A Permissivist Alternative to Encroachment

II.1 Reasons pragmatism

According to reasons pragmatism, practical considerations affect what we ought to believe by constituting distinctively practical (i.e., non-epistemic) reasons for or against belief. When a belief is (or has the potential to be) practically valuable, this value provides a (pro tanto) reason for belief. Likewise, when a belief is (or has the potential to be) practically costly, this cost provides a (pro tanto) reason against belief. Reasons pragmatism has the merit of providing a simple, intuitive explanation of how practical considerations affect what we ought to believe.

While the core claim of reasons pragmatism is that (actual or potential) benefits and costs of believing provide practical reasons for or against belief, this claim is compatible with different views about the relationship between practical and epistemic reasons for belief. Berker (2018) provides a helpful taxonomy of three different versions of reasons pragmatism. “Austere pragmatism” (Rinard 2019a; Mantel 2019; Maguire and Woods 2020) holds that practical considerations constitute the only genuine (or “authoritative”) normative reasons for belief, so epistemic reasons are at best “formally” normative (like the rules of etiquette or chess). “Interactionist pragmatism” (Reisner 2008; Leary 2017; Howard 2020) holds that practical considerations and epistemic considerations both constitute genuine normative reasons for belief, and that both contribute to determining what we all-things-considered ought to believe. “Separatist pragmatism” (Feldman 2000; Kauppinen 2023) holds that practical and epistemic considerations both constitute reasons for belief, but that they cannot be compared or weighed against each other to determine an all-things-considered verdict about what we ought to believe.

Of these versions of reasons pragmatism, only austere pragmatism requires denying that there are genuinely normative epistemic reasons for belief. We take austere pragmatism to be a highly revisionary view, and we will assume here that the most plausible form of reasons pragmatism will accept that there are both practical and epistemic reasons for belief.³ We also reject strict versions of separatist pragmatism—on which practical and epistemic reasons can never combine to determine what we all-things-considered ought to believe—for reasons that will become clear in §5.3.

Despite its simplicity and intuitive appeal, reasons pragmatism faces several significant challenges. The most prominent objection holds that, at least in many cases, it seems psychologically impossible to believe on the basis of practical considerations.⁴ Suppose someone offers you $1,000,000 to believe that 2+2=5. While this financial incentive may motivate you to try to bring it about that you believe that 2+2=5 (e.g., by taking a drug that will induce this belief), it cannot directly motivate you to believe that 2+2=5. So, proponents of this objection conclude, the financial incentive cannot constitute your motivating reason for belief (i.e., the reason for which you believe). A plausible necessary condition on normative reasons for belief is that it must be possible for them to serve as motivating reasons that directly figure in our deliberation about what to believe. Practical considerations such as financial incentives seem not to satisfy this condition, which suggests that they cannot be normative reasons for belief. At most, proponents of this objection argue, practical considerations are reasons to bring it about that you have a certain belief.

Second, because interactionist and separatist versions of reasons pragmatism hold that both epistemic and practical considerations constitute genuine reasons for belief, they create the possibility of normative conflicts, whereby one epistemically ought to believe p, but practically ought not to, or vice versa. It may not be so objectionable or counterintuitive to countenance conflicts in exceptional circumstances in which holding a patently epistemically irrational belief would be extremely (dis)valuable (e.g., in Threat). But it would be more worrisome to accept a) routine, pervasive conflicts, and b) conflicts that seem counterintuitive and distinctively troubling (e.g., the verdict that epistemic rationality requires holding racist beliefs in cases like Wedding). Moreover, Basu and Schroeder (2019) argue that not only is it objectionable to posit conflicts between moral and epistemic norms (whereby, for example, epistemic norms require a morally prohibited belief), but it is likewise unpalatable to allow a lack of coordination between moral and epistemic norms (whereby epistemic norms even permit a belief that morality prohibits). It is inappropriate, they claim, to apologize for holding an epistemically justified belief. So, they infer, epistemically justified beliefs cannot be morally wrong. Reasons pragmatism allows for conflicts—perhaps pervasive and especially troubling conflicts—and it certainly does not guarantee coordination. This, for some, is a reason to reject the view.

A third (and related) challenge—which arises primarily for interactionist versions of reasons pragmatism—is that it is difficult to provide a satisfactory account of how practical and epistemic reasons for belief interact to determine what we all-things-considered ought to believe. Selim Berker (2018) has most forcefully articulated this challenge. Berker observes that practical and epistemic reasons exhibit different weighing behaviors. While equally balanced practical reasons in favor of two competing alternatives generate a permission to choose either alternative, equally balanced epistemic reasons (according to Berker) generate a requirement to suspend judgment. In light of this difference, it is difficult to see how practical and epistemic reasons combine to render verdicts about what one all-things-considered ought to believe. Separatist pragmatism, of course, avoids this problem. But it faces a different worry:⁵ that by declining to offer a verdict about what we all-things-considered ought to believe in cases of conflict between practical and epistemic domains, it can in many cases offer no helpful guidance about what to believe.

II.2 Pragmatic encroachment

While proponents of reasons pragmatism have responses to these challenges,⁶ the difficulties are sufficiently significant to motivate exploring alternative accounts of how practical considerations affect what we ought to believe. The most prominent such account is pragmatic encroachment.⁷ Whereas reasons pragmatism holds that practical considerations constitute non-epistemic reasons for or against belief, pragmatic encroachment holds that practical considerations bear on what we ought to believe by affecting epistemic justification. According to the most common version of encroachment, practical considerations shift the threshold for how much evidence is necessary to epistemically justify belief (or for a belief to constitute knowledge). In at least some cases in which believing (or relying on one’s belief) is actually or potentially practically costly, the evidential threshold for epistemic justification increases (i.e., more evidence is needed to epistemically justify belief). And according to some versions of pragmatic encroachment, in some cases in which not believing is (actually or potentially) practically costly, the evidential threshold for epistemic justification decreases.⁸

Pragmatic encroachment has the advantage of avoiding the three problems facing reasons pragmatism. First, because it holds that practical considerations shift the threshold for how much evidence is needed for epistemic justification—rather than directly constituting practical reasons for or against belief—pragmatic encroachment provides a way for practical considerations to affect what we ought to believe without serving as the basis for belief. Second, by holding that epistemic norms are themselves sensitive to practical costs or risks, proponents of pragmatic encroachment can avoid the conclusion that there are pervasive and troubling conflicts between epistemic and practical norms. And third, pragmatic encroachment provides a straightforward account of the interaction between practical and epistemic considerations: instead of combining with epistemic reasons, practical considerations bear on what we ought to believe by affecting how strong the epistemic reasons must be to epistemically justify belief.

However, pragmatic encroachment faces problems of its own.⁹ First, Worsnip (2021a) argues that it is very difficult for proponents of encroachment to provide a principled explanation of which practical considerations encroach on epistemic justification. While proponents of encroachment have typically wished to include the practical considerations in cases such as Train Case 2, Wedding, Wine Stain, and (perhaps) PhD as encroaching on epistemic justification, they have typically wished to exclude bribes, threats, or Pascalian considerations about one’s eternal salvation or damnation from encroaching on epistemic justification.¹⁰ But, Worsnip argues, none of the most promising principles to which proponents of encroachment have appealed successfully distinguishes between those practical considerations that do encroach on epistemic justification and those that do not, such that they exclude all types of bribes, threats, or Pascalian considerations. Without such a principle, proponents of encroachment are arguably committed to the counterintuitive claim that sometimes bribes, threats, or Pascalian considerations affect the epistemic justification of our beliefs.

Relatedly, proponents of pragmatic encroachment have a difficult time explaining how certain moral costs of belief make a difference to epistemic justification. Fritz (2020) holds that traditional accounts of pragmatic encroachment can explain how the potential moral costs of relying on one’s belief affect epistemic justification. (He calls views on which only these moral costs make a difference “moderate moral encroachment.”) But he contends that proponents of what he calls “radical moral encroachment”—which holds that moral costs of beliefs themselves affect epistemic justification—cannot explain why these costs affect epistemic justification, rather than constituting practical reasons against belief.¹¹ One reason why this is particularly hard is that in many cases in which a belief—say, a racist belief—is itself morally wrong, this does not seem to merely raise the threshold for how much evidence one needs to permissibly believe, but rather to make the belief prohibited no matter how much evidence one has.¹² But if a belief is prohibited no matter how much evidence one has for it, it is hard to see how the relevant prohibition counts as epistemic.¹³

III.1 Permissivism about outright belief

Roughly, permissivist views in epistemology claim that, in some cases, one’s evidence is such that more than one of a range of incompatible doxastic attitudes toward a proposition would be epistemically justified (i.e., epistemically permissible) to hold. So, a permissivist about credence might hold that one’s evidence can be such that either a 0.7 credence in p or a 0.8 credence in p would be epistemically justified. The version of permissivism we will be making use of here, however, applies to coarse-grained doxastic attitudes toward a proposition p: believing p (i.e., having an “outright” belief in p), suspending judgment about p, and disbelieving p (i.e., having an outright belief in its negation). The claim is that there are situations in which one’s evidence is such that more than one of these attitudes would be epistemically justified (i.e., epistemically permissible) to hold.

For our purposes, we need take no stand on whether one’s evidence is ever such that believing p and disbelieving p would each be epistemically justified. Rather, we aim to motivate only the claim that there are cases in which believing p and suspending judgment about p would each be epistemically justified.¹⁴ Moreover, the version of permissivism we aim to motivate is both “intrapersonal”—holding that a single agent can be epistemically justified either in believing p or in suspending judgment about p—and “synchronic”—holding that this agent can be epistemically justified either in believing p at time t or in suspending judgment about p at a single time t.¹⁵

Thus, the view is as follows:

Intrapersonal, Synchronic Permissivism about Outright Belief. There are cases in which an agent’s evidence is such that (i) they would be epistemically justified in believing p at t; and (ii) they would be epistemically justified in suspending judgment about p at t.

III.2 Motivating permissivism

There are several motivations for accepting permissivist views. Permissivism about credence is often motivated via appeal to the intuitive implausibility of the claim that one’s evidence makes only one precise point-valued credence permissible. For example, it may seem absurd to say that one’s evidence makes credence 0.8136 permissible, but neither credence 0.8137 nor credence 0.8135. The facts about evidential support, it may be thought, are (at least typically) just not so fine-grained. And both permissivism about credence and permissivism about outright belief are sometimes motivated via appeal to the thought that there is sometimes more than one reasonable way to evaluate a particular body of evidence: when one’s evidence is complex and involves many different considerations, there is no uniquely correct way to weigh all those different considerations.

Here, however, we want to focus on a different way to motivate permissivism, one that is specific to permissivism about outright belief which will be of particular use for the permissivist-pragmatist hybrid we defend.¹⁶ Suppose that the probability of p on your evidence is exactly 0.9, and suppose that you have correctly determined this and, correspondingly, have formed a credence of 0.9 in p. You still face a further question about whether to (outright) believe p or not.¹⁷ Our claim is this: in this kind of situation, often neither believing p nor suspending judgment about p would constitute an epistemic mistake.

To bring this out, let us begin with an interpersonal case. Suppose that Manchester City, the top-ranked team in the English Premier League, are scheduled to play Norwich City, the bottom-ranked team.¹⁸ Liam and Noel are both rabid Manchester City fans. Together they have built a complex regression model drawing on a huge amount of data to assign probabilities to different outcomes for football (soccer) games. And Liam and Noel—rationally, let’s suppose—base their credences in the different outcomes solely on the model’s outputs. Their model assigns to a Manchester City win a probability of 0.9 in their game against Norwich City, so they both have a rational credence 0.9 that Manchester City will win.

But now suppose that Liam and Noel differ in their attitudes toward epistemic risk. At least when it comes to football games, Liam is relatively risk-seeking in his outright belief-forming practices. This does not involve being overconfident in the sense of overestimating the probabilities of Manchester City wins; his credences in those probabilities are perfectly epistemically rational. Rather, his risk-seeking involves being inclined to outright believe a proposition about who will win a football game when there is fairly strong, but far from completely infallible, evidence in its favor: evidence that would make, for example, a 0.9 credence rational. Consequently, Liam outright believes that Manchester City will win their game against Norwich City. By contrast, Noel is relatively risk-averse in his outright belief-forming practices. He is not generally inclined to believe a proposition about who will win a football game on the basis of evidence that would make a 0.9 credence rational. Consequently, Noel does not outright believe that Manchester City will win their game against Norwich City, but instead suspends judgment.

Here is our judgment, which we invite you to share: neither Liam nor Noel is making an epistemic mistake. There is a range of different permissible attitudes toward epistemic risk,¹⁹ and neither Liam nor Noel falls outside the permissible range.²⁰ We can think of this in terms of the twin “Jamesian” epistemic goals of believing the truth and avoiding error.²¹ As many epistemologists have emphasized, when it comes to propositions for which one has strong but not infallible evidence, these goals trade off against each other: believing gives one a good shot at believing the truth, but carries a risk of error; whereas suspending judgment guarantees avoidance of error, but precludes one from believing the truth. (Relatively) risk-seeking epistemic agents like Liam and (relatively) risk-averse epistemic agents like Noel assign different weights to believing the truth as opposed to avoiding error. Liam puts more weight on believing the truth than Noel does, and Noel puts more weight on avoiding error than Liam does. We find it extremely plausible that there is no uniquely correct way, at least from an epistemic point of view, to weigh the value of believing truth versus that of avoiding error. If this is right, it helps to explain why Liam and Noel are both epistemically justified in their doxastic attitudes.

As we said, this is an interpersonal case, and it might be objected that it does not support intrapersonal permissivism. Given Liam’s attitude toward epistemic risk, it might be said, he is epistemically justified only in believing that Manchester City will win; and given Noel’s attitude toward epistemic risk, he is epistemically justified only in suspending judgment.²² Thus, while this is a case in which the same evidence makes different attitudes permissible for different agents, it isn’t one in which a body of evidence makes more than one attitude permissible for the same agent. But we think this is the wrong way to think about things. It would be entirely permissible for Liam to have Noel’s attitude toward epistemic risk (and correspondingly, suspend judgment); likewise, it would be entirely permissible for Noel to have Liam’s attitude toward epistemic risk (and correspondingly, believe). Thus, Liam is permitted either to believe or to suspend judgment. It is just that he would need to do the latter in concert with having a different attitude toward risk.²³ We therefore think the case for interpersonal permissivism here plausibly extends to intrapersonal permissivism.

It might now be worried that, because it takes time to revise one’s attitude toward risk, Liam is permitted to suspend only once he has revised that attitude, and consequently, the case supports only diachronic and not synchronic permissivism.²⁴ But this objection too is misguided. First, the claim that Liam is permitted to suspend only once he has revised his attitude toward risk is, we think, mistaken. Rather, our view is that at the very time t that Liam has a risk-seeking attitude, he is permitted to either {have a risk-seeking attitude, believe} or {have a risk-averse attitude, suspend}.²⁵ That he has the former combination of attitudes at t does not render the latter combination impermissible at t (even if it means that he can feasibly adopt it only at a time after t).²⁶ Second, even if one finds the first response unpersuasive, the case still establishes synchronic permissivism. Again, let t be a time when Liam has the risk-seeking attitude. Let t1 be a time far enough after t that Liam has had time to revise his attitude toward risk by t1. Since it would also be permissible for Liam not to revise his attitude toward risk by t1, there is a time—namely, t1—when it would be permissible for Liam to either {have a risk-seeking attitude, believe} or {have a risk-averse attitude, suspend}. But then synchronic permissivism is true, since this thesis requires only that there is at least one time such that more than one outright doxastic attitude would be permissible to have at that time.

The case of Liam and Noel brings out the plausibility of the claim that often, when one has strong but not infallible evidence for a proposition, one’s evidence permits either believing or suspending judgment. As the case illustrates, one thing that influences whether one believes or suspends judgment in such a case is one’s attitude toward epistemic risk. But this general attitude toward epistemic risk may not be the only thing that might properly influence whether one believes in permissive cases.

We now turn to the idea that considerations concerning the practical costs and benefits of believing might do so. This will introduce the reasons-pragmatist aspect of our permissivist-pragmatist view, and allow us to contrast it with pragmatic encroachment.

III.3 How permissivism makes room for pragmatism

The idea of the permissivist-pragmatist view is simple: when there is more than one epistemically permitted doxastic attitude, practical considerations, including moral ones, can come in to determine which of these epistemically permitted doxastic attitudes one all-things-considered ought to have.

It is vital to distinguish this view from pragmatic encroachment. On the pragmatic encroachment view, practical considerations affect what one is epistemically justified in believing, or what one knows. This is not so on the permissivist-pragmatist view. On the latter view, practical considerations make no difference to what one is epistemically justified in believing or what one knows; they only influence what one all-things-considered ought to believe.²⁷ However—crucially—at least in permissive cases, they can do this without making it the case that one all-things-considered ought to believe something that one is epistemically unjustified in believing, since they select between the epistemically permitted options.

We can model all this a bit more precisely. As we saw in §2, on the most popular version of pragmatic encroachment, at least some actual or potential practical costs of believing p make it the case that one requires more evidence for epistemically justified belief in p than one would otherwise require, were those costs not present. Suppose that one has strong but far from infallible evidence for p. Let the “default case” be the one in which there are no special practical costs of believing p, and the “costly case” the one in which there are (actual or potential) costs (of the kind the pragmatic encroacher thinks makes a difference). Simplifying somewhat,²⁸ we can represent the pragmatic encroachment view pictorially as follows:

The diagonal black shading represents the extent to which one’s evidence probabilifies p (filling to the top of the cup would represent its giving p probability 1). The dotted line is the threshold for epistemically justified belief. Below it, the background is colored red to indicate that any amount of evidence below this threshold would not suffice for epistemically justified belief; above it, it is colored green to indicate that any amount of evidence above this threshold suffices for epistemically justified belief. According to the pragmatic encroachment view, the threshold rises in the costly case compared with the default case. This means that although one’s evidence probabilifies p to the same extent, in the costly case one is no longer justified in believing p.

By contrast, our view (again, simplifying somewhat) can be represented as follows:

Here, we have a zone of epistemic permission, shaded yellow, between the threshold for permitted belief and that for required belief. Pretty strong but far from infallible evidence—as in the case depicted here—will exceed the former threshold but not the latter. The epistemic status of belief in p does not change between the default and costly case: in both cases it is epistemically permitted but not required, and in both cases the thresholds for both epistemically permitted and epistemically required belief remain the same. So there is no pragmatic encroachment on epistemic justification.

However, let’s now introduce a different kind of threshold, which we will call one’s “personal psychological threshold” (PPT) for belief in a proposition.²⁹ One’s personal psychological threshold is the threshold of apparent evidence above which one will actually believe the proposition in question. Now, on a non-permissivist view, one epistemically ought to match one’s personal psychological threshold to the threshold for epistemically justified belief. By contrast, given permissivism, it is epistemically permissible to set one’s personal psychological threshold anywhere between the threshold for epistemically permitted belief and that for epistemically required belief (inclusive of these thresholds): that is, anywhere in the yellow zone above. However, it does not follow that it is always all-things-considered permissible to set one’s personal psychological threshold anywhere in this zone. Rather, practical considerations might make it the case that one all-things-considered ought to set one’s personal psychological threshold at a relatively high (or low) level within this zone.³⁰

Thus, there is a difference—albeit not one in epistemic status!—between the default and costly cases. In the latter case, one (all-things-considered) ought to set one’s personal psychological threshold at a relatively high level within the yellow zone. We can thus specify our view of the costly case in greater detail:

Here, the orange-colored area represents the zone within which belief is epistemically permissible, but all-things-considered prohibited—and given the amount of evidence in this case, that is the status which believing p has. By contrast, in the default case, one is (all-things-considered) permitted to set one’s personal psychological threshold lower within the epistemically permitted zone.

As a result, the permissivist-pragmatist view has some important similarities with pragmatic encroachment. Both views agree that purely evidential considerations underdetermine where to set one’s personal psychological threshold.³¹ And both views say that practical considerations can help to determine where one’s personal psychological threshold should be. However, our view demonstrates that, by accepting epistemic permissivism, it is possible to embrace both claims without allowing any pragmatic encroachment on the epistemic: that is, without allowing that practical considerations make any difference to the epistemic status of one’s beliefs.

Finally, let us compare our view with one combining reasons pragmatism with a non-permissivist view of epistemic justification. On a non-permissivist view, the thresholds for epistemically permitted and epistemically required belief are the same: below this threshold one is epistemically prohibited from believing, while above it one is epistemically required to believe. Thus, if practical considerations can shift the personal psychological threshold that one all-things-considered ought to have, they must move it away from the threshold for epistemically permitted and required belief:

The divergence of the personal psychological threshold one all-things-considered ought to have from the threshold for epistemically permitted and required belief opens up a zone, here colored purple, in which belief is epistemically required and yet all-things-considered forbidden. Indeed, on this view, any case in which practical considerations make a difference to what one all-things-considered ought to believe must be one in which they either make it all-things-considered prohibited to have an epistemically required belief, or make it all-things-considered required to have an epistemically prohibited belief. By contrast, while our view does not definitively rule out some such cases (see §5.2), it opens the way for practical considerations to influence what one all-things-considered ought to believe without doing this (and, as we have already shown, without encroaching on the epistemic).

One final clarification about our view. We have talked of practical considerations making a difference to what one ought to believe. Do they do so by constituting reasons for or against belief? One might suggest that they don’t. As we have proposed, we can think of practical considerations as influencing the personal psychological threshold that one ought to have. So perhaps, it might be suggested, they are (directly) reasons for adopting a particular personal psychological threshold rather than being reasons for (or against) belief as such. Now, this suggestion may be right. However, there are grounds to doubt the metaphysical distance between adopting a personal psychological threshold and believing (or suspending). Most ordinary people do not explicitly think of themselves as having a threshold of probability above which they are willing to believe a proposition. Consequently, a personal psychological threshold is best understood dispositionally or counterfactually in terms of what one would believe or not believe given different (perceived) amounts of evidence. Thus, in adopting a personal psychological threshold such that one’s (perceived) evidence for p exceeds that threshold, one arguably thereby believes p: if one does not believe p, one has not really adopted the threshold in question. Thus, at least in some cases, practical considerations plausibly constitute reasons to adopt a particular personal psychological threshold and thereby believe (or suspend).

IV.1 Train Cases 1 and 2

We will now demonstrate how the permissivist-pragmatist account can handle the cases described in §1. Let’s start with the Train Cases, in which the potential costs of relying on a belief intuitively affect what you ought to believe. Fantl and McGrath originally employed these cases to argue for pragmatic encroachment. And pragmatic encroachment provides a straightforward explanation of the difference between these two cases: while you are justified in believing that the train will stop in Foxboro in Train Case 1, in Train Case 2 the practical risks of relying on that proposition render belief epistemically unjustified. Proponents of pragmatic encroachment typically explain this by positing something like the following principle:

Epistemic justification-reliance link: If S is epistemically justified in believing p, then S is justified in relying on p in S’s reasoning.

Applying epistemic justification-reliance link to Train Case 2 generates the following explanation: since you are not justified in relying on the proposition that the train will stop in Foxboro, by modus tollens you are not epistemically justified in believing that the train will stop in Foxboro.

Although epistemic justification-reliance link is a normative principle stating a sufficient condition for being justified in relying on a proposition, it is often defended by appeal to a descriptive view on which believing that p involves relying on p in one’s reasoning, or being disposed to rely on it.³² If this is right, costs of relying on p are ipso facto costs of believing p.

Because the permissivist-pragmatist view denies that practical considerations affect epistemic justification, and because practical considerations obviously do affect which propositions one is justified in relying on, proponents of permissivist pragmatism cannot accept epistemic justification-reliance link. However, they can (if desired) accept the following alternative:

All-things-considered permission-reliance link: If S is all-things-considered permitted to believe p, then S is justified in relying on p in S’s reasoning.

Notice that appeal to a descriptive connection between belief and reliance does not favor epistemic justification-reliance link over all-things-considered permission-reliance link. Even if costs of reliance are ipso facto costs of believing, this does not show that they are costs of believing that affect one’s epistemic justification, as opposed to affecting whether one is all-things-considered permitted to believe p by constituting non-epistemic reasons against believing. And by appealing to all-things-considered permission-reliance link, the permissivist pragmatist can explain why you ought not to believe that the train will stop in Foxboro in Train Case 2: since you are not justified in relying on the proposition that the train will stop in Foxboro, by modus tollens you are not all-things-considered permitted to believe this proposition.

But it does not follow that you aren’t epistemically permitted to believe that the train will stop in Foxboro. Rather, the permissivist pragmatist can diagnose the case—and other similar cases used to motivate pragmatic encroachment—as an epistemically permissive one: that is, it is a case where you are epistemically permitted to either believe or suspend judgment. Like the case of Liam and Noel from §3.2, Train Cases 1 and 2 are scenarios in which you have pretty strong but far from infallible evidence that justifies a reasonably high credence in the proposition that the train will stop in Foxboro. While we share Fantl and McGrath’s intuition that it is permissible to believe in Train Case 1, it seems to us that you also don’t make an epistemic mistake if you, more cautiously, suspend judgment. That is, the evidence seems to permit either doxastic attitude. The permissivist pragmatist says that, likewise, neither believing nor suspending judgment constitutes an epistemic mistake in Train Case 2; rather, believing constitutes a prudential mistake.

Thus, the permissivist pragmatist can provide the following overall verdicts about the cases. In Train Case 1 (the low-stakes version), either believing or suspending is epistemically permissible, and either believing or suspending is practically permissible. Thus, all-things-considered, belief is permitted (but not required). To put things in terms of our model from §3.3, you are practically permitted to adopt a personal psychological threshold anywhere within the epistemically permitted zone, and thereby either to believe or to suspend judgment. In Train Case 2 (the high-stakes version), it is still the case that either believing or suspending is epistemically permissible, but, due to the potential cost of believing falsely, believing is practically impermissible (whereas suspending judgment is practically permissible). Thus, belief is all-things-considered impermissible, and suspending judgment is all-things-considered required. In terms of the model, you practically ought to adopt a relatively high personal psychological threshold, near the “top” of the epistemically permitted zone. Given that you have pretty strong but far from infallible evidence for the proposition that the train will stop in Foxboro, the evidence probabilifies this proposition to a degree below the high personal psychological threshold that you all-things-considered ought to adopt. You therefore all-things-considered ought to suspend judgment.

Is this permissivist-pragmatist diagnosis of the case at a disadvantage compared with the pragmatic encroachment diagnosis? The answer depends on whether there is a clear, pretheoretical intuition not just in favor of the coarse-grained verdict that you ought not to believe in Train Case 2, but in favor of the finer-grained verdict that you ought not to believe because you are epistemically unjustified in believing. For our part, we do not have the more fine-grained intuition. Since the intuition that belief is impermissible is driven by the potential practical costs of relying on one’s belief, we think it is at best unclear whether the prohibition in question is epistemic, and we are suspicious of claims to be able to intuit that it is.³³ The link between belief and reliance to which pragmatic encroachers appeal is intended to provide a powerful theoretical case for pragmatic encroachment beyond appeal to intuitions about cases. Yet it is only if this link is developed in terms of epistemic justification that it supports pragmatic encroachment over the permissivist-pragmatist view. And it is not clear what, beyond a similar appeal to questionable fine-grained intuitions about cases, justifies developing the link between belief and reliance in terms of epistemic justification, rather than in terms of all-things-considered permissibility.³⁴ Absent such a justification, the pragmatic encroacher’s diagnosis of Train Case 2 does not have an advantage over the permissivist pragmatist’s diagnosis.

Of course, one might not have either the coarse-grained or fine-grained intuition—particularly if one sees belief and reliance as not being all that tightly connected, such that one could believe without incurring the risks of reliance.³⁵ For the sake of argument, we have been granting to the pragmatic encroacher that there is such a link, and thus that we wish to accommodate the verdict that you (at least all-things-considered) shouldn’t believe in Train Case 2. But the permissivist-pragmatist view is also entirely compatible with a view on which there is no such link, and Train Case 2 thus would not be a case where significant practical costs or reasons come into play: there you are all-things-considered permitted to believe, but you shouldn’t rely on your belief if you do so. By contrast, pragmatic encroachers need a principle like epistemic justification-reliance link to explain why there is a difference in justification for belief between Train Case 1 and Train Case 2, and hence to establish that pragmatic encroachment occurs. We take this difference to be a point in favor of the permissivist-pragmatist view: it is consistent with but doesn’t depend on accepting a normative link between belief and reliance.

IV.2 Wedding, Wine Stain, and PhD

Let’s turn to how permissivist pragmatism addresses cases such as Wedding, Wine Stain, and PhD, in which moral costs associated with belief or suspension themselves—independently of the costs of relying on a belief—plausibly affect what we ought to believe. Proponents of pragmatic encroachment can say that the fact that holding a belief will (actually or potentially) wrong someone makes it such that more evidence is needed to epistemically justify belief, and that the fact that suspending judgment will (actually or potentially) wrong someone makes it such that less evidence is needed to epistemically justify belief. Some proponents of pragmatic encroachment claim that their view provides the best way to handle these cases since (as discussed in §2.1) it avoids positing a conflict between epistemic and moral norms in these cases.

However, another way to avoid positing a conflict between epistemic and moral norms in these cases is to hold that—like the pair of train cases—Wedding, Wine Stain, and PhD are all permissive cases.³⁶ Again, we find it very plausible that all of these cases are indeed permissive: in each case, you have pretty strong but far from infallible evidence, and you do not seem to make an epistemic mistake either by believing or by suspending judgment. We thus disagree with Gendler (2011), who seems to imply that refraining from believing in cases such as Wedding constitutes ignoring base rates about the proportion of service-industry workers who are Black, and thus involves epistemic irrationality. While base rates should be encoded in Andrew’s credence,³⁷ having a rational high credence while also suspending judgment does not constitute ignoring base rates in any good sense. Hence, the base rate information does not epistemically require Andrew to outright believe that John is a waiter.³⁸

If Wedding, Wine Stain, and PhD are permissive cases, our permissivist-pragmatist view applies straightforwardly. In Wedding and Wine Stain, belief is epistemically permitted but morally prohibited, whereas suspension of judgment is the only attitude both epistemically and morally permitted; hence, one all-things-considered ought to suspend judgment. In PhD, belief is epistemically permitted but morally required, whereas suspension of judgment is epistemically permitted but morally prohibited; hence, one all-things-considered ought to believe. Notably, in each case, there is an option that is both epistemically permitted and morally permitted (or required). We can consequently avoid a conflict between epistemic and moral norms in these cases without appealing to pragmatic encroachment.

Whether the more specific personal psychological threshold model developed in §3.3 can be applied in these cases may depend on whether beliefs can morally wrong only when false (Schroeder 2018, 2021) or even when true (Basu 2019b, Fabre 2022). If beliefs can wrong only when false, then the risk of doxastic wronging can affect what one ought to believe by shifting the personal psychological threshold one all-things-considered ought to adopt. But if beliefs can wrong even when true, such beliefs are plausibly morally impermissible no matter how much evidence one has. It may then be misleading to say that one should raise one’s personal psychological threshold in such a case: rather, one shouldn’t believe no matter how much evidence one has.

However, the permissivist pragmatist might reply that in cases where a belief wrongs even if it is true, one should simply adjust one’s personal psychological threshold to whatever level is required (given one’s evidence) to preclude belief. Since a personal psychological threshold is not a threshold for epistemic justification, this move does not seem obviously unprincipled to us.³⁹ In any case, regardless of whether the PPT model applies here, we can still preserve the core of the permissivist-pragmatist view in these cases: when two different doxastic attitudes are both epistemically permissible, moral considerations can come in to determine which of the two one should all-things-considered have.

IV.3 Threat

Thus far, we have diagnosed the cases in which practical considerations intuitively bear on what we ought to believe as plausibly epistemically permissive. But there are other cases in which practical considerations may bear on what we ought to believe—such as Threat—in which belief is clearly epistemically impermissible.

Because permissivist pragmatism is an account of how practical considerations affect what we ought to believe in epistemically permissive cases, permissivist pragmatists are not committed to accepting or denying any particular view of how practical considerations affect what we ought to believe in cases like Threat in which belief is epistemically prohibited. It is consistent with permissivist pragmatism to hold that there is a genuine normative conflict, which can either be left unresolved, be resolved in favor of the practical norms, or be resolved in favor of the epistemic norms. But it is also consistent with permissivist pragmatism to adopt a more restricted form of reasons pragmatism on which practical considerations do not constitute reasons for belief when belief is epistemically prohibited. This position would preclude conflicts between moral and epistemic norms.

V.1 Psychological impossibility

As mentioned in §2.1, those who wish to bring out the purported psychological impossibility of believing on the basis of pragmatic considerations usually appeal to cases like being offered $1,000,000 to believe that 2+2=5, or to believe that the sky is green. These cases have a striking feature: they involve being offered an incentive to believe something for which one obviously has inadequate evidence. It is quite plausible that practical incentives cannot motivate one to believe something that one judges oneself to have inadequate evidence for. And it is tempting to infer from this that practical considerations can never motivate one to believe, and thus can never constitute normative reasons for belief.

However, this line of thought overlooks the possibility of permissive cases. It is far from obvious that practical considerations cannot motivate us to believe (or suspend) in cases in which we take the evidence to merely permit, but not require, believing. In fact, numerous philosophers have argued that we do in fact have voluntary control over our beliefs in such cases.⁴⁰ Suppose, again, that one judges oneself to have pretty strong but not completely infallible evidence for p, and that one judges that either believing or suspending on p would be epistemically permissible. One nevertheless faces the question of whether to believe p or suspend judgment about whether p. We think that it is quite plausible that, in such a case, one can choose to believe p or to suspend judgment about whether p on the basis of practical considerations.

Indeed, we must have this ability if it is possible to respond to pragmatic and moral considerations in cases such as Train Case 2, Wedding, Wine Stain, and PhD—and it does indeed seem possible. For instance, it’s very natural to say that one can choose to suspend judgment on whether one’s alcoholic spouse has relapsed—as opposed to settling on the belief that they have done so. This is particularly natural if one thinks that belief is tightly connected to closing inquiry, and suspension of judgment to keeping inquiry open.⁴¹ One can choose to inquire into the matter more rather than making up one’s mind and believing now. Similarly, we think that when one already has a pretty high credence that one’s friend will complete their PhD, and judges that this is at least epistemically permissible to believe, one can choose to make up one’s mind and believe that one’s friend will successfully complete their PhD, rather than continuing to entertain doubts and inquire further.⁴² In our view, what explains the possibility of doing these things—as compared with cases in which one is, say, bribed to believe something absurd—is that the former cases are (at least implicitly) recognized by the agent to be permissive.⁴³

V.2 Conflicts and coordination

As we have already observed, going permissivist about epistemic norms at least reduces the prevalence of conflicts between epistemic and moral norms. For example, it allows us to say that there is no conflict in cases such as Wedding, Wine Stain, and PhD. In all these cases, there is a way to satisfy both the epistemic and moral norms.⁴⁴

However, in these cases permissivist pragmatism does not deliver coordination between epistemic and moral norms: in all of them, there is an attitude that is epistemically permissible but morally impermissible. For example, in Wedding, believing is epistemically permissible but morally impermissible. How troubling should that be? In our view, not very, given that believing p is not epistemically required, and there is an alternative—viz., suspending—that is both morally and epistemically permissible. The phenomenon of an option being permitted by one set of norms but prohibited by another is perfectly familiar. It arises whenever one set of norms leaves some choice between multiple options open, and another set of norms narrows that choice down. For example, it occurs whenever there are two morally permissible options, one of which is required of you given prudential considerations, or two prudentially permissible options, one of which is morally required of you. This is simply not as troubling as outright conflict.

What, then, of Basu and Schroeder’s argument for coordination? Their key premise is that it is inappropriate to apologize for holding an epistemically justified belief. But if ‘justified’ means permissible, we think this premise is too strong. It is plausible that it is inappropriate to apologize for holding an epistemically required belief. For if a belief is epistemically required, a conscientious follower of the evidence has, in effect, no choice but to hold it. But it is much less plausible that it is (always) inappropriate to apologize for a belief that is merely epistemically permitted. For here it is not as if a conscientious follower of the evidence has no choice but to hold the belief: it is open to them to suspend, and if the moral considerations require doing so, that is what they should do.

In this subsection, we have so far been discussing only epistemically permissive cases in which the moral norms require one of the epistemically permitted options (and prohibit the others). But we must acknowledge that our view does not demonstrate that all cases will be like this. It does not rule out the possibility of outright conflicts between epistemic and practical (including moral) norms in non-permissive cases, or in permissive cases where moral considerations favor an option that is not among the permitted ones.

However, we do not think this constitutes a severe problem for our view in particular, for two reasons. First, the onus is on the opponent of our view to provide a case in which (a) our view is committed to saying that morality prohibits a doxastic attitude that is categorically epistemically required, and (b) the verdict that epistemic and practical norms conflict is not intuitively plausible. We submit that such cases have not been forthcoming in the literature so far. Examples such as Wedding and Wine Stain involve morality prohibiting a doxastic attitude that is epistemically permitted, but not required, and so they fail to satisfy (a). On the other hand, cases such as Threat more plausibly satisfy (a),⁴⁵ but in our view they are very plausibly viewed as cases of conflict between moral and epistemic norms, and thus do not satisfy (b).

Second, pragmatic encroachment also does not categorically rule out conflicts between practical (including moral) and epistemic norms.⁴⁶ This is because there are plausibly limits on how low the threshold for epistemically justified belief can go, even given encroachment. If one’s evidence for a proposition is too weak to meet even the lowest possible threshold, even the encroacher will have to say that belief is epistemically unjustified, no matter how strong the moral reasons are. That opens the way for conflicts. In fact, it is not obvious to us that there are any cases in which we are committed to a conflict but the pragmatic encroacher is not. This is because, we suspect, whenever the evidential situation makes it plausible for the pragmatic encroacher to claim that a morally required attitude is epistemically justified given the moral considerations, it likewise makes it plausible for the permissivist to claim that the morally required attitude is among the epistemically permitted options (even if it isn’t the only epistemically permitted one). Thus, this purported advantage of pragmatic encroachment over reasons pragmatism has been neutralized.

V.3 Interaction

Let us turn next to the closely related issue of how epistemic and practical considerations interact (or don’t). Recall that “interactionist pragmatists” hold that epistemic and practical reasons combine to produce verdicts about what one all-things-considered ought to believe, while “separatist pragmatists” hold that they do not. Both views have their problems: interactionist pragmatists face the difficult task of explaining how seemingly incommensurable considerations weigh against each other, while separatist pragmatists fail to issue clear guidance about what one ought to believe.

With permissivism on the table, a strict separatist view—on which epistemic and practical considerations never interact to produce all-things-considered verdicts—becomes implausible. Suppose that two doxastic attitudes are each epistemically permitted, but only one is morally permitted.⁴⁷ It is incredibly natural to say that you all-things-considered ought to have the attitude that is both epistemically and morally permitted. Similarly, if two doxastic attitudes are both morally permitted but only one is epistemically permitted, it is incredibly natural to say that you all-things-considered ought to have the attitude that is both epistemically and morally permitted. Thus, there are at least some verdicts about what one all-things-considered ought to believe. Moreover, there is no mystery about how the moral and epistemic considerations combine to produce these all-things-considered (ATC) verdicts. We can map how they do so, on our view as specified so far:

The highlighted boxes are those that open up only once we recognize the possibility of epistemically and/or morally permissive cases. If we ignore this possibility, it looks like we only get clear all-things-considered verdicts when the moral and epistemic considerations happen to yield the same verdict. Given this, one might hold that moral and epistemic considerations never interact in any significant way: they either happen to concur, or they conflict. But once we acknowledge the possibility of permissive cases, we see more substantive interaction, whereby the moral and epistemic statuses of a belief are not the same, yet the belief has an all-things-considered status. This suffices, we think, to render strict separatist pragmatism very unattractive.

Of course, we are still left with the question of what to say in the boxes marked with ‘??’. These are cases in which epistemic and moral norms conflict. We take no stand about whether or how epistemic and moral norms interact in such cases. One of us is inclined to embrace a kind of restricted separatism which says that there is no all-things-considered verdict about these cases—while noting that such cases will be significantly rarer on our view than on a non-permissivist view, and that moral and epistemic considerations do interact in permissive cases. (This position is intermediate between the separatist and interactionist views.) But one could also marry our account with one of the existing accounts of how epistemic and practical considerations weigh against each other (Reisner 2008; Howard 2020), or with a view on which, in conflict cases, either practical considerations always take precedence or epistemic considerations always take precedence. In any case, attention to permissive cases at least opens a way for epistemic and practical considerations to interact in a non-mysterious, utterly unproblematic way in a range of cases.

V.4 The problems for pragmatic encroachment

While our view is in a good sense a version of reasons pragmatism, it is not a version of pragmatic encroachment because it denies that practical factors affect epistemic justification. Thus, it avoids both the problems raised for pragmatic encroachment. First, there is no need to delimit which practical considerations can make a difference to what one ought to believe. For while it doesn’t seem like bribes, threats, or Pascalian considerations are the right kind of things to make a difference to epistemic justification, it is far less obvious that they can’t (in principle) make a difference to what one all-things-considered ought to believe. We can therefore allow that, at least within the range of epistemically permissible options, any kind of practical considerations can help determine which of the epistemically permissible options one all-things-considered ought to adopt. Second, we do not need to explain why the wrongness of a belief itself, independently of the costs of relying on it, makes a difference to epistemic justification, as we deny that practical considerations affect epistemic justification. Rather, we say that the wrongness of a belief itself makes a difference to what one all-things-considered ought to believe, within the range of epistemically permitted options.