“Believing at will is possible”−or is it? Some remarks on Peels’s “truth depends on belief” cases and voluntariness

Abstract This article discusses Rik Peels's response to Williams's argument against voluntary belief. Williams argues that voluntary beliefs must be acquired independently of truth-considerations, so they cannot count as beliefs after all, since beliefs aim at truth. Peels attempted to reply by showing that in cases of self-fulfilling beliefs, a belief can indeed be voluntarily acquired in conditions which retain the necessary truth-orientation. But even if we make two crucial concessions to Peels’s proposal, his argument ultimately fails. The first concession is that beliefs can be weakly voluntary-namely, we can acquire them at will though we do not preserve them at will but on the basis of evidence. Conceding this, however, only lands us in the “acquisition problem”: how a belief can be acquired qua belief when we still do not think we have justification for it. This leads us to the second concession: that knowing in advance that a certain belief is self-fulfilling provides us with such a justification. However, this concession lands us in the ultimate obstacle: that, precisely because such a justification is available both before and at the moment of forming the belief, the cognitive perspective of the subject is identical at both moments, which obscures what it even means to say that at a certain moment she started to have a belief.


Introduction and outline
In this article we discuss Rik Peels's reply to the classical arguments about the alleged conceptual impossibility of believing at will. We will try to show that such a reply fails even if we concede two crucial moves. First, contra Williams, Peels assumes that voluntary beliefs do not need to be voluntarily preserved, only voluntarily acquired. Second, Peels considers that, in scenarios involving self-fulfilling beliefs, simply knowing that whatever such proposition that we choose to believe will turn out to be true gives us, by itself, enough epistemic justification to believe it. We will try to argue that his proposal fails in regard to the problem of determining what it is, exactly, to believe a proposition p at the same time at which we know that we have no better evidence supporting p than not-p.
Our text will be organized in the following way: Believing at will is possible 3 Manuscrito -Rev. Int. Fil. Campinas, v. 46, n. 2, pp. 1-39, Apr.-Jun. 2023. -In section 2, we will offer a historical reconstruction of the problem as it took place before the 20 th century, in particular with reference to the psychological doxastic involuntarism we find in Pascal and in Hume.
-In section 3, taking Bernard Williams's argument as a turning point, we will enter the discussion whether there is a conceptual impossibility of believing at will, stemming from the fact that we could not view as a belief a state which we can hold vis-à-vis any preferred proposition irrespective of considerations concerning its truth.
-In section 4, we will introduce a first concession to Peels's proposal in order to show that, even after such a concession, problems remain. We begin by presenting Barbara Winters's proposal, according to which we may distinguish between the conditions for the acquisition and for the preservation of a belief. As Winters points out, a subject might initially acquire a belief in a non-truth-oriented way, but later preserve it on the basis of (what she considers to be) evidence. Thus, contra Williams, beliefs could be voluntary in a weak sense. The proposal of Mark Johnston, which shows us how the evidence can be systematically connected with a belief previously acquired at will, is a further step in the solution of the "preservation problem". We will then try to show that in spite of this solution, the "acquisition problem" remains unsolved-namely, how a subject can acquire a belief for which she knows she has no justification. This will lead us to introduce Peels's proposal and ask whether it can deal with the acquisition problem in a more successful way.
-In section 5, we will show that a key element which distinguishes Peels's proposal (following Velleman's) from Winters's and Johnston's is that the subject S knows that the belief which she will voluntarily acquire is self-fulfilling, and that this knowledge provides S with sufficient justification for choosing the belief in question and not only retrospectively but also in advance. This proposal was met with criticism by Gregory Antill, who challenged the justificatory weight which this prospect of self-fulfillment can have. According to Antill, for a subject to be justified both for acquiring p and for acquiring not-p, she should have evidence that both p and not-p are true, which is impossible. On the contrary, our strategy will be to point out that even if we grant Peels that anticipatory knowledge of the self-fulfilling character of a belief provides S with justification for acquiring the belief (which is our second key concession), S's possession of such a justification for her belief turns out to be a poisoned chalice. We will argue that precisely because S already has at t0, before forming the belief that p, exactly the same justification for p which she will have when forming the belief at t1 (and which does not exclude the possibility that not-p is the case), it is not clear what it even means to claim that S has at t1 a belief which she does not have at t0. In other words: if the cognitive perspectives of a subject in two different moments are identical, there is no reason to claim that the beliefs the subject has at one moment are different from those she has at the other.

Historical framework of the problem: Pascal and Hume
In his posthumously published book Pensées (1670/2008), Blaise Pascal postulated the philosophical argument known as "Pascal's wager" in which he argued about why it is practical for human beings to believe in the existence of God. Aware of the possibility that there are people who do not believe in God but would want to do so in view of the convenience it would represent, he proposed a solution that involved emulating those who do believe. He encouraged the non-believers to behave as if they believed in order to, eventually, come to believe: Follow the way by which they began: by behaving just as if they believed, taking holy water, having masses said, etc. That will make you believe quite naturally, and according to your animal reactions (Pascal et al., 1670(Pascal et al., /2008). Now, why did Pascal appeal to this indirect recourse? Why did he not simply suggest that non-believers choose to believe in God? Despite not having developed a theory of belief, Pascal in this passage posed a question that would continue to be debated for centuries: can we believe at will?
Pascal adhered to the tenet that we cannot directly choose what to believe, and that consequently we need to acquire beliefs via a detour. However, he did not approach to the question of why it is impossible to directly choose what to believe. The debate about the (im)possibility of believing at will and whether this (im)possibility is psychological or conceptual was more directly addressed in David Hume's early proposals on the concept of belief (1732/1960, 1777/1975). For Hume our inability to believe at will is purely psychological. As we dive into the texts to elucidate Hume's reasons for arguing that it is impossible to believe at will, we find: Nature, by an absolute and uncontrollable necessity has determined us to judge as well as to breathe and feel; nor can we any more forbear viewing certain objects in a stronger and fuller light, upon account of their customary connection with a present impression, than we can hinder ourselves from thinking as long as we are awake, or seeing the surrounding bodies, when we turn our eyes towards them in broad sunshine (Hume, 1732(Hume, /1960.

Williams's "classical" argument
More than 200 years later, in what would become a foundational text of the current debates on doxastic voluntarism, Bernard Williams (1970) advanced an argument against the conceptual possibility of voluntary belief. He presented it in just a few lines: Why is this? One reason is connected with the characteristic of beliefs that they aim at truth. If I could acquire a belief at will, I could acquire it whether it was true or not; moreover I would know that I could acquire it whether it was true or not. If in full consciousness I could will to acquire a 'belief' irrespective of its truth, it is unclear that before the event I could seriously think of it as a belief, i.e. as something purporting to represent reality. At the very least, there must be a restriction on what is the case after the event; since I could not then, in full consciousness, regard this as a belief of mine, i.e. something I take to be true, and also know that I acquired it at will. With regard to no belief could I know -or, if all this is to be done in full consciousness, even suspect -that I had acquired it at will. But if I can acquire beliefs at will, I must know that I am able to do this; and could I know that I was capable of this feat, if with regard to every feat of this kind which I had performed I necessarily had to believe that it had not taken place? (Williams, 1970, p. 148).
Since he thinks -and this is his key point-a belief must be truth-oriented, doxastic voluntarism faces a dilemma: either a person who voluntarily acquires the belief that p needs to forget that she has acquired the belief in this manner, in which case the allegedly voluntary activity of believing becomes weirdly dissimilar from every other voluntary act, or she bears in mind that she has acquired the belief that p in a voluntary manner (thus irrespective of truth considerations), in which case the state of "belief" she finds herself in cannot actually be described as having a belief: she will know that her "belief" does not "purport to represent reality". Now, the argument seems to hinge on a relationship, which we need to handle with care, between a subject's current state of (alleged) "belief" that p (which requires that, for this subject, p appears as a true proposition) and the subject's immediately previous process of acquiring the (alleged) "belief" that p. The point seems to be that if the subject knows that this state of "belief" was acquired in a voluntary way, then she will know that the process was not aimed at selecting true propositions, and then she will be suspicious of the truth-value of p itself, and therefore will not be able to take herself to actually believe that p. His argument needs to take the following (not very convincing) form: 1) If I know that (in the past) I formed the belief that p in a voluntary way, then I know that (now) I do not have good evidence for believing p is true.
2) I know that (in the past) I formed the belief that p in a voluntary way. Therefore 3) I know that (now) I do not have good evidence for believing p is true. But 4) If I know that I do not have good evidence for believing p is true, then, ipso facto, I do not believe that p. Therefore 5) I do not believe that p This seems to be, in fact, the only way to account for the fact that Williams is convinced that a subject unaware of how she formed her "beliefs" would find herself in a better position to view them as, precisely, beliefs. His point seems to be that I cannot take myself to believe that p if I have serious doubts concerning the truth-value of p itself, and that I cannot help to have such serious doubts insofar as I am well aware of the "lowly origins" of my propositional attitude (my "belief") towards p.

The first concession: voluntarily acquiring but not voluntarily preserving a belief
Now, to make sense of Williams's argument, we have already needed to take into account that he seems to assume that the voluntary acquisition of a belief requires also the voluntary preservation of that belief. This would liken voluntary beliefs with what we might call "strongly voluntary states". In this sense, a state is "strongly voluntary" insofar as we not only enter it by an act of our will, but our remaining in that state requires a continuous decision to do so. A "weakly voluntary state", on the contrary, would be one in which we enter by an act of the will, but in which we remain irrespective of anything we may subsequently decide. A simple example is the following: in a country which has the institution of divorce, entering the state "being married" makes it a strongly voluntary state (I not only acquire the status "being married" voluntarily but also remain in it voluntarily), whereas, in a country which lacks that institution, entering the state "being married" makes it a weakly voluntary state (once I acquire the state "being married" voluntarily, I have no choice but to remain in it). In the case of belief, what we have is that, if states of belief are "strongly voluntary", then, once we have voluntarily acquired a belief that p (and, in turn, this involves having acquired it irrespective of any truth considerations), we must be able to retain or relinquish the belief in question in an equally voluntary way. If we stick to the belief that p, it will be because we are continuously choosing to do so, not because we perceive that reality, so to say, imposes itself on us appearing as if p were the case. When Williams argues that we cannot take ourselves to have chosen a belief, that we need to have somehow forgotten that choice, or else the state we find ourselves in would not be experienced by us as one of belief (as "purporting to represent reality"), his point seems to be that belief states, if voluntary, are strongly voluntary. But this point could be contested: we can imagine belief states as "weakly voluntary"; that is, states which we can acquire by an act of the will but such that their preservation is not dependent on anything we choose−that is, once having chosen the belief that p, we will continue to believe that p because reality appears to us as if p were the case. In this weak construal of voluntariness, our decision to continue to believe is neither necessary not sufficient for the preservation of the belief, and this is the kind of voluntariness which we can reconcile with the indispensable role of evidentiary considerations 1 . This would involve less voluntariness than is at stake in the stronglyvoluntary scenario but would be voluntary anyway. As we anticipated, this distinction will turn out to be crucial for criticisms of Williams's proposal.

Winters's precedent
This point is central in Winters's criticism of Williams's argument: "a belief can be sustained by considerations other than those which brought it into being". For example, she goes on, though she "may have first learned the location of the Louvre from a grade schoolteacher", she may since have been to Paris and observed it herself. Consequently, her belief now rests on her "personal observations and not on the teacher's authority". She goes on to apply the distinction to the case of voluntarily acquired beliefs: "it may be possible for me to think I still hold a belief originally acquired independently of truth considerations, if I think that I now believe it for different, truth-related reasons" (Winters, 1979, p. 246). Winters appeals to an example of a subject who is asked to decide to believe if "a gentleman or a hungry tiger waits in the next room", and a moment later decides to enter the room empty-handed instead of with a loaded gun. "We might conclude", Winter comments, "that she does believe that a gentleman is in the room. But we can take her to believe that she has this belief only by assuming that she is taking something as grounds for its truth, e.g., the lack of growling from behind the door, or that she believes that she once had such evidence and has forgotten it […]. Only through such suppositions can we think she regards herself as having succeeded in believing that there is a gentleman in the room" (Winters, 1979, p. 250). In other words, the subject "might arrive at this belief at will, but she cannot believe that she believes the gentleman is in the room while simultaneously believing that her belief is sustained at will" (Winters, 1979, p. 253). Belief can thus be weakly voluntary: after the acquisition of a belief that p, our will, so to say, "passes the torch" to evidential considerations on which we will hinge in order to preserve the belief 2 .
2 Something analogous can be said about Bennett's counterexample about the "Credamites" community: a member of the community can indeed will himself to believe a proposition and typically later forgets that he has acquired the belief in this way. Now, if he later happens to accept "that he has willed himself to have the belief" he will usually lose the belief "because of his knowledge of how he got it". But the nuance is crucial: "Usually -but sometimes he retains the belief because since voluntarily acquiring it he has encountered evidence for it" (Bennett, 1990, p. 93). So: the point still holds that no one, even a Credamite, can retain a belief that p without having come to believe that there is evidence supporting p. But this dissociation between the conditions of acquisition and those of sustainment certainly point beyond what Williams had explicitly considered. Analogously, Scott-Kakures introduces a similar distinction between conditions for the acquisition and for the preservation of a belief when he speaks of "anomalous belief transitions": "At time t an individual does not believe that p; indeed he may think that the evidence against p is impressive. But at t+1 the individual does believe that p; and he has come to believe that p without first having or coming to have other beliefs which provide reason for believing that p. […] What is important about these cases is that, though the belief state is caused, there are no causes which are also reasons for coming to believe that p. But, it is equally important to recognize that, though such belief state transitions are not accomplished by first coming to have reason for believing that p, it might, nonetheless, be the case that once the belief So, summing up, the scenario after Winters's remarks is the following: as was already the case in Williams's proposal, a voluntary belief has to be acquired irrespective of truth considerations but, at the same time, beliefs are defined as aiming at truth; so, this mismatch between the form of acquiring a belief and the truth-orientedness of beliefs in general is solved by a dissociation between how beliefs are acquired and how they are preserved: beliefs are only "weakly voluntary"; they can be acquired at will but need to be preserved on a non-voluntary basis; therefore, voluntarily acquired beliefs are acquired by means of a basic act but do not need to be acquired as truth-oriented; their truth-orientedness is acquired later, "outside" the basic act of their acquisition.

The preservation problem in Winters and Johnston
Now, although Winters mentions the general idea that a subject could somehow obtain justification which lets her preserve the belief which she previously acquired irrespective of truth considerations (and, as we saw, the person in the "tiger vs. man" example is supposed to preserve her belief by "taking something as grounds for its truth, e.g., the lack of growling from behind the door, or that she believes that she once had such evidence and has forgotten it"), Winters is produced it is had for reasons" (Scott-Kakures, 1994, pp. 80-81. Emphasis in the original).
does not provide us with a systematic, non-contingent, connection between the initial belief and the grounds which (may) appear later. This aspect of her proposal can be regarded as solved when an element of self-fulfillment enters the scene: the beliefs which we choose may be such that their mere occurrence would render them true (or very probably true). This element appears in a proposal by Mark Johnston, who, like Winters, attempts to solve Williams's concerns about truth-orientedness. Johnston writes: If someone offers me a million dollars if I can get myself to believe that I will be a millionaire, and I succeed and am about to be given the money, then although I know I acquired the belief at will, I have come to possess sufficient evidence in favor of its truth. I had evidence that if I acquired the belief at will then it would be a true belief. By acquiring the belief at will I simultaneously come to have evidence of its truth. This sort of complication can arise in any case in which acquiring a belief makes it more likely that the believed proposition is true (Johnston, 1988, p. 68. Emphasis ours).
Note that this kind of scenario only refers to what would count as justification once we have already formed the belief that we will receive the money. By appealing to self-fulfilling beliefs, Johnston takes a step towards presenting in more concrete terms Winters's point that a belief acquired without justification might later be preserved by it-in this case, justification provided by the existence of the belief itself. But neither Winters nor Johnston show us how the belief in question could be acquired qua belief if, at the moment of its alleged acquisition, we know that we lack justification for it. Johnston even seems to emphasize the difficulty in this step, by the wording "if I can get myself to believe"-which suggests that this possibility is not obvious. And the obstacle does not seem to be merely psychological but, all over again, the conceptual problem of truth-orientedness of anything we might consider a belief. Let us now turn to this.

A bad analogy and the perseverance of the acquisition problem
It is important, at this point, to return to Winters's proposal and argue that there is an important difference between the Louvre case and the "gentleman or tiger" case: the different kinds of considerations at stake in the first example (teacher's authority or personal perceptual experience) were both, from the cognitive perspective of the subject, truth-oriented, and none of them was, therefore, incompatible with the idea that the propositional attitude acquired on their basis was a belief. In the second example, on the contrary, what is at stake is the idea that a belief is acquired in such a way that the subject does not view as truth-oriented, and, whereas this problem may be solved at the moment of the belief preservation, it is not clear how this could be a belief preservation, i.e., how the original act of acquisition could be seen as the acquisition of a belief. Therefore, in Winters's reconstruction, a belief voluntarily acquired by means of a "direct" or "basic" act is not, strictly speaking, acquired as a belief, because, at the moment of its acquisition, it is not truth-oriented. The subject can only later, in a reflective approach, take herself to have a belief, insofar as she has preserved the "belief" in question by means of considerations which are, now, of evidentiary character.
Mutatis mutandis, the "acquisition problem" we are now considering converges with the remarks by Scott-Kakures, according to which we could not voluntarily acquire a belief that p because that would involve that, at a first moment, we view it as an epistemically unjustified proposition, and, at a second moment, we believe the opposite. This, claims Scott-Kakures, means that in order to carry out our plan to believe that p we need to have abandoned the cognitive perspective from which we conceived the plan−in roughly the same way in which, in order to carry out the plan to forget a proposition p, we need not to keep the plan in mind when we are executing it (Scott-Kakures, 1994, p. 96). And if we need not to bear in mind our plan when we carry it out (if there has to be a "cognitive fissure" between our cognitive perspective when we entertain the plan and our cognitive perspective when we carry it out), then we are not actually following a plan−i.e., we are not doing something voluntary. In the context of the present discussion, the point is: how can we say that we are voluntarily acquiring a belief at a moment in which we know we are not acting in a truth-oriented way?
Now, all of this would not constitute a difficulty if voluntary acquisition of beliefs were not viewed by Winters as a basic or direct act: if believing at will meant that, at a moment t1, we do something as a consequence of which at a later moment, t2, we come to hold a belief, then it would not be a problem if only at t2 we thought we had justification for what we believe. The obstacle emerges from the fact that, right from the beginning, performing the decision to believe must involve being in a state of belief 3 . So, to sum up: Winters and Johnston attack the step (1) in our reconstruction of Williams's argument by questioning the assumption that, if a belief was acquired at will, then its preservation must also be at will. They therefore take voluntary beliefs to be so in a weak, not a strong, sense (in the case of Johnston, in a very weak sense, because our choice is from the beginning restrained by the need that the voluntarily chosen belief be self-fulfilling). Now, even if we concede that a belief can be voluntary only in a weak sense, their proposal seems to fail because they cannot explain how a subject might voluntarily acquire a belief without thinking she has justification for it. Now, as we will immediately see, Rik Peels also objects to step (1) and, to do so, he resorts to a thought experiment. Does his proposal founder at the same obstacle as Winters's and Johnston's? In order to answer this question, let us now turn to his experiment.

Peels's thought experiment
Peels introduces a thought experiment (Peels, 2015, p. 6) in which (a) we can freely choose one between two incompatible beliefs and (b) either of these two beliefs, once chosen, will end up proving true, since they are self-fulfilling beliefs. More concretely, in Peels's thought experiment, a medical doctor, Dr. Transparent, promises (in a trustworthy way) to pay 10 dollars to anyone who, having come to his laboratory, has, precisely, the belief that he will pay her 10 dollars. Given that Dr. Transparent has devised a sophisticated apparatus for mind-reading, whose effectivity has survived the most demanding analyses by peers, we accept that he actually can read our minds and that, if we are in fact able to form the belief in question, then that very belief will come true because Dr. Transparent will "read" it and, accordingly, pay us the promised sum (Peels, 2015, p. 6). If the example appears to us too outlandish, Peels invites us to replace it with the supposition that we have a trusting friend that will pay us the sum in question if we sincerely tell her that we have formed the required belief (Peels, 2015, p. 9). In any case, the story goes on, a person meets Dr. Transparent, deliberately forms the belief "This person will give me $10", and, sure enough, such belief is detected and this leads the subject of the experiment to receive the 10 dollars.
The crucial point is that the subject of the experiment knows all of this: as soon as she forms the belief "Dr. Transparent will give me $10", she knows that Dr. Transparent has been able to "read" her belief and, consequently, will pay her the promised sum. Therefore, she immediately acquires evidence supporting the belief she has just formed. Consequently, and against the claim made by Williams (among others) that an attitude that we form in this fashion could not be perceived by ourselves as a belief, as "something purporting to represent reality", Peels agrees that "by 'belief'" we must "mean something like being convinced that a particular proposition p is true or that p is the case" (Peels, 2015, p. 16), but insists that, in the case of Dr. Transparent, "my resulting state of mind is truly a belief: I am convinced that it is true or that it is the case that I will receive $10" (Peels, 2015, p. 7. Emphasis in the original).
If this is so, then, Peels's "truth depends on belief" (TDB) case seems to follow a timeline like this: at a moment t0 we do not yet believe that p or that not-p (either "Dr. Transparent will give me 10 dollars" or "Dr. Transparent will not give me 10 dollars"), and we know that the truth-value of both propositions is undetermined (which is why, according to Peels, we can choose any of them: it is "up to us" which to believe"); at a moment t1 we voluntarily form a belief (say, "Dr. Transparent will give me 10 dollars"). As "coming to believe that p" is, according to Peels, a basic action, it is already at t1 that we believe this; i.e., it is not the case that we do something whose result is believing that p; at a moment t2 we continue to believe that p (which we already believed at t1); Dr. Transparent reads our mind, and we can reflect on the belief which we formed a moment before (and which, we predicted, will trigger the process resulting in our receiving the promised sum); at a moment t3, Dr. Transparent fulfills his promise and pays us, thus making our belief true.

Is Peels' proposal amenable to the same objection as Winters's and Johnston's? Once again on the problem of acquisition
Following Peels's experiment timeline, the evidence that I came to believe that p, being reflective evidence, which involves my previous mental states, can only emerge at t2. Therefore, if, to initially form the belief in question, S needed to have this very piece of evidence, then S would obviously be unable to find such evidence at t1 and therefore could never form the belief that p. Peels even highlights that, at the moment of forming the belief, S has not acquired new justification supporting it-which he makes clear by means of a comparison with so-called "Feldman cases" (following Feldman, 2000). In the latter, a subject S "voluntarily" acquires a belief p by means of making it the case that p, because S brings about a change in the world which she will immediately be able to track 4 . Peels wants to emphasize that TDB cases, on the contrary, are not like that. In his own words, In Feldman cases, my belief automatically tracks the evidence I have, and I acquire or abandon a belief merely by changing my evidence base. Not so in TDB cases. In those cases, my evidence remains the same: I know that, if I come to believe that I will receive $10, I will receive it, and I know that, if I do not believe that I will receive it, I will not receive it. Maybe there is some way I could change that evidence. But the point is this: Whether or not I can change the evidence, given the evidence I have, it is up to me whether I choose to believe that I will receive $10 or to believe that I will not. What is unique for TDB cases in comparison with Feldman cases, then, is that I do not come to hold a belief by changing my evidence base (Peels, 2015, p. 12. Emphasis ours).
Therefore, according to Peels, people in TDB cases, such as Dr. Transparent's experimental subjects, simply do not change any evidential base. Now, up to this point, it seems to be that Peels's proposal could be subjected, at least preliminarily, to the same objection we have mobilized against Winters's: even if, after forming the belief "Dr. Transparent will give me $10", we can reflectively discover that Dr. Transparent now has 4 "I have nonbasic voluntary control over whether the lights in my office are on. All I have to do is move in a certain way to get the lights on or off. And I can do this. The next step of the argument notes that my belief about whether the lights are on tracks their actual state almost perfectly. As a result, I have a similar amount of control over whether I believe that the lights are on" (Feldman, 2000, p. 671). something to "read" in our minds, and therefore we have justification to believe that he will very probably give us the money, it is not clear at all how, a moment before, when we still had no justification to believe, what we formed might have been a belief nonetheless. In fact, from our cognitive perspective before forming the belief, it was also possible that he would not give us the money. We needed to form a belief, "Dr. Transparent will give me $10", for which we knew that we had no justification, in order to have, a moment later, reflective evidence of the fact that we had given Dr. Transparent something to read. But then, just as in Winters's "tiger vs. man" example, we seem to have landed in a situation of the type "I know there is no justification for believing that p but I still believe that p". In other words, it is not the case that, in TDB cases, we come to believe that p by a basic or direct act, in spite of the fact that Peels shared with Winters the demand that a voluntary belief should be acquired in this direct way 5 . In fact, the moment from which we can be said 5 In order to prove his point, Peels (2015, pp. 4-6) analyzes eight conditions−which he takes up from work by Williams (Williams, 1970), Scott-Kakures (1994), Bennett (1990), Setiya (2008), Hieronymi (2008) and Winters (1979)−that have been taken to be requisites for voluntary belief: C1: S knows that she has the ability to believe at will; C2: S knows that she has exercised that ability in a particular case with regard to a specific proposition p; C3: S still rationally believes that p after realizing all this; C4: S's coming to believe that p is a basic or direct act, that is, something S does without doing it by something else; C5: S believes that p at least partly for practical reasons, that is, non-evidential; reasons; C6: S believes that p as the result of an intention to believe that p: C7: S has settled the question of whether p is true by having settled the question of whether it is worthwhile to believe that p; C8: S believes that p independently of any truth-considerations. He claims that the first seven conditions must indeed be satisfied for a situation to be classed as believing at will, but, as we anticipated, to actually believe that p is the moment from which we can, upon reflection, find that we have justification for believing p instead of not p. And this, in turn, only takes place from the moment in which we realize that we have done something ("choosing to believe" that p) which will trigger Dr. Transparent "reading" of our belief, and consequently his decision to give us money. Before that moment, however, if we reflect as to whether p, we are well aware that we have no justification for believing it−a state which can hardly be called one of belief. Therefore, the belief will only emerge indirectly, as the result of a reflection on a state which is not itself a belief.
At this point, however, it should be clear that our preliminary reconstruction of Peels's proposal (along with our equally preliminary objection against it) does not do justice to a crucial trait which differentiates it from Winters's and Johnston's. Namely, that, in Peels's case, the subject S in TDB cases not only forms a self-fulfilling belief but knows beforehand precisely this character of the belief in question, which, according to Peels, provides S with epistemic justification for whichever self-fulfilling belief she will form. In other words, although S does not yet possess specific evidence that p is the case, she does possess, according to Peels, some, more general, justification: knowing that both p and not p are self-fulfilling beliefs.
Whereas Winters needed to dissociate the non-truthoriented acquisition of voluntary beliefs from their truthoriented preservation, so that the conditions of the acquisition would in a way "pass the torch" to evidentiary considerations (and Johnston remained at least ambiguous in this regard), Peels thinks that, insofar as we are dealing with self-fulfilling he argues that the eighth condition -the one that is raised by Williams as the main obstacle-should not be regarded as necessary.
beliefs which we know to be so, he can show us a scenario in which the acquisition of a belief is truth-oriented even if it is voluntarily chosen. In other words, Winters agreed with Williams on the demand that voluntary beliefs should be acquired irrespective of truth considerations. Peels does not share this demand. In his proposal, then, beliefs, as in Williams, Winters, and Johnston, are defined as aiming at truth; a voluntary belief is not (unlike the cases of Williams and Winters) acquired irrespective of truth considerations; given that there is no mismatch between acquisition and preservation, there is no need of "torch passing" between the will and evidentiary considerations; the belief which is voluntarily acquired is, at the very moment of its acquisition, allegedly truth-oriented−and so we can speak of a direct or basic act of acquiring a truth-oriented belief.
Peels's proposal thus differs from Winters's in that the scenario he introduces is one in which voluntary acquisition of beliefs can be, ab initio, truth-oriented. Let us turn now to this difference, in order to assess whether it saves Peels from the problem of the status of the propositional attitude we voluntarily adopt. Briefly, what Peels will point out is that justification for the belief we will voluntarily adopt is provided by the fact that we know that the belief in question is self-fulfilling. Accepting this is a second key concession we will make to Peels's argument. However, as we will try to show, this is still not enough to make his argument work.

The second concession: the alleged justification provided by the self-fulfilling character of a belief
Peels allegedly solves the "acquisition problem", as we anticipated, by claiming that the acquisition of a self-fulfilling belief, given that we know that it is self-fulfilling, is a truthoriented process. To understand this, we need to take into account the fact that the relationship between will and evidentiary considerations in a TDB scenario is twofold: on the one hand, once the belief is held, the subject cannot help but preserve it, because new evidence will have emerged which proves the belief in question to be true; on the other hand, even at the moment of its acquisition, the specific belief which is chosen is not imposed by such considerations, but the subject can choose either a belief or its negation without ceasing to aim at truth precisely because she knows that any belief she chooses will be self-fulfilling and thus "safe": either p or not-p will turn out to be true once chosen 6 .
The difference between Winters and Peels regarding belief acquisition seems to be, then, what does the trick. Even if, by Peels' own admission, the evidence base at t0 for S is undecided between that proposition and its negation, and she voluntarily chooses her belief on the basis of this equipollence, perhaps it might be argued that, at t1, S has some kind of justification for believing that proposition, enough for her to actually be able to start to believe that proposition. As we saw in section 4.5 by means of a comparison with "Feldman cases", in a TDB scenario "I do not come to hold a belief by changing my evidence base", i.e., by acquiring evidence that p ("Dr. Transparent will give me $10") or evidence that not p ("Dr. Transparent will not give me $10"). On the contrary, the situation remains undetermined between p and not p. But this does not stand in the way of S's having more general and indirect justification: that whatever she chooses to believe will turn out to be true. Let us explore this possibility.

Peels's recourse to self-fulfillingness as a "prospective" justification for a belief
When, in section 4.3, we mentioned that even a solution for the "preservation problem" still did not solve the "acquisition problem", we noted, in passing, that our concern converged with Scott-Kakures's remark that, at the moment of deciding to believe that p, we need to be inclined towards adopting a belief which appears to us as epistemically unjustified. This is the point, then, to tackle what Peels points out as a reply to Scott-Kakures in this regard: according to Peels, from the cognitive perspective of the subject who is about to form the belief "Dr. Transparent will give me $10" (that of moment t0, which should extend without "fissures" to moment t1), the belief in question is not "epistemically unjustified". In his own words, worth quoting in extenso: one might be tempted only to consider whether, as things stand before forming the belief, one has sufficient evidence for the proposition p that one will receive $10. One does not. Before one has decided whether or not to believe that one will receive $10, one's evidence base does not render that proposition likely to be true. What is relevant in this context, though, is whether or not the belief that one will receive $10 is justified, that is, whether or not if one were to form that belief one's evidence base would render that belief likely to be true. And here, it seems, the answer has to be positive: if one were to believe that one will receive $10, one's evidence base would render it sufficiently likely that one will receive $10, for one will receive $10 if one believes that one will. […] Since what is relevant in this context is justification for a belief, we are allowed to admit, say, the belief that p into our evidence base. Since believing that p is epistemically justified and I know that it is-for I know that, if I form it, it is true-I can intentionally form the belief that I will receive $10 (Peels, 2015, p. 15. Emphasis ours).
So, according to Peels, at the moment when I form my belief and start believing (t1), I know (because, in fact, I already knew it a moment before, at t0) that the belief is justified, in the sense that "if one were to form that belief one's evidence base would render that belief likely to be true". This may appear as a peculiar construal of the notion of epistemic justification, because typically when we reflect about the justifiedness of a belief we do not ask how well it would be justified if we held it, but, given the self-referential character of the evidential base for the belief now at hand (i.e., the fact that having the belief becomes justification for the belief itself), it does not seem inappropriate to take this condition into consideration. In fact, and as we previously mentioned, even without reference to this kind of self-referentiality, some counterfactual analyses of the notion of knowledge reflect on precisely the problem whether a belief is such that, if the subject held it, then it would probably be true (which would make it a safe belief), or if, on the contrary, the belief is such that we may probably hold it without its being true (thus being an unsafe belief). From this point of view, Peels's point is that in TDB cases we have paradigm instances of safe beliefs, and that the subjects can know precisely this immediately before forming the beliefs in question. At t0, the subject of Dr. Transparent's experiment knows that, if she formed the belief that she will receive $10, the belief would very probably be true−and therefore she knows beforehand that the belief would be justified, and can at t1 form the belief in question.
Just like the recourse to a scenario of self-fulfilling beliefs was already present in Johnston, the defense of the justifiedness of acquiring this sort of belief is not entirely a novelty of Peels's proposal. Concerning the adoption of selffulfilling beliefs, Velleman had written: Surely, the rules of justification are designed to provide a method of maximizing the proportion of truths to falsehoods among one's beliefs […] This purpose may well require a rule that one shouldn't retain a belief unless one has evidence of its truth. But does it require a rule that one shouldn't form a belief without prior evidence? I say no-at least, not if one has evidence that the belief would be true if one formed it. Why would rules designed to help one arrive at the truth forbid one to form a belief that would be true? What errors would one be avoiding by refusing to form a belief that wouldn't be erroneous? (Velleman, 1989, p. 63) Allegedly, then, voluntarily chosen beliefs are beliefs right from the moment they are acquired. In Peels's description, we can neither sustain nor even acquire a belief without aiming at truth, though we are aiming at truth in a rather roundabout way: we know that whatever choice we make we will end up having a "safe" belief. An accurate comparison between the authors, then, would have to take a form like this: Let us analyze the scenario. Whereas Williams's arguments considered, and rejected, the possibility of voluntary beliefs in a strong sense, which involved that both the acquisition and the preservation of beliefs were voluntary, Winters turned our attention to the possibility of voluntary beliefs in a weak sense. In her proposal, the preservation of voluntary beliefs depended on evidentiary considerationsthough contingent ones. In Johnston's move towards selffulfilling beliefs, the evidentiary considerations justifying the preservation of a voluntarily acquired belief are of reflective character: our belief that p becomes our justification to believe that p (and, incidentally, this turns the voluntariness in question very weak, because our choices are limited to a narrow scope of self-fulfilling beliefs). This still only tackled, however, the "preservation problem"-not the "acquisition problem": even Johnston is not explicit as to what would justify us in advance in voluntarily forming a certain belief. In his proposal, our belief is self-fulfilling, but he does not emphasize that this self-fulfilling character function as the basis for the subject's belief. However, as we tried to show, Peels (as, before him, Velleman) specifically considers what our previous justification would be for forming the belief. He finds this justification in the knowledge we may have, in advance, of its self-fulfilling character. Let us now see how this proposal has been received, as a nexus to our own assessment of it.

The reception of Peels's proposal. Nexus to the new objections
This proposal has turned out to be convincing for some authors, who find that Peels has in this manner managed to show conditions which, though highly peculiar, effectively sidestep the obstacles for the conceptual impossibility of believing at will. According to Samuel Montplaisir, insofar as we can meet the conditions "i., that the fact of believing p be directly related to the truth conditions under which p is true, ii., that the truth conditions of belief be related to the agent's will, and, iii., that i. be a known fact on the part of the agent" (Montplaisir, 2019, p. 82. Emphasis ours), we can indeed believe at will. Therefore, Montplaisir goes on, the fact "that one cannot believe at will is not a conceptual truth characterizing the notion of belief, but a truth about our present psychological conditions" and about the fact that such voluntary belief demands "conditions that hardly ever arise in our daily lives" (Montplaisir, 2019, pp. 81-82). Similarly, the self-fulfilling character of the beliefs at stake in Peels's example would, according to Joshua Shepherd, accomplish the task of overcoming Williams's concern about truth-orientedness: all Peels needs to show, Shepherd writes, is that an agent could form a belief-like cognitive state at will without violating conceptual constraints on what it is for a cognitive state to be a belief, or on what it is for an agent to be a believer. The chief constraint featuring in Williams's argument, recall, was that belief aims at the truth. And, as Peels observes, an agent who realizes that if she forms the belief that p then p will be true, and that if she fails to form the belief that p then p will not be true, is an agent who realizes she is rationally permitted to believe p or not. Assuming it is somehow psychologically possible for this agent, all she then needs to do is exercise the relevant ability, and form the belief at will (Shepherd, 2018, p. 325. Emphasis ours).
Insofar as the hypothetical agent can indeed realize that her potential beliefs will be true, these scenarios, according to Shepherd, succeed in showing that "believing at will is conceptually possible", although only "in very special circumstances" (Shepherd, 2018, p. 326). Commenting on these remarks, Russell Varley writes, in turn, that whereas proposals such as Peels's "have some merit, their weaknesses rest on the highly unusual properties they are required to possess in order to achieve their goals" (Varley,n.d.,p. 49)which means that, again, no doubts are raised by the strictly conceptual claim that TDB scenarios are indeed intelligibly described as cases of voluntary belief. Unlike Montplaisir, Shepherd and Varley, however, Gregory Antill's detailed analysis of self-fulfilling beliefs rejects taking this character as justificatory for the acquisition of the beliefs in question. Antill suggests an interesting distinction: "in adopting a belief that p, there are two different possible sets of considerations that might be required for an agent to satisfy the aim of belief: considerations sufficient to show that the agent will acquire a true belief and considerations sufficient to show that the believed proposition is true". Antill concedes that, "in normal circumstances", a set of considerations cannot "be sufficient to show that you would be believing something true without also being sufficient to show that the proposition to be believed is true" (Antill, 2016, p. 50). If I want to show that I would have a true belief if I believed, say, that the solar system has eight planets, what I need to do is to show that the proposition "The solar system has eight planets" is true. The question is, however, how these two aims relate in the specific case of self-fulfilling beliefs.
Antill does not entertain doubts about what the outcome is if we consider the aim of belief in the first sense-that of believing something true. "The fact that a belief that p is selffulfilling", he agrees, "is sufficient to show that in believing p, you would be believing something true, and it is also sufficient to show that in believing not-p, you would be believing something true". However, he goes on, things are different "on the second reading of the aim of belief, on which sufficient reasons for belief must suffice to show that the proposition to be believed is true": Could the fact that a belief that p is self-fulfilling be sufficient both to show that p is true and to show that not-p is true? I think the answer is surely no. Nothing is sufficient to show that, since p and notp cannot both be so. Since p and not-p cannot both be so, your reasons, insofar as they are sufficient to show p true, will thereby be sufficient to show notp false (Antill, 2016, p. 52).
This observation is surely unassailable, but it is not clear whether it is pertinent as a criticism of Peels's proposal (or Velleman's). Antill is showing that if, in order to have a justification to believe whichever alternative p or not p, we needed to have evidence that each of these options is true, we certainly could not attain such a thing-and the fact that the beliefs in question are self-fulfilling does not make any difference in this regard: given the principle of noncontradiction, two incompatible beliefs cannot both be true. But Peels (and Velleman) could insist that their point is precisely that, for us to be justified to believe either p or notp, we do not need to have evidence that both propositions are true. On the contrary, according to the kind of proposal we are considering, we can appeal to a permissivist concept of justification, which endorses the choice of any option as authorized, without presenting any of them as obligatory. (In fact, permissivism would not be, from the beginning, a remotely suitable strategy if we thought that we could show one of the propositions in question to be true and the other to be false). The justification for permissivism itself as a truth-conducive strategy, in turn, is supposed to stem from the fact that, in the context of self-fulfilling beliefs, the propositions believed will prove to be true-once we have acquired a belief in them. A proposal such as Peels's (and Velleman's) can hardly be attacked for being indifferent to the "second reading of the aim of belief": in such a strategy, adoption of self-fulfilling beliefs appears as justified precisely because the propositions we will find ourselves believing cannot fail to be true. What is at stake cannot be whether or not we care about the truth of the proposition we consider as a candidate for our belief, but, instead, whether or not we think the safety of the belief in question is a sufficient justification for it.
The problem, as we will try to argue, is not really whether at t0 S has a justification to believe that p ("Dr. Transparent will give me $10"), a justification which, according to Peels, is provided by the fact that S knows that, if she forms the belief, it will be true. In fact, we can make this crucial concession, and a crippling problem for Peels's proposal will remain: whether S's cognitive perspective at t1 is sufficiently different from that at t0, so to authorize us to claim that at t1 she already believes that p whereas at t0 she did not. We will try to argue that this is not the case.

Breaking the tie: the problem of equipollence
Our point is: having a belief that p is having a certain perspective on the world; it is the world appearing to us as if p. Even if, by hypothesis, we can imagine adopting a belief at will, actually having the belief involves, correlatively, being in a cognitive perspective such that we take p to be the case. It does not matter if at t1, say, S repeats with her inner voice "p, p, p"this certainly would not be enough for belief attribution. S's evidence base when S believes that p (at t1) and when S does not yet believe that p (at t0) cannot, therefore, be exactly the same. There is a clear difference between S's cognitive perspective at t0 and that at t2, because at t2 she can reflect on the fact that, a moment before (t1) she formed the belief that p. But at t1 S cannot avail of such reflective evidence, and therefore not p must still appear as justified (in a permissivist sense), just like a moment before, and as epistemically on a par with p. What matters in Peels's case is that at t1, S allegedly has formed the belief that p ("Dr. Transparent will give me…") and has not formed the belief that not p. But, in fact, as we will try to show, nothing in S's cognitive perspective seems to be relevantly different between t0 and t1-so it is not clear how we might say that at, t1, S began to believe that p. Let us try to tackle this issue step by step.
Let us start by showing something Peels himself would have to acknowledge: why at t0 S cannot yet be said to believe that p (instead of not-p).
A first important point we need to make is that belief that p seems to involve having better evidence to justify believing that p than we have for believing not-p (which, we will try to argue, S cannot have). Now, why is this? Why should the evidence supporting a belief be, so to say, in competition with belief supporting another? When I believe a proposition p and do not believe a proposition q, p and q being logically independent from one another, my belief that p does not imply anything about my alleged justification (or lack thereof) for believing that q. Perhaps at the moment I simply have not thought about q and have no attitude towards it. In fact, if in the future I acquire evidence that inclines me to believe that q, this does not need to affect my attitude towards p. However, this cannot be the case when both propositions are not logically independent but mutually contradictory. This is precisely what happens with p and notp. Given that they are incompatible, if at t0 we justifiedly believed that p, we would need to have justification to believe that p instead of not-p. So, getting words down, we have a first step: 1) If, in Peels's TDB case, S at t0 believed a proposition p instead of another proposition not-p then S would need to believe that she has some evidence supporting p which is better than the evidence which she has supporting not-p. (Premise). 2) In Peels's TDB case, at t0, S knows that her justification for believing p ("Dr. Transparent will give me $10") is that it is self-fulfilling. (Premise).
But the problem is that this condition applies not only to p, but, symmetrically, to not-p as well. Thus, 3) In Peels's TDB case, at t0, S knows that her justification for believing not-p ("Dr. Transparent will not give me $10") is that it is self-fulfilling (Premise).
Note that we are not denying that, at t2, when we can reflectively note the belief we have just formed, not-p will not be justified anymore: given that, a moment before, we formed the belief that p, it is no longer self-fulfilling to form the belief not-p. Dr. Transparent is already aware that we formed the belief that p and will therefore give us the promised $10. The problem is, on the contrary, that at t0 itself this reflective evidence cannot, obviously, be available.
Therefore, 4) In Peels's TDB case, at t0, S knows that her justification for both p and not-p is that they are self-fulfilling beliefs (From (2) and (3); introduction of conjunction). Trivially, 5) If, for two propositions, a subject's justification for believing them is knowing that they are both selffulfilling beliefs, then she cannot think that the kind of evidence supporting one of them is not better than the kind of evidence supporting the other.
Therefore, 6) In Peels's TDB case, at t0, S cannot think that the evidence she has for believing that p is better than the evidence that she has for believing that not-p.
So: it is clear that, at t0, S does not believe either p or notp-so far so good, because this is precisely what Peels writes. Now, let us examine if there are any differences between S's cognitive perspectives at the different moments. According to Peels, the situation is the following: Now, this (alleged) asymmetry between t0 and t1 looks very odd. What would explain that, given that S does not think, at t1, that she has more evidence supporting p than supporting not-p, we attribute to her a belief that p? If at t1 we may say that S actually believes that p, in spite of the fact that she is well aware of the equipollence between p and not-p, why not say that she believes at t0 as well? The answer cannot be that the difference is that at t1 S chose to believe that p (something Therefore, the claim that at t1 S begins to hold a belief which she did not have at t0 is not tenable.

Summing up
In this article, we examined Rik Peels's response to Williams's argument against the possibility of believing at will. According to Williams, voluntary beliefs must be acquired independently of truth-considerations, and therefore cannot be considered beliefs, as beliefs are truthoriented.
Our starting point was Barbara Winters's proposal that the conditions for acquiring and preserving a belief can be different. Winters argues that a subject may initially acquire a belief in a way that is not truth-oriented, but later preserve it based on what she considers to be evidence. However, we pointed out that this proposal fails to show that we can choose a belief through a direct basic act by virtue of what we have called the " acquisition problem". A belief could at most be preserved as a belief at the moment we realize there is evidence to support it, but given that, by hypothesis, at the moment of acquisition it is not supported by evidence, it cannot be voluntarily acquired qua belief. We will have to refer to that state as some other kind of propositional attitude. Despite the differences that separate one proposal from the other, Peels's ends up facing the same hurdle. Specifically, Peels fails to show how we could move from not believing, at time t0, in a certain proposition p, but knowing that we have a permissivist justification to believe it, to actually believing it at time t1, when our evidential base has not changed at all. Again, if it were enough to have an attitude towards p at t1 that was not actually a belief (i.e., if it were sufficient to repeat "p, p, p" in our mind for Dr. Transparent to "read" that state and thereby effectively trigger the process that will result in us receiving the money or to initiate any other process that does not require better justification to accept p than to accept not-p), then we could indirectly come to believe p from the moment (t2) at which we reflect on the act we performed a moment before (at t1)-an act which was not itself an act of belief. But, once again, this was not Peels's purpose.
Even granting two crucial concessions (that we can call a belief voluntary even when it is only so in a weak sense, and that if S has anticipatory knowledge of the self-fulfilling nature of a belief, then it can provide justification for S to acquire that belief), Peels's argument ultimately fails to show how if the cognitive perspective of the subject is identical before and after the belief is formed, she can still form the belief that p.