Believing for truth and the model of epistemic guidance

ABSTRACT

Belief is said to be essentially subject to a norm of truth. This view has been challenged on the ground that the truth norm cannot provide guidance on an intuitive inferentialist model of guidance and thus cannot be genuinely normative. One response to the No Guidance argument is to show how the truth norm can guide belief-formation on the inferentialist model of guidance. In this paper, I argue that this response is inadequate in light of emerging empirical evidence about our system of belief-formation. I will then motivate an alternative response and present, in rough outline, a viable, reason-responsive model of epistemic guidance on which the truth norm can guide.

KEYWORDS:

• Epistemic guidance
• the truth norm of belief
• reason-responsiveness
• epistemic normativity

1. Introduction

It is a platitude that beliefs can be correct or incorrect, and it is widely accepted that truth is the standard of correctness for belief. A belief that p is correct if and only if p is true. Call this claim Doxastic Correctness.¹ Many have appealed to Doxastic Correctness and its normative import to do philosophical heavy-lifting. Philosophers of mind use it to individuate belief –belief is the type of attitude that is essentially subject to a norm of truth.² Some theorists of truth appeal to the normative import of Doxastic Correctness to argue that truth is a substantive property, based on its role as the norm of belief and assertion.³ Epistemologists use the presumed normative import of Doxastic Correctness to provide a rationale for truth-conducive accounts of justification.⁴ It has also been used to distinguish the ‘right kind of reason’ from the ‘wrong kind of reason’ for belief – only truth-related considerations are the right kind of reasons for belief.⁵

The list goes on, but does Doxastic Correctness have the kind of normative import to do the philosophical heavy-lifting that many have taken for granted? One widely embraced and frequently invoked idea across normative philosophy is that norms provide guidance.⁶ In recent discussion, the idea that the truth norm can guide has been challenged (e.g. Glüer and Wikforss 2009, 2010, 2013). On a standard inferentialist model of guidance, to be guided by a norm of the form ‘$\phi$ if and only if C’, is for one to first form a belief about C and then $\phi$ (or not $\phi$) on the basis of the belief about C. But it seems impossible to be guided by the truth norm on the inferentialist model. For to form a belief about C, in this case, is to form a belief about whether p is true. But to form a belief about whether p is true is just to form a belief about p. So, it seems that to be guided by the truth norm to form a belief about p, one already has to form a belief about p, which is the very same thing one seeks guidance for in the first place. Thus, it is impossible for one’s belief-formation to be guided by the truth norm of belief. Or so the objection goes.

The aim of this paper is to defend the truth norm of belief in response to the guidance challenge. Instead of trying to account for how the truth norm can guide on the inferentialist model, I will argue that the inferentialist model is not the only way to understand guidance. I will propose an alternative, reason-responsive model of guidance, according to which, to be guided by a norm is to respond to the relevant normative reasons. The truth norm of belief can guide our belief-formation.

The plan is as follows. I will begin by presenting Glüer and Wikforss’ No Guidance argument and briefly consider a popular response to the argument which maintains the inferentialist model of epistemic guidance (Sect. 2&3). I will argue that this response is inadequate in light of emerging empirical evidence about our system of belief-formation (Sect. 4). I will motivate an alternative response to No Guidance which rejects the inferentialist account of epistemic guidance (Sect. 5). I will then present, in rough outline, a reason-responsive model of epistemic guidance. By appealing to this model of epistemic guidance, I will offer a full response to the guidance challenge (Sect.6).

2. The No Guidance argument

Glüer and Wikforss (2009, 2010, 2013) start by sketching what they claim to be an intuitive picture of guidance. They claim that any generic guiding rule has the form:

(R) Do X when C.

To be guided by R, S must first form a belief as to whether the trigger condition C is met (2010, 758, 2013, 83). Call this the Doxastic Constraint. Second, S’s doing X must be in virtue of her acceptance of R: ‘to be guided by a norm or rule in our performances intuitively requires that R influences or motivates, or provides reasons for these performances.’ (2010, 758). In other words, if S’s doing X is guided by R, then S’s belief that C obtains, and her acceptance of R must motivate S’s doing X (and justify S’s doing X if R is a norm). Call this the Reason Constraint.

On this intuitive picture of guidance, we can make sense of how a rule like ‘buy low and sell high’ can guide. To be guided by that rule is for S to first come to have a belief as to whether the price is low or high, and for that belief and S’s acceptance of the rule to explain and rationalize S’s selling or buying things. However, Glüer and Wikforss argue, we cannot make sense of how the truth norm of the form:

(T) Believe that p when p is true,

can guide our belief formation on this intuitive picture of guidance, since the Reason Constraint cannot be satisfied. They argue:

‘First, the norm cannot influence belief formation, since in order to be guided by [the truth norm] one already had to form a belief as to whether p. The very question we wanted guidance on—whether to believe that p—needs to have been answered before any such guidance can be received. Thus, [the truth norm] always comes ‘too late’, as it were, to do the guiding work it is supposed to do. Second, [the truth norm] never makes a difference in the following sense, either: it never “tells me” or “provides me with a reason” to believe anything other than what I have come to believe anyway.’ (Glüer and Wikforss 2013, 84, their emphasis; see also 2010, 759)

One natural way to make sense of their argument, is to interpret Glüer and Wikforss as endorsing an inferentialist model of guidance.⁷ The inferentialist model of guidance is attractive because it captures what it is for the subject to act in virtue of being guided by R as opposed to merely happening to do X when C. The intuitive idea is that the ‘in virtue of’ relation can be cashed out in terms of S’s performing an inference. To be guided by the truth norm, on the inferentialist model of guidance, is to infer in accordance with the truth norm.

If one tries to infer in accordance with the truth norm, we see that the input belief required by the Doxastic Constraint is the very same belief that one seeks guidance for. Guidance from a norm is supposed to be normative: your response should gain some normative support in virtue of being guided. On the inferentialist model of guidance, the fact that you make a correct inference in accordance with the norm in question should provide you with some justification for doing X. But problems arise when one tries to make an inference in accordance with the truth norm. Inferring a belief in accordance with the truth norm does not seem to provide (any additional) justification for one’s belief since it seems to involve inferring a belief that p on the basis of the very same belief that p. But most beliefs – if not all beliefs – cannot gain any normative support in such inferences beyond what they already enjoy. So, one may argue, inferring in accordance with the truth norm cannot provide justification for believing p.

Moreover, it seems that inferring a belief in accordance with the truth norm cannot be the motivating reason for which you come to have that belief. If I ask you why you believe that the Pfizer Covid-19 vaccine is effective, and your answer is that you are guided by the truth norm, i.e. you infer the belief after you formed the belief that the Pfizer Covid-19 vaccine is effective, you do not really explain to me the motivating reason for which you believe that the Covid-19 vaccine is effective. Being guided by the truth norm does not seem to be the motivating reason for which you form the belief that the Covid-19 vaccine is effective. It seems that the Reason Constraint cannot be satisfied in the case of following the truth norm. For any S, p, S’s inferring p in accordance with the truth norm provides neither motivation nor justification for S’s believing p. We can reconstruct the No Guidance argument as follows:

(P1) A genuine norm is essentially capable of guiding – one must be able to act in response to that norm.

(P2) For any S, p, if S’s belief that p is guided by a norm, then S must infer the belief that p in accordance with S’s representation of that norm, which provides both motivation and justification for S’s believing that p.

(P3) For any S, p, it is impossible for S to believe that p by making an inference in accordance with the truth norm, which provides both motivation and justification for S’s believing that p.

(C1) Therefore, for any S, p, it is impossible for S’s belief that p to be guided by the truth norm. (P2 & P3)

(C2) Therefore, the truth norm is not genuinely normative. (P1 & C1)

In this paper, I will discuss two possible responses to No Guidance. I will argue that rejecting P3 is inadequate in light of emerging evidence about our belief-forming system. I will then motivate an alternative, more demanding response to No Guidance, which is to reject P2.

3. Rejecting P3

One popular response to No Guidance is to deny P3 and argue that making an inference in accordance with the truth norm can provide both motivation and justification for one’s belief (c.f., Steglich-Petersen 2010, 2013).

Consider first why we might think inferring in accordance with the truth norm can provide normative support for one’s belief in additional to what it already enjoyed prior to the inference. We can make a distinction between reasons and enabling conditions under which one can respond to a reason. Your belief that the price is low is an enabling condition for you to act on that reason, but that enabling condition, i.e. your belief, itself is not the reason in favour of buying things for which you act under the guidance of ‘buy low and sell high’. The reason in favour of buying things for which you act is the fact that the price is low. Likewise, the input belief about whether p is true merely plays an enabling role for one to apply the truth norm. As an enabling condition, the input belief that p is not itself the reason in favour of believing p for which one believes p under the guidance of the truth norm. We should not conflate the conditions under which the truth norm recommends beliefs with the psychological states one must have in order to apply that norm. The condition under which the truth norm recommends belief is the fact p, not the belief that p. With this clarification, one may argue, to infer in accordance with the truth norm is not to form a belief that p on the basis of the very same belief that p, but to infer a belief from a fact. Thus, making an inference in accordance with the truth norm can provide normative support for believing that p.

Consider next why we might think inferring in accordance with the truth norm can provide a motivating reason for belief. One might think that Glüer and Wikforss have a too narrow understanding of what it takes for a norm to influence or motivate one’s belief-formation. To see this, Steglich-Petersen (2010, 2013) invites us to imagine S, who adopts a rule of pleasure, which tells her to believe that p just in case believing p is pleasant. To be guided by the pleasure rule, S would first try to determine whether believing p is pleasant and make an inference about whether to believe that p in accordance with that rule. Had S instead been guided by the truth norm, she would have directed her efforts to determine whether p is true. Now, suppose that p is true but believing p is unpleasant. If S were to be guided by the truth norm, she would believe that p. If S were to be guided by the pleasure rule, she would refrain from believing p. By contrasting the two cases, we can see that S’s belief can be influenced and motivated by her making an inference in accordance with the truth norm, since S could have ended up with a different belief had she adopted a different rule such as the rule of pleasure. Whether one makes an inference in accordance with the truth norm or the pleasure rule does make a difference as to whether one would have that belief or not.

These considerations raise serious doubts about P3. However, in the next section, I will argue that emerging empirical evidence on our system of belief-formation poses a problem for this general line of response. Since the problem is not simply that making an inference in accordance with the truth norm cannot provide motivation and justification for one’s belief, but rather, forming a belief by making an inference in accordance with the truth norm does not seem to be something we can (or usually) do given the kind of belief-forming system we have, empirically speaking.

4. Spinozan system of belief-formation

Empirical studies of belief in fields such as delusion, self-deception, heuristics, biases, cognitive dissonance, and conspiracy theories have been growing at a rapid speed in the past decades.⁸ The emerging scientific picture of our belief-forming system is one that is fast, automatic and subject to many psychological and social factors often out of our direct control. It is a picture at odds with the traditional Cartesian picture held by many epistemologists, which is the backbone of the inferentialist model of epistemic guidance. On the Cartesian picture of belief-formation, one first entertains a proposition p and then either arrives at the belief that p or rejects it.⁹ It is during this evaluative process one may seek guidance from epistemic norms, which tell us whether to form the belief that p or to reject it. The emerging scientific picture of belief, however, aligns much better with the so-called Spinozan account of belief-formation (Egan 2008; Gilbert 1991; Gilbert, Tafarodi, and Malone 1993; Mandelbaum 2014). On the Spinozan account, one forms beliefs more or less automatically when representational contents are made available through cognitive mechanisms such as perception and they are not products of a norm-invoking, effortful inference at the person-level. Only a relatively small number of beliefs will go through a further evaluative process during which some of the beliefs may be rejected, provided that they are subject to an unfavourable verdict. In this section, I will first argue that our belief-forming system is likely to be Spinozan for a number of reasons. I will then explain in more detail why this poses a problem for the common response to the No Guidance argument we considered above.

4.1. Supporting evidence for the Spinozan account of belief-formation

To begin with, it would seem that from a purely evolutionary point of view, it is more likely that our belief-forming system is Spinozan as opposed to Cartesian. Beliefs play important action-guiding roles when paired with the relevant desires, and it is essential for us to have efficient belief-forming system that can help us navigating in a fast-changing world. Given our limited cognitive resources, it is a good thing to have a belief-forming system which can quickly and reliably add information to our stock of beliefs without taking extra cognitive efforts. If we had to undergo an evaluative process to turn the information we acquired through, say, perception into belief, then we would not be able to take up perceptual beliefs as efficiently and reliably, especially when we are put under significant cognitive load, e.g. when we are multi-tasking, as we often do. Furthermore, going through an evaluative process takes up additional cognitive resources. So, one may expect that a well-adapted effective mental economy is one where we apply evaluative processes selectively to only a small number of beliefs, e.g. beliefs flagged as suspicious or important.

Of course, there is a risk of feeding a significant number of false beliefs into our mental economy with the Spinozan system of belief-formation in place, since beliefs are added to our stock of beliefs without first being subject to evaluation and scrutiny. However, it is plausible that, the cost of failing to add information to our belief stock, or delaying the uptake, may well be higher than the cost of occasional uptake of false beliefs. When I see that a car is coming towards me at a high speed, there is a much higher chance to survive if I act on the unevaluated perceptual belief and run to safety, than having a belief-forming system which takes time to first evaluate the perceptual experience before delivering it as a belief which then guides me to run to safety.

In addition to this admittedly somewhat suggestive evolutionary consideration, there are important empirical results that provide evidence for the Spinozan hypothesis (Gilbert 1991; Gilbert, Tafarodi, and Malone 1993; Hasson, Simmons, and Todorov 2005; Masip, Garrido, and Herrero 2006; Skurnik et al. 2005; Unkelbach 2007). For example, Gilbert, Tafarodi, and Malone (1993) designed six experiments which placed one group of participants under a cognitive load condition – a disabling performance constraint – to make it harder for the participants to go through an evaluative process and compared these with a control group of participants who were not placed under a cognitive load. Broadly speaking, a Cartesian account of belief-formation would predict that participants under a cognitive load condition are less likely to arrive at a doxastic state compared to the control group and that the cognitive load will affect the forming and the rejection of belief equally. In contrast, a Spinozan account would predict that additional cognitive load will disproportionally affect a subject’s rejection of a belief rather than the acquisition of a belief. In each of the six experiments conducted by Gilbert, Tafarodi, and Malone (1993), researchers found that participants who were under cognitive load were much worse at rejecting falsehoods and arrived at more false beliefs than those who were not under cognitive load, but that both groups were almost equally good at arriving at true beliefs, inconsistent with the prediction of the Cartesian account.¹⁰

Furthermore, on the Cartesian account, what underlies our belief-formation is a controlled and reflective evaluative process. However, it is widely acknowledged that what often underlies our belief-formation in decision-making in a wide range of contexts is not a controlled and reflective evaluative process but heuristics and biases, which allow us to make quick and effective judgments with relatively low cognitive costs.¹¹ The Spinozan account of belief-formation can be seen as part of dual-process theories of the mind, which are default-interventionist in structure (Evans and Stanovich 2013; Kahneman and Frederick 2002). According to dual-process theories, there are two types of processes underlying thinking and decision-making. Type 1 process (e.g. automatic and reflexive) generates default responses on which subsequent Type 2 process (e.g. reflective and controlled) may or may not intervene. The empirical evidence on heuristics and biases casts doubt on the Cartesian account. The Spinozan account, by contrast, is easily compatible with the evidence that heuristics and biases underlie our belief-formation. On the Spinozan account, one forms beliefs more or less automatically through Type 1 process when representational contents are made available through various cognitive mechanisms, which can be susceptible to the influence of heuristics and biases. Such beliefs may subsequently be subject to evaluation through Type 2 process.

Together with heuristics and biases, the Spinozan hypothesis offers a better explanation of a much broader scope of phenomena such as self-deception, conspiratorial thinking and the easy uptake of misinformation when one’s epistemic environment is impoverished. In paradigm cases of self-deception, for example, self-deceptive subjects form beliefs insensitive and irresponsive to the evidence available to them even if they do subscribe to epistemic norms and are in a position to apply them correctly (e.g. they have the cognitive capacities to make inferences and possess the relevant evidence).

Let me illustrate this point with an example. Consider Lena, whose beliefs are by and large sensitive and responsive to evidence. However, she is self-deceptive when it comes to her career prospects. She believes that her debut novel is an enormous success, that she will receive a fellowship from a prestigious university, and so on, even though she has plenty of evidence suggesting the contrary. On the Cartesian picture, it is puzzling why self-deceptive subjects like Lena, who are otherwise responsible epistemic agents, would persistently fail to evaluate a large number of propositions concerning her career prospects. If an evaluative process underlies each belief-formation, why would Lena come to believe that her novel is an enormous success and that she will receive a fellowship, when she has plenty of evidence that the novel has not been well received? Likewise, it is puzzling, why many epistemic agents, who are otherwise responsible epistemic agents, would persistently fail to correctly evaluate a large number of propositions concerning conspiracy theories and other misinformation.

Now, of course, this is not to say that there can be no viable psychological and philosophical explanation as to why an otherwise responsible epistemic agent may persistently fail to evaluate her beliefs correctly on the Cartesian account. We might as well just bite the bullet and accept that we are often very bad at forming beliefs concerning certain subject matters even if we are in general quite capable of forming beliefs responsibly. My point here is a contrastive one. The Spinozan account offers a more attractive explanation of the phenomena of self-deception, conspiratorial thinking and the easy uptake of misinformation. Let me explain.

On the Spinozan account, evaluation does not occur prior to belief-formation. Self-deceptive, conspiratorial and misinformed beliefs are added to the subject’s belief stock in automatic, reflexive processes susceptible to the influence of various heuristics and biases.¹² These self-deceptive and conspiratorial beliefs are likely to be ‘protected’ from undergoing a further evaluative process due to various social and psychological factors, e.g. it is painful for Lena to believe that her debut is a total failure and beliefs in conspiracy theories often help alienated individuals to regain a sense of control in a fast-changing world.¹³ If self-deceptive and conspiratorial beliefs do not typically undergo an evaluative process in the Spinozan belief-forming system of those epistemic agents who are under certain social and psychological conditions, that explains why those epistemic agents who are otherwise responsible epistemic agents would end up with a large number of self-deceptive or conspiratorial beliefs. It also explains why the uptake of misinformation can be easy and widespread among epistemic agents whose epistemic environment is impoverished. On the Spinozan picture, beliefs are added to our belief stock from our information ecosystem in automatic, reflexive processes, and only a relatively small number of them will go through a further evaluative process. So, for those epistemic agents living in an impoverished epistemic environment filled with false information, there is no surprise that they might form a large number of false beliefs despite most of them are responsible epistemic agents and subscribe to epistemic norms.

The considerations I presented above, taken together, make a compelling case for the Spinozan account of belief-formation. The aim of our discussion of the dispute between the Spinozan and Cartesian system of belief-formation, however, is not to dismiss one and endorse the other. Rather, my goal is to show that the empirical evidence for the Spinozan account of belief-formation poses a problem for the common response to the No Guidance argument we considered above.

4.2. Systems of belief-formation and guidance

How does this emerging picture of our belief-forming system bear upon the debate on the No Guidance argument? In the previous section, we considered a response to No Guidance by rejecting P3 while maintaining the inferentialist model of guidance. But the emerging evidence on our belief-forming system suggests that this general line of response is inadequate. For a deeper challenge for normativists is not simply that if one were to infer in accordance with the truth norm whether that can provide both motivation and justification for one’s belief. Rather, the issue seems to be that given our belief-forming system is Spinozian in character, how belief-formation, quite generally, can be the product of making an inference in accordance with the truth norm. And if not, belief-formation cannot be guided by the truth norm on the inferentialist model of guidance. This, I believe, is a serious and underdiscussed challenge for normativists who want to reject P3 and show that the truth norm can provide guidance on the inferentialist model.¹⁴

Let us unpack this challenge more carefully and consider two possibilities. First, if our entire belief-forming system is Spinozan, then no belief can be formed as the product of making an inference in accordance with the truth norm. Rather, they are formed in automatic, reflexive processes when representational contents are made available through cognitive mechanisms such as perception. On this picture, only a small number of beliefs undergo a further evaluative process, during which an effortful, norm-invoking inference at person-level may occur.¹⁵ So, if our entire belief-forming system is Spinozan, no belief-formation can be guided by the truth norm, or indeed by any norm whatsoever, although some of our belief-revision may be guided by epistemic norms.

Indeed, this consideration has motivated some to endorse the view that all epistemic norms are norms of belief-revision (e.g. Helton 2020). We cannot properly assess this position here, but I think it is fair to say that most philosophers, including the critics of the truth norm, reject global scepticism about epistemic norms of belief-formation. The main takeaway point, in any case, is that no epistemic norm can guide belief-formation on the inferentialist model of guidance if our entire belief-forming system is Spinozan, which is unwelcome news for normativists who accept the inferentialist model of guidance.

But of course, whether our entire system of belief-formation is Spinozan is a debatable empirical matter. As I acknowledged earlier, the evidence we considered is not conclusive against the Cartesian account of belief-formation. This leads us to the second possibility, given the evidence, that our belief-forming system being largely but not entirely Spinozan. Now, even if only a large part of our belief-forming system is Spinozan, we are still forced to conclude that a sizable proportion of our belief-formation cannot be guided by the truth norm (or any other epistemic norms) on the inferentialist model of guidance. Only a small number of beliefs formed through a Cartesian evaluative process can be guided by the truth norm (i.e. if they are the products of inferring in accordance with the truth norm in cases of doxastic deliberation). However, this is a bad result for normativists who would like to use the truth norm to do the sort of philosophical heavy-lifting we sketched in Sect.1. If our belief-forming system is largely Spinozan, then it is not the case that for any S, p, S’s belief that p can be guided by the truth norm, as far as belief-formation is concerned. The truth norm can at best be a norm governing a small number of beliefs formed in the Cartesian system and a small number of beliefs revised in the Spinozan system. Most beliefs are not subject to or guided by a norm of truth. Normativists would therefore have to give up on the claim that belief is essentially subject to and guided by a norm of truth.

Thus, rejecting P3 while maintaining an inferentialist model of epistemic guidance is not a satisfying response in light of emerging empirical evidence about our belief-forming system.

5. Motivating an alternative response to No Guidance

An alternative, more demanding response to No Guidance argument is to reject P2.¹⁶ It is a more demanding response since, first, the inferentialist model of guidance is popular among philosophers. Rejecting this intuitive model seems an unappealing and costly move. Second, normativists who adopt this response would have to provide an alternative model of epistemic guidance and show that the truth norm can guide belief-formation on that model. Neither task is easy. In the remainder of this paper, I want to make some headway on both fronts.

5.1. The problem of unworthy guidance

Despite the popularity of the inferentialist model of guidance, recent work on inference and normative theories has casted some doubts on this model. One potential difficulty for the inferentialist model of guidance is what I will call the problem of unworthy guidance.

It is a familiar thought in ethics that one’s action can be morally correct while lacking moral worth. The most famous example is Kant’s shopkeeper who treats his customers fairly but only out of self-interest. The shopkeeper does the morally right thing but his action has no moral worth: he has no genuine concern or respect for moral laws (Kant 1998, 4:397).

One natural explanation for the lack of moral worth of the shopkeeper’s action is in terms of normative guidance. The shopkeeper’s action lacks moral worth because his action is not guided by the relevant moral norms but by self-interests.

On the inferentialist model, to be successfully guided by a moral norm is to make a correct inference in accordance with that norm. However, the inferentialist model faces a challenge: There are cases where the subject makes the correct inference in accordance with the relevant norm but nevertheless their response lacks the kind of worth that norm-guided responses deserve.

Consider Gendo who recently made a substantial donation to the World Food Programme. Suppose that he did so by making a correct inference in accordance with the norm of beneficence. However, Gendo does not think there is any moral reason to be beneficent and he believes that all moral talk is nonsense. For Gendo, a ‘norm’ of beneficence is just like any other societal rule that he follows when doing so is expedient for him to navigate through the world. It seems that Gendo’s action lacks moral worth because he shows no concern or respect for morality. On the inferentialist model of guidance, however, Gendo’s action is guided by the norm of beneficence. Since responses guided by a moral norm has moral worth, it follows that Gendo’s action has moral worth. But this contradicts the intuitive verdict that Gendo’s action lacks moral worth.

It seems then, the inferentialist model of guidance allows for unworthy moral guidance, i.e. cases where one’s responses are guided by moral norms but are not worthy. To avoid the problem of unworthy guidance, something has to go: Either the intuitive understanding of moral worth in terms of normative guidance is mistaken, or the inferentialist model of guidance is incorrect.

There is an analogous problem of unworthy guidance in epistemic cases, or so I shall argue. Just like an action can be morally correct while lacking moral worth, a belief can be correct while lacking epistemic worth (Whiting 2020). Consider Lena, who believes correctly that her debut novel is an enormous success. But Lena is an optimist. She would believe that her debut novel is an enormous success no matter how little evidence she has. Had her novel been a failure, she would still believe that it is an enormous success. It seems then, in this case, Lena’s belief lacks epistemic worth for she has no concern or respect for truth. Although she believes the right thing, she does not do so in a way that deserves credit for believing what is right.

One natural explanation for the lack of epistemic worth of Lena’s belief is in terms of normative guidance. Lena’s belief lacks epistemic worth because her belief is not guided by the relevant epistemic norms but by her blind optimism.

On the inferentialist model, to be successfully guided by an epistemic norm is to make a correct inference in accordance with that norm. However, it seems that there are cases where the subject makes the correct inference in accordance with the relevant norm but nevertheless their response lacks the kind of worth that norm-guided responses deserve.

Suppose that Gendo forms his beliefs correctly by making inferences in accordance with the truth norm of belief (leaving aside, for now, the controversy regarding whether this is possible). However, Gendo does not think there is any epistemic reason to believe the truth. For Gendo, a ‘norm’ of truth, too, is just like any other societal rule he follows when doing so is expedient for him to navigate through the world. It seems that Gendo’s beliefs lack epistemic worth because he shows no genuine concern or respect for truth.¹⁷ On the inferentialist model of guidance, however, Gendo’s beliefs are guided by the truth norm and thereby have epistemic worth. But this contradicts the intuitive verdict that Gendo’s beliefs lack epistemic worth.¹⁸

It seems then, the inferentialist model of guidance allows for unworthy epistemic guidance, i.e. cases where one’s responses are guided by epistemic norms but are not worthy. To avoid the problem of unworthy epistemic guidance, something has to go: Either the intuitive understanding of epistemic worth in terms of normative guidance is mistaken, or the inferentialist model of guidance is incorrect.

There are, of course, ways to respond to the problem by offering alternative accounts of moral and epistemic worth. This is not the place to assess such responses. The point here, is that the problem of unworthy guidance puts pressure on the inferentialist model of guidance for those who accept the intuitive understanding of moral and epistemic worth in terms of normative guidance.

5.2. The problem of vicious regress

A more devastating problem for the infertialist model of guidance is the problem of vicious regress raised by Boghossian (2008). The problem arises because inference itself is often understood as a kind of transition guided by rules of inference. This rule-guided conception of inference in combination with the inferentialist model of guidance leads to vicious regress. To illustrate, consider Boghossian’s ‘Email Rule’:

(Email Rule) ‘Answer any email that calls for an answer immediately upon receipt!’ (Boghossian 2008, 481)

On the inferentialist model of guidance, being guided by the Email Rule involves an inference in accordance with the Email Rule: I must first form a belief that the condition for applying that rule is met (i.e. that I received an email that calls for an answer), and then draw the conclusion with an intention to do what the rule requires me to perform and act accordingly.

There is a problem, however, since inference itself is guided by a rule of inference. Making an inference in accordance with a rule is to be guided by a rule of inference like the following:

(Inference Rule) If ‘Do X when C’ and C, do X.

Now if the inferentialist model of guidance is correct, being guided by the Email Rule involves making an inference in accordance with the Email Rule, which in turn involves being guided by the Inference Rule. If being guided by the Inference Rule itself requires yet another inference guided by the relevant rule of inference involved, then on the inferentialist model of guidance, making that inference would again require making another inference … Therefore, the combination of the inferentialist model of guidance and a rule-guided conception of guidance would lead us to an infinite vicious regress. To avoid the regress, something has to go: One should either reject the rule-guided conception of inference or the inferentialist model of guidance.

The problem of vicious regress has motivated some philosophers to reject the rule-guided conception of inference and propose that instead we should understand inference, for example, as a distinctive kind of response to some informational states that produces a conclusion (Siegel 2019); as a transition guided by rules built into the architecture of cognitive systems (Quilty-Dunn and Mandelbaum 2018); as an act of attaching inferential force to an argument (Hlobil 2019); or as responses regulated by some inferential pattern (Boghossian and Wright (ms)). It is beyond the scope of this essay to assess these alternative conceptions of inference. The point here is that the problem of vicious regress puts pressure on the rule-guided conception of inference as well as the inferentialist model of guidance. While much of the existing responses to the problem of vicious regress focus on revising our conception of inference, it is worth highlighting that by moving away from the rule-guided conception of inference, we in effect also move away from the inferentialist model of guidance found in the No Guidance argument, according to which, being guided by a norm involves a rule-guided inference which provides both motivation and justification for one’s belief-formation.

Our discussion shows that despite the popularity of the inferentialist model of guidance, it has been challenged in the wider literature. Thus, Glüer and Wikforss cannot simply take the inferentialist model of guidance for granted. A lot more need to be said to shore up their inferentialist conception of epistemic guidance. In fact, there are good reasons to reject P2.

Glüer and Wikforss might reply that there simply isn’t any alternative to the inferentialist model of guidance, for their model is the only one that can distinguish between genuine guidance and mere conformance. In the absence of an alternative, normativists have not fully met the guidance challenge. In the next section, I shall propose a non-inferentialist model of guidance on which the truth norm can guide belief-formation.

6. Guidance as reason-responsiveness

Contemporary normative theorizing has taken a turn to reason. It is an increasingly popular view that normativity is a matter of normative reason (e.g. Parfit 2011; Raz 2010; Schroeder 2007; Skorupski 2007; Dancy 2004; Scanlon 1998). A natural move for normativists then, is to develop an account of epistemic guidance on a reason-based normative framework. In this section, I will suggest that to be guided by an epistemic norm is for one to respond to the relevant epistemic reasons in forming one’s beliefs (REG). By appealing to REG, normativists can meet the guidance challenge: The truth norm of belief is a genuine norm capable of providing guidance.

6.1. REG

Let me begin by spelling out the reason-responsive view in more detail. The basic idea is simple: For one’s belief-formation to be guided by the truth norm is for one to form a belief that p in response to the reason that p is true.¹⁹ Let us begin our discussion with a characterization of the reason-responsive model of epistemic guidance (REG):

(REG) For any S, p, S’s forming a belief that p is guided by an epistemic norm according to which R is a normative reason to believe that p, just in case S forms the belief that p in response to R, provided that R is available to S.

Three clarifications are in order. First, I take normative reasons to be facts that count in favour of various responses. The fact there is a climate change is a normative reason to cut our carbon footprint. The fact that Smith has an alibi is a normative reason to believe that Smith is innocent. Normative reasons should be distinguished from motivating reasons. Motivating reasons are considerations to which an agent responds, which may or may not be facts that count in favour of that response.²⁰ For example, the motivating reason for which Sam got drunk at a party was the peer pressure to drink more. But the fact that there was peer pressure to drink more does not count in favour of getting drunk at a party, and thus not a normative reason to get drunk.

Second, a normative reason R to believe a proposition is available to S only if S has some epistemic access to R.²¹ I assume that S has epistemic access to R just in case S is in a position to know R through perception, intuition, testimony, memory, reasoning or other sources of knowing a fact. So, for example, the fact that the pavement is wet is a normative reason available to S only if S is in a position to know that the pavement is wet, e.g. by seeing that the pavement is wet, remembering that the pavement is wet and so on. The question of reason-responsiveness does not arise if R is not available to S, for one cannot respond to a normative reason outside one’s ken.

Third, I understand reason-responsiveness in dispositionalist terms. To believe that p in response to a normative reason R is for S to manifest some dispositions to use R as a normative reason to believe that p. A disposition, as commonly understood, can be characterized as a function mapping input conditions to the corresponding output conditions.²² For example, solubility can be characterized as a function mapping some input condition (e.g. being placed in water) to some output condition (e.g. being dissolved). Anger can be characterized as a function mapping some input conditions (e.g. being insulted, being hit) to some output conditions (e.g. face turning red, fists clenching). If an object O has a certain disposition with input conditions Xs and output conditions Ys then the following counterfactual is usually true of O: If O were Xs, then O would be Ys.²³ An epistemic disposition to respond to a normative reason R can be characterized as a function mapping certain input conditions about the external world, i.e. reasons, to S’s doxastic attitudes. S believes p in response to R just in case S’s believing p manifests some of S’s epistemic dispositions mapping some input conditions about R to output conditions about S’s believing p.²⁴ To fix ideas, I will follow Lord’s (2018:, 138–139) formulation of the dispositionalist account of responding to a normative reason:

(DRR) S forms a belief in response to a normative reason R just in case S’s believing p manifests a disposition to believe that p when R constitutes a normative reason to believe that p and to revise the belief that p when R ceases to be a normative reason to believe that p.

There are two ways in which R may cease to be a normative reason for believing p. First, since reasons are factive, R ceases to be a reason when it goes from obtaining to not obtaining. For example, the consideration that the pavement is wet ceases to be a normative reason in favour of believing that it is raining when the pavement is not in fact wet. Second, R ceases to be a reason when the favouring relation is disabled and thus R no longer counts in favour of believing p. For example, the fact that the cup appears red to me ceases to be a reason for me to believe that the cup is red when the lighting condition in my room is not normal and thus the appearance of the red cup no longer counts in favour of believing that the cup is red.

Note that most proponents of DRR take the reason-responsive disposition to be a primitive notion which resists further reductive analysis. That being said, it is also widely accepted that the manifestation of a disposition involves modal robustness. If S’s believing p manifests the relevant reason-responsive disposition to believe that p, then the following counterfactual is usually true: In the absence of overriding normative reasons, S would not believe that p if R were to cease to be a normative reason for believing p and S would believe that p if R were to obtain, provided that S has epistemic access to R. For example, given that I would not believe that the cup is red if the consideration that the cup appears red ceases to be a reason for me to believe that the cup is red (e.g. when the reason does not obtain or is disabled) and I would believe that the cup is red when the cup appears red constitutes a reason for me to believe that the cup is red (e.g. when the reason obtains and is not disabled), my belief manifests the disposition to respond to that particular perceptual evidence as a reason for belief. The falsity of the counterfactual usually indicates that my belief does not manifest the relevant reason-responsive disposition. Although the reason-responsive dispositions are not analysed in counterfactual terms, counterfactuals can be a good guide to whether one manifests the relevant dispositions or not (see also, Lord 2018, 138 fn.24).

With these clarifications in place, we can now flesh out REG in terms of DRR:

(REG-Fleshed Out) For any S, p, S’s forming a belief that p is guided by an epistemic norm according to which R is a normative reason to believe that p, just in case S’s believing that p manifests a disposition to believe that p when R constitutes a normative reason to believe that p and to revise the belief that p when R ceases to be a normative reason to believe that p, provided that R is available to S.

6.2. Why REG?

For our purposes, we need not give a full defence of REG here. It suffices to say that REG is an attractive alternative to the inferentialist account of guidance.

First, REG can distinguish between genuine guidance and mere conformance. The infernetialist model is not the only one that can capture this distinction. According to REG, to be guided by an epistemic norm is for S to form a belief in response to the relevant epistemic reasons in a way that manifests S’s reason-responsive dispositions. For example, Lena the optimist who believes in conformance with the norm of evidence that her novel is an enormous success is not guided by the norm of evidence. This is because although Lena’s evidence happens to support her belief, she would still believe that her novel is an enormous success had her evidence been flimsy. Lena’s belief does not manifest the relevant disposition to respond to her evidence as a reason for belief. Merely forming a belief that p in the presence of evidential reasons without manifesting the relevant reason-responsive dispositions, on REG, does not count as being guided by the norm of evidence.

Second, REG enables us to explain how an epistemic norm can guide belief-formation in a way compatible with both Cartesian and Spinozan account of our system of belief-formation. Recall that, if the inferentialist model is correct and our entire belief-forming system is Spinozan, no belief-formation can be guided by a norm. Even if only just a part of our belief-forming system is Spinozan, we are still forced to conclude that a sizable proportion of our belief-formation cannot be guided by a norm on the inferentialist model of guidance. Only a small number of beliefs formed through a Cartesian evaluative process can be guided. By contrast, the relevant reason-responsive dispositions, in principle, can be manifested in both a controlled, reflective process where one first entertains a proposition p and then either arrives at the belief that p or rejects it, as well as in an automatic, reflexive process where beliefs are added to one’s stock of beliefs without undergoing evaluation. Every belief can be guided by epistemic norms on REG, even if our entire system of belief-formation is Spinozan.

Third, REG offers a natural explanation as to why guided beliefs are epistemically worthy. A guided belief has epistemic worth because it is a belief formed in response to the relevant epistemic reasons available to S such that S manifests her reason-responsive dispositions. Recall that Gendo has no genuine concern for truth. For him, all normative talk is nonsense and he does not treat truth as a normative reason for belief. He believes whatever that is expedient for him to navigate through the world. The inferentialist model has trouble explaining why Gendo’s belief formed by inferring in accordance with the truth norm lacks epistemic worth, since correctly guided beliefs should be epistemically worthy. REG, by contrast, can deliver the right verdict. Gendo’s belief-formation does not manifest his disposition to respond to truth as a normative reason for belief. This is because he would still believe that p had p been false, as long as doing so is useful for him to navigate through the world. On REG, Gendo’s belief is not guided by the truth norm and thus his belief lacks epistemic worth. So, REG avoids the problem of unworthy guidance.

Fourth, on REG, guidance is explained in terms of reason-responsiveness. According to the account I presented above, reason-responsiveness consists of the manifestation of S’s dispositions to believe that p when R constitutes a normative reason to believe that p and to revise the belief that p when R ceases to be a normative reason to believe that p. Recall that, the inferentialist account of guidance runs into trouble because guidance is understood in terms of inference but making an inference is itself understood as a transition guided by the rule of inference. REG, by contrast, avoid the problem of vicious regress.

For these reasons, REG is an attractive alternative to the inferentialist model of guidance. By appealing to REG, normativists can offer a full response to the guidance challenge.

6.3. Meeting the guidance challenge

We are now in a position to articulate a novel account of what it is to be guided by the truth norm in belief-formation, applying REG to the truth norm of belief:

(REG-Truth) For any S, p, S’s forming a belief that p is guided by the truth norm just in case S’s believing that p manifests a disposition to believe that p when the consideration that p is true constitutes a normative reason to believe that p and to revise the belief that p when the consideration that p is true ceases to be a normative reason to believe that p, provided that p is available to S.

Whether truth as a reason for belief can be disabled is more controversial. One might insist that, necessarily, that p is true, if obtains, is a reason for believing that p and it cannot be disabled. The favouring relation is constitutive of the attitude type of belief. I discuss this issue elsewhere in more detail (Wei 2022). But for now, let us just consider two plausible candidates of disabling conditions. First, the favouring relation between the truth of p and the belief that p may be disabled if believing that p constitutes doxastic wrongdoing. For example, believing certain things about a marginalized community, even when they are true, can hurt. They often amplify an unjustified narrative about the community and put members of that community in harm’s way (Basu 2018). It may be argued that in such cases, the truth of p is disabled from constituting a normative reason for believing that p. Second, the truth of p may cease to be a normative reason for believing that p if it is not possible for ordinary epistemic agents like you and me to form a belief that p. For example, there are complex propositions that any ordinary epistemic agent with limited cognitive capacity to form doxastic attitudes about. The fact that we cannot form beliefs about those complex propositions due to our limited cognitive capacity disables the considerations about their truth from constituting normative reasons for us to believe those propositions.²⁵

Assuming what I said above is plausible, what does it take for S’s believing that p to manifest the disposition to respond to truth as a reason for belief? Recall that the reason-responsive disposition is a primitive notion typically understood as involving modal robustness. As such, counterfactuals can be a good guide to whether one manifests the relevant disposition or not. If one manifests the relevant disposition to believe that p in response to the normative reason that p is true, then the following counterfactual is usually true: In the absence of overriding reasons, S would not believe that p if the consideration that p is true were to cease to be a reason for believing that p (e.g. when the consideration fails to obtain or is disabled) and S would believe that p if the consideration that p is true were to constitute a normative reason (e.g. when the consideration obtains and is not disabled), provided that S has epistemic access, i.e. is in a position to know that p is true. The falsity of the counterfactual usually indicates that S’s believing that p does not manifest the relevant reason-responsive disposition and is not guided by the truth norm of belief.

On this account, then, our beliefs, including perceptual beliefs (that are likely to be formed via the Spinozan system of belief-formation) can be guided by the truth norm of belief. For all it takes for an epistemic agent to be guided by the truth norm is to manifest the disposition to respond to truth as a reason for belief. Now, there is a further question as to whether such dispositions are acquired or hard-wired. Some normativists (e.g. Wedgwood 2007) have argued that belief essentially consists of the disposition to respond to truth as a reason for belief. My view is that such dispositions are acquired through social interaction. Either way, for any epistemic agent, any p, it is possible for one to manifest the disposition to respond to the truth of p as a reason for believing that p. The truth norm of belief can guide belief-formation on REG. So, the No Guidance objection according to which the truth norm is not genuinely normative fails.

7. Conclusion

To sum up. I argued that rejecting P3 is an inadequate response to No Guidance in light of emerging empirical evidence about our system of belief-formation. I motivated a more demanding response and proposed an attractive, reason-responsive model of guidance. Of course, more should be said about REG and how the view can be defended, but that story has to be reserved for another occasion. At the very least, REG is an attractive alternative to the inferentialist model of epistemic guidance. By appealing to REG, I have shown that normativists can provide a full response to the guidance challenge. Truth is a genuine norm capable of guiding our belief-formation.

Acknowledgments

I am grateful to Philip Ebert, Paolo Savino and an anonymous referee for this journal for their helpful comments and discussion. This work was supported by the Arts and Humanities Research Council AH/T002638/1 ‘Varieties of Risk’, 2021–2022 Analysis Trust Research Studentship, Royal Institute of Philosophy 2020–2021 Jacobsen Studentship, the Aristotelian Society 2020–2021 Student Bursary, and the John Templeton Foundation PhD Studentship under Grant 58450.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Arts and Humanities Research Council: [Grant Number AH/T002638/1]; John Templeton Foundation: [Grant Number 58450].

Notes

1 See, e.g., Boghossian (2003), Engel (2013), Gibbard (2003, 2005), Shah and Velleman (2005), Velleman (2000), Wedgwood (2002, 2007, 2013).

2 See, e.g., Shah and Velleman (2005), Velleman (2000), Wedgwood (2002, 2007).

3 See, e.g., Dummett (1959), Engel (2002), Wright (1992).

4 The two dominant theories of justification, reliabilism and evidentialism, both appeal to the normative relation between truth and belief. On a reliabilist account, a belief is justified if and only if that belief is reliably produced. A reliable belief-forming process is one that produces many more true beliefs than false beliefs (e.g., Goldman 1979, 2008). On an evidentialist account, a belief is justified if and only if that belief fits one’s evidence (e.g., Feldman and Conee 1985). A consideration E counts as evidence for a belief that p if p is more likely to be true when E obtains than when E does not obtain. For further discussion concerning the truth connection, see, for instance, Conee (1992).

5 See, for instance, Hieronymi (2005), Reisner (2009).

6 See, e.g., Williams (1981), Darwall (1983), Nagel (1989), Gibbard (1990), Hursthouse (1999), Cruz and Pollock (1999). Notice, however, not all philosophers agree that genuine norms must be guiding (see, e.g., Ginsborg 2012; Hlobil 2015). In this paper, I follow most parties in the debate and take it for granted that normativity is prescriptive in nature and a principle is normative if and only if it can guide.

7 See, for instance, Boghossian (2008), Wright (1989, 2007).

8 For an overview, see Bortolotti (2009) on delusion, Von Hippel and Trivers (2011) on self-deception, Gilovich, Griffin, and Kahneman (2002) on heuristics and biases, and Douglas et al. (2019) on conspiracy theories.

9 For the ease of presentation, I take ‘rejecting’ a proposition p to include two possible doxastic states: One may either suspend judgment about p or one might arrive at the belief not-p.

10 It is debatable, however, whether affirming a statement p is a good indicator that the participant holds the belief that p, and the extent to which the structure of the task affects the result. For further criticism, see Street and Richardson (2015). Nevertheless, these data provide some prima facie evidence against the Cartesian account.

11 There is a wide range of evidence suggesting that there are various heuristics underlying our belief-formation (Tversky and Kahneman 1974), such as the availability heuristics (mental process used to make judgment about probability based on the ease with which the relevant content can be brought into mind), representativeness heuristics (mental process used to make judgment based on stereotypes and prototypes), anchoring and adjustment heuristics (mental process used to make judgment depending heavily on an ‘anchor’, e.g., an initial piece of information offered or an antecedent belief). There are also a wide range of biases underlying our belief-formation, such as confirmation bias, a tendency to form beliefs that support or confirm the beliefs one already has (Klayman and Ha 1987); partisan bias, a tendency to form beliefs based on one’s partisan identity (Van Bavel and Pereira 2018); self-enhancing bias, a tendency to form self-enhancing beliefs (Alicke and Sedikides 2009); and many more.

12 The Spinozan picture fits well with deflationist theories of self-deception, according to which self-deception consists of biased belief-formation influenced by mental states such as anxiety (Barnes 1997), desire (Mele 2001) or pretense (Wei 2020).

13 For an overview of the various factors underlying conspiratorial thinking, see, for example, Douglas et al. (2019).

14 Let me be clear. How beliefs are formed, empirically speaking, does not directly bear upon the normativity of belief-formation. The challenge here, is that the Spinozan account of our belief-formation, supported by psychological data, poses a problem for normativists who address the guidance challenge while maintaining the infernetialist model of guidance.

15 Note that some philosophers have a more permissive conception of ‘inference’ which concerns transition of states at the sub-person level. The intuitive interpretation of the inferentialist model of guidance, however, concerns norm-invoking inference at the person level. We will return to this point shortly.

16 Hlobil (2015) takes up a similar route in defence of content normativism, the view which says that the correctness conditions of contents are normative. Some of the points we make overlap.

17 An anonymous referee for this journal points out that in the epistemic cases, forming a correct belief by making an inference in accordance with the truth norm might suffice to render that belief epistemic worthy. After all, Gendo, unlike Lena, gets his belief right not by chance or luck, but in virtue of making a correct inference. His belief deserves credit. I do not deny that there is a sense in which Gendo does better than Lena. However, it is plausible that if showing respect for morality is an important aspect in our assessment of moral worth, so is showing respect for truth in our assessment of epistemic worth. Whether one’s belief is motivated by a concern for truth or mere utility does seem to make a difference to the epistemic worth of that belief.

18 Some might insist that one cannot be said to be making a correct inference in accordance with a norm if that person does not recognize the normative force of the norm. Since Gendo does not recognize the beneficence norm or the truth norm as genuinely normative, he cannot be said to be making correct inferences. So, Gendo’s responses are not norm-guided on the inferentialist model and thereby lack moral and epistemic worth. But this response seems ad hoc. There is no independent reasons why performing an inference requires one to appreciate the normative status of the rule in question.

19 I focus on belief-formation because this is what causes the most trouble for the normativists given the evidence about our belief-forming system.

20 There is an ongoing debate between externalism and internalism about the ontology of motivating reasons, whether they are facts, mental states, or something else. I am taking it for granted that at least some motivating reasons are facts, i.e., normative reasons. For a recent discussion, see, e.g., Alvarez (2018).

21 Just what exactly this access amounts to is controversial. Other popular candidates include, for example, knowing the fact (e.g., Williamson 2000) and having reflective access to the fact (e.g., Audi 1993).

22 Some philosophers reject the functional analysis of dispositions. For example, Vetter (2014) argues that a disposition does not involve input conditions.

23 Notice that this is not a counterfactual account of dispositions. The counterfactual is a useful but fallible guide to whether O has the relevant dispositions.

24 Since normative reasons are facts and whether they obtain is a matter external to S, REG is an externalist account about epistemic guidance. While the epistemic agent must have some epistemic access to the reason in question, there needs not be a stronger ‘conceivability’ condition such as understanding R, or believing R, etc., for S to manifest her reason-responsive dispositions. What mental state the epistemic agent S has with respect to R does not matter in the reason-responsive model of epistemic guidance. S is guided by the norm in so far as they manifest the right sort of dispositions where the input condition concerns the obtaining of R. An anonymous referee for this journal points out that one might manifest the relevant dispositions while completely fails to understand R. It seems problematic to say that in such cases one is guided by a norm. I have two things to say in response. First, the availability condition (i.e., the agent must have epistemic access to a reason R to respond to it) rules out the kind of case the referee describes here. If one completely fails to understand R (say, one lacks the relevant conceptual resources), then one does not have epistemic access to R. R is not a reason one can respond to and hence one cannot be said to manifest the sort of reason-responsive dispositions required to be guided by the norm in question. Second, if by understanding the referee means a belief/knowledge in R or/and why R is true, then I would say that such understanding is not required for one to respond to a reason. On my view, if I do have epistemic access to R, the robust dispositions I manifest suffices to capture what it is for me to form a belief ‘qua’/in virtue of responding to R, even if I do not have a belief about R or understand why R is the case.

25 I discuss such cases in more detail elsewhere (Wei 2019).