1 Introduction

The phenomenon of transparency may be stated as follows: once you settle whether p, you have thereby settled whether to believe that p (Shah, 2003, p. 448; Shah & Velleman, 2005, p. 499). In other words, the question of whether to believe that p is immediately and non-inferentially replaced and answered by the factual question of whether p. Shah and Velleman argue in detail that the best explanation for this phenomenon is that, in addition to being “causally regulated by truth-regarding considerations” (Shah, 2003, p. 448), which can be cashed out descriptively, belief is also constituted by a normative standard of correctness.

In parallel to this debate, those subscribing to belief normativism have tried to formulate the constitutive norm which is supposed to link belief and truth (Boghossian, 2003; Wedgwood, 2013; Raleigh, 2012; Whiting, 2010; McHugh, 2012; Greenberg, 2018): yet formulating the truth norm has proved to be more difficult than it seemed at first blush. There are different contenders for the truth norm of belief: some of the formulations are permissive, i.e., they state the conditions under which you may believe something; other formulations are injunctive, i.e., they state the conditions under which you ought to believe something. In responding to objections and challenges, many formulations have been suggested under each heading. However, as has been convincingly shown by Bykvist and Hattiagandi (2013, p. 114) and more recently and in more detail by Sullivan-Bissett and Noordhof (2020), the permissive version of the truth norm is wholly incapable of explaining the transparency phenomenon. This means that what was originally considered to be support for belief normativism has, due to its alleged explanatory inadequacy, become a stumbling block for a popular version of belief normativism.

This paper offers a critical re-assessment of the transparency phenomenon through the lenses of the extant versions of the truth norm of belief. As I shall argue, the problem is not limited to the permissive version of the truth norm: indeed, on scrutiny it is not clear that even the injunctive versions of the truth norm can really explain transparency, albeit for a different reason. I then reconsider the transparency phenomenon and argue that it consists of two aspects that should be (but have not been) explicitly kept apart, which I call the question-shifting aspect and the answer-shifting aspect, respectively. The question-shifting aspect is well explained by positing a conceptual truth norm of belief, irrespective of how the norm is formulated, whereas the answer-shifting aspect that poses problem for various formulations of the truth norm is explanatorily empty and in need of no explanation.

The paper proceeds as follows: In Section “The argument from transparency”, I briefly discuss the argument from transparency. In Secion “The travails of transparency for truth norms”, I discuss the problem that the permissive version of the truth norm is shown to have for explaining transparency, and I also argue that the injunctive versions of the truth norm are not without problem here. In Section “Transparency reconsidered”, I put forward my interpretation of the transparency phenomenon by distinguishing between the question-shifting and answer-shifting aspects. There I argue that it is the answer-shifting aspect that poses a problem for the various formulations of the truth norm, while, on a closer look, the answer-shifting aspect proves to be explanatorily empty and in need of no explanation. The question-shifting aspect, however, reveals something substantive about the relationship between belief and truth which is very well explained by the truth norm. I then conclude briefly.

2 The argument from transparency

The transparency phenomenon arises in the context of doxastic deliberation. This is the mental activity we engage in when we question whether to believe that p. For example, when we hear some new and strange fact, we naturally engage in doxastic deliberation to see whether or not to believe it. This question is transparent to the factual question of whether p, in the sense that if you settle whether p, you have thereby settled whether to believe that p (Shah, 2003, p. 448; Shah & Velleman, 2005, p. 499). Moreover, when we engage in doxastic deliberation, we intuitively concede that the only question that needs to be addressed is this factual question, and this shows that no consideration other than truth (like, for example, usefulness) is relevant to doxastic deliberation (Shah, 2003, p. 463). Furthermore, intuitively “the transparency transition from one question to another is immediate, unavoidable and non-inferential” (Steglich-Petersen, 2006, p. 510), which means that no mediate premise (like adopting a certain aim) is required for this transition to happen.

Teleologists think that belief is a mental state that is necessarily regulated by truth, in the sense that forming, revising, and extinguishing belief is regulated and controlled by the aim to accept the truth, or by sub-intentional truth-tracking mechanisms which can be understood descriptively. But a problem arises if we think that belief’s relation to truth can be captured merely by this feature, which Shah (2003) call “the teleologist’s dilemma” (p. 460). This dilemma is stated as follows: we have mental states properly called beliefs which are formed through processes that are not truth-responsive, for example wishful thinking. If we tighten the truth-regulation feature, we cannot explain how it is that belief can also be formed through these processes. If on the other hand, we loosen the truth-regulation feature to also include beliefs whose formation is not regulated by truth, then we cannot explain the strong intuition we have concerning the transparency phenomenon, according to which it is only truth-related considerations that are relevant to the question of whether to believe that p.

Here is where normativism enters the scene. If, further to being regulated by truth, belief is also conceptually governed by a normative standard of correctness, according to which a belief is correct if and only if its content is true, then we can neatly escape the teleologist’s dilemma: when you engage in doxastic deliberation, namely consider whether to believe that p, you are applying the concept of belief, and since the truth norm conceptually governs belief, you feel the intuitive pressure to concede that this question is to be replaced and answered by the factual question of whether p. But doxastic deliberation is not the only way that belief is formed and changed. There are also non-deliberative, sub-conscious and unconscious mechanisms through which new beliefs enter our minds. Since in these cases, unlike the case of doxastic deliberation, we may not consciously consider the concept of belief, the other factors can also influence the formation of belief. Therefore, Shah and Velleman’s solution is to loosen the regulation feature to encompass those beliefs that are not formed via the truth-tracking mechanism, while explaining the transparency phenomenon through a normative factor, beside the descriptive regulation. Of course, when considered deliberatively, the beliefs that are not responsive to truth are evaluated as incorrect and in need of revision by the conceptual truth norm of belief. It is a general merit of normativist accounts that they can very well explain the deviances as norm violations without denying that a norm is in force, and this is exactly how Shah and Velleman deploy the constitutive norm of belief to explain transparency.

Shah and Velleman have worked out and defended this argument in detail: these subtleties, however, are not important for the argument of this paper, and I hope the gist of the argument is clear enough. Unsurprisingly, along with Shah and Velleman, normativists have embraced this argument and considered explaining transparency as a virtue of their account. This doesn’t mean that no other proposal for explaining transparency has been suggested. Steglich-Petersen (2006), for example, thinks that transparency can be explained without appealing to a norm, but rather through an aim we necessarily adopt when we engage in doxastic deliberation. More recently, Sullivan-Bissett and Noordhof (2020) have suggested that transparency is explained by the standard of correctness considered non-normatively (p. 1225). It is not the aim of this paper to discuss these alternative proposals and to see whether they are plausible enough in comparison with the normative account Shah and Velleman put forward, although it is clear that such proposals are not congenial to the normativist project, which posits that there is something normative about the relation of belief to truth. However, a group of philosophers have argued that, contrary to what is claimed by Shah and Velleman, a certain plausible formulation of the truth norm of belief is not capable of explaining transparency. But why is this so? To this topic we now turn.

3 The travails of transparency for truth norms

3.1 Preliminary remarks

When Shah (2003) and Shah and Velleman (2005) put forward their argument, the debate on how to formulate the truth norm of belief was still in its beginnings. In view of the progress made in this regard in the past decade or so, we now have a clearer view of the relationship between transparency and the various formulations suggested for the truth norm. Variants of the truth norm have been suggested in response to objections and challenges raised, yet the transparency phenomenon and the ability of the new formulations to account for it has not been among the considerations driving these reformulations. Before seeing whether and how the various versions of truth norm explain transparency, we should briefly consider these versions.

Perhaps the first norm that comes to the mind when we think about the relationship between belief and truth is this: one ought to believe that p if and only if p is true. This is the formulation discussed by Paul Boghossian (2003) and is also mentioned in passing by Shah and Velleman (2005, p. 520). It is clear this norm is injunctive, because it mandates you to form a belief under a specific condition. However, as noted by Boghossian (2003) himself, this norm is not tenable because it is too strong. According to this norm, if p is true, then we are obliged to believe that p. But there are an infinite number of truths that we do not consider. There are also overcomplex true propositions that, given our cognitive limitations, cannot be considered by us, let alone be believed. Since ought implies can, and since it is beyond our ability to believe these truths, this norm is not acceptable (p. 37). In response to this challenge, Boghossian (2003) amends this norm as follows: one ought to believe that p only if p is true. Now from the truth of p, no obligation ensues. The only norm in force is that if you believe that p, p ought to be true. However, this norm creates a new trouble, because logically it is equivalent to: if p is false, it is not the case that you ought to believe that p, which means that you are also permitted to believe not p. This norm is too weak: we expect the truth norm to forbid belief in a false proposition (Bykvist & Hattiagandi, 2013, p. 108). For this reason, Ralph Wedgwood (as cited in Bykvist & Hattiagandi, 2013, p. 108) thinks that we should stick to the biconditional formulation of the truth norm, but inserts a further condition at the beginning of the norm: if you consider p, then you ought to believe that p if and only if p is true. Since this norm is preceded by the condition of considering p, no violation of the ought-implies-can principle occurs. If I do not consider p, then no obligation would be imposed me. More recently, Alexander Greenberg (2018) has defended a new injunctive norm: If you adopt a doxastic attitude towards p, then you ought to believe that p (in contrast to disbelieving or suspending belief) if and only if p is true (p. 3288). Like Wedgwood’s proposal, Greenberg’s norm does not require us to believe all truths, since there are many truths that we have no doxastic attitude towards.

The story does not end here. The injunctive versions of the truth norm, which oblige the subject to believe that p, are caused trouble by what Annandi Hattiagandi and Krister Bykvist (2013, pp. 107–113), following Roy Sorensen (1988), call blindspots; true propositions that are not truly believable. For example, the proposition that it is raining and nobody believes that it is raining can be true. Suppose that this is true. Then, according to Wedgwood’s norm, if we consider this proposition, we ought to believe it. But as soon as we believe it, it becomes false and now we are forbidden to believe it. The blindspots show that there are conditions under which the injunctive norms are not satisfiable, and since ought also implies satisfiability, the injunctive norms are not acceptable. The popular strategy in response to this challenge is to qualify the truth norm to exclude blindspots. For example, if you consider p and p is truly believable, then you ought to believe that p if and only if p is true (Bykvist & Hattiagandi, 2013, p. 110; Raleigh, 2013, p. 268). This strategy is usually charged with being ad hoc because the modification is just made to evade the problem, and can claim no independent rationale (see for example, Whiting, 2010, p. 219). However, ad hoc modifications do not make something false, even if they are unappealing.

Another group of normativists have suggested that the truth norm of belief should be spelled out permissively (Shah & Velleman, 2005, p. 519; Whiting, 2010; Raleigh, 2013; Kalantari & Luntley, 2013). As Whiting (2010) puts it, you are permitted to believe that p if and only if p is true. Unlike the injunctive norm, this norm does not oblige you to form a particular belief. Therefore, no violation of the ought-implies-can principle happens here. In the case of blindspots, since this norm can be also satisfied by not believing the truth in question, no violation of ought-implies-satisfiability is present here.Footnote 1 However, we should note that the permissive versions of the truth norm are quite strict when it comes to falsehood, because they are logically equivalent to: if p is false you ought not believe that p. Therefore, while the permissive version is liberal when it comes to believing a truth, it is strict in forbidding belief in a false proposition. The permissive version of the truth norm is popular because it seems capable of solving the problems that beset the injunctive versions in one fell swoop, without inserting complex conditions and seemingly ad hoc modifications. However, there are also serious reservations about them. For example, since this group of norms never obliges you to believe something, they are compatible with taking a Pyrrhonian stance and suspending belief about everything, even those propositions that are trivial or for which you have conclusive evidence (Raleigh, 2013, pp. 260–264). To many this is a high cost to be paid. But the standard response by the advocates of the permissive version of the truth norm is that forming new beliefs is motivated by practical considerations and there is no theoretical requirement in this regard (Raleigh, 2013, p. 263).

Here we may leave the debate on how to formulate the truth norm, and turn to the important question of whether and how the tenable versions of the truth norm can explain transparency.

3.2 From truth norm to transparency?

3.2.1 The permissibility norm

Are the truth norms discussed above capable of explaining transparency? Let me start with the permissive versions of the truth norm, whose problem in this regard is already noted in the literature. As Bykvist and Hattiagandi (2013, p. 114) put it, the permissive version of the truth norm will:

fail to explain the phenomenon of ‘transparency’, which Shah and Velleman claim is a signal virtue of the normative account of belief … how could a commitment to the avoidance of false belief explain transparency? Since [Doxastic Permission] cannot explain why you ever have a positive reason to form a belief, it cannot explain why, if you settle the question whether p is true, you thereby settle the question whether to believe that p. To settle the question whether to believe that p, you need a further reason, over and above the fact that believing that p would be compatible with [Doxastic Permission], that tips the balance in favor of believing that p over suspending judgement. This is because, even if you have settled whether p is true, suspending judgement that p is also compatible with [Doxastic Permission].

The force of this remark is clear. If belief is constituted by a permissive norm, then as far as this norm is concerned, by settling whether p, you are not required to believe that p; you are also free to suspend judgment with respect to p. In other words, there is a long way from settling whether p to settling whether to believe that p, as far as the permissive truth norm is concerned. Therefore the permissive norm does not explain transparency, and since we intuitively concede that transparency constitutes an “immediate, unavoidable and non-inferential transition”, the failure of the permissive norm undermines its viability: so if somebody, for whatever reason, thinks that the permissibility norm is the only plausible option we have, then the whole thesis of belief normativism would be in danger. This critique has been recently revived in more detail by Sullivan-Bissett and Noordhof (2020). As they put it, “It being permissible to believe that p is far cry from it being settled for you to believe that p. The norm tells you that you are permitted to believe that p, but that is all it tells you … The permissibility norm is thus not strong enough to generate Transparency” (p. 1215).

Ironically, Shah and Velleman (2005) themselves think that the permissive norm is preferable to the injunctive norm. Their reason for favoring the permissive norm is that they think it better captures the normative standard of correctness governing belief: “The standard of correctness implicit in the concept of belief, according to our view, is a biconditional norm: a belief is correct if and only if it is true. But correctness is itself a permissive rather than injunctive notion. A norm of correctness forbids the holding of belief that would be incorrect, but it merely permits the holding of correct belief. One is not required to hold every belief that would be correct” (p. 519).Footnote 2 They however recognize no explanatory inadequacy here. This is so because they think that in the context of doxastic deliberation we are “committed to forming the belief if it would be correct” and this commitment mandates “a belief in p if and only if p is true” (Shah & Velleman, 2005, p. 519). Nevertheless, as Sullivan-Bissett and Noordhof (2020) rightly point out, it is not clear how this might work. In doxastic deliberation we are thinking whether to believe that p. The other doxastic attitudes we may adopt towards p are disbelief and suspension of belief. For example, if I cannot settle the question of whether p, suspending belief would be the best option. If the range of options we have towards p includes suspension of belief, then even in the context of doxastic deliberation, contrary to what Shah and Velleman claim, the permissive norm cannot mandate the belief in p. Are the injunctive norms better in this respect? As I shall argue shortly, this is not as clear as it is usually supposed to be.

3.2.2 The injunctive norm

It is usually thought that this problem of explanatory inadequacy only concerns the permissibility norm. After all, subject to certain conditions, the injunctive norms mandate forming a belief that p if and only if p is true. Therefore, it seems clear that transparency is well explained by this group of formulations. On closer scrutiny, however, the case is not so straightforward and we should, I claim, have serious reservations about this. Recall that according to transparency if I settle whether p, I have thereby settled whether to believe that p. The truth norm is supposed to explain how it is that we intuitively concede this immediate and non-inferential transition. The explanation goes as this: in doxastic deliberation we are consciously applying the concept of belief and therefore cognize the norm that constitutes it. Now when we settle whether p, the norm we have in our mind through entertaining the concept of belief mandates believing that p.

But there is a problem here. By settling whether p, on one side of the transition, we are considering p as true, namely we check p and find it to be true. But look, for example, at Wedgwood’s formulation of the truth norm: if you consider p, you ought to believe that p if and only if p is true. The condition of application of this norm is considering a proposition, p, that is true. It is crucial to note the difference between considering a proposition p as true and considering a proposition p that is true. While in the latter, the subject has only considered p, in the former the subject has also affirmed the truth of p by his/her own lights. Put differently, while in the latter the only contribution of the subject is the act of considering or regarding a proposition, in the former further to considering or regarding a proposition, the subject also affirms its truth and this is why it is stated as considering (or regarding) p as true.Footnote 3 As such, the former condition is stronger and outstrips the latter. If this is so, we cannot be sure that the immediate transition is made thanks to the norm, because there is more at play here than the application condition of the injunctive truth norm. Therefore, it is still open to doubt whether it is really the truth norm that is responsible for this immediate transition.

Now, consider Greenberg’s (2018) recently defended injunctive formulation of the truth norm. According to him “if one has some doxastic attitude about p—i.e. if one believes, disbelieves, or suspends judgment about whether p—then one ought to believe that p if and only if p is true” (p. 3279).Footnote 4 In doxastic deliberation, namely in thinking whether to believe that p, we are clearly thinking about which doxastic attitude to adopt towards p. Therefore, this norm seems to be particularly suitable to explain transparency, because, when we are questioning which doxastic attitude to adopt towards p, and we determine that p is true, then, according to this norm, the only option available to us is to believe that p. But we should note again that in settling whether p, we are surpassing the application condition of this norm. The application condition of this norm is having a doxastic attitude towards p that is true. But, as I noted, by settling whether p we consider p as true, and considering p as true is a cognitive attitude about p. Therefore, it is open to doubt that it is really this norm that explains the transition we have in transparency. One can argue that since the argument from transparency is an argument from best explanation, this objection is rather weak because requiring that we should be sure that the norm explains the transition is too demanding and it would suffice if the norm is the best explanation of the transition. In other words, the objection would be forceful only when we can also find a better explanation for the phenomenon. In the next section, I will argue that there is a better way to understand the putative transition without appealing to a norm.

There is another problem that transparency poses for the injunctive norms. As we saw, in response to the blindspots, the injunctive norms have usually undergone modifications to exclude them. For example, consider the norm that if you consider p and it is truly believable, then you ought to believe that p if and only if p is true. This norm, as far as it goes, can sidestep the problem of blindspots. But if transparency is a guide to the truth norm of belief, then such qualifications to the truth norm of belief cannot be accepted. This is so because the “immediate, unavoidable and non-inferential transition” we intuitively concede between settling whether p and settling whether to believe that p, leaves no room for the modifications that the most tenable injunctive versions of the truth norm contain. Upon settling whether p, you have thereby settled whether to believe that p, and no further condition (like being truly believable) is present here. Therefore, if transparency is supposed to be explained through the conceptual truth norm governing belief, then the tenable injunctive formulations fall short in serving this purpose.Footnote 5

I hope my objection is clear. Although the injunctive norms are not subject to the specific problem that beset the permissive norm, they have their own difficulties. First, in transparency, we already outstrip the application condition of this group of norms and for this reason it is not clear that the norms in question can be held responsible for the immediate leap we intuitively concede here. In other words, one can coherently deny or at least doubt that the injunctive norm is responsible for this transition. Moreover, the direct link we intuitively feel between settling whether p and settling whether to believe that p leaves no room for the modification that such norms have undergone to face the challenges. These are important problems, and although I admit they are not as serious as the problem that the permissive norm has here, I think they betray a more fundamental problem about the argument from transparency. If we expose this problem, we can solve it in such a way that transparency can be explained by the truth norm without making reference to the specific formulations suggested for this norm, whether permissive or injunctive. Let’s see how this can happen.

4 Transparency reconsidered

4.1 Preliminary remarks

In light of the discussions of the previous section, I think it is clear that normativists have to critically re-assess transparency and its relationship to normativism. The argument from transparency has been attacked by non-normativists. For example, Steglich-Petersen (2006) argues that transparency requires no norm and can be “explained by the aim one necessarily adopts in deliberating about whether to believe that p” (p. 499). Sullivan-Bissett and Noordhof (2020) believe that transparency should be explained by the correctness standard of belief, interpreted non-normatively, namely the mere classification of belief as either true and false without any normative import (p. 1225). As is clear, these attacks to transparency belong to the non-normativist camps who want to resist normativism about belief. On the other hand, normativists have usually accepted the argument from transparency because, they thought, it serves what is dear and near to their hearts. But considering the difficulties that transparency poses to the tenable formulations of the truth norm of belief, normativists can no longer postpone the task of responding to these problems.

The way non-normativists have reacted to the transparency argument is not one that is available to normativists. For example, it is common ground among belief normativists that the standard of correctness governing belief is a normative standard (Gibbard, 2003). Therefore, Sullivan-Bissett’s and Noordhof’s proposal cannot be accepted by them. In a similar vein, for normativists truth is not linked to belief through the aim one adopts, as Steglich-Petersen holds. So, normativists must think of another way to resolve the tension between transparency and belief normativism. It is the task of this section to provide such a response. As I shall argue shortly, there is a way to keep the best of the transparency argument for normativism without letting it do any harm to the formulations of the truth norm offered by normativists. The first step in doing so is to distinguish between two aspects of transparency which should be (but have not been) explicitly distinguished and addressed separately; what I call the answer-shifting aspect and the question-shifting aspect. The answer-shifting aspect concerns the immediate transition we intuitively concede between settling whether p and settling whether to believe that p. The question-shifting aspect is about the immediate transition we endorse between the question of whether to believe that p and the factual question of whether p. While I believe that the question-shifting aspect reveals something important about the normative nature of belief, the answer-shifting aspect is explanatorily empty. Let’s discuss these aspects in turn.

4.2 Answer-shifting aspect

If you settle whether p, you have thereby settled whether to believe that p. This is what I call the answer-shifting aspect of transparency, because it means that the answer we give to a particular factual question immediately provides the answer to the question of whether to have a particular belief. Note that it is this transition between the answers that troubles the formulations of the truth norm. According to the permissive norm, if p is true then you may believe it, and what Bykvist and Hattiagandi (2013) and Sullivan-Bissett and Noordhof (2020) have rightly pointed out is that as far as this norm is concerned, from settling whether p, settling whether to believe that p does not ensue. In the case of injunctive norms, one of the problems that I raised was that when you settle whether p, you have thereby settled whether to believe that p, and this direct and immediate transition acknowledges no modification or exception, such as the ones normativists advocating the injunctive norms have been forced to insert in order to exclude cases such as blindspots. Moreover, I argued that by settling whether p, by providing an answer to whether p, one has already outstripped the application condition of the injunctive norm and this makes it unclear that the injunctive norm can be taken to be responsible for the transition. I hope it is clear from these remarks that it is the transition between the answers which is the point of tension between transparency and the truth norm of belief.

But how can normativists resolve this tension? In what follows, I defend a particular version of the view that this answer-shifting aspect is explanatorily empty and, as such, requires no explanation, be it normative or otherwise. In other words, contrary to appearances, there is no transition here and therefore this is not a genuine phenomenon disclosing an important feature of belief that is in need of explanation. The suspicion that there is something circular or empty about transparency has already been raised in the literature. The way I argue for explanatory emptiness, however, is different from these claims and is not subject to the problems that beset them. Therefore, before discussing my proposal, some discussion about these previous attempts is in order.

Shah and Velleman (2005) have addressed the objection that there is something circular about settling whether to believe that p by settling whether p: as they put it, “ascertaining whether p is true would entail arriving at a belief with respect to p, as an intermediate step in deliberating whether to believe it. And believing that p cannot be an intermediate step in deliberating whether to believe that p” (pp. 519–520). But why should ascertaining whether p necessarily entail arriving at a belief with respect to p? Sullivan-Bissett and Noordhof (2020) explain the same objection in terms of Kathrin Glüer’s and Asa Wikforss’s (2009) No Guidance Argument, and their account is illuminating here. As they put it, one might think that transparency is explanatorily empty “because it can seem that there is no further question for the deliberator after they have established that it is permissible to believe that p since, having arrived at the conditions under which it is so—settling that p—they already believe it. If the norm cannot guide because ascertaining its application conditions just is to form the belief said to be permitted, it might look like Transparency does not need explaining after all” (p. 1225).

What Sullivan-Bissett and Noordhof are getting at is that the specific way that the truth norm is formulated is not important, because, if the norm is supposed to guide belief formation, the subject has first to establish that the application condition of the norm is met and this requires a belief. Therefore, the answer-shifting transition is empty because there is no denying that as far as the permissive norm is concerned, by settling whether p, it would be permissible to believe that p (even if it is not settled whether to believe that p). But in establishing that it is permissible to believe that p, the agent has first to establish that its condition of application is met, which requires a belief that p. Therefore, even for the permissive norm, settling whether p along with establishing the permissive norm, yields the belief that p. Of course, this is not limited to the permissive norm and includes the injunctive norm as well. So, the answer-shifting aspect is not a challenge for the permissive version of the truth norm because if the permissive norm is supposed to be guiding here, it requires a belief which makes the transition explanatorily empty.Footnote 6

However, I believe that there are two reasons why normativists should not accept this answer. First, even if this response can establish that transparency is explanatorily empty, it makes belief normativism fall prey to the No Guidance argument. According to the No Guidance argument, the only sense in which we can say belief is constituted by a norm is that belief is formed by following a certain norm. If belief formation is merely in accordance with a rule, it would not be genuinely normative. Now, they go on, belief formation cannot result from following a rule because the only plausible account of rule-following we have requires a belief that the application condition of the rule is met and a pro-attitude to follow the rule. But on this account, belief formation requires another belief (about the conditions of application of the rule in question), and now a regress is forthcoming (Glüer & Wikforss, 2009, pp. 55–63). This is a formidable challenge, and the way Sullivan-Bissett and Noordhof (2020) formulate the explanatory emptiness of transparency clearly falls prey to the No Guidance objection. Therefore, this does not seem to be a promising strategy for normativism to solve the tension between transparency and the truth norm of belief; this would be jumping from frying pan to fire. Of course, normativists have suggested responses to the No Guidance objection. According to a now-popular response, the normativity of belief does not mean that belief formation is the result of following a rule. Rather, it means that belief is conceptually assessable by a norm (Jarvis, 2012; Hlobil, 2015; Tracy, 2020; McHugh & Whiting, 2014, p. 698). If we accept this conception of belief normativism, the No Guidance argument would be disarmed. For normativists who find this response convincing, as I do, the way Sullivan-Bissett and Noordhof argue for explanatory emptiness would no longer be available.

Second, there is another less pressing problem with this way of arguing for explanatory emptiness. As we saw, transparency is shown to be explanatorily empty because as we establish that it is permissible or obligatory to believe that p, we have to ascertain that the application condition of this norm obtains, which requires a belief that p. But here the explanation of transparency is not really normative. The norm does have a role in showing why we have transparency, but the explanation is not really normative. It is only because every norm has an application condition and we ought to have a belief about the application condition in order to follow a norm that transparency is explained. The normative force of the norm has no hand in explaining transparency. Rather, it is another feature of the norm, namely its descriptive condition of application, that explains transparency. Given this, one might wonder whether the non-normative suggestions about how to explain transparency might be taken more seriously here.

So, if this attempt is unsuccessful, how else can we show that the answer-shifting aspect is explanatorily empty? There is a simple way to do it. To put it bluntly, settling whether p is just to believe that p. In other words, if you settle whether p, if you answer the question of whether p affirmatively, you already believe that p. If this is so, then there is nothing to be explained as regards why settling whether p also settles whether to believe that p. Note that unlike what Shah and Velleman and Sullivan-Bissett and Noordhof have claimed, this way of arguing for explanatory emptiness does not hinge on following a certain norm, which in turn entails forming a belief about the application condition of the norm in question. Rather, here there is nothing to be explained because settling whether p is just believing p, period. If this response is right, then the “immediate, unavoidable, and non-inferential” transition we intuitively concede between the answer we give to whether p and the answer we give to whether to believe that p, is nothing mysterious, because in fact there is no transition here; the appearance is false.

But why is it that this simple option is not considered by other philosophers? Two reasons come to mind. First, as I shall argue in the next sub-section, we have to distinguish between the transition we have between the factual question of whether p and the question of whether to believe that p, on the one hand, and the answer we give to question of whether p and the question of whether to believe that p, on the other. These two aspects are distinct. But the answer we give to the factual question, i.e., settling whether p, is just believing that p. Answering the factual question requires a contribution on the part of the agent which is not present in the mere act of thinking about the factual question. Failing to distinguish between the question-shifting aspect and the answer-shifting aspect is, I think, the first obstacle to realize that settling whether p is just believing that p. Since it is perhaps thought that since the two questions are clearly distinct, the answers should be similarly distinct. More on this in the next sub-section.

The second reason, I believe, is rooted in some presuppositions about the nature of belief. Settling whether p is regarding p as true or considering p and finding it true. But belief is usually considered to be more than the mere regarding-as-true. Velleman (2000) nicely summarizes this way of thinking:

Whatever regarding-as-true turns out to be, it will still be involved in more than believing, since it will be involved, for example, in supposing or assuming, and in propositional imagining as well. These attitudes are cognitive, like belief, rather than conative, like desires. To imagine that p is to regard p as describing how things are, not a prescribing how they should be. Imagining is therefore a way of regarding a proposition as true—or, to introduce a term, a way of accepting a proposition. The question remains how belief differs from imagining and the other cognitive attitudes. (p. 250)

This way of thinking is also supported by the observation that although all of these cognitive attitudes contain the element of regarding-as-true, belief is subject to the truth norm while imagining or supposing is not. Therefore, it is thought, there ought to be something more to belief that is responsible for its normativity. If we accept this line of reasoning, then my proposal about explanatory emptiness might seem to be in trouble because it might be thought that there is a further step between settling whether p and believing that p.. However, I think that our intuitions run contrary to this line of reasoning. Vahid (2006), for example, defends a deflationary conception of the aim of belief which I find tenable. For him, “To believe p is to regard p as true for its own sake” (p. 324), without any further aim or purpose. In other words, to believe p is “to regard it as true seriously” (p. 326). Now what distinguishes belief from other cognitive attitudes which seem to also contain regarding-as-true is its very purity or aimlessness. For example, while (propositional) imagining is regarding as true for the sake of fantasy, or assuming is regarding as true for the sake of argument, belief is regarding as true for its own sake. One might object that if belief normativists accept this proposal, they are obliged to explain the fact that belief is conceptually subject to a truth norm but imagining or supposing is not. This is not, however, strange. The norm-neutrality of the cognitive attitudes other than belief can be understood through the additional elements contained within them. For example, imagining is not subject to a truth norm because it is regarding as true for the sake of fantasy. But believing is governed by a truth norm because it is regarding as true, seriously without any further element. This is, I hope, uncontroversial. Adding a further condition to a normative state or act can change its normative status or override its normative force. For example, many believe that the speech act of assertion is subject to a truth norm. If you assert something, it ought to be true. But asserting is putting forward as true. Now, joking is asserting with a further purpose (that of having fun, we might say), and this further element can neutralize the truth norm governing assertion. I think this deflationary conception of belief aiming at truth is plausible and is also compatible with normativism. I also think that this is what Brandom (1994) has in mind when writing that “believing is taking-true” (p. 5). He then goes on to explain that belief, considered as taking-true, is governed by a truth norm: “A particular belief may actually relate in various ways to how things are, but its content determines how it is appropriate for it to be related, according to the belief—namely that the content of taking-true should be true” (p. 13).

Now, if we accept this conception of belief, then there would be no transition between settling whether p, which is regarding p as true seriously, and believing p. Note that unlike what Shah and Velleman and Sullivan-Bissett and Noordhof remarked, it is not the case that the belief is inserted as an intermediary between settling whether p and settling whether to believe that p. Rather, settling whether p is just believing p. This way of making sense of the answer-shifting aspect of transparency makes it explanatorily empty. This means that no explanation, be it normative or otherwise, is required here. Furthermore, the appearance of the direct and non-inferential transition we have here cannot tell for or against any particular version of the truth norm, because there is no transition to be explained. For example, consider the permissive norm. According to this norm, you may believe that p if and only if p is true. But if you settle whether p, you already believe it. Therefore, this cannot show that the permissive norm has explanatory inadequacy here. Moreover, this is in alignment with the points I made in the previous section about the injunctive norms of belief. As I argued there, settling whether p overreaches the application condition of the injunctive norms. It is now clear that this is so because settling whether p is just believing p. Moreover, the modifications made to the injunctive norms (like being truly believable) cannot square with transparency because by settling whether p you already believe that p and it would be already too late for these considerations to play any role.

This may be something of a relief for the advocates of particular truth norms who find their view to be in tension with, and unable to explain, transparency. If we stop here, then transparency can no longer be considered as a reason for belief normativism, although it also cannot be regarded as a stumbling block for normativism. However, discarding the argument from transparency wholesale would be throwing the baby out with the bathwater, because there is another aspect to transparency which finds its best explanation in the constitutive normativity of belief. This is so because one can still wonder how it is that a question about whether to form a particular mental state is immediately replaced by a factual question. This is an important question that, I think, needs to be addressed by thinking about the question-shifting aspect of doxastic transparency.

4.3 The question-shifting aspect

As I have argued so far, by settling whether p we thereby settle whether to believe that p, because the former is just believing that p. But here I want to argue that the immediate transition we intuitively concede between asking whether to believe that p and asking whether p requires a different treatment. How can it be, one might wonder, that the answer-shifting aspect is explanatorily empty but the question-shifting aspect is not. Here I want to show that treating these two aspects differently is indeed plausible.

In doxastic deliberation, namely in asking whether to believe that p, this question is immediately replaced by the factual question of whether p. This question-shifting aspect is an interesting and genuine phenomenon, because we can ask how it is that we are immediately led to ask the factual question, and only this question, upon thinking whether to believe that p. Not only do other questions, like whether believing that p is useful, have no place in doxastic deliberation, but also this shift in question is immediate and non-inferential. Given that our beliefs are also formed through processes that are not necessarily truth-conducive (like wishful thinking), a teleologist has to, but seemingly cannot, explain how it is that we have the strong intuition of question-shifting. And now the best explanation (and in this I agree with Shah and Velleman) is that there is a conceptual truth norm governing belief, and since in doxastic deliberation we are consciously applying the concept of belief we are immediately led to ask the factual question.

Recall that the truth norm of belief determines when we ought (not) or may (not) believe that p given the truth or falsity of p. According to the truth norm, irrespective of how it is formulated, it is primarily the truth of p which obliges or permits you to believe that p; and since it is a conceptual norm governing belief, it explains why in thinking whether to believe that p, we have to see whether we are permitted (not) or required (not) to believe that p. Since, according to the truth norm, the permission or obligation to believe that p hinges on the truth of p, the immediate question to be asked is whether p. These are clearly two distinct questions. Even on the deflationary account of belief I favored in the last sub-section, the question of whether to regard p as true is clearly different from the question of whether p. This therefore cries out for an explanation of why, by asking the former question, we are directly ushered to ask the latter question. This shift in question is compatible with all versions of the truth norm. If p is false, then according to the permissive version of the truth norm, you ought not believe it. So, it is understandable that you have to ask the factual question. Similarly, given that you have already considered this question, according to Wedgwood’s injunctive norm, if it is true you ought to believe it. Hence, it is again understandable that you are led to ask this question. The same holds for Greenberg’s norm; in asking whether to believe that p, you are asking whether to take the doxastic attitude of believing towards p. Now according to Greenberg’s norm, if you take a doxastic attitude towards p, you ought to believe that p (in contrast to disbelieve that p or suspending belief with respect to p) iff p is true. Therefore, by asking whether to believe that p, you are led to ask whether p.

Those favoring the transparency argument have not distinguished between the question-shifting and answer-shifting aspects of transparency, and this is unfortunate. As I argued, when you settle whether p, namely answer the question of whether p, there is a cognitive contribution on the part of the agent which amounts to regarding p as true which makes it, on my account, just believing that p. The same, however, does not hold for the question-shifting aspect. The mere act of questioning whether p does not require any belief about p and is distinct from the question of whether to believe that p. Therefore, although the question-shifting and the answer-shifting aspects of transparency are related, they are still distinct and require separate treatment. To better understand this point, consider the cognitive attitude of guessing. A guess is similar to a belief in many respects. For example, a guess is also correct only if it is true. Now, consider the question of whether to guess that p. This question is not immediately replaced by the question of whether p. After all, a guess is usually made in situations where evidence is not conclusive and there are several competing options. Therefore, it is not plausible to say that necessarily by asking whether to guess that p (rather than q), we have to ask whether p. Moreover, further to the truth of the content of the guess, the reward (or punishment) that would be ensuing following making a particular guess is also relevant to the question of whether to guess that p (in contrast to q). Therefore, it is not the case that the question of whether to guess that p comes down to the question of whether p. However, if you somehow succeed in settling whether p, you have thereby settled whether to guess that p. This means that if you find p true, then you have no other option than guessing that p. This, I hope, lends further credence to my claim that the question-shifting and the answer-shifting aspects, though related, need to be treated separately; they need not stand or fall together.Footnote 7

One may object that if what I said is correct then there is no doxastic deliberation. Because if settling whether p is just believing that p, then there is nothing genuinely called doxastic deliberation. However, my account does not discount the importance of doxastic deliberation. Thinking about what to believe is an important question that we keep asking ourselves in our daily life. If believing that p is regarding p as true, then doxastic deliberation is the question of whether to regard something as true. This is of course an important question. The question of whether p is true is conceptually distinct from this question, but settling the question of whether p is just regarding p as true. Consider the example that Shah and Velleman give to show that doxastic deliberation is possible. Suppose that you hear that Iraq has chemical weapons (Shah & Velleman, 2005, p. 502). When encountering this proposition, it is plausible to ask whether to believe it. And intuitively you are immediately led to ask whether it is true.

But as I argued, settling the factual question is not something other than believing (or disbelieving) that p. If you settle that Bush is right, you already believe that Iraq has chemical weapons. Or if you settle that Bosh is lying, you already disbelieve that Iraq has chemical weapons. But settling whether p is an important task; you have to look for considerations for and against the claim made to see whether they are strong enough. Therefore, doxastic deliberation is just the cognitive activity of determining the truth or falsity of a proposition. I have no problem with the idea that doxastic deliberation exists and is very important in our life. I just disagree with the specific way that Shah and Velleman interpret doxastic deliberation. Doxastic deliberation does not start from settling whether p and end in settling whether to believe that p; it is just settling whether p. I think my interpretation of transparency is faithful to our commonsensical understanding of doxastic deliberation.

If what I have said so far is correct, by distinguishing between the question-shifting and the answer-shifting aspects of transparency, the argument from transparency can be kept as a good reason to embrace belief normativism. However, we should keep in mind that it is only the question-shifting aspect that is in need of explanation and is explained by appealing to a truth norm of belief. On the contrary, the answer-shifting aspect is not a genuine phenomenon and, as such, needs no explanation, be it normative or otherwise. Fortunately, the tensions between the various versions of the truth norm and transparency only concern the answer-shifting aspect, which I have argued is explanatorily empty.

5 Conclusion

Once advertised as one of the virtues of the normativist accounts of belief, the phenomenon of transparency has become a stumbling block for a plausible version of the truth norm of belief, because it cannot explain this phenomenon and is, therefore, charged with explanatory inadequacy. In this paper, I argued that the problem is not limited to the permissive version of the truth norm, and also bedevils the injunctive versions, albeit in a different way. Instead of taking this to undermine belief normativism, I argued that transparency consists of two aspects which should be (but have not been) distinguished. What I called the question-shifting aspect is a genuine phenomenon which cries out for an explanation and is best explained by holding that belief is conceptually governed by a truth norm, irrespective of how this norm is spelled out. However, the charge of explanatory inadequacy of the truth norm concerns what I called the answer-shifting aspect of transparency, which I, drawing on the deflationary conception of the aim of belief, have argued is explanatorily empty and, as such, in need of no explanation. Therefore, the inability of truth norms of belief to explain this answer-shifting aspect does not tell against their viability. For this reason, my argument keeps the merit of the argument from transparency for belief normativism while exonerating the formulations of the truth norm from the charge of explanatory inadequacy.