1 Introduction

When Roderick (1956; 1959) introduced the notion of epistemic rationality, he argued on extensional grounds. Firth claimed that unless we introduced a distinctively epistemic sense of rationality, we would make extensionally false claims about rational belief. For example, we would say that agents should believe what will make them happy or well off, rather than what their evidence supports.

Epistemic purists follow Firth in positing a distinct type of epistemic rationality.Footnote 1Impurists follow Chisholm (1956; 1961) in denying the need to posit a new type of epistemic rationality. Impurists hold that the notion of epistemic rationality is unparsimonious (Papineau, 2013; Rinard, 2019a), potentially ill-defined (Cohen, 2016a, b; Chisholm, 1991) or a distraction from important normative phenomena (Dotson, 2019).Footnote 2

As time went on, Firth’s extensional argument began to be challenged. Impurists discovered that they could mimic most of the purist’s extensional judgments on Cliffordian grounds without positing a distinctively epistemic type of rationality. Debate shifted to edge cases where Cliffordian arguments disagree with traditional epistemic theories about what agents should believe. But in edge cases, impurists claimed, it is the impurist and not the purist who has the right of it. Impurists held that we may rationally believe against the evidence in order to be good friends (Keller, 2004; Stroud, 2006); maintain confidence and psychological health (McKay & Dennett, 2009; Rinard, 2019a); avoid racism (Basu, 2019; Dotson, 2019; Gendler, 2011); show faith in humanity (Preston-Roedder, 2013); or respond to external incentives for holding particular beliefs (Reisner, 2009; Rinard, 2019b).

The ‘new evidentialists’ (Rinard, 2015) held their ground on extensional questions. They argued that there could not, even in principle, be non-epistemic reasons for belief, and hence that it could never be rational to believe against the evidence. Non-epistemic reasons could not exist because they would not be motivating reasons (Kelly, 2002, 2003); would play the wrong role in doxastic deliberation (Shah, 2003, 2006); or would be wrong kinds of reasons for belief, and hence really reasons for action rather than belief (Way, 2012). Over time, these arguments met with resistance, and some evidentialists tried a new tack.Footnote 3

A group we might call the newer evidentialists cedes extensional ground to the impurist (Christensen, 2021; Maguire & Woods, 2020).Footnote 4 The newer evidentialists concede, at least for the sake of argument, that there can be non-epistemic reasons for belief, and that it can sometimes be rational to believe against the evidence on the basis of non-epistemic reasons. However, the newer evidentialists maintain that in addition to rationality simpliciter we must also posit a distinct notion of epistemic rationality which is not reducible to rationality simpliciter. Even if the impurist is correct about what rationality requires, the purist may yet be right about the mandates of epistemic rationality.

These concessions are welcome, but I think that the newer evidentialists may have given away too much ground. Once we grant that the impurist has given a correct account of rational belief, we deprive ourselves of the original extensional motivation for positing a distinctive type of epistemic rationality. After that, it becomes more difficult to motivate the need for a distinct type of epistemic rationality, and there is a real threat that the notion will not be needed at all.Footnote 5

To illustrate this threat, I survey two newer evidentialist views in detail and argue that they face opposite challenges. I argue that the first, due to Christensen (2021), may correctly describe what rationality requires, but fails to posit a distinctively epistemic type of rationality. The second, due to Maguire and Woods (2020), posits distinctively epistemic types of epistemic reasons and rationality, but struggles to explain why epistemic reasons are normative reasons and why epistemic rationality is a type of rationality. Together, these discussions suggest that the newer evidentialists’ extensional concessions may bring us closer to impurism than the newer evidentialists hoped to land.

Here is the plan. Section 2 presents Christensen’s view. Section 3 argues that this view is best understood as a type of global consequentialism on which evidence bears on rational belief in the same way that it bears on rational action. I argue that this view is plausible, but does not commit the impurist to recognizing a distinctively epistemic type of rationality. Section 4 presents Maguire and Woods’ view. Section 5 argues that this view commits us to denying a range of platitudes about normative reasons and rationality. I suggest that we should save the platitudes by redescribing Maguire and Woods’ view using other normative categories, unless Maguire and Woods can provide significant support for their view. Section 6 argues that this support has not been provided. Section 7 concludes with the suggestion that newer evidentialism may represent a wrong turn in purist theorizing. Purists would do better to retreat from the newer evidentialist program back towards more traditional arguments for purism.

2 Christensen’s evidence-relative view

Christensen (2021) aims to show that the concept of epistemic rationality is indispensable not only to the evaluation of rational belief, but also to the rationality of action. Christensen does this by criticizing two candidate impurist views and arguing that their failure pushes us towards a third view, which involves a type of epistemic rationality.

First, impurists might adopt a subjectivist view on which rationality is determined by agents’ background beliefs. On a consequentialist reading, this view holds:

Belief-Relative View (BRV): It is rational for an agent to believe that P just in case believing P (as opposed to suspending belief about P or disbelieving P) would have the best expected consequences, given the agent’s beliefs. (Christensen, 2021, p. 503).

However, Christensen argues, BRV gets the wrong answer when agents hold evidentially unsupported beliefs. Consider:

Good Ol’ Charlie: Every morning, Charlie eats a bar of soap. Without fail, it tastes terrible and makes Charlie sick. Today, Charlie is faced with an identical bar of soap. Charlie believes that from today onwards, the soap will taste great and improve his health. Charlie concludes that the soap will taste great today and eats it. (Christensen, 2021, p. 504).

The problem is that BRV seems to imply that Good Ol’ Charlie is rational in believing that the soap will taste great today, because based on his background belief that soap will taste good from today onwards, it will have the best consequences to believe that the soap will taste great today. To many ears, that sounds like the wrong result.

Impurists might respond by holding that rationality is determined, not by an agent’s actual beliefs, but rather by the beliefs it would be rational for her to hold.

Rational Belief-Relative View (RBRV): It is rational for an agent to believe P just in case it would be rational for that agent to believe that no alternative is a more effective means to their ends. (Christensen, 2021, p. 510).

On this view, Good Ol’ Charlie’s beliefs about soap are irrational. Although these beliefs have the best expected consequences relative to Charlie’s actual beliefs, they do not have the best expected consequences relative to what it would be rational for Charlie to believe, which is precisely the opposite.

However, Christensen argues, RBRV threatens circularity since RBRV helps itself to a prior notion of rational belief, but the impurist also uses RBRV to determine the rationality of beliefs. The impurist can retreat from circularity by allowing RBRV to go vacuous, or at least mostly vacuous, ruling out for example some patterns of attitudes which would come out as irrational on any application of RBRV. But this gives only a relatively thin notion of rational belief, analogous to accounts of structural rationality, whereas many impurists have wanted to develop a more robust positive account broadly analogous to theories of substantive rationality.

Christensen suggests we instead adopt an evidence-relative view on which rationality is determined by the beliefs that the total evidence supports:

Evidentially Supported Belief-Relative View (ESBRV): It is rational for an agent to believe P just in case their evidence supports the belief that no alternative attitude towards P is a more effective means to their ends. (Christensen, 2021, p. 513).

However, Christensen suggests that this view would involve recognizing a type of epistemic rationality:

It would go against fundamental motivations of avoiding reliance in our theorizing on a distinctive evidentially-based, non-pragmatic dimension of normativity. The rationality of a given belief, while not directly dependent on this evidential dimension of evaluation, would depend on it indirectly; the rationality of any belief would crucially turn on the purely evidential support for other beliefs. (Christensen, 2021, p. 513).

Here I break from Christensen. I am quite happy to adopt ESBRV. But l want to suggest that ESBRV should not be viewed as committing us to the existence of a distinctively epistemic type of rationality.

3 Evidence and epistemic rationality

The turn to ESBRV is meant to reveal the need for impurists to recognize a distinctively epistemic type of rationality. One possible response would be to deny ESBRV (Rinard, 2021). In this section, I want to urge a different response: ESBRV is at best orthogonal to the debate between impurists and purists, and at worst threatens to help the impurist. In favor of orthogonality, I argue in Section 3.1 that ESBRV does not commit impurists to positing a distinctively epistemic type of rationality, but rather supports a series of claims about evidence which impurists can accept, and which many prominent impurists already accept. Pushing further, I argue in Section 3.2 that ESBRV lends itself naturally to a global consequentialist unification of rational belief and action under a single account. This means that absent further reasons to posit a distinctively epistemic type of rationality, impurists may have some parsimony-based reason to recognize only a single type of rationality.

3.1 Evidence

One worry may be that ESBRV involves the notions of evidence and evidential support. But many impurists have long been comfortable with these notions. Most famously, Chisholm (1956) offered an outright conceptual analysis of what agents ought to believe which made ineliminable reference to the notion of which propositions agents have adequate evidence for. Firth (1956) introduced the notion of epistemic rationality directly in response to Chisholm’s analysis. What was at issue between Chisholm and Firth was not the existence of evidence, the notion of evidential support, or the relevance of evidence to what agents ought to believe. From the very start, these were points of common ground between purists and impurists.

Many modern impurists follow Chisholm in giving pride of place to evidence in their accounts. For example, one of the best-known instrumentalist accounts of how to weigh epistemic and non-epistemic reasons for belief proposes that the rationality of belief and action is determined by two factors: our reasons to pursue aims, and the likelihood that our beliefs or actions will fulfill those aims (Steglich-Petersen & Skipper, 2019). Steglich-Petersen and Skipper hold that likelihoods are determined exclusively by evidence in the standard way, but argue that purists err in refusing to countenance a second factor, the importance of aims themselves. This account gives pride of place to a normative notion of evidential support. In fact, it is very naturally read as an instrumentalist version of ESBRV.

Given this background, the problem cannot simply be that we must posit notions such as evidence and evidential support, or that we must rely on these notions in theorizing. That has often been a point of agreement between impurists and purists. Perhaps the problem is rather that ESBRV forces us to take the notion of evidence as normative rather than descriptive? This is a natural way to interpret Christensen’s worry about “reliance in our theorizing on a distinctive evidentially-based, non-pragmatic dimension of normativity” (Christensen, 2021, p. 513). And this reaction is shared by Rinard (2021), who argues that impurists can happily posit and utilize notions such as evidential support, so long as these notions are regarded as merely descriptive.Footnote 6 Could the problem then be that ESBRV pushes impurists to accept that evidential support is a normative notion?

The first thing to say is that it is hard to see how ESBRV could help in this regard. Not every term which appears in a normative theory is itself normative. From the fact that evidence bears on rational belief, it does not follow that evidential support is a normative notion. So it is not clear that the debate between BRV, RBRV, and ESBRV does much to settle whether evidence is normative.

Another challenge worth noting is that the relationship between evidence and epistemic normativity was meant to be a subject for debate. Evidentialism is one among many theories competing to be the correct account of epistemic normativity. If we simply define epistemic normativity to be the relationship of evidential support, non-evidentialist approaches such as classical versions of coherentism and reliabilism will become straightforward conceptual mistakes.Footnote 7 Hence this strategy will not recover a more robust type of epistemic normativity capable of supporting traditional debates.Footnote 8

More to the point, while some impurists may regard evidential support as descriptive, there is no need for impurists to deny that evidential support is a normative term. Many philosophers already accept and theorize about a wide array of normative terms including rationality, justification, duty, ought, praiseworthiness, fittingness, virtue, coherence, and thick evaluative terms. To be an impurist about epistemic rationality is not to deny that evidential support belongs on this list of normative terms. Impurists deny that there is a distinctively epistemic type of rationality, and perhaps also justification or ought, if these are distinguished from rationality (Littlejohn, 2012; Lyons, 2016; Siscoe, 2021). But we must be careful to separate debates about rationality from the further question of which terms beyond rationality count as normative rather than descriptive.

To illustrate the importance of this separation, consider debates about structural rationality. Defenders of structural rationality hold that we need to posit a distinct type of structural rationality which comes apart from familiar all-things-considered notions of rationality. Consider a simple version of structural rationality on which structural rationality is exhausted by coherence. Defenders of structural rationality might with some plausibility observe that coherence is a normative notion. For example, Alex Worsnip holds that states are incoherent when they display a lack of “fitting-togetherness” (Worsnip, 2021, p. 4). This notion of fitting-togetherness does not look like a descriptive notion. It looks irreducibly normative. So far so good.

But suppose I were to argue for structural rationality in the following way: coherence is a normative term, therefore we must posit a novel type of structural rationality which is identical with coherence. This argument has not, to my knowledge, been attempted, and with good reason. To say that a term is normative is not yet to say that it defines a novel type of rationality, or else we should also be forced to posit types of rationality corresponding to humility, blamelessness, and not being a ninny. But if we cannot infer the existence of a distinctive type of structural rationality from the fact that coherence is a normative term, then neither can we infer the existence of a distinctive type of epistemic rationality from the fact that evidential support is a normative term.

Why might someone want to deny that evidential support is normative? One reason to deny this would be reductive naturalism: the view that all normative properties are ultimately reducible to non-normative properties (Railton, 1986). For a reductive naturalist, the fact that evidential support features in the best account of rational belief would be a reason to produce a purely descriptive account of evidence. But there is no especially direct connection between impurism about epistemic rationality and reductive naturalism. Impurism is the metaepistemological view that there is no distinctively epistemic type of rationality, whereas reductive naturalism is the meta-normative view that there are no irreducibly normative properties. Reductive naturalism is compatible with impurism, but it is also compatible with purism. And in fact, if we are not reductive naturalists, then we may well be unsurprised by the fact that the correct account of rationality bottoms out in irreducibly normative notions such as evidential support. That is exactly the result which opponents of reductive naturalism expect.

Are there other reasons for impurists to deny that evidential support is normative? Perhaps there are, but at this point it is incumbent on Christensen to provide them.

So far, we have seen that the problem for impurists cannot be that they must rely on the notions of evidence or evidential support, since impurists have long made use of these notions. Nor can the problem be that evidential support is a normative concept, for that point does not follow from Christensen’s analysis, and poses no trouble to impurists unless they hold further views such as reductive naturalism. In general, the debate between impurists and purists is not a debate about whether evidential support is normative, but rather about whether evidential support corresponds to a distinctively epistemic type of rationality.

3.2 Parsimony

What is to stop us, Christensen asks, from positing a novel type of epistemic rationality? One traditional impurist answer is that parsimony may stop us. Until we have been given good reasons to posit a new type of rationality, we should avoid duplicating species of rationality without necessity.

To this, Christensen replies that there may be limits to parsimony: perhaps the notion of epistemic rationality cannot be dispensed with. Maybe so, but ESBRV does the purist no favors in this regard, for ESBRV paves the way towards a global consequentialist unification of rational belief and action. This account would be more parsimonious than a separation between rational belief and action, and if so, it looks like parsimony continues to favor the impurist.

To see the point, begin with two questions about rational action. First: by what criterion is the rationality of actions to be assessed? BRV evaluates beliefs by a consequentialist standard of promoting value, whereas RBRV and ESBRV evaluate beliefs by the instrumentalist standard of serving as means to an agent’s ends. For consistency, we will have to settle on one of these standards. As a card-carrying consequentialist, I will deliver a version of ESBRV which evaluates beliefs according to the value they promote, although instrumentalists are welcome to reinterpret my discussion in instrumentalist terms by restricting the value function to the agent’s own ends. This means I will also take consequentialism as a standard of rational action, and for simplicity I will take a maximizing standard on which rationality requires doing what is best.

Second, from what perspective is the rationality of actions to be assessed? Objectivists take that perspective to be the totality of facts, so that agents should do what will actually best promote the most value. Subjectivists take that perspective to be the agent’s beliefs, so that agents should do what they believe will best promote value. A recently popular perspective splits the middle. On information-sensitive views, agents should do what is best relative to a body of information (Charlow, 2013; Kolodny & MacFarlane, 2010; Silk, 2014). One natural candidate for the relevant body of information is the agent’s evidence: agents should do what the evidence suggests will promote the most value.Footnote 9

Combining a maximizing consequentialist criterion of correctness with an information-sensitive perspective on deontic modals yields a complete account of rational action:

Maximizing Information-Sensitive Act Consequentialism: For all agents S, acts X and times t:

  • Maximizing Criterion of Correctness: S is rationally permitted (required) to X at t just in case S’s X-ing at t is at least as good as (better than) every alternative to X available to S at t.

  • Information-Sensitive Value: The value of S’s X-ing at t is the expected value of X’s consequences, where the expectation in question is taken relative to evidential probabilities. That is, \(V(X) = \sum _w Pr_E(w)V(X,w)\), where \(Pr_E\) are the evidential probabilities generated by S’s total evidence at t.

But consequentialism is not just an account of rational action. Consequentialists have long had global ambitions, expanding their account to handle normative terms beyond rationality including rightness, virtue and blame, as well as evaluands beyond action, such as systems of rules, dispositions and character traits (Kagan, 2000; Parfit, 1984; Pettit & Smith, 2000). Could we deliver a similar account of rational belief?

Nothing simpler. Christensen’s ESBRV results from letting X range over doxastic attitudes rather than actions in Maximizing Information-Sensitive Act Consequentialism, giving:

Maximizing Information-Sensitive Belief Consequentialism: For all agents S, beliefs X and times t:

  • Maximizing Criterion of Correctness: S is rationally permitted (required) to hold X at t just in case S’s holding X at t is at least as good as (better than) every alternative to X available to S at t.

  • Information-Sensitive Value: The value of S’s holding X at t is the expected value of X’s consequences, where the expectation in question is taken relative to evidential probabilities. That is, \(V(X) = \sum _w Pr_E(w)V(X,w)\), where \(Pr_E\) are the evidential probabilities generated by S’s total evidence at t.

This is just the consequentialist analog of ESBRV, and if desired we can recover the original instrumentalist statement of ESBRV by letting the value function V encode the agent’s own ends.

What is the upshot of this discussion? My purpose is not to defend global consequentialism.Footnote 10 Indeed, I suspect that even many consequentialists would not go so far as to defend Maximizing Information-Sensitive Belief Consequentialism, for they would not want to defend ESBRV. Many consequentialists would be held back by the arguments traditionally cited against views such as ESBRV: that consequences could not be motivating reasons for belief (Kelly, 2002, 2003); would play the wrong role in doxastic deliberation (Shah, 2003, 2006); or would be wrong kinds of reasons for belief, and hence really reasons for action rather than belief (Way, 2012).

Rather, my point concerns parsimony. Once Christensen grants, for the sake of argument, that ESBRV is true, then it becomes very natural to fold ESBRV into a broader global consequentialist story about action. This story purports to use a single criterion to answer questions about rational belief and action. And most natural objections to this story will be off the table once we grant ESBRV. We cannot deny that consequences are reasons for belief, since ESBRV treats them as such. Nor can we raise traditional complaints about ethical consequentialism, since most such complaints would generalize against ESBRV.

Whatever its remaining faults, the global consequentialist story is unequivocally more parsimonious than one which posits two distinct types of rationality. Now of course, it could turn out that there are facts which the global consequentialist story cannot explain, in which case we would be forced to posit a new type of rationality. But the truth of ESBRV is a fact which the global consequentialist can happily accept, and will with some justification take as support for her global ambitions.

3.3 Taking stock

So far, we have considered Christensen’s argument that impurists should adopt ESBRV.Footnote 11 On this view, the rationality of beliefs is determined by the evidential expectation of the instrumental value of their consequences. We saw that the turn to ESBRV is mostly orthogonal to the debate between impurists and purists. Accepting ESBRV does not require impurists to posit a novel type of epistemic rationality, but only to posit descriptive or normative notions of evidence and evidential support. This is something that impurists can consistently do, and which many prominent impurists have already done. We also saw that ESBRV may help the impurist, insofar as ESBRV lends itself to a global consequentialist unification of rational belief and action, making the impurist’s account of rationality more parsimonious than the purist’s. Here we begin to gain evidence that extensional concessions such as ESBRV may be more favorable to the impurist than the newer evidentialists suppose.

4 The game of belief

Barry Maguire and Jack Woods try a different tack (Maguire & Woods, 2020; Woods, 2016, 2018). Maguire and Woods begin with a sharp distinction between two questions: what we just plain ought to believe, and what it is correct to believe according to the standards of correctness for belief.Footnote 12

For Maguire and Woods, what we just plain ought to believe is determined by authoritative normative reasons. These reasons determine what we ought to believe through weighing explanations of a standard type. Maguire and Woods hold that all and only practical reasons are authoritative reasons for belief. We have standing practical reason against holding any belief, due to the need to avoid cognitive clutter. We often have stronger practical reasons for holding evidentially supported beliefs, such as the need to guide action. These practical reasons can be outweighed by other practical reasons, such as the desire to think the best of ourselves and our kin. But importantly, for Maguire and Woods it is only practical reasons which determine what we just plain ought to believe. Epistemic reasons do not weigh against practical reasons and hence have no impact on what we just plain ought to believe.

However, we can also ask when a belief counts as correct according to the standards of correctness for belief. This need not be a practical question. Although practical considerations play a role in explaining why standards are in force and why the standards have the shape that they do, the standards of correct belief may themselves make mention only of evidence and other epistemic considerations. And indeed, Maguire and Woods hold that the standards of correctness for belief constitute an autonomous epistemic domain to which only epistemic considerations, and perhaps only evidence is relevant.

Because epistemic reasons do not participate in weighing explanations of what we just plain ought to believe, they cannot be authoritative normative reasons. Nevertheless, Maguire and Woods suggest, we can posit a new type of non-authoritative normative reasons, of which evidence for belief is a paradigmatic example. Evidential considerations are not authoritatively normative, because they do not bear on what we just plain ought to believe. But they are still normative reasons, insofar as they participate in an explanation of which beliefs count as correct under the standards of correct belief.

How do standards of correctness for belief come to be in force for us, if these standards are distinct from the question of what agents just plain ought to believe? Woods offers two versions of what he calls a quasi-conventionalist account, although we will see later that Maguire demurs.

Woods (2018) holds that systems of norms are in force when the community adopts what Hart (1961) called the internal point of view towards them, treating norms as rules to guide practical deliberation and hence as grounds for obligation and criticism. Woods holds that communities have respect-based reasons to obey standards which are in force for them. Individuals within those communities will have derivative respect-based reasons to obey communal standards, insofar as they are appropriately related to the fact that the community accepts these standards. And individuals will have instrumental reasons to obey the specific requirements of norms that are in force for them, because that is a necessary means to complying with their respect-based reasons.

Woods (2016) offers a different version of the same story. Woods holds that normally constituted individuals desire to avoid the liability to sanction, and not only its imposition. There is, Woods thinks, a Footian oddity to agents who desire only to avoid punishment for their transgressions, and are unmoved by the fact that they are liable to sanction (Foot, 1972). This means that normally constituted agents have desire-based reasons to obey communal standards, in order to avoid the liability to sanction under communal standards. As before, those desire-based reasons generate instrumental reasons to take the specific actions required by communal standards.

Summing up, Maguire and Woods concede that what agents just plain ought to believe is a practical matter, but maintain that there is also a distinctively epistemic standard of correctness which applies to belief. Evidence and other epistemic considerations are not authoritative normative reasons, but rather non-authoritative normative reasons, bearing on the correctness conditions for belief but not on what agents just plain ought to believe. Woods offers two related quasi-conventionalist accounts of how epistemic standards of correctness could come to be in force for us.

In the next section, I argue that this view commits Maguire and Woods to denying a range of platitudes about reasons and rationality. Given enough encouragement, we could be pushed to deny the platitudes. But absent sufficient encouragement, it would be better to redescribe the view in a way that preserves the platitudes.

5 Three platitudes about reasons and rationality

Impurists do not deny that purists are talking about something. Purists are hardly talking nonsense. Nor do impurists deny that purists are characterizing a normative notion. Purists are not descriptive psychologists. What impurists deny is that purists have characterized a novel type of rational belief.

Impurists think that purists may have confused rationality for one of several neighboring normative notions. Rather than positing a new type of epistemic rationality answering to a novel class of non-authoritative normative reasons, impurists suggest we should redescribe the purist’s view using familiar normative notions. Most obviously, we saw in Sect. 3 that purists could be characterizing the normative notion of evidential support, which answers to evidence. Purists could also be characterizing the notion of fitting belief, which answers to fit-making facts.Footnote 13 Indeed, several authors already defend a fittingness analysis of standards of correctness for belief (Berker forthcoming; McHugh 2014). Or perhaps purists are characterizing the notion of blameless belief, determined by blame-making facts (Boult, 2021; Brown, 2020).Footnote 14

These salvage proposals may seem like inconsequential semantic quarrels, but Maguire and Woods’ proposal reminds us why they are important. By positing a new class of non-authoritative normative reasons, Maguire and Woods commit themselves to denying a series of normative claims about reasons and rationality. These claims are platitudes in the sense that they are widely held and uttered, taken for granted in many epistemological debates, and felt by many philosophers to be firmly intuitive.

Given enough encouragement, we could be induced to give up on the platitudes. Philosophy can, at times, overturn common sense (Rinard, 2013). But we could also save the platitudes by relocating the purist’s theories away from normative reasons and rationality, and towards fittingness, evidential support, blameworthiness, or another normative category where the platitudes are no longer in force. Here are three examples of platitudes that Maguire and Woods may have to deny.

First, many authors hold that:

(Authority of Rationality Platitude) Rationality is authoritative. (Kauppinen, 2021; Kiesewetter, 2017).

Indeed, the need to explain the authority of rationality has been wielded as an argument for and against a variety of epistemological theories (Côté-Bouchard, 2016; Nolfi, 2021; Quinn, 1992). Maguire and Woods deny that epistemic rationality is authoritative, insofar as authoritative normative reasons determine the separate question of what agents just plain ought to believe, while epistemic rationality is determined by an autonomous domain of non-authoritatively normative reasons.

Now some epistemologists do push against the Authority of Rationality Platitude. Many theories of structural rationality follow John Broome (2013) in holding that rationality is one among many sources of requirements which together determine what agents ought to do and believe. This view would at least partly limit the authority of rationality. But these theorists have fought tooth and nail to preserve the authority of reasons:Footnote 15

(Authority of Reasons Platitude) Reasons of rationality are authoritative and participate in weighing explanations of what agents ought to do and believe. (Broome, 2013; Lord & Maguire, 2016).

Despite significant pressure against the claim that reasons of structural rationality are authoritative reasons which participate in weighing explanations of facts about the just plain ought (Kolodny, 2005), many defenders of structural rationality have felt that it would be going too far to deny that reasons of rationality are authoritative and weigh on oughts. Here Maguire and Woods are distinguished by their willingness to take a step which many participants in the structural rationality debate have seen as off limits.

A final platitude concerns the value of rationality:

(Value Platitude) Rationality has significant value. (Horowitz, 2014; Wedgwood, 2017).

The claim that rationality is valuable is often taken for granted, and used to support or attack theories throughout epistemology (Horowitz & Dogramaci, 2016; Steglich-Petersen, 2011; Schoenfield, 2019). But it is not easy for Maguire and Woods to recover the Value Platitude. Maguire and Woods express sympathy for a value-based approach on which we just plain ought to do what is best. But because epistemic reasons have no bearing on what we just plain ought to believe, epistemic reasons cannot in isolation indicate that a belief is valuable or disvaluable, or else epistemic reasons would weigh on just plain oughts by indicating what it is best to believe. So it looks like epistemic rationality will have no value of its own, and that when epistemic rationality comes apart from the just plain ought, it would be better to be irrational than to be rational.Footnote 16

So far, we have seen that impurists accuse purists of mistaking rationality for related normative categories such as fittingness and evidential support. Maguire and Woods lend fuel to this charge by denying three platitudes about reasons and rationality: that rationality is authoritative; that normative reasons are authoritative and participate in weighing explanations of just plain oughts; and that rationality has significant value. Because these claims are not platitudes when rationality and reasons are swapped for notions such as fittingness and fit-making facts, it is tempting to save the platitudes by claiming that Maguire and Woods’ real target is one of these other normative notions.

If Maguire and Woods want to retain their view and deny the platitudes, they need to offer a weighty argument in its defense. Maguire and Woods advance several arguments in favor of their view. Are these arguments substantial enough to save the view?

6 Motivating the view

Woods’ approach is grounded in a more general conventionalist story about most or all types of normativity, including promising (Woods, 2016), etiquette and morality (Woods, 2018). If true, that is an account with enough generality and explanatory payoff to justify bending a few platitudes. But is it true? In Sect. 6.1, I argue against Woods’ conventionalism, and in Sect. 6.2, I argue that a more plausible conventionalism may not favor the purist. Sect. 6.3 considers an argument which does not appeal to conventionalism.

6.1 Against bare conventionalism

Woods’ conventionalism is a species of what we might call bare conventionalism. It takes the very fact that society has adopted a convention to be sufficient for the convention to be in force. Bare conventionalism is often thought to suffer from serious challenges (Nieswandt, 2019; Scanlon, 1990). Here is one such challenge.

Suppose that society adopts a truly heinous convention. This convention requires kicking puppies, maintaining apartheid, or working to destroy humanity. Bare conventionalism says that this convention will be in force, grounding a type of obligation and criticism for members of society. We would have, for Woods, desire- or respect-based reasons to maintain apartheid, would count as violating normative obligations if we do not, and would be criticizable as a result. Call this the objection from heinous conventions.

To his credit, Woods is admirably clear in his desire to bite the bullet.

A referee complains that [my view] suggests Rosa Parks was obliged, in some subscripted sense, to sit in the back of the bus (and that she was criticizable for not doing so). This, again, is a feature, not a bug. Rosa Parks is liable to criticism for violating a norm presumably in force then. (Woods, 2018, p. 217).

We saw in Sect. 5 that Maguire and Woods bite a number of normative bullets by denying three platitudes about rationality and reasons. But I think that Woods may have bitten one bullet too many here. Even Maguire seems prepared to jump ship at this point.Footnote 17

What is wrong with claiming that Rosa Parks was obliged to sit in the back of the bus and liable to criticism for failing to do so? Besides its intuitive implausibility, this claim serves as a breakdown point for both of Woods’ views about how normative conventions come to be in force for us once they are adopted.

Woods (2018) holds that when communities adopt norms, individuals within the community inherit respect-based reasons to comply with those norms. But I don’t think we would like to claim that Rosa Parks had a respect-based reason to comply with Jim Crow norms. Parks’ act was an act of civil disobedience. The entire purpose of her disobedience was to convey symbolic disrespect for Jim Crow norms by refusing to comply with them. Parks did this because what she had reason to do was to disrespect Jim Crow norms. If Woods is to capture this fact, he will have to say that Parks had decisive reason to disrespect Jim Crow norms, but nevertheless had a respect-based reason to respect Jim Crow norms. But this is a mouthful. And in virtue of what had Jim Crow norms earned Parks’ respect? Merely by being mistakenly adopted by society as norms? Even if we grant that societal adoption gives society, qua group agent, reason to respect the norms it adopts, it is not clear that Parks would inherit that reason. Since Parks was not consulted in the adoption of Jim Crow norms, and since the whole purpose of these norms was to oppress Parks, it is hard to see how the mere adoption of Jim Crow norms could give Parks reason to respect them.

Woods (2016) holds that individuals desire to avoid the mere liability to sanction, and not just its imposition. This gives us a desire-based reason to comply with the norms in force in our community, in order to avoid sanction-liability. Woods’ claim is plausible when norms are just and fairly adopted. There would indeed be a Footian oddity to agents who cared only about the actual imposition of sanctions for promise-breaking, and not about the fact that breaking a promise made them liable to sanction. But there is no oddity in failing to care about sanction-liability under unjust norms. Parks was doubtless concerned about the actual imposition of sanctions, in the form of arrest, imprisonment or worse. But did she desire to avoid the mere liability to sanction under Jim Crow norms? Why would she? You do not have to be a Footian moral monster to be unconcerned with liability to sanction under unjust norms. All you must do is to reject the norms.

So far, we have seen that bare conventionalism encounters the objection from heinous conventions. Bare conventionalism says that heinous conventions are in force for us when they are adopted by society. I suggested that this is the wrong consequence, and also that it puts pressure on Woods’ explanations for how conventions come to be in force by generating respect- and desire-based reasons. Could Maguire and Woods escape the objection from heinous conventions by retreating to a more restrictive version of conventionalism?

6.2 Justified conventionalism

If the problem with heinous conventions is that they are unjustified, then we can escape the objection from heinous conventions by requiring conventions to be appropriately justified before they gain normative force. Call this justified conventionalism.

Justified conventionalism is a natural response to the problem of heinous conventions. In fact, we saw that Maguire breaks from Woods in suggesting the need for conventions to be justified (Maguire & Woods, 2020, p. 229), referencing a popular version of justified conventionalism due to Katharina Nisewandt:

If the rules of practice P say that X must \(\phi\) given fact F, and P is a justified practice, then, given F, X must \(\phi\) because of F. (Nieswandt, 2019, p. 17).

Could justified conventionalism save Maguire and Woods’ view? Perhaps. It is not my burden to survey all possible versions of conventionalism, nor could I do so in the space allotted. To a large extent, the reply here is that if Maguire wants to develop a more selective conventionalist view, it is incumbent on Maguire to tell us what that view is and how it accounts for epistemic normativity.

However, I do think that justified conventionalism risks taking us out of the frying pan and into the fire. Justified conventionalism purchases plausibility for the conventionalist view at the expense of making it mysterious why the justified conventions would be purely epistemic.

For example, Nieswandt lists four ways in which a convention P might be justified:

Promoting impartial good: The greatest good of the greatest number can be better furthered with P than without it.

Pareto optimality: P maximizes everyone’s impartial gains.

Ends, not means: P is required in order for us to treat others as ends in themselves.

Good human life: P is necessary for a good human life.

But it is not clear that any of these standards justifies adopting purely epistemic conventions. There is no clear argument that epistemic conventions are needed for living a good human life or treating others as ends in themselves. And although neither camp has a plausible claim to Pareto optimality, you might well think that admitting some non-epistemic considerations into our normative conventions would better promote the impartial good than would adopting purely epistemic standards. After all, Maguire and Woods take non-epistemic considerations to bear on what it is best to believe. And if the best conventions are not epistemic conventions, then justified conventionalism threatens to support impurism over purism.

There is room for purists to push back here. Although Maguire and Woods concede that it would sometimes be better to form beliefs which violate, rather than respect purely epistemic standards, it does not follow that purely epistemic standards could not be the best standards for promoting the impartial good. Perhaps strict epistemic standards are needed to instill a habit of forming beliefs based on evidence and careful reasoning. But if Maguire and Woods want to make some such claim, they need an argument. Impurists are happy to concede that the best standards to adopt should instill a healthy response for evidence, but wouldn’t it be better if those standards also recognized the importance of other goals, such as friendship and anti-racism? That is not a crazy view to hold.

Now Nieswandt was talking about the justification of moral conventions. Could the newer evidentialist hold that conventions for belief are justified in a narrower way? For example, if we could show that epistemic conventions are the best means to the promotion of accurate belief, would that be enough for these conventions to have genuine normative force, even if epistemic conventions are suboptimal for the purpose of promoting overall value? The problem with this move is that it assumes precisely what is under dispute: that standards acquire genuine normative force through their relationship to a restricted type of epistemic value, without regard for their practical, moral or all-things-considered implications.Footnote 18 It is exactly this claim which impurists deny.

So far, I have argued that Maguire and Woods’ view cannot be saved by appeal to conventionalism. Woods’ bare conventionalism faces the objection from heinous conventions. Justified conventionalism avoids the objection from heinous conventions, but struggles to explain why justified conventions would be purely epistemic. Is there anything which can be said for purism which does not rely on conventionalism?

6.3 Criticism

Maguire and Woods make one argument which may not depend on conventionalism.Footnote 19 This is the argument that there is an important sense in which we are aptly criticizable for violating epistemic norms. If it is apt to criticize us for violating epistemic norms, then it is natural to think that epistemic norms are genuine requirements of rationality.

Now some impurists will not grant that we are criticizable in any sense for violating epistemic norms (Rinard, 2021). But suppose we grant the datum. Does this require us to posit a new type of epistemic rationality? It does not.

The problem is that many of the impurist’s salvage proposals can also capture the datum. We can be criticized for beliefs which are unfitting or blameworthy. And it may be criticism enough to say of a belief that it is evidentially unsupported. So the mere fact that beliefs which violate epistemic standards are criticizable does not imply that they are criticizable for being irrational, rather than unfitting, blameworthy or evidentially unsupported.

Moreover, Maguire and Woods’ proposal threatens to overgenerate normative criticism. Maguire and Woods hold explicitly that there are chess-specific reasons for chess players to play well, and that there were Nazi party reasons for party members to obey Nazi doctrine. On their view, there is a type of normative criticism aptly directed at chess players who play badly in order to let a child win, or Nazi party members who hide Jews. But the first claim is at best not obvious, and the second threatens to force Maguire back into the bare conventionalist view about Rosa Parks. So I do not think that the impurist should be quick to concede any explanatory advantage to the purist in her account of apt normative criticism. We might do better to draw accounts of criticism-liability out of independently motivated accounts of rationality, blameworthiness and related terms.

I do not expect Maguire and Woods to accept the impurist’s salvage proposals. Doubtless each proposal faces some difficulties of its own. But we saw in Sect. 5 that Maguire and Woods’ proposal involves denying a range of platitudes about rationality and reasons. These are serious costs, and it takes a strong argument to justify imposing them. If we cannot justify these costs by appeal to metanormative conventionalism, and need not incur these costs in order to explain practices of normative criticism, then it would seem better to side with the impurist and avoid positing new species of rationality without necessity.

7 Conclusion

My project in this paper was to address the newer evidentialist program. The newer evidentialists concede to impurists that there is a coherent all-things-considered notion of rational belief whose extension is determined in large part by non-epistemic considerations. But the newer evidentialists maintain that there is still an important role in epistemology for a separate notion of epistemic rationality. I argued that existing newer evidentialists face opposite worries.

Christensen argues for the plausible claim that rationality is evidence-sensitive. This claim, however, does not assert the existence of a distinctively epistemic type of rationality. Christensen’s view is naturally read as a vindication of the global consequentialist claim that all rationality is a matter of promoting overall value. While Christensen’s view does show that rationality makes essential appeal to the concept of evidential support, the rational importance of evidential support is not what is in dispute between impurists and purists.

Maguire and Woods argue that the practical question of what we just plain ought to believe should be distinguished from the epistemic standards of correctness governing belief. I argued that the proposed standards are both distinctive and epistemic, but not best understood as a type of rationality or as determined by a new class of non-authoritative normative reasons. By redescribing epistemic standards in other normative terms, such as fittingness or evidential support, we can preserve a range of platitudes about reasons and rationality which would otherwise be violated by Maguire and Woods’ view.

This discussion illustrates the difficulty of the newer evidentialists’ attempt to cede the notion of rationality simpliciter to the impurist. Once we concede to the impurist that she is correct about rationality or the just plain ought, it becomes much harder to motivate the need for a separate, purely epistemic type of rationality.

One reaction to this discussion would be to retreat to the new evidentialist position on which there could not possibly be non-epistemic reasons for belief, and hence rational belief is a purely epistemic matter. Although this view has faced recent pushback, it is far from refuted and it would avoid the difficulties faced by the newer evidentialists.

Another reaction to this discussion could be to concede that the impurist was right. There were clear reasons why purists thought that a separate notion of epistemic rationality was needed: purists held that there is no coherent all-things-considered notion of rational belief; that there are no practical reasons for belief; and that impurists give extensionally incorrect verdicts about what it is rational for us to believe. But if we are ready to concede these points to the impurist, then we should be open to the idea that epistemic rationality is theoretically dispensable after all.

These are both excellent reactions. But it remains to be seen whether there is any room to split the difference between them. Once purists begin to cede extensional ground to impurists, it becomes difficult to motivate the need for a distinctively epistemic type of rationality. As a result, purists who are willing to make extensional compromises with impurists may have purchased a more wholesale form of impurism about epistemic rationality than they bargained for.