The weight of truth

Abstract

Belief is said to be subject to a norm of truth. A norm, intuitively, tells us what we ought to or may do. What sort of claim can truth make on us? On one standard view, the truth norm of belief is obliging. One ought to believe the truth and truth only. On another view, the truth norm of belief is permissive. One may believe the truth and truth only. Recently, it has been argued that the truth norm plays no interesting role in our normative theorizing for it issues excessive, unsatisfiable claims. This paper defends the truth norm of belief and proposes a novel answer to the question concerning its normative force on a reason-based framework. I argue that the normative force of the truth norm depends on the weight of truth as a normative reason for belief, which, just like that of any other normative reason, may vary across different contexts. I develop the idea that the weight of truth as a reason for belief is conditioned by what an epistemic agent can or cannot believe and modified by epistemic risk, among other things.

1 Introduction

For belief, the standard of correctness is truth.¹ The idea that belief is subject to a norm of truth is widely accepted and central to a variety of philosophical projects. A norm, intuitively, tells us what we ought to or may do. Recently, however, it has been argued that the truth norm of belief which says that one ought to believe p if and only if p is true issues excessive, unsatisfiable obligations (e.g., Bykvist & Hattiangadi, 2013; Glüer & Wikforss, 2013; Hattiangadi, 2010). These problems have motivated a permissive truth norm, according to which one is permitted to believe p if and only if p is true (e.g., Kalantari & Luntley, 2013; Raleigh, 2013; Whiting, 2010, 2012, 2013). However, as we shall see, there are a variety of cases in which the epistemic agent is arguably neither obliged nor permitted to believe truly and only truly. If so, what sort of claim can the truth norm make on us?

Although contemporary normative theorizing has taken a turn to reason, surprisingly few have approached the debate on a reason-based normative framework.² The aim of this paper is to propose an answer to the question concerning the normative force of the truth norm on such framework. I will argue that the normative force of the truth norm depends on the weight of truth as a reason for belief, which, just like that of any other normative reason, may vary across different contexts. Truth as a reason for belief functions holistically and can be conditioned and modified, or so I shall argue.

The plan is as follows. I begin by discussing three problems facing the requiring truth norm, which have been thought to favour the permissive truth norm. I argue that the permissive truth norm does not fare any better than the requiring truth norm (Sect. 2). I then offer a diagnosis on a reason-based normative framework. The requiring and permissive truth norms are what I will call invariantist views, for they claim that truth as a reason for belief has requiring/permitting weight across all contexts. I develop the idea that the normative force of the truth norm varies across different contexts for the weight of truth as a reason for belief is conditioned by what an epistemic agent can or cannot believe and modified by epistemic risk, among other things (Sect. 3). Finally, I contrast variantism favourably with two alternative views: hedged truth norm and perspectivism (Sect. 4).

2 The state of the debate

There are two standard views about the normative force of the truth norm in the current literature. According to the requiring truth norm (e.g., Gibbard, 2003, 2005; Wedgwood, 2002, 2007, 2013a, 2013c):

($\mathcal{T}_{\rm O}$) For any S, p, S ought to believe p if and only if p is true.

Versions of $\mathcal{T}_{\rm O}$ are in contrast with a weaker, permissive norm of truth (e.g., Kalantari & Luntley, 2013; Raleigh, 2013; Whiting, 2010, 2012, 2013):

($\mathcal{T}_{\rm p}$) For any S, p, S may believe p if and only if p is true.

$\mathcal{T}_{\rm P}$ seems just as normatively interesting as $\mathcal{T}_{\rm O},$ for it not only tells us what we are permitted to believe, it also tells us what we ought not to believe (Whiting, 2010, pp. 216–217). $\mathcal{T}_{\rm p}$ can play a significant role in our normative theorizing just like $\mathcal{T}_{\rm O}.$ In recent discussion, $\mathcal{T}_{\rm O}$ has been widely criticized on three grounds: (1) it makes excessive demands; (2) it has unpalatable consequences regarding true blindspot propositions; and (3) it clashes with our epistemic obligation to believe in accordance with evidence. These problems are thought to motivate $\mathcal{T}_{\rm p}.$ There are more sophisticated versions of $\mathcal{T}_{\rm O}$ that might tackle some of the objections, which will be discussed later in Sect. 4. In what follows, I will first argue that, upon a closer look, $\mathcal{T}_{\rm p}$ is not preferable to $\mathcal{T}_{\rm O}.$

2.1 The problem of excessive demands

One common objection to the requiring truth norm is that it makes excessive demands on us regarding what we ought to believe (e.g., Bykvist & Hattiangadi, 2007; Glüer & Wikforss, 2013). The claim that for any p, the truth of p is a requiring reason for us to believe p leads to an ‘explosion’ of epistemic obligations which are excessive given that we are ordinary epistemic agents with finite cognitive powers. Since there are infinitely many truths in the world (and some of which are far too complex for most humans to believe), we cannot, surely, believe every single one of them (Limited Capacity). As such, the requiring truth norm clashes with the intuitive principle ‘ought’ implies ‘can’ (OIC).³ Call this the problem of Excessive Demands.⁴

The permissive truth norm seems to avoid the problem all together, after all, it says that for any S, p, the truth of p permits S to believe p but falls short of requiring it. There is no ‘explosion’ of epistemic obligations. It is not the case that we are required to believe every truth that is out there. The permissive truth norm is compatible with OIC and Limited Capacity. Since both Limited Capacity and OIC are plausible, the fact that they are incompatible with $\mathcal{T}_{\rm O}$ but compatible with $\mathcal{T}_{\rm p}$ seems to be a good reason to reject the former and adopt the latter.

However, since may x = _def ¬ ought ¬x,⁵$\mathcal{T}_{\rm P}$ implies a requiring norm of falsity:

($\mathcal{F}_{\rm O}$) For any S, p, S ought to refrain from believing p if p is false.

$\mathcal{T}_{\rm P}$ faces an analogous problem of excessive demands, since $\mathcal{F}_{\rm O}$ is jointly incompatible with the following two claims:

(OIC):

For any S, φ, necessarily, if S ought to φ then S can φ.

(Limited Capacity*):

For some S, p, S cannot refrain from believing p when p is false.

Whiting (2013, pp. 125–126) argues that no critic of $\mathcal{T}_{\rm p}$ has shown that there are cases where if p is false, it is humanly impossible to refrain from believing that p. Furthermore, if it is humanly impossible to refrain from believing p, it is doubtful whether the attitude S has towards p is a genuine belief.

In Wei (2019), I offered several cases in support of Limited Capacity*. It is reasonable to anchor the notion of what is humanly possible or impossible to believe empirically, as supported by cognitive science and folk psychology. Very briefly, first, some beliefs might be so deeply integrated in our psychological make-up that we cannot refrain from having them. For instance, evidence suggests that patients with Capgras delusion are unable to refrain from believing that a close relative has been replaced by an impostor, often due to cognitive failure including abnormal perceptual experiences (as a result of a malfunctioning face recognition system) and possibly also with a deficit in their belief evaluation system.⁶ Now, of course, few of us suffer from clinical delusions, yet some of our core beliefs may be psychologically impossible to shake off in a rather similar way as a result of how we are hard-wired to perceive the world. In fact, many philosophical theories, if correct, would render some of our core beliefs false. For instance, if error theories about mathematics and ethics are correct, none of our mathematical and ethical beliefs are literally true. If the B-theory of time is correct, then the passage of time is an illusion and the present is not ontologically privileged. We cannot refrain from being disposed to act on beliefs about temporal experiences, that 2 + 2 = 4, or that torturing an animal for fun is wrong. Furthermore, some propositions are deeply integrated in our epistemic life, such as the so-called cornerstone propositions (e.g., that there is an external world, that I am not a brain in a vat are cornerstone propositions). If I were a brain in a vat, then those cornerstone propositions would be false. Yet, can I genuinely refrain from believing those cornerstone propositions? Perhaps in an epistemology seminar I can momentarily suspend judgments about cornerstone propositions while entertaining the sceptical scenarios. However, it is hard to imagine that we can refrain from believing cornerstone propositions if we were to live a normal life. It is not possible for me to not to be disposed to act and think as if it were true that there is an external world and that I am not a brain in a vat. If I did not believe that I am not a brain in a vat, I would not be able to have the ordinary empirical beliefs which are crucial for me to navigate through the world. Of course, the point here is not to claim that scepticism is true. Rather, the point is to emphasize that there are some propositions at the core of our belief system that we cannot refrain from believing as far as we are engaging in any believing at all and as we carry on our daily life, given the kind of creatures we are.

But is the attitude still a belief? Suppose that I cannot refrain from believing, say, that 2 + 2 = 4, or that time passes, even in the presence of overwhelming evidence for mathematical fictionalism and the B-theory of time, then, it may be argued that my attitudes are not that of belief. However, one cannot maintain this point without presupposing that it is constitutive of the attitude of belief that it is subject to a norm of truth. In other words, if an attitude is not subject to a norm of truth, that attitude cannot be counted as a belief. In (Wei, 2022), I considered several arguments for the constitutive interpretation of the truth norm and argued that they are unpersuasive. The onus of proof is on Whiting to show that my attitude cannot be that of a belief, even if the attitude plays the characteristic role belief plays in my mental economy.

The problem of excessive demands for $\mathcal{T}_{\rm p}$ might appear less worrisome in the sense that it relies on the assumption that some of our psychologically ingrained beliefs are false. However, the force of the point is that $\mathcal{T}_{\rm p}$ too issues demands exceeding our cognitive capacities.

2.2 The problem of blindspot propositions

The requiring truth norm has also been objected to on the ground that it requires us to believe true blindspot propositions, but such demands cannot be satisfied (e.g., Bykvist & Hattiangadi, 2007, 2013). Schematically, a blindspot proposition $\alpha$ is of the following form:

(Blindspot):

For any S, $\alpha$: necessarily, if $\alpha$ is true, then S does not believe that $\alpha$ and necessarily, if S believes that $\alpha$, then $\alpha$ is false.⁷

How do blindspot propositions pose a problem for the requiring truth norm? Suppose that the complex proposition ‘it is rainy and nobody believes that it is rainy’ is true. According to $\mathcal{T}_{\rm O},$ S is required to believe that ‘it is rainy and nobody believes that it is rainy’. But if S believes this proposition, then S believes both conjuncts, including the first conjunct ‘it is rainy’. Since S believes that ‘it is rainy’, the second conjunct is false. So, the complex proposition ‘it is rainy and nobody believes that it is rainy’ is false. So, S is required by the same norm not to believe that proposition! It seems then, with respect to a blindspot proposition $\alpha$, it is impossible for one to satisfy one’s epistemic obligation to believe $\alpha$ when $\alpha$ is true, while its being true that one has an obligation to believe $\alpha$. Bykvist and Hattiangadi argue that $\mathcal{T}_{\rm O}$ is false because it cannot be satisfied, and hence, according to them, cannot exist:

‘The problem here is not that the proposition cannot be believed, but that the obligation to believe p cannot be satisfied. So this is not a violation of the principle that ‘ought’ implies ‘can’. Rather, it is a violation of the principle that ‘ought’ implies ‘can satisfy’, which says that if you ought to believe that p, then it is logically possible for you to discharge or satisfy this ought. Or, more generally: ‘Ought’ implies ‘can satisfy’. If you ought to A, then it is logically possible for you to A while its being true that you ought to A. Now, the principle that ‘ought’ implies ‘can satisfy’ seems as plausible as the principle that ‘ought’ implies ‘can’. Just as one cannot have an obligation to do what is impossible to do, one cannot have an obligation that it is impossible to satisfy.’ (Bykvist & Hattiangadi, 2013, p. 109)

Since the requiring truth norm cannot be satisfied with respect to true blindspot propositions, the requiring truth norm is jointly incompatible with the principle of ‘ought’ implies ‘can satisfy’ (OICS) and the fact that there are true blindspot propositions. Call this the problem of Blindspot Propositions.

By contrast, the permissive truth norm seems to avoid the problem of Blindspot Propositions. After all, the permissive truth norm says that for any S, p, the truth of p permits S to believe p but falls short of requiring it. There is no epistemic obligation to believe true blindspot proposition: S is permitted to not to believe $\alpha$. The fact that $\mathcal{T}_{\rm O}$ but not $\mathcal{T}_{\rm p}$ violates OICS seems to be a good reason to reject the former and adopt the latter view (e.g., Whiting, 2010, pp. 218–220).

However, the permissive truth norm also faces an analogous problem of Blindspot Propositions. Suppose that the complex proposition ‘it is rainy and nobody believes that it is rainy’ is true. $\mathcal{T}_{\rm p}$ implies that, for any S, S may believe that ‘it is rainy and nobody believes that it is rainy’. But if S believes that proposition, then necessarily S believes both conjuncts, including the first conjunct ‘it is rainy’. If S believes that ‘it is rainy’, then the second conjunct is false. So, if S believes that ‘it is rainy and nobody believes that it is rainy’, this proposition is false. We have seen that the permissive truth norm also implies that for any S, p, S has an epistemic obligation to refrain from believing p when p is false. So, if the permissive truth norm is true, it says that it is permissive to believe true blindspot propositions. But if I were to believe those propositions, I would believe things that I am not permitted to believe. Bykvist and Hattiangadi explain why this result is unpalatable:

‘So what we have here is a permission which, if acted upon, would unavoidably change into a prohibition. And this seems fishy. Here’s a vivid illustration of the fishiness: imagine you went to a (fish!) restaurant that offers an all-you-can-eat buffet. You pay, and tuck in, but as you do, the waiters come running and explain that you are permitted to eat as much as you want only if you do not eat as much as you want, whereas if you do eat as much as you want, you are forbidden to do so.’ (Bykvist & Hattiangadi, 2013, p. 113)

Of course, proponents of $\mathcal{T}_{\rm p}$ could bite the bullet, after all, the fishy consequences are limited to blindspot propositions only. However, dialectically, this objection is significant, for it shows that $\mathcal{T}_{\rm p}$ does not enjoy an edge of advantage over $\mathcal{T}_{\rm O}$ when it comes to the problem of Blindspot Propositions.

2.3 The problem of conflicting norms

Another problem with the requiring truth norm is that in some cases, it makes demands on us that clash with our epistemic obligation to believe in accordance with evidence (e.g., Glüer & Wikforss, 2013; Hattiangadi, 2010). To fix ideas, consider the following norm of evidence:

(No Evidence):

For any S, p, S ought to refrain from believing p if S has no evidence supporting p.⁸

Since the requiring truth norm implies that for any p, S, S ought to believe p when p is true, it clashes with No Evidence in cases where p is true, but S has no evidence supporting p. In such cases, S is required to believe p by $\mathcal{T}_{\rm O}$ but S is also required to refrain from believing p by No Evidence. Suppose that I toss a fair coin. The coin has landed heads. But I have not revealed the coin to you. According to the truth norm, you ought to believe that the coin has landed heads. According to No Evidence, you ought to refrain from believing that the coin has landed heads for you lack evidence. In this case, it seems that you are both required to believe that the coin has landed heads and refrain from believing that the coin has landed heads. But this is not something you can do: you cannot both believe p and refrain from believing p. The requiring truth norm is jointly incompatible with OIC and No Evidence. Call this the problem of Conflicting Norms.

By contrast, the permissive truth norm seems to avoid the problem of Conflicting Norms. Since it says that for any S, p, the truth of p permits S to believe p but falls short of requiring it, it is not the case that S is required to believe p when p is true but unsupported by S’s evidence. You are not required to believe that the coin has landed heads by $\mathcal{T}_{\rm p}.$ There is no clash between $\mathcal{T}_{\rm p}$ and No Evidence. Since both No Evidence and OIC are intuitively plausible principles, the fact that they are incompatible with $\mathcal{T}_{\rm O}$ but compatible with $\mathcal{T}_{\rm p}$ seems to be a good reason to reject the former and adopt the latter.

However, $\mathcal{T}_{\rm p}$ also faces an analogous problem of Conflicting Norms. $\mathcal{T}_{\rm p}$ implies $\mathcal{F}_{\rm O}$: one is required to refrain from believing p when p is false. But the epistemic obligation to refrain from believing false propositions can come into clash with our epistemic obligation to believe in accordance with evidence. Consider the following norm of evidence:

(Evidence) For any S, p: S ought to believe p if S’s evidence supports p.⁹

Since the $\mathcal{T}_{\rm p}$ implies that for any p, S, S ought to refrain from believing p when p is false, it clashes with Evidence in cases where S’s evidence supports p, but p is false. In such cases, S is required to believe p by Evidence, and S is also required to refrain from believing p by $\mathcal{T}_{\rm p}$. But S cannot both believe p and refrain from believing p! $\mathcal{T}_{\rm p}$ is jointly incompatible with OIC and Evidence.

But of course, one might find Evidence too strong and opt instead for a weaker norm of evidence:

(Evidence*) For any S, any p: S may believe p if S’s evidence supports p.¹⁰

But Evidence* does not help proponents of $\mathcal{T}_{\rm p}$, on the contrary, it worsens the problem. For Evidence* contradicts $\mathcal{T}_{\rm p}$ in cases where S’s evidence supports p but p is false. To illustrate, suppose that Fin was hiking in the Cairngorms National Park and saw a large white object that looked like a mountain hare. Fin’s evidence suggests that he saw a mountain hare. But in fact, what Fin saw was the local farmer’s pet rabbit. In this case, according to $\mathcal{T}_{\rm P}$, Fin is not permitted to believe that he saw a mountain hare, since it is false. But according to Evidence*, Fin is permitted to believe that he saw a mountain hare since his evidence supports it. A contradiction.

Again, dialectically, this objection is significant, for it shows that $\mathcal{T}_{\rm p}$ is not preferable to $\mathcal{T}_{\rm O}$ when it comes to the problem of Conflicting Norms and so in that respect $\mathcal{T}_{\rm p}$ does not fare any better than $\mathcal{T}_{\rm O}$ which it aims to replace.

3 The weight of truth

The arguments we considered might move the reader to reject the truth norm of belief in favour of alternative norms of belief such as a norm of evidence or knowledge. But I think this is too quick. In this section, I offer a diagnosis of the debate on a reason-based normative framework. On this framework, the question of what sort of claim truth can make on us turns on the weight of truth as a reason for belief.

On a reason-based normative framework, normativity is a matter of normative reasons. Normative reasons are facts that count in favour of various responses. By responses, I refer to things that are responsive to reasons, such as actions, intentions and attitudes. The idea that truth is a norm of belief minimally entails that for any p, the truth of p is a normative reason that counts in favour of believing that p.

The idea that reasons have weight is familiar from our everyday talk about reason (see, e.g., Maguire & Lord, 2016). When one stands in relation to a reason, the extent to which a reason counts in favour of $\varphi$-ing comes in various degrees: the stronger the normative support it has for one to $\varphi$, the weightier the reason is. For example, the fact that you like ice cream is a reason for you to have ice cream for dessert, but your doctor’s advice is a weightier reason for you not to have ice cream for dessert.

We can distinguish between reasons that are weighty enough to require certain responses and reasons that are weighty enough to permit certain responses but fall short of requiring them.¹¹ Some reasons are weighty enough to issue a permission or even a requirement on an agent regarding what to do. A reason R is weighty enough to require S to $\varphi$, if S ought to $\varphi$ and is blameworthy for not $\varphi$-ing without good excuses when R obtains. By contrast, a reason R is weighty enough to permit (but falls short of requiring) S to $\varphi$, if S may $\varphi$ and S is not blameworthy for not $\varphi$-ing when R obtains.

On this reason-based normative framework, the requiring and permissive truth norms are what I will call invariantist accounts, for they claim that truth as a reason for belief has either a requiring or permissive weight across all contexts. According to $\mathcal{T}_{\rm O}$, truth has a requiring weight regarding what to believe: for any S, p, the truth of p is a requiring reason for S to believe p. But the view faces counterexamples: there are cases in which p is true but S is not required to believe that p. As we have seen, the alternative invariantist account, i.e., $\mathcal{T}_{\rm p}$, does not fare better. According to $\mathcal{T}_{\rm p}$, truth has a permitting weight regarding what to believe and a requiring weight regarding what not to believe: for any S, p, the truth of p permits S to believe p and requires S to refrain from believing not-p. But there are analogous cases in which truth either fails to be a permitting reason for one to believe p or fails to be a requiring reason for one to refrain from believing not-p. We discussed three kinds of cases in which truth fails to be a requiring/permitting reason for one to believe p:

• (a) Cases involving propositions which we cannot believe or refrain from believing;

• (b) Cases involving blindspot propositions;

• (c) Cases involving no or misleading evidence.

In these cases, the truth of p does not require one to believe that p or require one to refrain from believing that not-p. But it is too quick to infer from these cases that truth is not a norm of belief for it cannot make an interesting claim on us regarding what to believe. Rather,$\mathcal{T}_{\rm O}$ and $\mathcal{T}_{\rm p}$ are false because truth, as a normative reason for belief, does not have invariant weight across all situations, or so I shall argue. Whether truth has a requiring/permitting weight can vary depending on the circumstances. Call this view variantism.

In what follows, I will develop the idea that truth as a reason for belief functions holistically just like any other reason, and its weight varies depending on the presence/absence of conditions and modifiers. Variantism enables us to explain type (a)-(c) cases, without invoking alternative conceptions of ‘ought’ or ought-limiting principles.

Since the pioneering work by Dancy (2004), it is widely acknowledged that the weight of reasons is holistic and context-sensitive. A consideration for $\varphi$-ing can, in one context have a certain weight, and in another context a different one. For example, intuitively, the fact that climate emergency is an existential threat to humanity in some contexts can be a requiring, not merely a permitting reason for me to engage in civil disobedience. For suppose that the window of opportunity to avoid irreversible catastrophic climate consequences is closing, it may be argued that, in this context, my reason to engage in civil disobedience has a requiring weight, not merely a permitting weight.

What are the relevant factors that affect the weight of a reason in a context and is there a principled way to describe them? We can distinguish between two types of features: conditions and modifiers. Conditions can be divided into enablers and disablers. An enabler is a condition whose presence allows R to constitute a reason favouring $\varphi$-ing. A disabler, by contrast, prevents R from constituting a reason that favours $\varphi$-ing. Modifiers also come in two types: intensifiers and attenuators. An intensifier is a consideration which strengthens the normative support R has for $\varphi$-ing. By contrast, an attenuator weakens it.

Three questions arise concerning the move from holism to variantism. First, can truth, as a reason for belief be conditioned? If so, what sort of facts constitute conditions that enable/disable the reason to believe p and refrain from believing not-p when p is true? Second, can the reason be modified? If so, what sort of facts constitute modifiers that intensify/attenuate truth as a reason to believe p and refrain from believing not-p when p is true? On the view I favour (Wei, 2022), truth is a reason for belief because it is constitutive of a justified social practice that the truth of p counts in favour of believing p (and against believing not-p). Very briefly, truth as a reason for belief is grounded in what I called the T-practice, a social practice that generates knowledge to facilitate social cooperation. There are two possible ways in which a practice-based reason can be disabled. First, a practice-based reason to φ can be disabled for S if it is not possible for S to participate in the practice which grounds the reason to φ. Second, a practice-based reason to φ can be disabled for S if the practice which grounds the reason to φ is unjustified in a given situation. A social practice typically produces, distributes, or organizes some social goods which have a practical impact on the participants of that practice. Importantly, not all social practices can be justified by their social functions and many of them should be reformed or abolished altogether. History is full of examples of unjustified social practices. Some social practices discriminate and oppress particular groups of people, e.g., the practice of slavery, patriarchal gender roles, voter suppression, etc. Some social practices exploit nonhuman animals and natural resources, e.g., the practice of unregulated farming, deforestation, etc. An unjustified social practice performs its social function at the cost of causing substantial harm to participants involved in that practice. Unjustified social practices have no authority to ground any normative reasons. There can be no practice-based reason to $\varphi$ when $\varphi$-ing does not constitute a justified social practice.

Thus, practice-based reasons in general are not unconditional. They can be enabled/disabled by facts about what participants of that practice can or cannot do and facts about whether the practice is justified/unjustified in the situation under consideration. Truth, as a practice-based reason, is not unconditional. Truth as a reason for belief can be enabled/disabled by facts about what an ordinary epistemic agent can or cannot do with their epistemic lives and by facts about the justificatory status of the social practice which grounds truth as a reason in the situation under consideration. Accordingly, the weight of truth can vary across different contexts.

One example we discussed earlier concerns complex propositions that are impossible for any ordinary epistemic agent with limited cognitive capacity to form doxastic attitudes about. The fact that we cannot form beliefs about those complex propositions disables the T-practice-based reason for us to believe those complex propositions when they are true. For in that case, it is not possible for us to participate in the T-practice with respect to those propositions to generate knowledge which grounds the reason for us to believe them. In another example, a patient with Capgras delusion cannot refrain from believing that a close relative has been replaced by an imposter due to cognitive failure. Refraining from believing that the relative is an imposter is not something one who suffers from that medical condition can do. Again, it is not possible for them to participate in the T-practice with respect to those propositions to generate knowledge which grounds the relevant reasons. Therefore, the fact that one suffers from Capgras delusion and cannot refrain from believing that her relative is an imposter disables the T-practice-based reason for her to refrain from believing that proposition.

But let me be clear, disablers are not cheap to come by. One cannot $\varphi$ in a given situation if one does not have the ability or opportunity to $\varphi$ in that situation, broadly based on folk-psychology laws governing our attitudes and actions. There is a line to be drawn between things that are extremely difficult to believe under certain conditions and things that are impossible to believe under those conditions. A Trump supporter who has been fed lies and conspiracy theories about the presidential election does not have his T-practice-based reasons disabled. Even if it is difficult for him to come to believe the truth about the election, after all, his epistemic environment is impoverished, and he is emotionally manipulated. Nonetheless, he has fully functioning cognitive capacities to think and make correct judgments. He has the opportunity to reflect upon what is going on around him, to review the evidence and to reason. There is no reason to think that it is impossible for him to participate in the T-practice with respect to those propositions. Of course, there might be borderline cases where the line between what one can and cannot believe is harder to draw. But those cases need to be carefully assessed on a case-by-case basis with the relevant details filled in.

Practice-based reasons for belief can also be disabled if the T-practice which grounds practice-based reasons for belief is unjustified in a given situation. For example, believing certain things about a marginalized community, even when they are true, can often amplify an unjustified narrative about the community and put members of that community in harm’s way. If believing such propositions constitute doxastic wrongdoing, as some philosophers have argued (e.g., Basu, 2019), then believing those true propositions which constitute doxastic wrongdoing cannot be a justified social practice of belief management. The fact that believing p constitutes doxastic wrongdoing in a given situation can disable the T-practice-based reason to believe p when p is true in that situation. One merit of this view is that we can be neutral as to whether there are moral or pragmatic reasons for belief while maintaining that non-evidential considerations play a role in determining the weight of truth.

Truth as a reason for belief can also be modified. Notice first that desire-based and value-based reasons can be intensified/attenuated by facts concerning the extent to which the states of affairs constituted by φ-ing are desired or valuable, since both desired and valuable states of affairs are gradable, i.e., they can be more or less desired or valuable. For example, the fact that Davidson writes elegant prose intensifies my reason to read his work since doing so in the presence of that fact constitutes a more desired state of affairs than it would otherwise be in its absence. The same can be said for value-based reasons. For example, the fact that the pedestrian is an elderly lady intensifies my reason to help her since doing so in the presence of this fact constitutes a more valuable state of affairs than it otherwise would be in its absence.

A practice is not gradable in the same way as desires or values. Either φ-ing for R would constitute a justified social practice or it would not. But plausibly, we can distinguish between the centre and the periphery of a social practice relative to its function. The distance between the centre and the periphery is gradable and φ-ing for R in a situation can be closer or less close to the centre of the practice. Practice-based reasons can be intensified/attenuated by facts concerning how central φ-ing for R is within the practice relative to its function in a given situation.

We can apply this idea to the T-practice with the help of a simple diagram of concentric circles. Consider the following simplified diagram (Fig. 1).

Fig. 1

Practice-based Modifiers

Let C represent the centre of the T-practice. Let any point falling within the periphery of the T-practice represent a situation in which one believes p/refrains from believing not-p when p is true. We can model this type of modifier of T-practice-based reasons using the diagram. The fact which says that believing p/refraining from believing not-p when p is true in a given situation falls within the inner circle, i.e., in the gridded area, constitutes an intensifier. The shorter the distance to C, the greater the extent to which it intensifies the T-practice-based reason to believe p/refrain from believing not-p when p is true in the situation under consideration. The fact which says that believing p/refraining from believing not-p when p is true in a given situation falls outside the inner circle but within the periphery of the T-practice, i.e., in the dotted area, constitutes an attenuator. The longer the distance to C, the greater the extent to which it attenuates the T-practice-based reason to believe p when p is true in the situation under consideration.¹²

This, of course, raises the question about what constitutes the centre of the T-practice. As I suggested, the function of the T-practice is to facilitate knowledge production for social cooperation. At a first approximation then, we can say that if believing p/refraining from believing not-p when p is true in the situation under consideration plays a greater role in generating knowledge to facilitate social cooperation, then the closer it is to the centre of the T-practice. Of course, a good deal more can be said to make this characterization more precise.¹³ But I take the general idea to be intuitive enough.

Let’s apply this idea first to the example of a blindspot proposition such as ‘it is rainy but nobody believes that it is rainy’. Since blindspot propositions are necessarily false if one believes them, believing a blindspot proposition p when p is true, one might think, is at the very periphery of the T-practice, since forming necessarily false beliefs does little to facilitate social cooperation, knowledge production or maintenance. The fact that p is a blindspot proposition attenuates the T-practice-based reason to believe p when p is true. By contrast, when the survival and flourishing of our community hinge on us grasping the truth about p, even if p may be painful to accept or hard to uncover, believing p when p is true is at the centre of the T-practice. For example, when p concerns issues such as the COVID-19 pandemic, climate change, social justice, etc., the T-practice-based reason is intensified. In such cases, believing p and refraining from believing not-p when p is true is crucial to facilitate social cooperation, knowledge production and maintenance.

These examples are relatively simple and straightforward. But you might think there are much more complex cases where, for instance, believing p/refraining from believing not-p might be essential to promote social cooperation but threatens knowledge production.¹⁴ And in some cases, it is difficult to make a judgment about the importance of believing p/refraining from believing not-p when p is true in a given situation relative to the function of the T-practice. Such cases would require a much more worked-out theory, a task I have to leave to future work. Again, at this stage, it is important to emphasize a close case-by-case analysis. The quick sketch suffices to illustrate how the framework I just offered can model one way in which the T-practice-based reason can be modified, which is our focus here.

Another important type of modifier is the risk attenuator. The thought is this: a reason R to φ can give an agent S normative support for φ-ing in a given situation only if R in fact constitutes a reason for S to φ in that situation. For example, the fact that a project will lift one hundred families out of poverty is a reason to invest in that project. But in our world of uncertainty, there is a risk that the project will not lift one hundred families from poverty. The risk that the project will not lift one hundred families out of poverty attenuates my reason to invest in that project. According to an intuitive probabilistic account of risk, the risk of R being false is determined by the probability of R conditional upon the background evidence available to the agent in a given situation. The higher the evidential probability of R, the less risky is R.¹⁵ The risk that my reason to invest in the project does not obtain, i.e., the high probability that the project will not lift one hundred families from poverty given my evidence, attenuates my reason to invest in that project. We might call this type of risk, empirical risk. It is the risk of R being false in a given situation. If R were false, then R would not constitute a reason for φ-ing in that situation. The higher the risk, the less weighty R is.

Empirical risk can be contrasted with what we might call normative risk. The latter is the risk that R fails to count in favour of φ-ing in a given situation, i.e., the risk of the grounding fact in virtue of which R counts in favour of φ-ing failing to obtain.¹⁶ If R did not count for φ-ing in a given situation, then R would not constitute a reason for φ-ing in that situation. For example, the fact that a project benefits future generations is a reason for investing in that project, grounded in the value-based fact that the welfare of future generations constitutes valuable states of affairs. But there is a risk that we are wrong about how much we should care about future generations. Perhaps, one may argue, that only the welfare of the present and near-future generations matters. The normative risk that the future-oriented project fails to constitute valuable states of affairs attenuates my reason to invest in that project. According to the probabilistic account of risk, the normative risk of R failing to count for/against φ-ing is determined by the probability that the grounding fact in virtue of which R counts for φ-ing obtains, conditional upon the background evidence available to the agent in a given situation. The higher the evidential probability for the grounding fact, the less (normatively) risky R is. The higher the normative risk, the less weighty R is.

Truth as a reason for belief can be attenuated by risk attenuators. Consider first the empirical risk that p is false. If p were false, the truth of p would not constitute a reason to believe p/refrain from believing not-p. For example, suppose that Fin was hiking in the Cairngorms National Park and saw a large white object. It was a snowy day and the ground was covered in white. Fin saw that the children of local farmers were out playing in that area. Fin has never seen a mountain hare in this part of the park. But luckily for Fin, the large object he just spotted was a mountain hare. In this case, the truth of the proposition is of high empirical risk. For given the evidence available to Fin, the evidential probability that Fin saw a mountain hare is low. The risk attenuates the weight of truth as a reason for Fin to believe that he saw a mountain hare.

Consider next the normative risk that the practice-based grounding fact in virtue of which the truth of p constitutes a reason does not actually obtain in a given situation. To illustrate, consider an example adapted from Basu (2019, pp. 915–916). Suppose that the conference at Aanya’s university has ended, and the participants are having dinner at a local restaurant. After a few drinks, Aanya gets up to use the restroom. As she returns to her table, one of the diners, Jim, asks her for another drink—he believes that Aanya works in the restaurant. With respect to their melanin levels, Aanya’s is more similar to those of the waiting staff than to her fellow philosophers. Aanya feels hurt and is upset. But as a matter of fact, Aanya does have a part-time job in the restaurant. Jim’s belief that Aanya works in the restaurant is true. In this case, although Jim has a reason to believe that Aanya works in the restaurant since it is true, intuitively, the weight of truth as a reason for belief is attenuated by a normative risk, or so I will suggest.

I noted earlier that the T-practice-based reason to believe p when p is true in a given situation can be disabled, if believing p constitutes epistemic wrongdoing and renders the T-practice unjustified in that situation. One example I provided, is a case where believing certain truths about a marginalized community amplifies an unjustified narrative about the community and puts members of that community in harm’s way. I suggested that in such cases T-practice-based reasons to believe those propositions are disabled. Now, in the present case concerning Aanya, no such disabler is present, for we lack conclusive evidence as to whether the moral implications of Jim’s belief render the T-practice unjustified. Nevertheless, given the pain Jim’s belief has inflicted on Aanya, there is a normative risk that truth does not in fact count in favour of believing that Aanya works in the restaurant and thereby attenuating the weight of that reason.

Variantism enables us to explain (a)–(c) type cases in an attractive and economical way. It is attractive and economical because with the holistic toolkit, we can deliver correct verdicts and a nuanced analysis of the normative import of truth in a given case. They are simply the results of the fact that truth, just like any other reason, functions holistically and has varying weight in different situations.

It will be helpful to go through some examples. Let us begin with an (a)-type cases involving propositions which we cannot believe or refrain from believing. Consider the following example:

(PASSING TIME) It appears to me that time passes. Days end, months go by and years flow. It is impossible for me to refrain from believing that time passes. However, as a matter of fact, the passage of time is illusory.

In this case, that the passage of time is illusory is not a requiring reason for me to refrain from believing that time passes, since it is incompatible with OIC. On variantism, the presence of a disabler, namely, the fact that it is impossible for me to refrain from believing that time passes prevents the truth from constituting a practice-based reason for me to refrain from believing that time passes and therefore the T-practice-based reason in PASSING TIME does not have any weight in making a claim on me. This explanation generalizes to other type (a) cases involving propositions which we cannot believe or refrain from believing where the T-practice-based reason to believe p/refrain from believing not-p when p is true is disabled.

Type (b) cases involve blindspot propositions. Consider the following example:

(RAIN) Ginger is at a conference with a group of philosophers. The talk is engaging, and the discussion is animated. Suppose that, as a matter of fact, it is rainy outside and no one in the room believes that it is rainy outside.

In this case, the T-practice-based reason to believe that blindspot proposition—it is rainy outside and no one in the room believes that it is rainy outside—cannot be a requiring reason, since it is incompatible with OICS. Nor is it a permitting reason since permission to believe a true blindspot proposition is normatively ‘fishy’. Again, variantism enables us to explain why in this case, the T-practice-based reason is neither a requiring nor a permitting reason.

According to variantism, the T-practice-based reason has neither a requiring weight nor a permitting weight because the normative strength of the T-practice-based reason to believe the true blindspot proposition in RAIN is attenuated. As I have argued, the fact that believing the blindspot proposition—it is rainy outside and no one in the room believes that it is rainy outside—is at the very periphery of the T-practice attenuates the normative strength of the T-practice-based reason to believe that blindspot proposition. Moreover, there is also an empirical risk that the T-practice-based reason in RAIN does not obtain. After all, even if the talk is engaging and the discussion is animated, there is a chance that someone looks out and arrives at the belief that it is rainy. The empirical risk further attenuates the T-practice-based reason to believe that blindspot proposition. Similar analysis can be provided to other type (b) cases involving blindspot propositions where the T-practice-based reason to believe a blindspot proposition is significantly attenuated by the fact that it lies at the periphery of the T-practice and its truth is risky.

Type (c) cases involve propositions unsupported by evidence. Consider the following example:

(MOUNTAIN HARE) Fin was hiking in the Cairngorms National Park and saw a large white object that looked like a snowman. It was a snowy day in April and the ground was covered in white. Fin saw that the children of local farmers were out playing in that area. Fin knows that mountain hares have never been seen in this part of the park. Suppose that, luckily for Fin, the large object he spotted was a mountain hare.

In this case, the claim that Fin saw a mountain hare is unsupported by his evidence. The T-practice-based reason for Fin to believe that he saw a mountain hare cannot be a requiring reason, since it is incompatible with the evidential norm which says that one ought to refrain from believing p if p is unsupported by one’s evidence and OIC. With variantism, we are able to explain, why in MOUNTAIN HARE, the T-practice-based reason does not have a requiring weight.

According to variantism, the empirical risk of the T-practice-based reason is significant given that the evidential probability that Fin saw a mountain hare is low. The risk attenuates the normative strength of the T-practice-based reason for Fin to believe that he saw a mountain hare. Similar analysis can be provided to other type (c) cases involving propositions unsupported by evidence. Now, whether Fin is permitted to believe that he saw a mountain hare depends on the balance of his reason since the truth of the proposition in this case by itself does not have requiring or permitting weight.

Let me briefly raise and respond to an objection before situating my view within the context of alternatives in the next section.¹⁷ A critic may suggest that the process of weighing the T-practice-based reason to believe p/refrain from believing not-p when p is true on the variantist account seems overly complicated, and few epistemic agents seem to be capable of engaging in such weighting processes correctly. And yet, most of us seem to be able to judge correctly what sort of claim truth can make on us in a given case. So, something must be wrong with the variantist account. In response, it is worth emphasizing that the version of variantism I develop and defend in this section is simply an application of the widely accepted view that reasons are holistic and context-sensitive. So if there is indeed a problem of intellectualization, it is a problem for everyone who accepts a holistic account of reasons, according to which how weighty a reason is in making a claim on us regarding what to do or believe depends on the circumstances in the way we sketched earlier. Moreover, from an evolutionary point of view, it is not surprising that human beings develop capacities to recognize and respond to the relevant conditions, modifiers and opposing reasons in a given situation and to weigh reasons in an efficient and reliable manner. After all, to be able to respond to the claims that reasons make on us is crucial to our survival and flourishing. Another potential objection comes from proponents of alternative views who claim that their views offer equally good explanations of cases (a)-(c). In the next section, I will address this worry by contrasting it favourably with two alternative views. It should be stressed here that the goal is not to refute the alternative views, nor to demonstrate that variantism is the only game in town. As I have argued in this section, the variantist account of the truth norm, as an application of the independently attractive reason-based normative framework, offers an attractive response to the three problems discussed in Sect. 2. In highlighting the differences between variantism and two alternative views, I will further show the distinct virtues of variantism and strengthen its case.

4 Contrasting variantism with alternative views

4.1 Hedged truth norms

Some versions of the invariantist view of the truth norm remain unscathed by the arguments discussed in Sect. 2. For example, consider a hedged truth norm which says that the truth of p has a requiring weight except (1) when S cannot believe p or refrain from believing not-p; (2) when p is a blindspot proposition; and (3) when p is not supported by evidence available to S. The hedged truth norm vindicates the intuitive verdict that in TIME PASSING, RAIN and MOUNTAIN HARE the truth of the proposition involved does not have a requiring weight: these are cases to which the requiring truth norm does not apply.

One problem with the hedged truth norm is that it faces further counterexamples. Consider the following case:

(MOUNTAIN HARE*) Fin was hiking in the Cairngorms National Park and saw a large white object that looked like a mountain hare. It was a snowy day in April and the ground was covered in white. Fin believes that he saw a mountain hare. However, Fin knows that mountain hares have never been seen in this part of the park. Furthermore, it is unusual for a mountain hare to have white fur at this time of year. Luckily for Fin, the large object he spotted was indeed a mountain hare. Based on that belief Fin conducted further scientific research in that area and made new findings about mountain hares.

In this case, it seems that the T-practice-based reason for Fin to believe that he saw a mountain hare is not a requiring reason, but perhaps a permitting reason.¹⁸ He does not seem to commit any epistemic wrongdoing if he did not form the belief that he saw a mountain hare, after all, there is evidence suggesting that he did not see a mountain hare. But according to the hedged truth norm, the T-practice-based reason is a requiring reason. We can conceive of other cases in which the truth of a proposition falls short of having a requiring weight but are not included in the list of exceptions like (1)–(3).

Now, you might think this problem can be solved by adding further exceptions to the hedged principle. It is designed to include all potential exceptions so that it can explain away any putative counterexample. But there are good reasons to prefer variantism over a hedged view with a long list of exceptions.

To begin with, variantism is explanatorily more powerful. The hedged truth norm does not have the capacity to offer an analysis of the weight of truth in type (a)-(c) cases, for it does not apply in those cases. By contrast, variantism enables us to say what sort of claim truth can make on us regarding what to believe in those cases, and why.

There is also a concern about whether non-trivial principles can include exceptions. On the one hand, one might think it is not possible to draw a complete list of exceptions. And on the other hand, if the list of exceptions is open-ended, there is a further worry that a hedged principle including an opened list of exceptions seems incapable of doing any explanatory work.¹⁹

More generally, hedged invariantist truth norms lack motivations on a holistic framework. On a holistic framework, no reason has pre-fixed weight outside a given context—the weight of a reason is neither prior nor independent of other salient facts which obtain in the context. Whether a reason has the weight it has depends on what else is true in the situation. The weight of truth is neither determined nor explained by some invariantist norms, but depends on the relevant conditions, modifiers and its opposing reasons in the way I spelled out in the previous section. An invariantist view on a holistic framework does no real explanatory work.

Furthermore, on a holistic framework, the invariantist has the burden to show that the truth has requiring/permitting weight in all contexts, despite the infinitely many possible changes of circumstances. It calls for an explanation of why such an incredible coincidence should obtain. Variantism, by contrast, is simply the result of applying holism to truth as a reason for belief. So, if one accepts the holistic account of reason, as many do, it is difficult to see why one would prefer a hedged invariantist truth norm over variantism.

4.2 Perspectivism

Recently, an increasingly popular view in the literature of normativity is perspectivism. Perspectivists hold that what one ought to do or believe depends on one’s epistemic perspective. Objectivists, by contrast, hold that what one ought to do or believe depends on all the facts, irrespective of one’s epistemic perspective.²⁰ There are, of course, many forms of perspectivism. What interests me here is a form of perspectivism about the truth norm, which, as we will see, shares a number of features with variantism.

Perspectivism about the truth norm is the view that, roughly, whether the truth of a proposition is a requiring (or permitting) reason for one to believe that proposition is determined by one’s epistemic perspective, which is constituted by the total evidence available to one in a given situation (e.g., Conee & Feldman, 2004; Gibbons, 2013; McHugh & Way, 2017).²¹ It is easy to see the overlap between variantism and perspectivism. Both variantists and perspectivists reject unqualified invariantist truth norms such as the requiring and permissive truth norm. Both hold that the correct truth norm is context-sensitive. And both think that one’s epistemic perspective matters to what one ought to or may believe in a given situation. So, it will help to further highlight the distinct virtue of variantism by comparing it with perspectivism about the truth norm, which is in its close vicinity.

To facilitate our discussion, it will be helpful to put a more precise formulation of perspectivism on the table. We can focus on the following version of perspectivism:

(Perspectivism) For any S, p, the truth of p is a requiring reason for S to believe p if and only if S’s evidence decisively supports p; and the truth of p is a permitting reason for S to believe p if and only if S’s evidence sufficiently supports p.²²

As one might expect, perspectivism does well with respect to type (c) cases involving propositions unsupported by evidence, such as MOUNTAIN HARE. In MOUNTAIN HARE, Fin’s evidence does not sufficiently support the claim that he saw a mountain hare. So according to perspectivism, the truth that Fin saw a mountain hare is neither a requiring nor a permitting reason. It can also explain why in MOUNTAIN HARE*, the truth that Fin saw a mountain hare is a permitting reason for him to believe that. Fin saw a white object that looked like a mountain hare. But since Fin knows that mountain hares have never been seen in the area, Fin’s evidence seems to sufficiently but not decisively support the claim that he saw a mountain hare. So according to perspectivism, truth has a permitting but not a requiring weight.

It is less clear, however, that perspectivism can explain all type (a) cases involving propositions which one cannot believe/refrain from believing. A subset of type (a) cases involving true propositions decisively/sufficiently supported by one’s evidence provides counterexamples to perspectivism, because according to perspectivism, the T-practice-based reasons to believe p in those cases has a requiring/permitting weight. But we have seen, that when one cannot believe p, the T-practice-based reason to believe p is disabled and thus does not have a requiring/permitting weight.

Perspectivism also has difficulty explaining all type (b) cases involving blindspot propositions. A subset of type (b) cases involving true blindspot propositions decisively/sufficiently supported by one’s evidence also provides counterexamples to perspectivism, because according to perspectivism, those true propositions have a requiring/permitting weight. But we have seen, when p is a blindspot proposition, the T-practice-based reason is significantly attenuated. Thus, it does not have a requiring/permitting weight.

Perspectivists might respond to the putative counterexamples by tightening up their account. For example, consider the following revised version of perspectivism:

(Perspectivism*) For any S, p, the truth of p is a requiring reason for S to believe p if and only if the evidence possessed by S decisively supports p; and the truth of p is a permitting reason for S to believe p if and only if the evidence possessed by S sufficiently supports p.

According to one recent proposal of evidence possession, epistemic access to evidence is insufficient for one to count as possessing that evidence. An additional, dispositionalist condition must be satisfied. If one possesses evidence for p, then one must treat that evidence as a reason for believing p, which is to manifest a disposition to believe p when that evidence constitutes a reason for believing p (e.g., Lord, 2018; Sylvan, 2016). So, on this account, the truth of p is a requiring/permitting reason for S to believe p if and only if S treats the decisive/sufficient evidence available to S as a reason for believing p.

Now, perspectivists could argue that perspectivism* can explain all type (a) and type (b) cases. Consider type (a) cases involving propositions which one cannot believe. Perspectivists could argue that, even if one has epistemic access to sufficient/decisive evidence for a proposition in those cases, it is impossible for one to treat the evidence as a reason for believing that proposition since one cannot have the relevant disposition to believe something that is impossible for one to believe. Hence, it is impossible for one to possess such evidence. So according to perspectivism*, the T-practice-based reason to believe p in those type (a) cases involving propositions which one cannot believe does not have a requiring/permitting weight.

Considering type (b) cases involving blindspot propositions, perspectivists could argue that, even if one has epistemic access to sufficient/decisive evidence for a proposition in those cases, it is impossible for one to treat that evidence as a reason for believing a blindspot proposition since one cannot have the relevant disposition to believe something that, one knows, if she were to believe it, would be necessarily false. Hence, it is impossible for one to possess such evidence. So according to perspectivism*, the T-practice-based reason to believe p in those type (b) cases involving blindspot propositions does not have a requiring/permitting weight.

So, it seems that perspectivism* can do just as well as variantism in explaining all type (a)-(c) cases. Why then, should we prefer variantism? We cannot properly assess the dispositionalist account of evidence possession here, which is the key to the success of the revised perspectivist explanation of those cases.²³ But even if perspectivism* can successfully explain type (a)-(c) cases, there remains a crucial difference between variantism and perspectivism*. On variantism, as I have developed it, non-evidential considerations play important roles in weighting truth as a reason to believe p/refrain from believing not-p when p is true (i.e., as enabling/disabling conditions and as attenuating/intensifying modifiers), whereas on perspectivism*, only evidence can play such a role.

One important virtue of variantism, then, is that it can explain two further types of cases. Consider first type (d) cases involving true propositions that matter to us but are not well-supported by the evidence possessed by the agent. Variantism vindicates our intuitive judgment that the truth of the matter can make a strong claim on us regarding what to believe in type (d) cases. For example, the fact that p concerns issues such as the COVID-19 pandemic, climate change, social justice, etc., significantly intensifies the truth as a reason to believe p/refrain from believing not-p when p is true. Thus, according to variantism, in type (d) cases, truth has a requiring weight. Perspectivism* by contrast cannot explain type (d) cases. According to perspectivism*, the truth of a proposition makes no claim on what one ought to believe in those cases, since one does not possess the relevant evidence for those propositions.

Variantism can also explain type (e) cases involving true propositions that matter very little to us but are well-supported by the evidence possessed by the epistemic agent. Variantism vindicates our intuitive judgment that in such cases, the truth of the matter is unlikely to make a requiring claim on us regarding what to believe. For example, the fact that p concerns issues such as car numbers significantly attenuates truth as a reason to believe a proposition about car numbers. Believing the truth about trivial issues such as car numbers is close to the periphery of the T-practice since it plays a limited role in promoting social cooperation, knowledge production and maintenance. Thus, according to variantism, in type (e) cases, truth does have a requiring weight. Whether one is overall permitted to form a belief about car numbers depends on the balance of one’s reasons. The truth of the proposition by itself does not issue a permission or requirement for belief. Perspectivism*, by contrast, cannot explain our intuition about type (e) cases. Thus, even if perspectivism* about the truth norm shares a number of features with variantism and can deal with type (a)-(c) cases, there remains a crucial difference. Variantism has a distinct virtue in accommodating type (d) and (e) cases.

Perspectivists might respond by incorporating a non-epistemic perspective into their view to accommodate type (d) and (e) cases. For example, they may propose that whether the truth of a proposition is a requiring (or permitting) reason for one to believe that proposition is determined by one’s epistemic as well as non-epistemic perspectives. However, there are at least three difficulties in developing this line of response. To begin with, what constitutes a non-epistemic perspective? While epistemic perspective is constituted by the total evidence one possesses, it is much less clear what the relevant non-epistemic perspective consists of. Moreover, it is not clear how the two perspectives interact with each other. How exactly do the two perspectives play out in determining whether we have an obligation or a mere permission to believe the truth? Lastly, this perspectivist line of response seems ad hoc without offering independent motivations for incorporating non-epistemic perspectives.

By contrast, variantism developed in this paper offers a coherent framework in explaining how and why the weight of truth as a reason for belief varies. Non-evidential considerations can play a role in modifying the weight of truth because truth as a reason for belief is based on the T-practice and some non-evidential considerations determine how a given instance is situated in that social practice.

Even if these difficulties may be overcome, there remains a fundamental difference between perspectivism and variantism. On perspectivism, the normativity of truth is subject to one’s evidential (and perhaps also non-epistemic) perspective. When truth falls out of the relevant perspective, it plays no role in determining what one should or may believe. By contrast, on my view, truth is a normative reason for belief regardless of one’s perspective, although the weight of truth as a normative reason varies across different contexts. At a more fundamental level, I think perspectivism is mistaken because it fails to correctly locate the source of normativity. The source of normativity, on my view, is grounded in social practices, and therefore is not subject to individual epistemic agents’ evidential (and non-evidential) perspective.

In short, while I do not rule out the possibility of more sophisticated versions of the truth norm that might tackle the problems discussed here, as it stands, variantism offers a more attractive response to the weighting challenge. Given that variantism is independently well-motivated and offers a plausible explanation as to why truth is a normative reason in all those problematic cases, the arguments considered in this section further strengthen the case for variantism.

5 Conclusion

To sum up, in this paper, I offered a novel answer to the question concerning the normative force of the truth norm on a reason-based normative framework. On the view I developed, the normative force of the truth norm varies, depending on the weight of truth as a reason for belief. Truth as a reason for belief has varying weight just like any other normative reason, subject to conditions and modifiers. I spelled out in detail what could condition and modify truth as a reason for belief. I suggested that what an epistemic agent can believe, the justificatory status of the underlying T-practice and the epistemic risk involved can affect the weight of truth as a reason for belief.