how complex the reasoning is on the basis of which one infers p. While this issue is far from clear, one can describe clear cases. Suppose the connection between one's reasons and believing p is obvious, requiring, say, only a modus tollens inference. If I have been thinking about it for an hour, and I still have not revised, it seems pretty clear I am irrational.
What if I have not exceeded the amount of time I am allocated to make the inference, but I get distracted and stop considering whether p? By hypotheses, if I have not exceeded my allotted time, I am not irrational. Does the clock keep ticking? Surely that would be an odd result. I am considering whether p and I notice a large truck bearing down on me. I divert my attention away from reasoning about whether p to getting out of the way of the truck. Again, it would be harsh to say that although I saved my life, I became irrational for not believing p. What if I start deliberating again later? Does the clock reset? Or do I only have the remaining time from my previous deliberation?
These kinds of considerations are not pretty, but I fear we are stuck with them. One way out would be to deny that one is ever required to believe anything, or to take any doxastic attitude. Rather one's evidence determines the uniquely permissible attitude for you take towards p, and you are required to take this attitude only if you are taking some attitude or other toward p.Footnote 12 But even this view will not escape the problems. Suppose I already rationally believe p, and then I get a defeater of my reason to believe. Am I immediately irrational if I do not give up p? Again, reasoning from the defeater to giving up p will take some time. So the same difficult issues will arise.
We have seen that the permitted/required distinction differs structurally between believing and acting. In the case of acting the distinction involves a difference in the relative strength of ones reasons to do A and one's reasons not to do A. One is rationally permitted to do A just in case one's reasons to do A are at least as strong as one's reason not to do A, and one is rationally required to do A just in case one's reasons for doing A are stronger than one's reasons for not doing A. In the case of believing p however, the distinction need not involve a difference in the relative strength of one's reasons to believe p and one's reasons to believe ~p. Rather the difference hinges on whether I am attending to whether p.
Does attention play any role in determining whether an action is rationally required? Even if I have stronger reason to do A than not to do A, if I don't consider whether to do A, am I rationally required to do A? This seems to depend on how much is at stake. Suppose I am getting ready to cross a busy street. I have most reason to cross as quickly as I can. But halfway across I am distracted by a beautiful sunset and stop attending to whether I should keep moving. If because of this I stop walking and stare at the sunset, it seems clear that I am thereby irrational. So attending to whether I should cross is not required for me to be rationally required to cross. Presumably, the reason for this is that given the high stakes, attending to whether I should keep moving is itself something I am rationally required to do. Attending is, in the relevant sense, something that one does.
But suppose I am trying to decide what to eat for breakfast. I have both cornflakes and cheerios but I slightly prefer eating cornflakes to eating Cheerios, and so have slightly stronger reasons to eat cornflakes rather than Cheerios. While I have both Cheerios and cornflakes, I see some Cheerios first and failing to attend to my having cornflakes, proceed to pour myself a bowl of Cheerios for breakfast. Am I irrational? It seems not. Attending to all my options takes time and energy. Given that there is not much at stake in what I eat for breakfast, I am under no practical requirement to devote my attention to all my breakfast options. If this is correct, then the mere fact that I have stronger reasons to eat cornflakes does not mean that I am rationally required to eat cornflakes. But were I attending to there being cornflakes in the cupboard, I would be irrational for choosing the Cheerios. This suggests that in low stakes cases, having slightly stronger reasons to do A than not to do A rationally requires one to do A only if one is attending to A.Footnote 13
So what can say about the how rational permission differs from rational requirement in the cases of believing and acting? As I noted earlier, the difference in the case of acting requires a difference in the relative strength of one's reasons. While one is rationally permitted to do A if one has at least as strong reasons to do A as not to do A, one is rationally required to do A only if one has stronger reasons to do A than not to do A. In addition, in some low stakes cases, one is required to do A only if one is also attending to whether to do A.
In the case of believing, the difference between being permitted to believe p and being required to believe p does not depend on a difference in the relative strength of one's reasons to believe p and to believe ~p. One is required to believe p only if one is attending to whether p.Footnote 14
