1 Introduction

While we sometimes have direct control over our actions, we never enjoy such control over our beliefs. Take doxastic involuntarism as the view that we cannot form a belief as an intentional action. Involuntarism is virtual orthodoxy amongst epistemologists. In fact, it has seemed to many philosophers that involuntarism is not merely a contingent psychological fact Bennett (1990), but a conceptual feature of the nature of belief itself. A small sample of the philosophers who defend this view includes Williams (1970), Scott-Kakures (1994), Setiya (2008), and Hieronymi (2009). Despite these supporting arguments, some philosophers defend the view that involuntarism is not only contingent, but downright false. So take doxastic voluntarism as the view that we can exercise a degree control over at least some of our beliefs that is comparable to the control we have over any of our actions.

One strategy for defending doxastic voluntarism is to show that one might be able to believe at will in evidential situations in which more than one doxastic attitude is consistent with being fully rational. This strategy requires one to reject Uniqueness:

Uniqueness: For any body of evidence E, and proposition p, there is a unique doxastic attitude towards p that is consistent with being perfectly (epistemically) rational and having E as one’s total evidence.Footnote 1

Philosophers who deny Uniqueness endorse permissivism. Permissivists argue that, for at least some bodies of evidence, there can be more than one doxastic attitude that is rationally permissible for a subject to adopt conditional on their evidence. Permissivism seems to many like an attractive picture of rationality. It enjoys substantial intuitive and theoretical support (Schoenfield, 2014). Given the above definition, it seems plausible that permissivism can be recruited in support of doxastic voluntarism.

This strategic inference from permissivism to doxastic voluntarism raises a concern. If you’re like me, then you think favorably of permissivism but not of doxastic voluntarism. So, if you’re like me, you will likely be troubled to find that permissivism might lead to voluntarism. In what follows, I show that—on at least one recent construal of this inference—we needn’t worry. Permissivism is consistent with involuntarism after all.

The present paper examines an argument from permissivism to voluntarism recently advanced by Roeber (2019a, b). Roeber (2019b) defends what he calls the “argument from permissive situations.” Yet, despite what he says therein, the discussion of permissivism in Roeber (2019b) is not germane to the argument presented Roeber (2019a). My aim is to show that Roeber fails to distinguish between two different arguments for voluntarism, both of which which can be shown to fail by proper attention to different kinds of permissivism. Roeber considers a generic treatment of permissivism, but key premises in both arguments depend on different, more particular notions of permissivism. Attending to the distinction between single-agent and inter-subjective versions of permissivism reveals that the inference from permissivism to voluntarism is unwarranted.

2 The argument from permissive situations

Roeber frames his argument in terms of evidential fit. Some doxastic attitude A toward a proposition p fits a body of evidence E just in case E provides at least as much support for A as it provides for any alternative attitude. From this we can say that doxastic attitude A toward p uniquely fits E just in case E provides more support for A than any alternative attitude toward p, and A does not fit E just in case E provides more support for some alternative doxastic attitude toward p.

The relevant notion of support on this view takes the form of a “half normative” and “half psychological” compulsion (Roeber, 2019b, p. 2). This is cashed out in terms of the level of compulsion a cognitively healthy person who has E as her total evidence would feel towards doxastic attitude A if she were considering whether p. So, a doxastic attitude fits a body of evidence exactly when the evidence compels that attitude at least as strongly as any competing attitude. According to Roeber (2019b, p. 3):

Evidential Fit: Doxastic attitude A towards p fits E just in case a cognitively healthy human being who was considering the question whether p and who had E as her total evidence would feel a compulsion towards A that was at least as strong as any compulsion she felt towards any alternative doxastic attitude towards p.

Roeber then defines permissive situations in terms of this notion of evidential fit. Permissive situations are cases in which more than one doxastic attitude fits one’s total evidence equally well. Given his notion of fit above, we can say that

Permissive Situations: A cognitively healthy human being who has E as her total evidence is in a permissive situation with respect to p just in case she would feel an equally strong compulsion towards more than one competing doxastic attitude towards p, were she to consider the question whether p.

According to Roeber, if permissive situations of this kind are possible, then a cognitively healthy person can choose to believe at will. If more than one competing doxastic attitude toward p fits a body evidence E, then a cognitively healthy person considering whether p with E as her evidence would not feel more compelled towards one attitude rather than the other; she would feel equally compelled towards both. If this were the case, then she might be able to decide at will to adopt one attitude toward p over the other. Where there is no single-focused compulsion, there might be choice.

Uniqueness, however, says that it is always the case that some doxastic attitude A uniquely fits a body of evidence E. Uniqueness entails that permissive situations of this kind are impossible. Roeber’s strategy for defending doxastic voluntarism is to argue that permissive situations are possible and that it is therefore possible to believe at will. The argument from permissive situations can be stated as follows:

R1.:

Permissive situations are possible.

R2.:

If permissive situations are possible, then a cognitively healthy person can have as much control over some of her beliefs as she has over any of her actions.

\(\therefore \) R3.:

Doxastic involuntarism is false.Footnote 2

In what follows, I highlight a distinction between generic treatments of permissivism and single-agent accounts. Permissivism simpliciter merely requires that a given body of evidence fails to determine a unique rational attitude towards some proposition. This can be satisfied in cases where different agents rationally adopt different attitudes given the same body of evidence because they endorse different epistemic values and standards as truth-conducive. Single-agent accounts of permissivism demand something stronger. Single-agent accounts require that a given body of evidence fails to determine a unique attitude that a single agent can rationally adopt towards some proposition. While Roeber’s argument invokes generic language, R2 is plausible only if a single-agent account of permissivism is true. However, I argue that objections in favor of Uniqueness give us good reasons to reject single-agent accounts of permissivism in our interpretation of R1. But, modifying our account of permissivism to one that can countenance these objections will show that permissivism is consistent with involuntarism after all. Since R2 is plausible only if the single-agent version of permissivism is true, and since R1 is not plausible on that interpretation, it follows that R2 is true only if R1 is false (and vice versa). Therefore, the argument from permissive situations is unsound.

3 Permissivism

When we go about acquiring beliefs we typically look to our evidence for guidance. The permissivism/impermissivism debate concerns to what extent epistemic rationality is constrained by a given body of total evidence. Impermissivists think that, for a given body of evidence, there is a unique stance to take towards a proposition that is consistent with being fully rational. Permissivists think that our evidence sometimes grants more leeway. Philosophers inclined to endorse permissivism can opt to defend either of the following versions:

Doxastic Permissivism: For some body of evidence E, and proposition p, more than one doxastic attitude towards p is rationally permissible.

Credal Permissivism: For some body of evidence E, and proposition p, more than one credence in p is rationally permissible.

While both doxastic and credal variations imply that permissive situations are possible, it is important to note that these are independent theses. The exact relationship between doxastic and credal permissivism depends on the relationship between credences and beliefs. Since this relationship is a matter of contention, I will remain agnostic on it here. For the sake of argument, I will assume that it is possible to accept one while rejecting the other. This is significant since one must at least endorse doxastic permissivism for permissivism to be of service to doxastic voluntarism. This is because credal permissivism says that more than one credence in a proposition can be rational given a single body of evidence. Such wiggle room in rational credence is not sufficient to show that more than one doxastic attitude is rational, since doxastic attitudes might be overdetermined by credence value.

At this point, the invocation of “doxastic attitude” in doxastic permissivism requires a bit of clarification. By this, I just mean belief, disbelief, or suspension of judgment. So doxastic permissivism simply says that some bodies of evidence make adopting more than one of these competing attitudes towards a proposition equally rational. This tripartite notion of doxastic attitude brings out the need to distinguish between strong and moderate versions of doxastic permissivism:Footnote 3

Strong Doxastic Permissivism: For some body of evidence E, and proposition p, it is rationally permissible to believe p, but it is also rationally permissible to believe \(\sim {p}\).

Moderate Doxastic Permissivism: For some body of evidence E, and proposition p, it is rationally permissible to believe p, but it is also rationally permissible to suspend judgment on p.

It is not necessary for doxastic voluntarists to appeal to strong doxastic permissivism. Suppose you find yourself in a situation where you have rationally suspended judgment on a proposition p, but where belief that p is equally rational. It is easy to see how you might reasonably think that you could directly choose to believe p, since your evidence does not point conclusively towards belief or suspension. This is exactly what Roeber has in mind when he purports to adopt doxastic permissivism in support of R2 (cf. Roeber, 2019b, pp. 4–8). Thus, for convenience, I will use doxastic permissivism interchangeably with moderate doxastic permissivism, unless otherwise specified.

Yet, none of the variants of permissivism discussed above entail the truth of R2. There remains a further distinction to which Roeber does not attend. So far, I have discussed various subsets of a single, generic permissivist doctrine—one that will be of little help to doxastic voluntarists. Roeber implicitly adopts a stronger doctrine that I will call single-agent doxastic permissivism. Single-agent doxastic permissivism (or, SDP for short) is the denial of Single-Agent Uniqueness.

Single-Agent Uniqueness: For any body of evidence E, and proposition p, there is a unique doxastic attitude towards p that a single agent can take that is consistent with being perfectly (epistemically) rational and having E as one’s total evidence.

A bit more precisely, SDP means that for some body of evidence E, and proposition p, it is rationally permissible for a single agent S to believe p, but it is also rationally permissible for S to suspend judgment on p. Again, I will use SDP interchangeably with this slightly tempered “moderate” SDP, unless otherwise specified.

It should be clear that Roeber must deny Single-Agent Uniqueness in order to support R2. If permissive situations are to support voluntarism it needs to be the case that more than one doxastic attitude is rationally permissible for a single rational agent. However, permissivism can be true when permissive situations occur only inter-subjectively. Formulated generically, permissivism only says that a given body of evidence does not determine a unique rational attitude. This thesis is satisfied when, given identical evidence, two different agents with the same total evidence can rationally hold different views. Call this latter formulation inter-subjective doxastic permissivism, or simply inter-subjective permissivism. Inter-subjective permissivism is insufficient to show that R2 is true. So, Roeber must defend the stronger SDP. The plausibility of R2 depends on the plausibility of SDP. Likewise, the plausibility of Roeber’s argument from permissive situations depends on the plausibility of R1 interpreted as an endorsement of SDP.

4 Roeber’s defense of R1

Matheson (2011), White (2005, 2014), and others purport to defend Uniqueness. However, Kelly (2014), Schoenfield (2014, 2019), and Meacham (2014) all authors who defend permissivism rightly point out that arguments due to White and others merely defend Single-Agent Uniqueness, leaving their preferred brand of inter-subjective permissivism unscathed. This leaves prospects for SDP somewhat lacking. Nonetheless, Roeber himself thinks that all of the challenges to SDP can be met. I am not convinced that he is correct.

Take for instance what Roeber calls White’s “toggling argument” (2019b, pp. 8–9). According to White, we should reject permissivism because it entails the possibility of situations where you can arbitrarily switch back and forth between believing p and believing \(\sim {p}\) without any change in your evidence. The target of White’s argument is the strong version of single-agent doxastic permissivism mentioned above, but it can easily be co-opted to reject the more moderate SDP. The idea would be that, since belief in p and suspension of judgment on p fit your evidence equally well, you can rationally switch back and forth between believing p and suspending on p without any change in your evidence. Since this kind of arbitrary switching could not possibly be rational, SDP must be false.

We can sharpen this idea a bit with an example. Consider a true/false test where you are rewarded for each correct answer, punished for incorrect answers, and neither gain nor lose anything of value for each question left blank. For one question in particular, you find yourself in a permissive situation. While your total evidence rules out marking ‘F,’ your evidence supports equally well either marking ‘T’ or skipping the question. Suppose you decide to mark the answer ‘T’. Given that the situation is permissive, what is stopping you from erasing your answer? Either response would be rational given your evidence. If you do decide to erase your answer, then why not re-check the box marked ‘T’ again? If SDP is true, then you could rationally repeat this process ad infinitum without any change in your evidence. But, this kind of behavior would be positively irrational (cf. Roeber, 2019b, pp. 6–8). Evidently, we should rule against SDP.

But Roeber seems to think that this toggling argument does not in fact threaten SDP. Here is his reply:

The reason is simply that, on the stipulative definition of ‘fit’ that we have been working with throughout, there’s no reason to think you can rationally flip-flop between believing p and suspending on p, without any change in your evidence, if belief in p and suspension of judgment with respect to p both fit your total evidence. Consider the true/false test from §3. Your total evidence rules out checking the box marked ‘F’ while leaving you torn between skipping the question and checking the box marked ‘T’, but it doesn’t follow that you can rationally check the box marked ‘T’ and then erase your answer and then re-check the box marked ‘T’ and then re-erase your answer, round and round, without any change in your evidence. That behavior would be nuts, even though your total evidence supports skipping the question and checking the box marked ‘T’ equally well. (Roeber, 2019b, pp. 7–8)

This response seems unsatisfactory. Given the notion of fit outlined above, if SDP were true then both skipping the question and marking ‘T’ would fit your evidence equally well. If this were the case, then your net compulsion would neither favor marking ‘T’ nor skipping the question. In such a situation it would be epistemically rational to switch between marking ‘T’ and skipping the question without any change in your evidence. Importantly, even if you decided arbitrarily to mark ‘T’, you would still not feel any less compelled to erase your answer and skip the question. After all, your evidence still fits both decisions equally well! Nothing about deciding to mark ‘T’ changes your evidential situation. Since your evidence remains exactly the same there can be no corresponding change in your net compulsion. Roeber still has not told us why changing your answer would be epistemically irrational even though your evidence supports skipping the question and marking ‘T’ equally well and your healthy compulsion remains equally distributed over the two options.

It seems that the involvement of material reward in the case above skews our intuitions. Even if permissive situations of the kind Roeber has in mind are possible, it would seem practically irrational to repeatedly switch your answer ad infinitum without any change in you evidence. At some point you have to just cut your losses and pick one. Failure to do so would amount to forfeiting any further rewards you might accrue by answering subsequent questions correctly. As Tucker (2016, p. 4) suggests, cases like these militate against the tight connection between the quality of one’s options and the quality of one’s choice. Your evidence dictates that neither option available is better than the other, but—assuming that you are aiming at the most good you can get—the best way of pursuing your ends might be to choose to mark an answer arbitrarily. So, let’s say you arbitrarily mark ‘T’. This would be a pragmatic choice made on the basis of your aiming at the most reward possible, and even might be a rational choice. But, it surely cannot be an expression of your sincere belief that ‘T’ is correct. After all, your evidence still compels you just as strongly towards suspension of judgement as towards belief.

But the example in question is just a metaphoric one. Consider instead a case where you have suspended judgment with respect to p but believing p would be equally rational. Suppose I then offer you a magic, belief-inducing pill that will cause you to form the belief that p. Why shouldn’t you then just take the pill? Your response cannot be that you are not sure that p is true. Your evidence supports believing p just as well as suspending on p. It seems perfectly consistent with SDP that you could rationally pop as many magic pills as you want, toggling back and forth between belief and suspension as you please. Since we have already ruled that this would be epistemically irrational, it seems like we should reject SDP.

If this is right, then Roeber has not given us adequate reasons to think that SDP is safe from White’s toggling argument. In general, it seems likely that Single-Agent Uniqueness is true. So, if Roeber hopes to motivate premise R1, then he must appeal to a version of permissivism that is safe from arguments in favor of Single-Agent Uniqueness.

5 Roeber’s defense of R2 (or, more exactly, R2*)

After presenting the argument from permissive situations, and just before defending R1 at length, Roeber (2019b) remarks that R2 “requires a paper of its own” and notes: “I defend [R2] in Roeber (2019a).” Yet this is not the case. In that paper, Roeber defends a different premise altogether. What he says there is that “there are possible cases where a person thinks she is in a permissive situation with respect to p and she neither feels compelled to believe p nor feels compelled to suspend judgment with respect to p” (Roeber, 2019a, p. 839). Roeber calls this kind of situation, distinct from permissive situations, an equipollent case. He then goes on to argue that that if an agent finds herself in an equipollent case she can decide to believe at will.

Yet, being in an equipollent case does not entail that one is in a permissive situation, since thinking that one is in a certain situation does not entail that one is truly in it. Moreover, the possible truth of R2*, defended in Roeber (2019a), seems irrellevant to the possible truth of Roeber’s (2019b) original R2:

R2*::

If equipollent cases are possible, then a cognitively healthy person can have as much control over some of her beliefs as she has over any of her actions.

This is because permissive situations, as we saw earlier, involve equally strong compulsion towards more than one competing doxastic attitude towards p, while equipollent cases, as we have just seen, involve not feeling a compulsion either way. What kind of control we might have in one of these situations has no bearing on what kind of control we might have in the other. So Roeber’s shift from a discussion of the relation between permissive situations and doxastic voluntarism to a discussion of the relation between equipollent cases and doxastic voluntarism, despite his claim that they are connected, undermines his alleged defense of R2.

What we have, instead, is a separate argument, which we can call the argument from equipollent cases:

R1*.:

Equipollent cases are possible.

R2*.:

If equipollent cases are possible, then a cognitively healthy person can have as much control over some of her beliefs as she has over any of her actions.

\(\therefore \) R3. :

So doxastic involuntarism is false.

For the purposes of dislodging the orthodoxy of involuntarism, of course, any sound argument is a welcome recruit. So we can put aside the insufficient support for the premises in the argument from permissive situations. Perhaps the argument from equipollent cases will do the trick.

Roeber’s idea here can be traced back to William James. James (1897) famously argued that reason aims at a balance of two competing goals: acquiring true beliefs and avoiding false beliefs. Believing everything indiscriminately achieves the first goal, but not the second. Believing nothing achieves the second goal at the expense of the first. Presumably, we want to be somewhere in the middle; we should try our best to maximize our true beliefs while simultaneously minimizing false beliefs.

So, suppose you find yourself in an equipollent case. You strongly think that your evidential situation is permissive and that your evidence supports belief and suspension of judgment equally well. If you are epistemically courageous—that is, you weigh the importance of believing truth over that of avoiding falsehoods—then you can decide to believe at will, or so the thought goes.

For example, consider the fictional case of Jones’s murder trial. Having heard all of the evidence you feel neither compelled to believe that Jones committed the murder, nor do you feel compelled to continue suspending judgment. You are in an equipollent case. Roeber says:

By my own lights, I’m rationally permitted to pursue the acquisition of true belief by believing that Jones did it, and I’m also rationally permitted to safeguard against the acquisition of false belief by continuing to suspend judgment on this proposition. If I were intellectually cautious, I would pick the latter option. Since I’m intellectually courageous, I pick the former and then carry out my intention to believe that Jones did it by affirming that Jones did it—by consciously thinking to myself ‘Okay, Jones did it.’ (2019a, pp. 847–848)

It is hard to see why Roeber thinks this is so. Even unconstrained by rationality, it does not follow from the fact that I think that I am permitted to either believe that “Jones did it” or suspend judgment that I am capable of deciding at will to believe that “Jones did it.”Footnote 4 This much becomes clear when we consider versions of permissivism that are compatible with Single-Agent Uniqueness.

I have argued above that we have good reason to think that Single-Agent Uniqueness is true. Kelly (2014), Schoenfield (2014, 2019), and Meacham (2014) all defend versions of permissivism that are not threatened by the arguments in favor of Single-Agent Uniqueness. Of these accounts, I take Schoenfield (2014, 2019) to be the best representative of permissivism. Schoenfield largely defends a credal version of permissivism, but her view is amenable to doxastic permissivism as well.

On her account, an agent’s “cognitive system” is the system by which she responds to evidential inputs by forming doxastic attitudes. An agent endorses some set of cognitive properties that she deems truth conducive. Such properties might include forming beliefs on the basis of reliable perceptual evidence, trusting the testimony of experts, etc. To be more precise:

An agent endorses a set of cognitive properties, C, if she prefers, when her only goal is accuracy, any cognitive system that instantiates all of the properties in C, to a cognitive system that lacks some of these properties. (Schoenfield, 2019, p. 288, emphasis hers)

We can cash this out in terms of an inter-subjective account of permissivism. There is a sense in which one prefers their own set of epistemic standards and endorses them as what is most truth conducive, yet at the same time recognizes that other agents’ alternative standards might be equally rational. Rationality is constrained by general principles regarding what kinds of considerations count as reasons in support of what kinds of hypotheses. Nevertheless, given the various admissible sets of epistemic standards, these principles fail to always determine a uniquely rational doxastic attitude (cf. Schoenfield, 2014, p. 202). This is because, to judge that an agent S is rational in responding to a total body of evidence E by adopting a doxastic attitude A is to judge that adopting A in response to E is consistent with the instantiation of the set of cognitive properties C that S endorses (cf. Schoenfield, 2019, pp. 299–9).

This account of permissivism is consistent with involuntarism. Different subjects might instantiate qualitatively different systems for mapping evidential inputs to doxastic outputs. This does not entail that there is more than one rational output for a given set of cognitive properties C and body of evidence E. Permissive situations are possible, but only inter-subjectively. This means that the possibility of permissive situations does not entail that more than one doxastic attitude is rational for a single agent given the same total body of evidence. With this account in hand we can see that equipollent cases and Jamesian goals do not imply the denial of involuntarism, as Roeber seems to suggest.

Instead, cases like Jones’s murder trial can be resolved in one of two ways. Either Jamesian goals factor into your belief forming practices as context-sensitive features of your cognitive system, or they give you pragmatic reasons that factor into your practical deliberation. If Jamesian goals factor only as pragmatic considerations, then they cannot bear on the question of whether to believe that p. Belief that p can only be settled by alethic considerations–by reasons that bear on the truth of p. I have already argued above that one doxastic attitude uniquely fits a single agent’s available evidence, but even if you thought that this was not the case you would nevertheless fail to sincerely believe at will on the basis of reasons you do not take to settle the truth of whether p. Even if your estimate of your evidential situation leads you to feel no more compelled towards belief that p than suspension of judgement, it does not follow that you can make yourself believe on the basis of anything other than considerations that bear on whether p is true. So, if Jamesian goals are to bear on your beliefs at all, then they must do so as features of your cognitive system.

On an inter-subjective account of permissivism, the weighting of Jamesian goals can be thought of as one of the various properties of your cognitive system. As such, Jamesian goals factor into your belief acquisition only as properties that inform what you are inclined to think is true in a given evidential situation when your only goal is accuracy. This does not entail that you can choose to weigh one goal more than the other on a case by case basis. Merely, you find yourself with a context-sensitive set of epistemic standards which you deem to be truth conducive, and this might include the property of being more or less epistemically courageous when you find yourself in equipollent cases. As such, your belief that Jones is the murderer was formed automatically. Given that your epistemic standards for a given context include the property of generally being epistemically courageous in situations of that kind, the evidence simply compels you slightly more towards belief than towards suspension of judgment. The formation of such a belief is no more under your control than in situations when you find your evidence decisive.

6 Conclusion

I have argued that Roeber’s argument from permissive situations trades on a generic interpretation of permissivism. I take myself to have shown that R1 and R2 are incommensurable. R1 is plausible only on an inter-subjective account of permissivism, while R2 is plausible only on a single-agent account. Moreover, even if an agent thinks her evidence is permissive, it does not follow that she can decide to believe at will. Therefore, the argument from permissive situations is unsound.