Introduction
Imagine you are offered a huge sum of money on the sole condition that you form the belief that Buckingham Palace is in the United States. Unfortunately, no matter how much you would like to, it seems you cannot consciously form that belief just because you want to. This is, in a nutshell, the Explanandum, the phenomenon that I will try to make sense of in this paper.Footnote 1 I will look at three very promising accounts of this puzzling phenomenon and show that, unfortunately, they all face significant criticisms. Although these are not intended as knockdown objections, I offer an account that is preferable given its ability to avoid the pitfalls of these views, and is based on the following two widely discussed claims: first, in conscious reasoning, we generally treat our premises as reasons; and second, belief is a disposition to use a proposition as a default premise in reasoning.
1. The Minimal Explanandum
A number of different and competing definitions have been offered to try to capture the mental phenomenon I will call the ‘Explanandum’. Some prefer the label ‘doxastic involuntarism’, by which they usually mean that beliefs cannot be formed through an act of will (Williams Reference Williams1973). Others have focused their attention on the fact that the Explanandum takes place within doxastic deliberation or conscious reasoning (e.g., Shah Reference Shah2003; Steglich-Petersen Reference Steglich-Petersen2006). These scholars argue that the best way to capture what is going on in deliberation about belief is to look at the types of considerations that can be the reason for which we come to form or abandon a belief. R is a reason for which S (the agent) will φ only if R is able to consciously and directly cause φ when S deliberates about whether to φ, using R as a premise in their deliberation (as in Shah Reference Shah2006: 485). Epistemic considerations concerning the truth of proposition p can function as the reasons for which one forms the belief that p. By contrast, considerations that have nothing to do with the truth of p do not seem to be able to play the same motivating role. In deliberation, those ‘non-epistemic’ considerations cannot play the role of key motivators. Within the scope of this paper, I shall remain neutral regarding whether non-epistemic considerations could play some motivating role in forming the belief that p, for instance, if the subject already takes herself to have sufficient evidenceFootnote 2 for the truth of p. What I take for granted here is that non-epistemic considerations cannot per se deliberatively bring about a belief.Footnote 3
Minimal Explanandum:Footnote 4 epistemic considerations concerning p can motivate (qua reasons) an agent to believe that p. Considerations not concerning the truth of p cannot (qua reasons) motivate an agent to believe that p if she takes herself to lack sufficient epistemic reasons for believing that p.Footnote 5
Let me say a little more about the scope of the Minimal Explanandum. The truth of the Minimal Explanandum does not preclude the truth of a stronger claim, namely that epistemic considerations are the only considerations that could motivate conscious and deliberate belief-formation (Kelly Reference Kelly2002; Steglich-Petersen Reference Steglich-Petersen2009; McHugh Reference McHugh2013). Furthermore, it does not rule out an even stronger point, viz., the Transparency of Doxastic Deliberation (Shah Reference Shah2003). This is the idea that, when we deliberate about whether to believe that p, that question collapses into the question of whether p is true. On this view, only considerations relevant to the factual question about the truth of p can enter the deliberation about whether or not to believe that p. Thus, the role of the Minimal Explanandum is simply to set a minimum requirementFootnote 6 that any view needs to meet. In what follows, I will look at three explanations of this Minimal Explanandum and show that they are all lacking in some way. I then propose my own solution.
2. The Teleological View
Bernard Williams famously declared that “belief aims at truth” (1973: 136). On a common reading of this expression, there is literally something about belief that has the goal of truth. This view – not surprising called the Teleological View – boils down to the claim that “believing that p essentially involves having as an aim or purpose to believe p truly” (Steglich-Petersen Reference Steglich-Petersen2009: 395). Velleman (Reference Velleman2000) defends this view by first pointing out that belief is part of a larger group of cognitive attitudes, called acceptances (i.e., belief, supposition, imagining), which all present their content as true. These attitudes, he adds, furthermore share the ability to cause actions when combined with some suitable conative attitude.Footnote 7 This means that we cannot differentiate believing from, say, assuming simply based on the roles these have in causing action, since the two attitudes in fact share the same role as regards output. By contrast, what sets belief apart is its aiming at truth, which, for Velleman, means that belief is a truth-regulated acceptance: believing that p means accepting that p is true, and such a state is either produced by an intention to accept p only if true, or it is de facto regulated by a truth-tracking mechanism.
Unfortunately, this account faces two well-known objections which, I claim, undermine its ability to make sense of the Minimal Explanandum. There are two possible readings of belief's truth-directedness: one is that, for an acceptance to count as a belief, it needs to be mostly truth-regulated (strong reading). Alternatively, we may say that for an acceptance to count as a belief it needs to be only minimally truth-regulated. In this case, it is possible that non-epistemic factors may predominantly influence the formation of a belief (weak reading). This, however, raises the following dilemma (as exposed by Shah Reference Shah2003). If we adopt the strong reading, while we have a nice way to make sense of the Explanandum, we nonetheless end up with the very unintuitive proposal that our attitudes are never subliminally influenced by our biases, desires, and emotions. If, on the other hand, we adopt the weak reading, then we allow for the possibility that our beliefs may be predominantly influenced by non-epistemic factors, but we lose sight of the fact that, in deliberations, non-epistemic considerations cannot predominantly motivate us qua reasons to form a belief for which we do not have sufficient evidence.Footnote 8
The teleologist reacts to this by focusing on people's personal-level aim in forming beliefs, rather than their sub-personal goals (Steglich-Petersen Reference Steglich-Petersen2006; Whiting Reference Whiting, Littlejohn and Turri2014). However, the problem with that move is, as Owens (Reference Owens2003) has now famously argued, that it is a necessary condition for having a (personal-level) aim that this aim can be psychologically weighed against other aims one might have, and it does not seem that in deliberation, we are really weighing the aim of truth against any other aim. Note that the possibility of weighing opposing aims naturally includes that one might decide to silence one of them and go with the other. It is possible to concoct situations in which the agent's aim to form a belief for pragmatic reasons should be intuitively much stronger than her alleged epistemic goal. Rather catastrophically: if the only way to save humanity from destruction were to form the belief that Alfred Einstein was Austrian rather than German, then I am sure we would rather form that false belief than witnessing destruction. And yet, no matter how strong our motivation is, in doxastic deliberation, it cannot consciously trump the evidence (we think) we have. This suggests that deliberation's (alleged) epistemic aim does not really square with our intuitive notion of ‘aim’, according to which not only are aims weighed against one another, but also stronger aims prevail.Footnote 9
To help the teleological view, McHugh draws an analogy with the aim we have when playing sudoku puzzles: an “activity counts as an instance of doing a sudoku puzzle only if it is brought about and executed in a way that appropriately involves the aim of filling out the grid in the right way. (You can enter numbers into a sudoku grid without having that aim. But then you are not really doing the puzzle)” (McHugh Reference McHugh2012: 427). McHugh points out that other aims might interact with the constitutive aim of sudoku provided that your solution “strikes you as sufficiently likely to be right” (McHugh Reference McHugh2012: 434). But this is highly controversial, since two possible situations come to mind. I am having a sudoku competition with my young daughter owing to which I intentionally fill out some grids in the wrong way to give her an advantage. Even so, I still count as playing the game, while my desire to let her win trumps the constitutive goal of sudoku. Similarly, I decide to play sudoku to alleviate the pain of grading papers, so I am playing – distractedly – while also reading the papers. Thus, I have two conflicting goals: grading fairly and filling out the sudoku grid in the right way. Given that the first goal is more important, I write down numbers in the grid even if they only vaguely strike me as right (rather than sufficiently likely to be right). Distractedly playing sudoku (and knowing it) still counts as playing sudoku. In the case of belief, by contrast, this cannot happen, given the Minimal Explanandum.
The Teleologist could try to reply by arguing that there is something about the aim of belief that makes it un-weighable (Steglich-Petersen Reference Steglich-Petersen2009;Footnote 10 Whiting Reference Whiting, Littlejohn and Turri2014). While I am not able to tackle this proposal directly, it is worth noticing that any account that requires a special notion of ‘aim’ (in making sense of the aim of belief) lacks intuitive force and makes one wonder why we need an aim of belief at all. It seems reasonable to ask whether we would better off dropping the idea of any necessary epistemic goal that guides deliberative belief-formation. This is exactly what the Normative View proposes.
3. The Normative View
Contrary to the Teleological View, friends of the Normative View have argued that believers are motivated the way they are because truth is a constitutive normative standard for belief. Thus, when forming beliefs, we are necessarily bound by a norm of truth (Shah Reference Shah2003, Reference Shah2006).Footnote 11 This position again stems from Velleman's idea that the functional-motivational role of belief is not unique but shared with other cognitive attitudes, such as imagination and acceptance. Hence, in order to define ‘belief’, we need to appeal to its relation to truth, which is not shared by any other attitude. In fact, says the normativist, the Explanandum itself is the result of belief being a normative concept, which contains the following truth-norm:
(TN): you ought to believe that p only if p is true.
Nishi Shah explains that the conceptual nature of the norm of belief makes it the case that when we apply the concept belief in deliberation, we also necessarily apply this normative stricture to our belief-forming process. belief is a normative concept because “it is one of the conditions for possessing the concept of belief that one accept the prescription to believe that p only if p is true” (Shah Reference Shah2003: 470, my emphasis). This requirement takes precedence over any (non-alethic) norm that may apply to belief, and, as a result, we are motivated most prominently by considerations that can satisfy the truth-norm.
However, as intuitive and plausible as it may sound, this way of making sense of the Explanandum faces a number of well-known worries. It is indubitable that we often say that beliefs are correct when true, and we praise true beliefs and criticize false beliefs. From this, however, it is a stretch to go on to say that TN is a conceptual norm in Shah's sense, namely that if you fail to appreciate and consistently apply that norm, then you lack the concept belief altogether (Shah Reference Shah2003: 455). Indeed, it is important to clarify that the notion of ‘acceptance’ for Shah is very strong.Footnote 12 But it is also a necessary component of his explanation for why, in doxastic deliberation, we cannot be motivated by our desires and practical considerations.Footnote 13 Contra Shah, it seems that our way of applying the concept belief sometimes allows for cases in which we take moral or prudential norms to be more important than the available evidence. For instance, we might think that it is more important to believe that we will recover from a serious, life-threatening illness than forming a true belief about it. Intuitively, in a situation like that, we should believe that we will recover (even if the evidence may suggest otherwise). Shah's view excludes this possibility, which seems to me unnecessarily strong.
And even setting this point aside, Shah's notion of ‘acceptance’ requires being committed to a very demanding form of motivational internalism. On that view, accepting a norm means that one is necessarily motivated by it (Shah Reference Shah2003: 454–5). This is worrisome because, as Steglich-Petersen (Reference Steglich-Petersen2006: 507) puts it, “[i]t is doubtful whether a consideration which necessitates motivation should be considered a normative consideration at all.” If the Explanandum is the result of applying the norm of truth, then such a norm cannot be violated. However, it seems a requirement of following a norm that one has the freedom to violate it. Thus, the inescapability of doxastic deliberation itself raises concerns about the normative explanation. In addition, the form of internalism required to make sense of even the Minimal Explanandum seems highly implausible as a general account of how norms motivate us (McHugh Reference McHugh2013). The normativist may reply that the truth norm is special: it is as strong as it is because it is constitutive of the concept of belief. It is now clear that this reply will not do. It is often said that assertion is constituted by some further epistemic norm (e.g., truth, knowledge). However, contrary to what happens in the case of belief, the constitutive norm of assertion does not prevent me from consciously violating it when I make an assertion.Footnote 14 Clearly, if we had a view that could avoid such unappealing upshots, then that would be a preferable position to adopt.
4. The Functional View
In a recent paper (2018), Sullivan-Bissett puts forward a new explanation for why doxastic deliberation seems to be constrained by evidential considerations. Her focus is the so-called Transparency of Doxastic Deliberation which, as aforementioned, constitutes a stronger version of the same phenomenon we are trying to explain here. Contra Shah and Velleman (Reference Shah and Velleman2005), Sullivan-Bissett rightly points out that the motivational role of belief is unique, and she builds on this idea in order to provide an account of Transparency. She supports the view that belief's unique motivational role is to dispose an agent to act – across contexts – in ways that would satisfy her desires, if the content of the belief were true (Sullivan-Bissett Reference Sullivan-Bissett2018: 3463). She also discusses Transparency as a “causal factor” (realized in some neurological structure), which makes it the case that we are motivated by epistemic considerations when we deliberate. In addition, her view is that the presence of this causal factor is explained by the biological history and function of deliberation. Given the motivational role of belief, true beliefs allow us to satisfy our desires, find food, avoid danger, and so on. Accordingly, her point is that true beliefs and truth-tracking mechanisms are evolutionarily adaptive. As she puts it, “[o]ur mechanisms for belief-production … Normally produce beliefs with true contents. … Such mechanisms have proliferated because – given the essential nature of belief (its motivational role) – creatures with true beliefs will be more likely to survive” (Sullivan-Bissett Reference Sullivan-Bissett2018: 3467). The hypothesis she puts forward is that deliberation – like any belief-productive mechanism – was selected because focusing on the environment (and thus being transparent) made deliberation (more) likely to deliver true beliefs (Sullivan-Bissett Reference Sullivan-Bissett2018: 3453–76; my emphasis).
While this is an attractive proposal, worries remain. One problem is that this view explains why human doxastic deliberation normally seeks true beliefs (while trying to avoid false beliefs), but nevertheless fails to explain why human doxastic deliberation is always like that. The fact that some mechanism was biologically selected owing to how it functions does not guarantee that the mechanism will always function that way, for malfunction is never completely avoidable. Let us consider an example. It is conceivable that the human eye has the biological function of delivering veridical images. Another plausible assumption is that the current human eye was selected because, among other things, its components allowed for color-vision, which made images more precise. Some human eyes do not see colors, however. This means that a selected-for organ may lack the feature it was selected for, while still retaining its function. Likewise, even assuming that Transparency (or the Minimal Explanandum) is the feature that made the mechanism of deliberation selected-for, it still does not follow that deliberation is always transparent: it could happen that the “causal factor” responsible for Transparency may fail for some humans. However, a key point in the entire discussion about Transparency (and the Minimal Explanandum) is that it always occurs in human deliberation. It is unclear to me that the Functional View is able to make sense of that point. Indeed, I worry that the view may fail to explain why human doxastic deliberation can never be non-transparent (for a similar objection see also Noordhof Reference Noordhof2001: 255).
I now want to raise a second, independent, concern. It is unclear how well natural selection can explain the occurrence of Transparency (or of the Minimal Explanandum). In particular, I am not sure that true beliefs are consistently more adaptive than false beliefs. Notoriously, there are situations in which an overly sensitive perceptual system would be more evolutionarily advantageous than a reliable belief-producing system. This is because it would avoid predation-situations more often than an accurate detector (Nesse and Williams Reference Nesse and Williams1995). Similarly, while the assessments one makes about one's own abilities do shape one's actions, there is evidence that these are beneficial regardless of their truth value (Mandelbaum Reference Mandelbaum2019). It is thus an open question whether doxastic deliberation – in focusing on truth – is necessarily adaptive.Footnote 15
Moreover, even assuming that it is adaptive, it is nevertheless unclear that doxastic deliberation is adaptive owing to its truth-conducing character. The key point of doxastic deliberation is that it is motivated by what the agent takes to be evidence for the truth of a certain proposition. However, recent studies on confabulation and self-deception have shed important light on the fact that we often fabricate reasons for things we want to believe (Nisbett and Wilson Reference Nisbett and Wilson1977). We are moreover plagued by biases, such as the confirmation bias, to such an extent that a recent view about the nature of conscious reasoning claims that reasoning does not have the biological function of tracking the truth at all. On this view, the biological role of conscious reasoning is communicative and argumentative (Mercier and Sperber Reference Mercier and Sperber2011). Since deliberation is one form of conscious reasoning, it seems at least open to question whether or not Transparency and deliberation are really adaptive qua truth-conducive mechanisms (although they may be adaptive for other reasons).
Finally, even if we grant that transparent deliberation has been evolutionarily selected because it is truth-conducive, this evolutionary story would only explain why, in deliberating, we want to avoid obviously false beliefs. It would not explain why we cannot – in the absence of a clear answer to whether something is true – decide to believe what would make sense from a practical standpoint. Indeed, having some belief as opposed to a gap in our understanding of the world could provide a further evolutionary advantage. Allow me to explain. The case in point is when we have no decisive evidence one way or another when we are deliberating about what to believe (Ginet Reference Ginet1985). Transparency, and the Minimal Explanandum alike, indicate that deliberation never settles a belief for pragmatic reasons (at least when we lack sufficient evidence for that belief). I worry that the Functional view cannot explain why this happens. For, even if true beliefs are more adaptive than false beliefs, in cases in which we have no evidence one way or the other, there are no obvious evolutionary advantages in having transparent deliberation that settles for suspension of judgment. In other words, natural selection hardly explains why we cannot go ahead and believe what would make practical sense. This is especially worrisome in light of the claim that we often take ‘gaps’ in our knowledge and explanation of the world to be problematic. For instance, Gopnik's (Reference Gopnik, Keil and Wilson2000) findings indicate that, generally, humans find that lacking an explanation for an event “is, to varying degrees, an unsettling, disturbing, arousing experience. Conversely, finding an explanation for something is accompanied by a satisfaction that goes beyond the merely cognitive” (Gopnik Reference Gopnik, Keil and Wilson2000: 311). This motivation often generates confabulated explanations that – unbeknownst to the subject – are false, but which could nonetheless be advantageous qua gap-filling explanations (Sullivan-Bissett Reference Sullivan-Bissett2015; Coltheart Reference Coltheart2017: 67–8). This raises the worry that the Functional view is at best able to explain why, in doxastic deliberation, we want to avoid obvious errors. It fails, however, to bring to light the (alleged) evolutionary advantage of a deliberative process that strenuously prioritizes error-avoidance, even in cases of uncertainty.
5. Explaining the Explanandum
In the remainder of this paper, I defend a new position that explains the Minimal Explanandum by combining the uniqueness of the concept of belief (section 5.1) with widely recognized and intuitive features of our conscious reasoning processes (section 5.2). This position does not require invoking any specific norm or aim of belief, nor does it rely on the adaptive function of doxastic deliberation. By contrast, the very nature of our beliefs and reasoning processes explains why doxastic deliberation is shaped so uniquely.
5.1. The uniqueness of the concept of belief
Our common understanding of belief is that of a mental attitude, or state of mind, which plays an important role in guiding actions and forming other beliefs. Accordingly, it is often said in philosophy that believing that p means being disposed to use p as a premise in one's reasoning. Williamson (Reference Williamson2000: 99) points out that “one believes p outright when one is prepared to use p as a premise in practical reasoning”. Frankish (Reference Frankish2004) argues that flat-out believing is tantamount to having a disposition to take a proposition as a premise in all deliberations and reasoning. Thus, it is a hallmark of belief that it shapes our reasoning processes in this way.Footnote 16 Although this is quite a popular position, there are philosophers who have resisted the idea that the role of belief in shaping our actions and thought is in any way distinctive of belief as an attitude. For instance, Velleman (Reference Velleman2000) has forcefully argued that belief's dispositions are far from unique: imagination, acceptance, and supposition all behave like belief. In a co-authored paper, Shah and Velleman (Reference Shah and Velleman2005: 497–8) write:
The question is how to differentiate the concept of belief from the concepts of other attitudes … The answer cannot be that belief plays a distinctive motivational role, because the motivational role of belief is one that it shares with other cognitive attitudes. Assuming that p and supposing that p resemble believing that p in that they dispose the subject to behave as if p were true; and even imagining that p may resemble belief in this respect.
Notwithstanding many similarities between belief and other neighboring attitudes, claiming that they share the exact same role means overlooking the fact that these non-doxastic cognitive attitudes are attitudes that influence us in only limited, circumscribed contexts. If two children are playing cops and robbers, their imaginings will not affect what they do beyond the remit of the game. If, in a make-believe situation, we all imagine that our school is on fire, then we will as a result also imagine that we are in danger (Sinhababu Reference Sinhababu2013: 160–1). However, my imagining will not produce the belief that I am in danger, nor will it motivate me to actually call the fire department. This position is also found in Bratman's (Reference Bratman1992) theory of “acceptance in a context”. A lawyer may accept that her clients are not guilty and do so for professional reasons, without always believing that they are (Frankish Reference Frankish2004). While in her office her actions reflect what she accepts to be true, we do not expect this attitude to change what she believes, nor that it impacts on her behavior beyond her working hours.
By contrast, belief triggers new thoughts and behaviors by default (O'Brien Reference O'Brien2005; Van Leeuwen Reference Van Leuween2009). Believing is being disposed to use a proposition as a default premise, i.e., as a premise we readily use in day-to-day situations and in contexts that are not in any way special (Wedgwood Reference Wedgwood and Chan2013; Ross and Schroeder Reference Ross and Schroeder2014). To illustrate, if I believe that my Dad lives in California, whenever that information becomes relevant, I will be ready to use it to guide me (modulo irrationality, of course). If I am not disposed to use that information as my default starting point, then I do not really believe it. We can thus agree with Velleman that all cognitive attitudes may be dispositions to enter our reasoning processes, but that does not make them identical to belief (Bergamaschi Ganapini Reference Bergamaschi Ganapini2019).Footnote 17 Contrary to what Shah and Velleman state in the quotation above, there is a clear way to distinguish the concept of belief from the concept of neighboring attitudes. We can do this by looking at the sort of default role belief has in our reasoning. Furthermore, mastering the concept of belief will therefore require at least implicitly subscribing to the following:
Concept-Belief: Believing that p means being disposed to use p as a default premise in reasoning.
5.2. Two constraints on reasoning
As we have just seen, a fairly intuitive understanding of belief is that believing that p means being disposed to use p as a default premise in reasoning. In this section, I will show that we share important intuitions about reasoning too.Footnote 18 Let me proceed step-by-step. First, it is commonly assumed in the literature that reasoning is a “personal-level process” (Boghossian Reference Boghossian2014: 2). This means that reasoning is a process attributable to the whole person and it is not part of a sub-mechanism of the mind (e.g., perceptual module). Furthermore, reasoning is a conscious process, namely a process of which “we are, in some appropriate sense, aware” (Pettit Reference Pettit1993: 223–4; Boghossian Reference Boghossian2008: 483). Another point on which there is widespread agreement is that, in reasoning, the reasoner takes her premises to support her conclusion. For instance, according to Boghossian, reasoning requires the following conditions:
(Taking Condition) reasoning necessarily involves taking the accepted truth of your premises to support your conclusion, and drawing such a conclusion because of that fact.Footnote 19
The idea here is that reasoning requires seeing these premises as lending some support to the conclusion: “[a] transition from some beliefs to a conclusion counts as [reasoning] only if the thinker takes his conclusion to be supported by the presumed truth of those other beliefs” (Boghossian Reference Boghossian2014: 4).Footnote 20 Arguably, if I take a proposition p as supporting a certain conclusion φ, then I must be doing at least two things. First, I take p to be true; second, I take my concluding φ as the rational thing to do given the truth of premise p. In other words, I take p to be a reason to conclude φ. But what does ‘taking’ mean here? Some argue reasoning requires that the agent believes that the premises support the conclusion. Others are happy with less restive conditions than believing, choosing to talk instead of intellectual seeming (Tucker Reference Tucker2012; Leary Reference Leary2017). Notwithstanding the various ways of cashing out the notion of ‘taking’, I believe that all the views here converge on this minimal requirement for reasoning:
Negative Taking Condition: if you (consciously) reason from p to φ, you cannot at the same time (consciously) believe that concluding φ from p is irrational.
This condition states that, if the reasoner accepts the irrationality of concluding φ from p, this would undermine her ability to make the inference.Footnote 21 Now, here is one way in which concluding φ from p could be irrational: the agent has insufficient evidence for the truth of p and thus for taking p to be a reason to φ. Indeed, its irrationality is highlighted by the fact that reasoners can be criticized if they reason using premises they did not know were true.Footnote 22 I take this to be quite intuitive. Suppose, for instance, that Tommy decides not to bring his car in for the annual checkup and inspection. He explains his decision to a friend by mentioning that his car is in good shape and he does not see it breaking down in the near future. Tommy is open to criticism if it turns out that he did not have enough evidence to believe that his car was in good shape. This intuition squares nicely with the widely endorsed view that one can appropriately employ a proposition as a reason only if one knows it (Hawthorne and Stanley Reference Hawthorne and Stanley2008).Footnote 23 The idea here is that there are some epistemic conditions that are necessary (and possibly also sufficient) for a subject to be in a position to correctly take a proposition as a reason in support of some conclusion. This suggests the following Rationality Constraint on taking a proposition to be a reason:
Rationality Constraint: It is irrational to treat p as a reason to φ if one lacks sufficient evidence that p is true. Footnote 24
Combined with the Negative Taking Condition, the Rationality Constraint tells us that one cannot treat p as a reason to φ (and thus cannot reason from p to φ), while also being aware that one lacks sufficient evidence that p is true. Let us return to Tommy. He is adamant that he will not bring his car in for its annual inspection because, he claims, the car is doing just fine. His friend openly disagrees with this and pointedly recounts all the evidence suggesting that Tommy's car is in need of major repairs. At this point, Tommy indignantly admits to having little evidence to believe his car will not break down anytime soon but, surprisingly, he does not seem willingly to change his reasoning and actually reasserts his initial intentions. But how can he ground his decision on a proposition he himself sees as having little support? Doing so is in fact tantamount to saying “p therefore φ” while openly admitting “it is irrational to treat p as a reason to φ and thus to reason from p to φ”.Footnote 25 The self-defeating nature of this sentence is due to the fact that uttering it violates the Negative Taking Condition – a minimal requirement for reasoning. To make sense of Tommy's odd behavior, then, a natural explanation is that Tommy either does not genuinely believe he has little evidence for the claim that his car will not break down anytime soon, or he does not really take that proposition as his reason (and he is basing his conclusion on something else). But he clearly cannot treat that premise as a reason for his conclusion while also seeing it as insufficiently-supported by the available evidence.
5.3. Explaining the Minimal Explanandum
We are now in a position to account for the Minimal Explanandum without invoking any epistemic aim or norm of belief. In deliberation, we apply the concept of belief (Shah Reference Shah2003). That is, when deliberating, I am implicitly guided by the question of whether to believe that p, and thus I ask myself whether or not to form an attitude whose content p I will use as a premise (whenever p bears on the issue in hand). Now, that question cannot be answered, or even properly posed, if it also appears to me that I cannot use p as a premise. I cannot look into building a time-travel machine while knowing that time-travel is impossible. Similarly, it hardly makes sense for me to ask whether to form a disposition to do X, if I take it that I cannot do X to begin with.Footnote 26 Imagine Laura, deliberating about whether to believe she will get an A in her upcoming math test, which is key to passing the class. Her lack of evidence for believing that now becomes salient to her. And yet she knows she would be happier if she thought she was likely to get an A. If she were to form a cognitive attitude with that content, and did so primarily based on pragmatic considerations, then she would come to have that attitude while also taking herself to lack sufficient evidence for its content. Obviously, such a proposition – “I will pass the test” – could not be used as a default premise in her reasoning: using p as premise while taking it to be insufficiently supported by evidence violates the Negative Taking Condition, an intuitive condition for being able to reason at all. Hence, I contend that the attitude Laura has formed through deliberation, if any, simply cannot be a belief. Because we are constrained by how reasoning works and by the role of belief in reasoning, if we consciously form an attitude simply because we want to, then we are not doxastically deliberating.
This point bears some emphasis. When deliberatively forming a belief that p, we are implicitly guided by questions of whether to form a belief that p, namely a disposition to use proposition p as a default premise. This question may not be posed explicitly, but it is crucial for doxastic deliberation that this question is in the background. Once deliberation has begun, it becomes apparent that I lack evidence for p, and it also becomes clear that, given my current epistemic position, p would not be able to guide my reasoning now (owing to the two aforementioned constraints on reasoning). As a result, it also becomes apparent that, unless I acquire new supportive evidence about p, proposition p will not be used as a default starting point in my reasoning. The result for this is that I cannot form any disposition to use p as a default premise and thus I cannot make myself believe that p as a result of my doxastic deliberation. Hence, the Minimal Explanandum is the upshot of applying the concept of belief together with some constraints on using the content of belief as our default guide.Footnote 27
As I conclude, let me forestall two possible objections. First, a reader may at this point wonder how my view really differs from a version of the Teleological View. Indeed, such a reader may worry that my view is ultimately committed to the claim that agents – when doxastically deliberating – aim at forming beliefs that are sufficiently epistemically supported, as these are the beliefs they will be using in future reasoning. But such talk of aims is superfluous here, and really confuses, rather than clarifies, my point. I do not think any real aim is needed here: when we doxastically deliberate, we apply the concept of belief, and that automatically poses some constraints: once it becomes apparent that we lack sufficient epistemic support for the truth of proposition p, it is that realization that undermines the possibility to believe what we want.Footnote 28
A second objection goes as follows. Readers may worry that my view – in denying that the Explanandum results from the aim of truth (or any other epistemic aim) – falls short of the intuitive claim that agents usually strive for beliefs that are true. For, if and when we discover that we reasoned towards a false belief, we usually regret it and deem this to be a failure. Hence, the aim of ‘truth’ must be key in doxastic deliberation. Against this objection, my view capably explains why false beliefs are problematic, and why forming a false belief constitutes a failure. Beliefs are our dispositions to use propositions as default premises. Since, in reasoning, we treat our premises as reasons and thus true, this means that a false belief falls short in what it is supposed to do qua premise, viz., to be a reason. And deliberatively forming a belief that turns out to be false would constitute a failure insofar as the goal of belief-formation through deliberation is to form premises for future reasoning. If we use a false belief as a premise in reasoning, then there is a failure of our reasoning because we have treated a proposition as a reason when it was not. What distinguishes my position from the widely accepted, alethic-teleological view is that, on my view, the value of truth is not intrinsic to the attitude of belief and is not necessary to explain the Minimal Explanandum. That does not mean, however, that my account of the Minimal Explanandum is incompatible with, or cannot explain, why we care for truth when forming beliefs via a deliberative process.
6. Conclusion
It is a well-established fact about belief that we cannot deliberatively believe something simply because we want to. In this paper, I have argued that this psychological impossibility is the result of the nature of our conscious reasoning and the default role our beliefs have therein. This means that we do not need to invoke the norm or aim of truth to explain why we cannot manipulate our beliefs at will. By contrast, there is a simpler story: once it becomes apparent that I lack evidence for p, the now-overt impossibility of using p as a default premise is enough to undermine for me the (conscious and deliberate) formation of the belief that p.Footnote 29
