This is a pre-publication version of a paper forthcoming in Dialogue. Please cite the official version – Online First version available on Cambridge Journals Online, DOI: 10.1017/S0012217315000487. Epistemic Deontologism and Strong Doxastic Voluntarism: A Defense PATRICK BONDY Cornell University ABSTRACT: The following claims are independently plausible but jointly inconsistent: (1) epistemic deontologism is correct (i.e., there are some beliefs we ought to have, and some beliefs we ought not to have); (2) we have no voluntary control over our beliefs; (3) S’s lack of control over whether she φs implies that S has no obligation to φ or to not φ (i.e., ought-implies-can). The point of this paper is to argue that there are active and passive aspects of belief, which can come apart, and to argue that deontological epistemic evaluations apply to the active aspect of belief. RÉSUMÉ: Les idées suivantes sont raisonnables prises indépendamment, mais inconsistantes quand on les considère ensemble : (1) le déontologisme en épistémologie est correct; (2) on n’exerce pas de contrôle volontaire sur ce que l’on croit; (3) si S n’a pas de contrôle sur le fait qu’elle φ, elle n’a aucune obligation de faire φ ou de ne pas faire φ (c’est-à-dire que le devoir implique le pouvoir). Cet article vise à soutenir que nos croyances ont des aspects actifs et passifs, lesquels peuvent être séparés, et que les évaluations épistémiques déontologiques s’appliquent aux aspects actifs de la croyance. The debate in the literature on the problem of doxastic voluntarism and the ethics of belief is driven by two basic, resilient intuitions. The first is that our beliefs seem to be the proper objects of deontological epistemic evaluation: there are some things we epistemically ought not to believe, some we may but need not believe, and (perhaps) some we ought to believe. What makes our epistemic obligations deontological is the apparent fact that we are responsible for our beliefs, in the sense that, when we violate our epistemic duties, we are blameworthy for doing so. The second intuition is that beliefs are not under our direct voluntary control; forming a belief isn’t something that we can just do, simply by deciding to do so. Each of these intuitions seems to be independently plausible, but taken together, they appear to entail the falsity of the principle that ought-implies-can. The problem is to decide which of these three ideas to give up. The point of this paper is to draw attention to the fact that belief has both passive and active aspects, and to argue that the active aspect of belief is what makes it subject to deontological epistemic evaluations. Belief is a multi-faceted phenomenon: believing that p typically involves, among other things, both having a feeling of truth toward or confidence in p, and a willingness to employ p in one’s theoretical and practical deliberations. For example, when I look in my mug and see that it still contains some delicious coffee, I typically feel it to be true that there is coffee in my mug (passive), and I am typically willing to take it as a premise in my deliberations that there is coffee in my mug (active). These two aspects of belief almost always go together, but they can come apart. For example, suppose that Claire was raised in a racist environment, and she has always both felt it to be true and taken it as a premise in her deliberations that people with her skin colour are superior to all others. But when she goes off to college, Claire learns that there is no evidence supporting her racial superiority. She is unable to shake the feeling of truth which she has when she thinks about the proposition that she is racially superior (deeply ingrained beliefs are often difficult to shake), but she commits herself to not employing that proposition in her future deliberations about what to think and how to treat people. Because the active and passive aspects of belief can come apart in this way, we do not need to give up any of the intuitions with which we began: we can grant that ought-implies-can without giving up on the deontological approach to epistemic evaluation. 1. The Anti-Voluntarist Argument The main argument against the claim that our beliefs are proper objects of deontological evaluation can be put as follows: P1. If deontological terms of appraisal are applicable to an agent’s φ-ing, then it is under the agent’s direct voluntary control whether she φs. (A strong version of ought-impliescan.) P2. It is never under a person’s direct voluntary control whether she has a given belief. C. So, deontological terms of appraisal are never applicable to beliefs. This argument needs a few quick clarifications. Direct voluntary control contrasts both with lack of control and with indirect control. Subject S lacks any control over whether she φs just in case there is nothing S can do just by willing it which would have an effect on whether she φs. S has direct voluntary control over whether she φs just in case she can φ just by willing it, without any intermediate intentional steps. Direct voluntary control applies to what are sometimes called ‘basic’ actions. S has indirect voluntary control over whether she φs just in case there is something ψ which S can do, distinct from φ, just by willing it, which can bring it about that, or have an effect on whether, S φs. That is, there is some ψ over which S has direct voluntary control, which she can do with the intention to try to bring it about that she φs. So any time S has indirect voluntary control over something, that is because she has direct voluntary control over something else. For example, I can stand just by willing to do so, but I can only turn off the lights by willing to stand, walk to the light switch, and turn it to the ‘off’ position. I normally have direct control over whether I attempt to stand to turn off the lights; I normally have indirect control over whether I turn off the lights. The issue about voluntary control over our beliefs is about whether we have either direct, indirect, or no voluntary control over them, and about which sort of control, if any, is necessary for us to count as being responsible for the beliefs that we have. The issue is not about whether we have libertarian free will. We could have direct voluntary control over our beliefs even if we lacked libertarian free will. (I have direct voluntary control over whether I move my hand toward my coffee cup right now—I could move it there if I wanted to, just by willing to do so, and that is true whether or not I have free will in the libertarian sense.) P1 and P2 of the anti-voluntarist argument together entail C, so the arguments in the literature are over whether P1 and P2 are correct. Some people have appealed to constitutive features of belief having to do with its truth-aim, in support of the claim of P2.1 The idea is that the fact that belief constitutively aims at truth entails that nothing could be both adopted directly See Williams (1973) for the classic formulation of this argument. Many people have objected to Williams’s line of argument (e.g., Shah (2002) and Ginet (2001)). Frankish (2007) repairs some of its flaws and defends an updated version of the argument as an objection to what he calls ‘strong voluntarism.’ 1 by a decision and still be a belief; a mental state adopted by a subject directly by a decision to adopt it need not be adopted for evidential (i.e., truth-related) reasons, and the subject would know that, which means that the mental state would fail to be constitutively aimed at the truth.2 William Alston endorses both P1 and P2.3 He accepts the ought-implies-can principle, so he accepts P1, and he argues for P2 on the grounds that beliefs just don’t seem like they change in light of decisions to change them. For example, it doesn’t seem that any decision of mine will be able to generate in me a sincere endorsement of the proposition that I am currently sitting in a café in Paris, when I so manifestly am not in a café in Paris. Richard Feldman defends P2 for the same reason, but he rejects P1 because he thinks that we can be responsible for things that are out of our control, when we undertake social roles which come with responsibilities. According to Feldman, being a believer is a role that comes with epistemic responsibilities.4 Sharon Ryan rejects both P1 and P2.5 Regarding P1, she argues against several different versions of the ought-implies-can principle, arguing that, although it seems initially plausible, no version of the principle survives critical scrutiny. Regarding P2, she concedes that we cannot decide to believe for any arbitrary reason at all, but she argues that the ability to φ for any reason at all is not a condition of having control over whether to φ. It follows that the phenomenological evidence adduced in support of P2 by Alston and Feldman does not show that we cannot decide to believe, when we decide for the right reasons. When we have what appear to us to be good (evidential) reasons for belief, we can voluntarily believe what our reasons support, even if we are unable to refrain from forming the belief, given our recognition of what our reasons support.6 Others7 have argued against P2, agreeing that many beliefs do not seem to be subject to decisions about whether to believe them, but arguing that there do appear to be cases where people come to have beliefs by deciding to do so, especially when the evidence is insufficient to force belief one way or the other.8 So we may distinguish three sorts of doxastic voluntarism, which might be defended in response to the anti-voluntarist argument: Weak Voluntarism: We can never decide to believe what we regard to be unsupported by our evidence, but when we believe in accord with our apparent evidence, we do so for reasons we recognize as good. When we form beliefs, we therefore do so intentionally, and voluntarily. Intermediate Voluntarism: When the evidence is compelling either for or against a proposition, we are incapable of deciding to believe or disbelieve it. But when the evidence is inconclusive, we can decide to believe, disbelieve, or suspend judgment. 2 The aim-of-belief thesis is both unclear and controversial. Although I am inclined to reject it, I do not mean to take a stand on that thesis here; I only mean to point out that some people have taken it to be sufficient for showing that we never have any direct control over our beliefs. 3 Alston (1988). 4 See Feldman’s (2000) and (2008). 5 Ryan (2003). 6 Steup (2000; 2008) argues along similar lines. 7 E.g., Ginet (2001), Frankish (2007), and Weatherson (2008). 8 Mourad (2008) similarly argues that we are sometimes able to decide to form beliefs indirectly but immediately, so that there is no temporal lapse between the decision and the belief-formation, but Mourad doesn’t think that these are decisions to believe. Strong Voluntarism: We are always capable of deciding what to believe. Even when the evidence very strongly supports p, we are capable of deciding to believe not-p. Weak and Intermediate Voluntarism have each been ably defended, and of course there are those who deny all forms of direct voluntarism. But Strong Voluntarism has had precious few friends among epistemologists of the past hundred years or so. My aim in this paper is to defend a version of the thesis of Strong Voluntarism. I grant that P1 of the anti-voluntarist argument is correct for the purpose of this paper: if a subject is responsible for her beliefs, then she must have direct voluntary control over them.9 (The formulation of the ought-implies-can principle no doubt needs some refinement and defense, but I assume here that that that can be provided.) The trouble with the involuntarist argument is that P2 is too simple. There is a sense in which it is obviously true, and a sense in which it is obviously false. The passive aspect of belief is not typically under our direct voluntary control, but the active aspect is, and it is its active aspect which renders belief subject to deontological epistemic evaluation. The two aspects usually go together, but, as in the case of Claire, which we saw at the beginning of this paper, the two aspects can come apart. (We’ll get further into the divergence between the active and passive aspects of belief in Section 2.) When they do come apart, the active aspect remains subject to epistemic evaluation. The passive aspect of belief is not usually normatively evaluable in any deontological sense;10 because it can arise unbidden and in a manner entirely out of a person’s control, it seems incorrect to say that a person is in general responsible for that aspect of her beliefs. But none of that poses a problem for the applicability of deontological terms, because they aren’t meant to apply to the passive aspect of belief in any case. Once we clearly distinguish the active and passive aspects of belief, the problem of doxastic voluntarism and the applicability of deontological terms of epistemic appraisal simply vanishes. I should point out that the anti-voluntarist argument above is cast in terms of direct control, not because I think that indirect control is insufficient for responsibility, but because I think that it is relatively uncontroversial that we do often have indirect control over our beliefs, via the evidential standards to which we hold ourselves, the practices of inquiry which we undertake, the intellectual authorities we pay attention to, the evidence we ignore, and so on. Indirect control is sufficient to save some kinds of deontological evaluations of beliefs from ought-implies-can objections. But there are epistemic obligations which indirect control over our beliefs does not secure. For example, there appear to be synchronic epistemic obligations to adopt or abandon beliefs in response to the available evidence, which apply to us even when we are unable to indirectly influence ourselves into adopting or abandoning the beliefs in question in a reasonable amount of time, or when the indirect means of control which we exert are impotent with respect to the kind of belief in question. However, even if I am mistaken about the extent of the epistemic responsibilities which indirect control can secure, it is clear enough that, if we have direct control over all or most of our beliefs, then that will be enough to save deontological epistemic evaluations from the charge 9 Direct voluntary control is needed for responsibility for our beliefs because, although indirect voluntary control can account for some, or perhaps many, of our epistemic obligations, it cannot account for all of them. I come back to this point shortly. 10 The passive aspect of belief may of course be normatively evaluable in non-deontological senses. One might say, with Feldman (2008), that S ought to believe that p given evidence E, meaning that believing p would be the epistemically best thing to do given that S has E, but without meaning to imply that S is either praiseworthy or blameworthy in complying or failing to comply with that epistemic ‘ought.’ of violating the ought-implies-can principle. That makes the question of the extent of our direct control over our beliefs an interesting one. My aim in this paper is to argue that we do have direct control over the active aspect of our beliefs. 2. Belief and Acceptance The concept of belief is complex.11 For one thing, to believe p is in some sense to take p to be true. For another, when S believes that p, S is disposed to feel it true that p, and to feel confident that p is true. And beliefs play a functional role in the cognitive economy: they are often arrived at automatically in response to apparent evidence, but they are also often arrived at via inference. We can also take our beliefs as premises in deliberation, to infer further beliefs and to negotiate our way in the world. Belief-attribution also plays an explanatory role: the attribution of beliefs and desires (or beliefs and plans, intentions, etc.) typically provides a rational explanation of action. And, of course, beliefs can be justified or unjustified, and in at least many paradigmatic cases, the epistemic status of beliefs is determined by the available evidence. This cluster of features typically all go together. I will normally feel confident that p is true when my apparent evidence obviously supports p, and if there’s no available counterevidence, my belief that p will be justified by that evidence. And when I feel confident that p is true, I will normally take it that p is true, and I will employ p as a premise in my theoretical and practical deliberation. As a result, together with my other beliefs, plans, desires, and so forth, my taking p to be true helps me to negotiate my way in the world in a rational way. And someone watching me negotiate the world will normally be able to rationally explain my actions, in part, by attributing to me the belief that p. But sometimes these features come apart, as we saw in the case of Claire. Claire can’t help feeling as though it is true that she is part of a superior race, but she refuses to take that proposition as a premise in her deliberation. Cases like this make it natural to wonder how we ought to regiment our terminology. We might say, on the one hand, that Claire has a certain feeling about the relative worth of people of different races, but that she refuses to believe it. Or we might instead say that Claire continues to believe that she is racially superior, but that she refuses to treat her belief as a reason in further deliberations. Which description of the case we adopt is unimportant; what is important is to ensure that, however we describe the case, we respect the distinction between the active and passive features of Claire’s mental life, and the way in which those aspects have come apart. A helpful distinction to draw here is therefore L. Jonathan Cohen’s distinction between what he calls ‘belief’ and ‘acceptance.’ For Cohen, the belief that p is a disposition “normally to feel it true that p and false that not-p, whether or not one is willing to act, speak, or reason accordingly,” whereas “to accept that p is to have or adopt a policy of deeming, positing, or postulating that p—i.e., of including that proposition or rule among one’s premises for deciding what to do or think in a particular context.”12 Beliefs have to do with how we are disposed to feel 11 Or, better: the concept of belief is a concept of a complex state. Perhaps belief itself is a simple referential concept, or perhaps it contains multiple elements, or some descriptive content. I am not worried about the semantic content of ‘belief,’ so much as I am about features of the state of believing. 12 Cohen (1992, p. 4). Note that it is possible to reject Cohen’s distinction between belief and acceptance, by rejecting the claim that there is such as a thing as a feeling that a proposition is true. I think that the feeling of truth is a familiar feeling, a sort of confidence, or feeling of agreement, that a proposition is true. At the very least, to have a feeling of truth toward a proposition involves not feeling the irritation of doubt (cf. Peirce (1877). To deny that we have a feeling of truth seems to me like denying that there is such a thing as a feeling of comfort. about the truth of propositions in normal circumstances, whereas acceptances have to do with what we are willing to use in our reasoning. This distinction cuts our cognitive lives at the relevant joint for our purpose here. Of course, this way of talking about distinct states of believing and accepting that p has an admittedly artificial feel to it; what Cohen calls ‘belief’ and ‘acceptance’ are both aspects of the complex that we usually think of as making up the state of belief. I do not mean to endorse Cohen’s view that belief just is the passive disposition to feel it true that p, but his way of talking is a convenient way to label the active and passive elements of the set of features which make up what we normally think of as belief. So, in what follows, in spite of its artificial feel, I adopt Cohen’s terminology, and talk about the passive feeling or disposition to feel it true that p as the belief that p, and the active aspect of taking p up as a premise or rule for theoretical or practical deliberation as the acceptance of p.13 There are a number of related but importantly different takes on the belief-acceptance distinction in the literature. I do not take Cohen’s distinction to be superior, or even a rival, to other similar distinctions; Cohen’s is just a more useful conceptual scheme for the purpose of a discussion of the voluntariness and epistemic evaluability of cognitive phenomena. But it will no doubt help to clarify the idea at hand by contrasting it with some other belief/acceptance distinctions in the literature. Keith Lehrer draws a belief/acceptance distinction very similar to Cohen’s, according to which beliefs are involuntary propositional attitudes, but where we can accept propositions in response to evidential considerations even when we do not believe them.14 Both Lehrer and Cohen take belief to be a passive phenomenon, and they both take an acceptance to be something which a subject can acquire by deciding to do so, when the evidence seems to support a proposition which a subject is unable to come to believe. But the difference between Lehrer’s and Cohen’s belief/acceptance distinctions is that, on Cohen’s picture, but not on Lehrer’s, we may accept propositions for non-evidential reasons as well. Another belief/acceptance distinction is Michael Bratman’s.15 Bratman distinguishes between propositions which are taken for granted in the cognitive background of deliberation across all contexts, and what is taken for granted only in particular contexts of deliberation. The former cognitive items are what Bratman calls ‘beliefs’; the latter, ‘acceptances.’ On Cohen’s picture, on the other hand, any proposition which is taken for granted in the cognitive background of deliberation, whether it be a fixed element across all relevant contexts, or an element which is only present in the background of certain specific deliberative contexts, is an accepted proposition. And if the thought of an accepted proposition is accompanied by an appropriate feeling of agreement in normal circumstances, then that proposition is also believed. Keith Frankish draws a belief/acceptance distinction similar to Bratman’s.16 For Frankish, S believes that p just in case S is disposed to employ p as a premise in contexts in which S cares But if I am wrong about this, and there is no distinctive feeling of truth that we have, then it seems to me that the justification for P2 in the anti-voluntarist argument, i.e., the feeling of being unable to change our beliefs at will, becomes drastically reduced—for if there is no feeling of truth, then it is just not clear that belief is something which we are unable to alter by a direct act of will. 13 In talking about acceptances as ‘active,’ I do not mean that one always or even usually must make a conscious mental act of accepting that p. Most acceptances are automatic. The point is, rather, that one is always in a position to be able to accept or to refrain from accepting that p, should one choose to do so. 14 Lehrer (1997; 2011). 15 Bratman (1992). Cf. also Cresto (2010). 16 Frankish (2007). only about having true premises, whereas S accepts that p just in case S is willing to use p as a premise in specific or short-lived practical/institutional contexts, where truth is only one among many of the constraints on our premises. Frankish allows that we may accept that p directly and immediately by deciding to do so, but that acceptance does not imply belief. I come back to Frankish’s conception of belief and acceptance in Section 4. Finally, it might help to look at Tamar Gendler’s distinction between belief and alief, which bears some surface similarities to the distinction I am borrowing from Cohen. Belief, for Gendler, is the propositional attitude of thinking that p is true, and being willing to take p as a premise in deliberations, and so on. (Paradigmatic) alief, on the other hand, is “a mental state with associatively linked content that is representational, affective and behavioral, and that is activated (consciously or nonconsciously) by features of the subject's internal or ambient environment.”17 An alief is a state which is activated in some way, typically outside of a person’s control, which involves having a representation of some sort, an emotional response to that representation, and some behavioural dispositions. For example (to take an example from personal experience, which closely parallels one of Gendler’s): when I was young, I once went up in the CN Tower, and walked on the small glass floor they had installed, which stands over a straight drop of 342 meters. I certainly believed that the glass floor was as safe as the rest of the tower to walk on. But I was nevertheless very nervous, and it took a few seconds before I was willing to walk on it. Gendler would say that I believed that the floor was perfectly safe—I thought it was true that it was safe, and I was willing to walk on it, after all—but I alieved (roughly) that it was unsafe. I had an alief with the content “Big drop! Unsafe! Back away!” Alief bears some similarities to what Cohen calls ‘belief’ (call this C-belief), and what Gendler calls ‘belief’ (G-belief) bears similarities to what Cohen calls ‘acceptance.’ Aliefs are not under our voluntary control, and they involve an emotional or feeling component, just like Cbeliefs. And both aliefs and C-beliefs are capable of causing us to act in ways contrary to our better judgment. And, just like acceptances, G-beliefs are what we take up as premises in our deliberations. But there are important differences between the alief/G-belief and the Cbelief/acceptance distinctions. For one thing, neither G-beliefs nor aliefs are subject to our voluntary control, whereas acceptances are. Furthermore, aliefs are activated in response to specific features of circumstances, whereas C-beliefs are what a subject is disposed to feel about the truth of a proposition in normal circumstances. Although it’s not easy to say with precision what sorts of circumstances count as normal, I think it’s fair to say that my current circumstances, as I sit at my computer and reflect on the prospect of walking on a glass floor, should count as normal circumstances. In these circumstances, I do not as a matter of fact have any negative feelings about the truth of the proposition that the glass floor of the CN Tower is perfectly safe to walk on. So it’s fair to say that I C-believe that proposition. But I am also fairly certain that, were I now in the CN Tower, about to walk on the glass floor, I would have a jolt of adrenaline and feel very nervous about it. So it seems fair to say that I alieve that the glass floor isn’t safe to walk on. So this is a case where alief and C-belief diverge. So despite some surface similarities, Gendler’s distinction isn’t the same as Cohen’s, and I take it that Cohen’s distinction is the relevant one to draw for the purpose of highlighting the active features of our mental lives, given the active aspect of acceptances, and the passive aspect of both aliefs and G-beliefs. 17 Gendler (2008, p. 642). Now it is obvious that belief and acceptance in Cohen’s sense typically go together. When I pick up my coffee cup and look inside, I am presented with some visual evidence which automatically gives rise in me to the belief that there is some delicious coffee left in my cup. And at the same time I also accept that there is coffee in my cup, and I proceed to take a drink on that basis. That is the typical sort of case: the vast majority of things that we accept, we feel to be true, and the vast majority of things we feel to be true, we accept. Indeed, there is good reason for being disposed generally to accept the things which we feel confident to be true: given that our feelings of the truth of propositions are typically generated by more or less reliable processes, if a subject finds herself with a belief, she will usually thereby have a prima facie reason for thinking that her belief is true.18 But, as we saw above with the case of Claire who refused to accept her racist beliefs, belief and acceptance can come apart. Other cases which illustrate the divergence between belief and acceptance include the following: (1) A high school student, upon learning about Special Relativity for the first time, might accept that it is true on the basis of her teacher’s instruction, and yet she might still be stuck with the feeling that Newton’s picture of the world is true. Feelings about the fundamental nature of things are not typically easy to shake off, after all, even given good arguments. So this student accepts what she does not believe, in Cohen’s sense (that Special Relativity is true), and she believes what she does not accept (that Newton’s theory is true). (2) A logician might work through each step of a difficult proof and come to accept that it is valid, even if the complexity of the proof prevents him from feeling it to be valid; indeed, he might accept that it is valid even though he began the process with the feeling that the proof is invalid, and even if he continues to have that feeling. (This example is adapted from one of Cohen’s.) (3) A delivery driver might wonder whether he remembered to close the back door of his truck before leaving his last stop. He might reflect that he’s forgotten to close it once or twice before, and that he has no clear memory of having closed it this time. But, because it is impractical to stop one’s truck 10 times in a day to check the back door, he might simply accept that he did remember to close the door. Yet he might not come to feel a sense of agreement with the proposition that he remembered to close the door. In that case, he accepts something that he does not believe. 19 (This example is autobiographical.) (4) A person might go on an airplane for the first time, even though she believes that the plane will crash. She accepts that it will not crash—she bought the ticket and went on 18 This is not to endorse the extreme thesis that having a belief that p always gives a subject a reason to believe that p; that thesis is, I think, obviously false. But it seems to be true that, other things being equal, finding oneself with the belief that p is indirect evidence for the truth of p, for any subject who is self-aware enough to realize that most beliefs are formed on the basis of evidence. See Foley (2001) for a good discussion of intellectual self-trust. 19 It might be objected that the driver only acts as if he believes that he’s closed the back door, but he doesn’t really take that proposition to be true—at best, he hopes it’s true. But, given that this example is autobiographical, I can tell you confidently that there was a time when the driver, upon reaching his next stop and seeing that the door was still open, cursed himself for an idiot and judged that he had been wrong about his having closed the door. And it’s only possible to make an idiotic mistake about whether you’ve closed the door if you’ve adopted a doxastic attitude with respect whether you’ve closed the door. So I think it is plausible to say that the driver did adopt a doxastic attitude about it, not merely an attitude of hoping that he’d closed the door. the plane, after all—but she has a terrible fear of flying, which prompts her to feel it to be false that the plane will safely make it to its destination, whenever she thinks about it. In this case, she accepts that the plane will be fine, but she does not believe it. In each of these cases, passive and active aspects of people’s mental lives come apart. Adopting Cohen’s terminology, we can call these aspects ‘belief’ and ‘acceptance’ respectively. Or, if we want to do so, we can just keep talking in terms of active and passive representational mental states, and reserve the term ‘belief’ for the generic state of taking-to-be-true, when both the active and passive features are involved. But, for convenience, I adopt Cohen’s terminology here, since it cuts the aspects of our mental lives at the relevant joint, I think, for purposes of epistemic evaluation. 3. The Applicability of Deontological Terms of Evaluation We do not have direct voluntary control over the things that we believe, since feelings or dispositions to have feelings are not things that we can simply call up or dispel at will. We do, however, have direct voluntary control over the things that we accept. That is sufficient for the applicability of deontological epistemic terms, since the proper objects of epistemic evaluation are acceptances. As we have just seen, belief and acceptance almost always go hand-in-hand. I both feel to be true and accept as true the proposition that I am sitting in my living room, and this is a very typical, if uninteresting, feature of my cognitive life. I suspect that the fact that belief and acceptance rarely come apart is one main reason why the distinction between the two has not been more prominent in the literature on voluntarism. Most of the time, there is no point in distinguishing the active and passive features of belief, and it is harmless to think of belief tout court as held responsibly or irresponsibly, and as subject to deontological epistemic evaluations. But that way of thinking is only appropriate in normal cases, where the active and passive features of belief do not come apart. Another likely reason for not focusing on the distinction between belief and acceptance in the context of a debate about the legitimacy of deontological epistemic terms of appraisal is that, although it is fairly clear that we do have these active and passive cognitive phenomena, and it is easy to see, once it is drawn to our attention, that they can come apart in some cases, there is a default assumption that epistemic evaluations must always attach to beliefs, and never to acceptances without belief. In other words: in order to be a candidate for epistemic evaluation, a mental state with content p must at least involve a credal feeling with respect to p. The commonly-held view is roughly this: if you offer me a million dollars to believe that there are no cars driving outside my window, when I can hear them, I will helplessly continue to believe that the cars are there, and that belief will continue to be justified by the evidence of my senses. I can, of course, cease to accept that the cars are there, in the sense that I can commit myself in my theoretical and practical deliberations to the truth of the claim that there are no cars driving outside my window, but although my acceptance may be pragmatically justified, it is not epistemically justified (nor is it epistemically unjustified), because it is not a belief, and beliefs are what we evaluate epistemically.20 20 Can I really accept that there are no cars driving outside my window, when I can hear them? Yes, but it is not easy. It will require quite a commitment on my part, to revising a number of other things that I accept—propositions about the possible causes of engine sounds through a window, about the kinds of sounds that I can reliably But there is good reason for taking the proper objects of epistemic evaluations to be acceptances rather than beliefs. When acceptance and belief come apart, it is acceptances, not beliefs, which get taken up in our reasoning about what to think and to do. Now, a plausible principle about the justification of action is the following principle of the Necessity of Justified Premises (NJP) for deliberation: NJP: An action is only practically justified on the basis of a process of deliberation about whether to perform it, if the premises which are taken to be true in the process of deliberation are epistemically justified.21 It might seem that the NJP principle is open to counterexamples, in cases where a subject S must choose between incompatible actions A and B, but S has no evidence regarding which action is likeliest to succeed. For example: Susan gets lost driving at night, and has no cell phone. Her car breaks down at an intersection which has an old sign indicating that there are two 24 hour gas stations, five kilometers down the road in both directions. Unfortunately, there is also a ‘closed for renovations’ sign, which clearly had been meant to cover one of the signs, but which has slipped and is now hanging awkwardly off the sign-pole. So Susan has to pick a direction in which to walk, but she has no way to tell which direction will get her to an open gas station. It seems that she can be practically justified in just picking a direction, in spite of the fact that the deliberation results in her taking up a course of action which she is not epistemically justified in thinking will be successful. Plausible as this case is, it is not, in fact, a counterexample to the NJP principle. For Susan’s practical deliberation would have gone something like: “Well, this is the worst. I’ve got to walk, in order to get a tow truck, but there’s no way to tell which way to go. I’ll just go left.” And, although she lacks epistemic justification for thinking that her decided-upon course of action will succeed, the two premises of her practical deliberation here—that she needs to walk to call a tow truck, and that there’s no information about whether going right or left will get her to where she can call one—are both epistemically justified, which is precisely what NJP requires. Susan’s decision to go left is based on her acceptance of the proposition that she must go either right or left, but not on the acceptance of the proposition that she must go left. Of course, she hopes that going left will get her where she wants to go, but she does not need to accept that it will do so in order to be justified in going left. distinguish, and so on. Deciding to accept that p is usually not a simple, one-off decision that a person may legitimately simply make and then ignore; if a person decides to accept that p when there is good evidence at hand for not-p, further revisions of that person’s cognitive picture will be required, in order to bring the set of things she accepts into a coherent enough picture to enable her to successfully navigate the world again. Now, some things might seem to be beyond our power to intend, such as intending to jump in front of a car on the highway. Perhaps there are, similarly, some things we cannot bring ourselves to take up in our practical deliberations, such as the proposition that one ought to jump in front of a speeding car. Perhaps this shows that we do not have the degree of voluntary control over our acceptances that I am suggesting. My reply to cases such as this is to say that one can, after all, intend to jump in front of a speeding car; it’s just that one (typically) will not form that intention except under extreme conditions. I come back to this point at the end of the paper. 21 See my (2010) article for discussion of justification and truth as constraints on the acceptability of premises. Notice, by the way, that the necessity of having justification for one’s premises, in order for one’s deliberations to be rationally acceptable, is not controversial. The controversy is about whether more is needed, too. (See, for example, Hawthorne and Stanley (2008).) Now, if the NJP principle is correct, then the fact that it is acceptances which enter as premises into the process of practical deliberation entails that acceptances must be able to be epistemically justified, if it is possible for actions to be justified on the basis of practical deliberations. But surely actions can be justified on the basis of practical deliberations. So acceptances are proper objects of epistemic evaluation. Further support for the applicability of epistemic evaluations to acceptances rather than beliefs comes from considering cases like Claire’s. Recall that Claire has been raised with the deeply-held belief in her racial superiority; later, in spite of finding that there is no reason to believe that any races are superior to any others, she is unable to avoid the feeling that she is racially superior, when she thinks about the question. But she refuses to use the proposition that she is racially superior in her deliberations about what to think and to do. In this case, Claire is responding to the evidence that there are no plausible theories of human nature according to which there are any superior races, by modifying what she accepts about the status of different races. And that is exactly what she epistemically ought to do. The fact that she is unable to shake off her deep-seated prejudicial feeling of superiority is merely an unfortunate fact about her, over which she has no direct control. But she is able to control whether she takes up that feeling of racial superiority in her future deliberations. Of course, although Claire does not accept her racial superiority, her feeling of superiority might still indirectly affect her decisions and actions, even if not through her conscious deliberative processes.22 And if her feeling of superiority were to cause her, in spite of herself, to make racist decisions, we would likely think that her decisions would be appropriately blameworthy. This might seem like a problem for my account, because on the picture of epistemic blameworthiness I am defending here, it is acceptances rather than beliefs which are blameworthy. So I need to account for the blameworthiness of Claire’s racist decisions, without attributing the blameworthiness of her decisions to an epistemically blameworthy belief which caused them (because it’s not beliefs that are blameworthy). The way I would account for the blameworthiness of Claire’s racist decisions is not in epistemic terms at all. Claire fulfills her epistemic obligations by refusing to accept her racial superiority. The blameworthiness of Claire’s racist decisions does not derive from the blameworthiness of her feeling of superiority, since feelings are not truly blameworthy; it derives, instead, from the fact that she fails to live up to her commitment to treating people as equals (and from the fact that that is a morally important commitment to fulfill), even if she does not realize that she has failed in this way. She is blameworthy for this failure because she could and should have lived up to her commitment to treating people equally.23 22 If so, then as Gendler (2008) would say, Claire alieves that she is racially superior. Perhaps it will be objected that, if Claire cannot recognize that her racist beliefs are influencing her actions and decisions, then we violate the ought-implies-can principle if we hold her morally responsible for those decisions, since she can only block the influence of her beliefs if she knows about the influence of her beliefs. I accept this objection, but I deny that Claire is incapable of recognizing that her racist beliefs are influencing her decisions. She is a person who is careful enough and who cares enough to be able to recognize when her decisions are reflecting her former racist values. She is therefore rationally blameworthy for allowing beliefs which she doesn’t accept to influence her deliberations. But maybe I’m wrong about that: maybe Claire really just isn’t capable of recognizing that her decisions are influenced by her racist beliefs. In that case, Claire really isn’t capable of limiting the influence of her racist beliefs on her decisions, and I am willing to say that she is, therefore, blameless for allowing those beliefs to influence her decisions. Her decisions are still unfortunate, or bad, or suboptimal, but not blameworthy. 23 The argument in support of the applicability of deontological epistemic terms of evaluation to acceptances rather than beliefs, then, is (i) that we have control over acceptances, so applying deontological evaluations to acceptances does not violate the ought-implies-can principle; and (ii) acceptances are what guide our theoretical and practical inferences, and the premises of theoretical and practical deliberation are proper objects of epistemic evaluations. Beliefs, on the other hand, are not under our direct control, and they are relevant to our deliberations only when we accept what we believe. Although beliefs can cause us to act in certain ways even when we do not accept them, and we can be blameworthy for being influenced by our beliefs in such cases, the blameworthiness derives from our failure to live up to our commitments, rather than from a failure to control our beliefs. I want to emphasize, however, that, although it is useful for keeping the active and passive aspects of belief distinct, I am not seriously committed to this terminology of acceptances and beliefs. I am just as happy to say that Claire refuses to believe that she is racially superior to anyone else in spite of her feeling of racial superiority, as I am to use Cohen’s terminology and say that Claire refuses to accept her belief that she is racially superior. The terms we use here are less important than the fact that there is a kind of cognitive phenomenon which is both active and representational, which I am following Cohen in calling acceptance, and that that is what is properly subject to epistemic evaluation. And because acceptances are under our voluntary control, the involuntarist argument for the inapplicability of deontological epistemic terms to our cognitive picture of the world simply does not go through. Before closing this paper, now, I want to address two important objections to my claim that, because we have direct voluntary control over the things that we accept, what we accept is subject to deontological epistemic evaluation.24 4. Insincere Acceptance The first objection is that, if we accept things that we do not believe, our acceptance is insincere. Insincere acceptances are not really proper objects of epistemic evaluations because we do not, after all, think that their propositional contents are true, and it is only what we think to be true that can be epistemically justified or unjustified. Or, in other words, acceptance without belief is only pretend, and therefore not really up for epistemic evaluation. This objection misconstrues the nature of a sincere acceptance. To sincerely accept that p is not to feel that p is true; it is to genuinely commit oneself to taking p as a premise in one’s practical and theoretical reasoning. And taking p to be true for the purpose of reasoning about what to think and to do seems to me to make acceptances exactly the kind of mental states that we ought to subject to epistemic evaluations. There is, of course, a form of pretend acceptance, but it is not the result of accepting what one does not believe. It is, rather, the result of claiming to accept a proposition which one is unwilling to take as true in one’s theoretical and practical deliberations. The objection can be pressed, however, by pointing out that a defense lawyer, for example, can believe that her client is guilty, but accept that he is innocent for the purpose of 24 A third possible objection, which I have come across but do not take to be a serious worry, is that because acceptances are responsive to practical reasons, they are not belief-like enough to be epistemically evaluable. The response, of course, is that, although acceptance is not constitutively governed by epistemic reasons, it is still normatively governed by such reasons. We are capable of governing our acceptances in accord with our epistemic reasons, and we epistemically ought to do so. conducting a trial.25 It seems correct to say that, in a case like that, the lawyer accepts her client’s innocence, at least for certain professional purposes. But, at the same time, it seems incorrect to epistemically evaluate the lawyer’s acceptance, since she is not trying to get it right whether her client is innocent when she accepts that he is. She is only performing her professional duty, which requires her to accept it. So we ought to say that her belief is epistemically justified, while her acceptance is only practically or institutionally justified. According to Frankish, the distinction between belief and acceptance comes down to the contexts in which we are willing to rely on a proposition which we take to be true. When we are willing to rely on the truth of a proposition only in narrow social or institutional contexts, we can be said to accept that proposition. When we are willing to rely on the truth of a proposition in an open-ended fashion in our future theoretical and practical deliberations in which we are concerned with having only true premises, we believe it.26 If this is the correct way to distinguish belief and acceptance, then surely it is beliefs rather than acceptances which come up for epistemic evaluation, on Frankish’s account. To the extent that Frankish is stipulatively defining his use of the terms ‘belief’ and ‘acceptance,’ I have no objection to his usage. But drawing the distinction this way does not track the divergence between the active and passive features of beliefs, as Cohen’s does. For example, in the case of Claire, she has a passive feeling of racial superiority, but she is unwilling to rely on the truth of the proposition that she is racially superior in her future deliberations in which she cares about having only true premises. There is a natural sense in which she refuses to accept (take up in her reasoning) what she believes (has a feeling of truth towards or agreement with), and it seems to me that it is what she accepts that we ought to evaluate epistemically, because it is only what she accepts that she is able to regulate in accord with the evidence as it appears to her, and it is what she accepts that will be taken up in her future deliberations. Still, the case of the lawyer who appears to believe in her client’s guilt but accept his innocence needs to be accounted for. Distinguishing belief and acceptance as Cohen does, it is clear that the lawyer believes her client to be guilty, because she is disposed to feel the proposition that he is guilty to be true. And there is a sense in which she accepts her client’s innocence, a sort of restricted, acceptance-in-a-professional-context sense. But there is also a wider sense of acceptance, the sense of an open-ended willingness to take up the proposition that her client is innocent in her future theoretical and practical deliberations, and in this sense she does not accept her client’s innocence. So one might wonder: if the lawyer accepts in one sense/context, and does not accept in another sense/context, that her client is guilty, which acceptance ought to be subject to epistemic evaluation? My inclination is to say that it is the wider sense of acceptance that matters; I am inclined to say that that is what she really accepts, and the acceptance in the narrower context is insincere. It seems to me that there are certain sorts of institutional contexts in which we allow people to accept propositions without subjecting their acceptances to epistemic evaluation, as long as their acceptances do not fly in the face of very strong evidence. Lawyers in court-room settings are a case in point: we do not expect a lawyer to accept her client’s guilt, unless there is conclusive evidence of his guilt, in which case we expect her to accept his guilt and attempt a different sort of defense, or try to get as light a sentence as possible. But I need not commit myself to that view here, since one could easily hold that each of the lawyer’s acceptances is sincere, and properly subject to epistemic evaluation. The lawyer 25 26 Frankish (2007, p.536). Frankish (2007, pp. 536-537). accepts her client’s innocence, for one set of deliberative situations, and his guilt for another set, and at most one of these acceptances (probably the acceptance of her client’s guilt) is epistemically justified. That is not to say that the lawyer’s acceptance of her client’s innocence is overall inappropriate in the context of providing a legal defense; she has excellent institutional reason for that acceptance. But the appropriateness of accepting her client’s innocence on the basis of that institutional reason might not free her acceptance from epistemic censure, if there really is very good evidence indicating that he is guilty. 5. Practical Payoffs The second objection I want to address comes from an argument of Feldman’s.27 When a subject is capable of φ-ing as a direct result of practical deliberation about whether to φ, Feldman says that the subject has ‘response control’ over whether he φs. Feldman argues that normal humans do not have response control over their beliefs. In Feldman’s terms, the view I am defending in this paper is that typical humans have response control over what they accept. An argument Feldman employs in arguing that we do not have response control over our beliefs is that we can imagine being offered a sum of money for forming an innocuous but evidentially unsupported belief, without gathering evidence first. Feldman’s example is that, on a random winter day in Rochester, where it snows about half the time, a colleague might offer him some reward for forming the belief, without first looking out the window, that it’s snowing outside. Feldman imagines himself deliberating about the practical and epistemic reasons for and against having the belief, and concluding that he would, in fact, rather have the belief that it’s currently snowing, and collect the offered payment. But he cannot act on the conclusion of that practical deliberation; he cannot get himself to have the unsupported belief, just by deciding to do so. I have been arguing that, although we cannot decide for practical reasons to feel some random innocuous false or evidentially unsupported proposition to be true, we can accept that it is true in our present and future deliberations. For example, though I am unable to feel it to be true that it’s snowing outside right now in Ithaca without looking outside, I can commit myself to the truth of that proposition. (It’s been snowing a lot here lately, but not every day, so my evidence doesn’t clearly militate one way or the other.) If I were to accept that it is snowing right now, and commit myself to employing that proposition in my reasoning about what to think and to do, it would no doubt have some practical consequences: if I were to decide to go outside, I would put on boots and a coat instead of shoes and a jacket; if my mother were to ask me how the weather is, I would probably tell her that I’m assuming that it’s snowing and that I’m dressing accordingly; and so on, with other similar small consequences. Since these are very small consequences, though, and since I can think of no truly important practical or even epistemic effects that my accepting that it is snowing outside could plausibly have, a cash payment might be enough to get me to accept that. If a small payment is not enough, surely I can imagine a larger one which would do the job. So acceptances, I am arguing, are under our voluntary control. The objection that I have in mind now is that, if acceptance is as much under our voluntary control as I have been saying, then it should follow that a large enough payoff would be able to induce us to accept even such obviously false propositions as that the earth is hotter than the sun, or that humans are all 27 See Feldman (2008). omnipotent.28 But no payoff that I can think of is enough to get me, at least, to accept those propositions. It is possible that an eccentric but exceedingly wealthy person might offer me a billion dollars to accept that the earth is hotter than the sun. Such a payoff is about as good a practical payoff as I might hope for; surely, it seems, if I could accept propositions just by deciding to do so, then I would decide to accept that one. But the fact is that I am unwilling to accept that the earth is hotter than the sun, not even for a billion dollars. Doesn’t that show that acceptance is not as free as I have been saying here? No, it doesn’t show that. It is not the case that, if I could accept propositions in response to practical reasons, I would accept any proposition whatsoever in order to achieve an astronomical reward. My current belief (and acceptance) that the earth is cooler than the sun is very fundamental to my cognitive picture of the world. If I were to accept that the earth is hotter than the sun, my cognitive picture would become quite incoherent, and I am uncertain how I could go about revising the things that I accept to make it coherent again. In particular, I am not certain that the resulting picture of the world would be accurate enough to keep me alive and in good health. (If I were to change the things that I accept about heat, for example, I might end up burning to death. Or I might end up funding dangerous climate-change-denial research and advertising campaigns, and in my own way help bring about an environmental catastrophe.) So it simply is not clear that any cash reward would be large enough to induce me to give up any of my fundamental acceptances about the world; in spite of the improved health of my bank account, it just is not clear that the resulting state of affairs would be better overall. Still, perhaps there can be even bigger rewards, which come not in the form of my own personal enrichment, but in the form of the avoidance of serious calamities. Suppose, for example, that my nemesis has a string of nuclear bombs planted across the globe, which he will set off unless I accept that the earth is hotter than the sun. Surely that is a large enough reward to get me to accept that proposition. But, surely, I can’t bring myself to accept that proposition, can I? So acceptances are not as much under our voluntary control as I have argued. I agree that, in a case like this, the reward is large enough to induce me to accept that the earth is hotter than the sun. But it no longer seems clear to me that I would not accept that proposition in this extreme case. Of course, it would take work to bring my cognitive picture into a coherent enough state to be able to effectively engage with the world, once I’ve accepted that proposition, but I see no reason to think that I would be unable to accept it. I would not feel it to be true; I would feel very confident of its falsity; I would be aware of very good evidence for its falsity; it would be an extremely unjustified belief. But none of these things preclude my accepting that the earth is hotter than the sun, in the sense of being willing to take that proposition as true in my future theoretical and practical deliberations. So, while it is true that, if acceptances are under our direct voluntary control, then a large enough practical reward should be able to get me to accept just about anything, it is not clear that an extremely large practical reward is incapable of getting me to accept just about anything. 28 Another related objection would be: if acceptances are under our voluntary control, then we must be able decide to accept that p for a predetermined amount of time. (I can decide now to accept that it’s snowing outside for the next three hours, for example.) But nothing that we can decide to do for a predetermined amount of time is sufficiently belief-like to be a candidate for epistemic evaluation. So, if acceptances are belief-like enough to be properly subject to epistemic assessment, then acceptances aren’t really under our voluntary control. The response to this objection is to accept its initial conditional, but to deny that the possibility of accepting a proposition for a limited time undermines the legitimacy of deontological epistemic evaluations of our acceptances. Acceptances are still what get taken up as premises in deliberation, so (given the NJP principle) they are prime candidates for epistemic evaluation. 6. Conclusion I conclude, then, that because acceptances are under our direct voluntary control, and acceptances are the cognitive phenomena which are taken up as premises in theoretical and practical deliberations, acceptances are proper objects of deontological epistemic evaluations. The ought-implies-can objection is therefore not a serious obstacle to a deontological approach to epistemic evaluation. Acknowledgments: Thanks to Carl Ginet, Benjamin Wald, Dustin Olson, and two anonymous reviewers for this journal for their very helpful discussion and comments on previous drafts. References Alston, William 1988 The Deontological Conception of Epistemic Justification. Philosophical Perspectives. vol. 2: Epistemology, 257-299. Bondy, Patrick 2010 Truth and Argument Evaluation. Informal Logic. vol. 30, no. 2, 142-158. Bratman, Michael 1992 Practical Reasoning and Acceptance in a Context. Mind. vol. 101, no. 401, 1-15. Cohen, L. Jonathan 1992 An Essay on Belief and Acceptance. Oxford: Oxford University Press. Cresto, Eleonora 2010 Belief and Contextual Acceptance. Synthese. vol. 177, no. 1, 41-66. Feldman, Richard 2000 The Ethics of Belief. Philosophy and Phenomenological Research. vol. 60, 667695. Feldman, Richard 2008 Modest Deontologism in Epistemology. Synthese. vol.161, 339-355. Foley, Richard 2001 Intellectual Trust in Oneself and Others. Cambridge: Cambridge University Press. Frankish, Keith 2007 Deciding to Believe Again. Mind. vol. 116, 523-547. Gendler, Tamar 2008 Alief and Belief. The Journal of Philosophy. vol. 105, no. 10, Part Two, 634-663. Ginet, Carl 2001 Deciding to Believe. In Mattias Steup, ed. Knowledge, Truth, and Duty. Oxford: Oxford University Press. 63-76. Hawthorne, John, and Jason Stanley 2008 Knowledge and Action. The Journal of Philosophy. vol. 105, no. 10, 571-590. Lehrer, Keith 1997 Self-Trust: A Study of Reason, Knowledge, and Autonomy. Oxford: Oxford University Press. Lehrer, Keith 2011 Evidentialism and the Paradox of Parity. In Trent Dougherty, ed. Evidentialism and its Discontents. Oxford: Oxford University Press. 55-68. Mourad, Ronney 2008 Choosing to Believe. International Journal for Philosophy of Religion. vol. 63, 55-69. Peirce, Charles 1877 The Fixation of Belief. Popular Science Monthly. Reprinted in H. S. Thayer, ed. Pragmatism: The Classic Writings. Indianapolis, IN: Hackett, 1982. 61-78. Ryan, Sharon 2003 Doxastic Compatibilism and the Ethics of Belief. Philosophical Studies. vol. 114, 47-79. Shah, Nishi 2002 Clearing Space for Doxastic Voluntarism. The Monist. vol. 85, 436-445. Steup, Matthias 2000 Doxastic Voluntarism and Epistemic Deontology. Acta Analytica. vol. 15, 25-56. Steup, Matthias 2008 Doxastic Freedom. Synthese. vol. 161, 375-392. Weatherson, Brian 2008 Deontology and Descartes’ Demon. The Journal of Philosophy. vol. 105, no. 9, 540-569. Williams, Bernard 1973 Deciding to Believe. In his Problems of the Self: Philosophical Papers 19561972. 136-151. London: Cambridge University Press.