The prima facie irrationality of believing for the sake of indirect epistemic benefits

There are several authors who grant the possibility of there being indirect epistemic benefits as a result of a person holding a particular belief in particular circumstances, while at the same time insisting that holding the belief in question on the basis of that reason would not be epistemically rational. These authors thus take the view that indirect epistemic reasons ought to be characterized as mere pragmatic reasons for belief. Consider, for example, a type of case discussed by Marc Moffett (Reference Moffett2007) and Catherine Elgin (Reference Elgin, Feldman and Warfield2010). Suppose an agent working in a research community, A, believes some idiosyncratic thesis, P, and suppose that the community's epistemic aims (e.g. acquiring knowledge in the relevant domain) are likely to be more effectively advanced as a result of this agent holding this unorthodox belief.Footnote ⁵ In discussing this kind of case, David Christensen says that A believing P in view of these epistemic benefits isn't epistemically rational; if such belief is rational at all, he says, it's merely practically rational. Although ‘there may well be important epistemic benefits’ that result from the agent's eccentric patterns of belief, he says, ‘the patterns of belief are no more epistemically rational for all that’ (Christensen (Reference Christensen2007), 216). Richard Feldman makes a similar point about the epistemic benefits that result from people holding firm in their beliefs in the face of widespread disagreement. ‘It may be that the search for the truth is most successful if people argue for the things that seem true to them’, he says, ‘but one can do that without being epistemically justified in believing that one's view is correct’ (Feldman (Reference Feldman, Goldman and Whitcomb2011), 157). While both authors are happy to say that the attainment of valuable epistemic outcomes as a result of A believing P can provide A with some kind of a reason to believe P, their point is that this won't be an epistemic reason.

Berker agrees, and he uses an example from Roderick Firth (Reference Firth1981) and Richard Fumerton (Reference Fumerton and Steup2001) to elicit intuitions about epistemic rationality in support of his view (we've tweaked his example below).

Now, is it epistemically rational for S to adopt a belief in God in this case, in view of the epistemic benefits that stand to be gained? Berker says: obviously not (Berker (Reference Berker2013), 364). Indeed, the obviousness of the irrationality of believing for this reason is used by Berker to illuminate what he sees as a larger set of problems that arise in formulating norms of epistemic rationality in consequentialist terms. Epistemic consequentialists will treat the rationality of a belief about P as hinging on whether the possession of this attitude conduces to true belief in general, not just whether it is an optimal appraisal of P's likely truth value, in view of the evidence. This then licenses the kind of trade-offs in which S's evidentially unsupported belief can be endorsed as rational because it will causally lead to people coming to have true beliefs later on. But on Berker's view, such trade-offs are ‘beyond the pale’ of epistemic rationality (ibid., 363). When assessing the epistemic merits of a belief, he says, it is strictly irrelevant whether that belief conduces to the attainment of true belief and the avoidance of false belief in propositions other than the one whose status is the focal point of the appraisal (ibid., 365). To treat such reasons as epistemic reasons, according to Berker, is to fail to respect the ‘separateness of propositions’. It is to think of our judgements about whether it's rational for A to believe P as being grounded in very general desiderata – e.g. people believing more truths overall – whereas, so he maintains, such judgements categorically must be grounded in how A has assessed her first-order P-relevant evidence.Footnote ⁶

In an article that explores the prospects for modelling epistemic rationality on decision theory, Hilary Greaves arrives at a similar conclusion, suggesting that some element of our intuitive notion of epistemic rationality is ultimately incompatible with a consequentialist formulation of epistemic norms of belief. Any formulation of epistemic norms which encodes the notion that it's rational to hold a belief that maximizes epistemic utility overall will – absent the introduction of an ad hoc side constraint – endorse the rationality of accepting ‘epistemic bribes’, in which a person is induced to reject belief in a proposition for which she has conclusive evidence in order to promote global epistemic benefits. According to Greaves, then, we have

In different ways, both of these authors are suggesting that our intuitive notion of epistemic rationality entails that what one ought to believe about a given proposition is entirely determined by one's reasons for thinking that that particular proposition is true or false. If we start to allow that people's doxastic attitudes may be adopted for strategic reasons, with a view to increasing the preponderance of true beliefs per se, we will violate the constraints that are imposed by that particular conception of epistemic rationality. Granted, we might have all-things-considered reasons to violate those constraints and adopt doxastic attitudes for strategic reasons. But our notion of epistemic rationality will not – by its nature cannot – endorse such infringements.

Organizing our intuitions about the atheist scientist

The intuitions that Berker is seeking to elicit in his discussion of Atheist Scientist are difficult to resist. It does seem paradigmatically epistemically irrational for S to believe in God in order to win research funding that is expected to result in epistemic benefits further down the line, and that's true even if those benefits are of immense significance by S's own lights. But there is an oversupply of reasons that favour Berker's intuitive verdict. So before we grant Berker's claims about the duty to respect the separateness of propositions, and about the untenable character of consequentialist approaches to epistemic rationality, we first have to identify the other sources of irrationality that would be manifested in S's belief in God in Atheist Scientist, and try to bracket-off their influence on our wider judgements about the sense in which it is or isn't apt for people to hold their beliefs in a way that's sensitive to the consideration of indirect epistemic benefits.

One problem with S believing in God is that this seemingly must involve her holding a belief about P that conflicts with her own appraisal of the P-relevant evidence. S is an atheist in the first place, so we may surmise, because she sees this as the doxastic stance vis-à-vis God's existence that the evidence recommends. If that is S's judgement, and if S then adopts a belief in God based on a strategy which is aimed at maximizing valuable epistemic outcomes globally, then she will, at a minimum, be left with conflicting beliefs about (i) the question of God's existence and (ii) the probative force of the evidence vis-à-vis God's existence.

Another related problem with S believing in God in Atheist Scientist is that it seems that this will involve some kind of disingenuous willing of belief. If S's appraisal of the relevant evidence leads her to think that ‘God exists’ is likely false, then – given a plausible view about belief's non-voluntary nature – this simply entails, ipso facto, that S doesn't believe in God. Maybe S can pretend to believe in God, and cajole herself into thinking this pretence represents her beliefs. But this will be intellectually fraudulent, and in any case, it will leave S's real beliefs untouched. S's belief in God would have to be irrational, then, in Berker's example, since the only way for her to induce this belief in herself would be via some sort of self-delusion.Footnote ⁷

It doesn't follow from either of the above points, though, that indirect epistemic reasons for belief are not epistemic reasons, in this case or in others. Consider a modified version of Berker's example.

Suppose, then, that in Permissive Scientist, S takes account of the valuable epistemic consequences that are much more likely to ensue for herself or others if she believes in God (and wins her research grant etc.), and that consideration of this matter compels her – in a Pascalian spirit – to undertake practices that are likely to result in her undergoing a change in her views, from what is, by her lights, one rationally permissible doxastic attitude given the evidence (namely, non-theism) to what is, by her lights, an alternative rationally permissible doxastic attitude given the evidence (namely, theism). Granted, this doesn't settle the question of whether the considerations whose normative force S is responding to in this should ultimately be conceived of as epistemic or pragmatic reasons for her to favour theism. Our point here is just that if there is something epistemically unacceptable about S's favouring of theism in Permissive Scientist, the unacceptability can't be due to S's belief conflicting with her appraisal of the evidence's probative force, nor due to S having disingenuously willed herself to believe, since neither factor figures in the scenario. It may be a rationality-defeating feature of a belief – perhaps a decisively rationality-defeating feature – that it conflicts with the agent's assessment of the probative force of the evidence. However, it is consistent with this to allow that, wherever this disqualifying condition doesn't hold, beliefs that are informed by a strategic consideration of what's likely to maximize indirect epistemic benefits are, or at least can be, counted as epistemically rational.Footnote ⁹

At this point one might object that although S doesn't think theism is false, in light of the evidence, in Permissive Scientist, she also doesn't think it's true. Perhaps, then, Berker's complaint about believing against the evidence could be reformulated, so that the claim is that it's irrational for S to try to get herself to believe something she doesn't think is true, in light of the evidence.Footnote ¹⁰ It seems to us, however, that for this objection to really work, one would have to deny Permissivism outright. Permissivists defend the plausible view that in at least some situations there is more than one rationally permissible doxastic attitude that can be held towards a given proposition, given a certain body of evidence. By definition, the Permissivist regards it as rationally permissible for an agent with the relevant evidence, in such a case, to hold any of the permissible attitudes. In Permissive Scientist, then, it would misrepresent S's perspective to say that she's nurturing belief in a proposition whose truth she doesn't accept. The more apt description would be to say that she's nurturing a belief which, given the relevant evidence, (i) might well be true, by her own lights, and (ii) whose rational permissibility she has affirmed all along. The only way to insist on the irrationality of these kinds of judgements is to reject the Permissivist's characterization of S's initial appraisal of the evidence. Granted, someone may doubt that the evidence is ever ‘permissive’ when it comes to belief in God. But our point – that it's possible to be sensitive to indirect epistemic reasons, without disingenuously believing against one's own appraisal of the evidence – should still hold, as long as there are some permissive cases.Footnote ¹¹

To characterize indirect epistemic reasons for belief as epistemic reasons is not to say that people should delusionally gull themselves into having beliefs that conflict with their views about the evidence's probative force. It's just to say that where an agent identifies some spectrum of rationally permissible doxastic attitudes in response to some body of evidence, and where she has some ability to influence what doxastic attitude she holds among that range of options, then, to the extent that she has reason to believe that her holding doxastic attitude X will have better global epistemic consequences than her holding doxastic attitude Y, she has a reason – one that may be aptly characterized as an epistemic reason – to favour X over Y.Footnote ¹²

Two kinds of epistemic reasons

We aren't trying to deny that there is an important difference between a person believing P because her appraisal of the evidence indicates that P is true, and a person believing P because indirect epistemic benefits are in the offing if she does so. It's entirely reasonable, moreover, to want the difference between these two kinds of reasons for belief to be reflected in our classificatory terminology. What we oppose is the framing of this as a distinction between epistemic and pragmatic reasons. In a case like Permissive Scientist, the valuable outcomes whose realization gives S a reason to believe P are, after all, epistemically valuable outcomes.Footnote ¹³ Where such cases are in the frame of discussion, we think that the discussion will be better served by a distinction between direct and indirect epistemic reasons. This terminology provides a better indication of the nature of the conflicting desiderata that are in play in a case like Permissive Scientist. The choice that S is facing in this case isn't whether to ignore the requirements of epistemic rationality for the sake of pragmatic considerations. Rather, the choice she's facing is whether to deprioritize the demands of direct epistemic rationality, and instead to prioritize indirect epistemic considerations. In such a scenario, if S did successfully nurture a belief in God, then even if we still wanted to criticize the (direct) epistemic rationality of her beliefs, we also can (and should) leave ourselves room to concede that there are reasons favouring S's theism that are epistemic reasons of a certain sort – again, indirect epistemic reasons – and not merely pragmatic reasons.

In recent work defending the ‘equal weight view’ about the epistemic significance of disagreement,Footnote ¹⁴ Jonathan Matheson posits a distinction between ‘non-instrumental’ and ‘instrumental’ epistemic reasons that is similar to our proposed distinction between direct and indirect epistemic reasons. One of the concerns that arise for the equal weight view, in so far as it recommends suspension of judgement about a range of issues, is that it may lead towards general scepticism. In response to this worry, Matheson suggests that there can be instrumental epistemic reasons for remaining steadfast in one's views in the face of peer disagreement, the consideration of which should arrest this lapse into scepticism. For example, with regard to some types of research or inquiry, progress is ‘more likely to happen when individuals believe their controversial views and when there is a diversity of opinions being defended’ (Matheson (Reference Matheson2015), 145). The reasons this consideration generates are, according to Matheson, instrumental epistemic reasons for belief, because

Matheson's distinction between synchronic and diachronic ways of conceiving of epistemic justification is a useful one. Synchronic justification ‘concerns the current time and what doxastic response right now would do you best with respect to these goals right now’; diachronic justification, by contrast, ‘concerns later times and what doxastic response right now would do you best with respect to these goals in the long run’ (ibid., 146). The equal weight view is the appropriate view on the epistemic significance of disagreement, Matheson suggests, if and in so far as we're thinking about epistemic justification synchronically. But it's consistent with this to say that synchronic and diachronic factors can aggregate into all-things-considered epistemic reasons, and that – because of the indirect epistemic benefits that disagreements can contribute to – the all-things-considered epistemic reasons that are in play in cases of peer disagreement sometimes (or often) recommend a steadfast (i.e. scepticism-resistant) response to peer disagreement. Formulated in our preferred terminology, Matheson's claim is that consideration of the indirect epistemic reasons that obtain in a particular case can affect what's epistemically rational for someone to believe in the face of peer disagreement.

In response to all this, someone might concede that there is some sense in which indirect epistemic benefits generate epistemic reasons of a certain kind, while at the same time insisting that responding well to direct epistemic reasons (as we call them) is really what epistemic rationality is all about. Our claim in the previous section was that in a scenario like Atheist Scientist, it's possible to bracket off the main factors that lead us to want to deny the epistemic rationality of nurturing beliefs that promote indirect epistemic benefits. Being sensitive to the possible epistemic benefits of one's beliefs doesn't necessarily lead to self-deception or internal conflict. But even if this is right, the critic may still say that the person who forms a belief in P based on her appraisal of the first-order, P-relevant evidence, while paying no mind to other considerations – in Berker's terms, the person who honours ‘the separateness of propositions’ – shows a kind of dutiful respect for the evidence that is especially meritorious. But we don't think that appealing to a duty to respect the evidence brings things to a justificatory bedrock. One can still ask why it's meritorious to be dutiful in this way. And one plausible explanation is that epistemically valuable outcomes (e.g. people believing more than falsehoods) will be more effectively realized overall if categorical doxastic precepts – such as: ‘form beliefs based on what the first-order evidence shows, and nothing else’ – are accepted and followed.Footnote ¹⁵ In so far as something like a rule-consequentialist interpretation of strict evidentialist norms is viable, there isn't any need to insist on a stark divide, as Berker does, between (i) belief-formations that give some consideration to possible indirect epistemic benefits, and (ii) belief-formations that ignore this consideration. Nor is there any need to denounce the former as being beyond the pale of epistemic rationality. Granted, this is precisely the kind of epistemic consequentialism that Berker critiques. The point is that once we have bracketed off the various complicating factors discussed earlier, the remaining intuitive force animating Berker's separateness of propositions norm is amenable to a rule-consequentialist interpretation. And because of this, we have the option of categorizing indirect epistemic reasons as a bona fide species of epistemic reasons for belief, rather than being forced to relegate them into an entirely different (i.e. pragmatic) normative genus.

We've suggested that intuitions about the deontological nature of epistemic norms are amenable to a rule consequentialist interpretation. Much more would need to be said, though, in order to assert this confidently. Thinking of parallels in ethical theory, some authors argue that rule-consequentialist norms boil down to deontological norms (e.g. Howard-Snyder (Reference Howard-Snyder1993)), while others argue that rule-consequentialism collapses back into act-consequentialism (Smart (Reference Smart1956)). If we're after a consequentialist epistemic framework, but one which makes room for non-consequentialist evidentialist norms – by ‘going indirect’, à la rule consequentialism – then we'll need to think about how such critiques of rule consequentialism play out in the epistemic domain. We might also have to consider whether deontological understandings of epistemic norms rely on something analogous to the idea of ‘agent-relative reasons’ in ethics. Just as some ethicists deny that any of us have impersonal reasons to promote the ethical good (e.g. Nagel (Reference Nagel1986), Darwall (Reference Darwall2006)), so too one might deny that we have impersonal reasons to promote the epistemic good. If this is how we conceive of the roots of non-consequentialist epistemic norms, then elaborating the kind of view we've sketched here will involve saying something about what it means to approach epistemology in an agent-neutral frame of mind – ‘from the point of view of the universe’, to borrow that well-worn and evocative phrase.Footnote ¹⁶ An assessment of the general prospects for epistemic consequentialism is a task that several other authors have taken up recently.Footnote ¹⁷ But any detailed elaboration of the above is beyond the scope of the present article. In any case, we acknowledge that in our defence of why indirect epistemic reasons for belief can properly be regarded as epistemic reasons, and not mere pragmatic reasons, we've assumed that an epistemic consequentialist ultimately will have satisfactory positions to take in response to issues and questions like the ones we have mentioned here.

An argument from interpersonal normative parity

Up to this point we've been responding to different kinds of worries, all of which seem to favour the view that we should conceive of indirect epistemic reasons as mere pragmatic reasons for belief. But we think there is also an interesting positive case to be made for thinking of indirect epistemic reasons as bona fide epistemic reasons for belief. Indirect epistemic reasons, we want to suggest, exhibit a type of ‘interpersonal normative parity’ that is characteristic of epistemic reasons for belief in general, but not of pragmatic reasons.

First, consider how pragmatic reasons for belief get their normative purchase on agents. Where R is a pragmatic reason for A to believe P, this by itself does not make it the case that R is also a pragmatic reason for person B to believe P. R's normative force only carries over like this if it happens that B shares the aims or interests by virtue of which R has its pragmatic normative force for A. So, for example, suppose A has a pragmatic reason to believe P because this makes it easier for her to get along with her family. This pragmatic reason – A getting along with her family – will have little or no purchase on B, insofar as getting along with A's family (presumably) doesn't figure among B's aims or interests. In this sense pragmatic reasons for belief only exhibit a relatively weak and relatively contingent form of interpersonal normative parity.

By contrast, where evidence E constitutes a bona fide epistemic reason for A to believe P, it also, ceteris paribus, constitutes an epistemic reason for B to believe P. Or to express this in a more qualified way, where evidence E is an epistemic reason for A to believe P, the question of whether E is also an epistemic reason for B to believe P doesn't hinge on what A's and B's respective aims and interests are, nor the extent to which they coincide. For example, suppose A cares a great deal about people having true beliefs on the question of God's existence, whereas B doesn't care about this at all. Irrespective of this disparity in A's and B's interests and concerns, the relevant epistemic considerations (e.g. a priori and a posteriori arguments for and against theism) constitute reasons for B to believe that God does or doesn't exist, just as much as they constitute epistemic reasons for A to believe this. Epistemic reasons for belief thus exhibit a robust form of interpersonal normative parity – robust in the sense that the kind of demands that they generate for agents are ostensibly impervious to disparities in the aims and interests of different agents. We aren't prepared to stake a claim here about the weight of epistemic reasons for belief. We aren't committed to any particular view about the nature of the blame that one is liable to receive as a result of, say, failing to interpret one's evidence appropriately. All we're saying is that one cannot evade the demands of epistemic reasons by renouncing an interest in those demands. We're answerable to norms of rational belief regardless of whether we want to be.Footnote ¹⁸

Having observed this general difference in the interpersonal normative parity of epistemic reasons and pragmatic reasons, our claim is that it's a consideration in favour of the appropriateness of classifying indirect epistemic reasons as epistemic reasons, rather than pragmatic reasons, that these kinds of reasons for belief exhibit the more robust form of interpersonal normative parity. In circumstances where more truths will be believed in domain D provided that someone believes P, both A and B are given an indirect epistemic reason to believe P. And this interpersonal normative parity obtains even if it happens to be the case that A is deeply concerned about the prevalence of true belief in domain D, whereas B doesn't care at all about this. As an illustration of this, consider the story of Twelve Angry Men. One dissenting juror believes the defendant to be innocent, even though all the other jurors believe that the defendant is guilty. This one juror's dissent forces the jury to discuss and reassess the evidence, and as a result of this more truths about the details and circumstances of the crime are (eventually) believed by all the jurors. In a parallel scenario, in which the social epistemic dynamics of consensus and dissent were operating in a similar way (and recognized as such), but where none of the jurors initially had any doubts about the guilty verdict, every juror would have an indirect epistemic reason to believe that the defendant was innocent, because any of them believing the defendant to be innocent would ultimately result in more truths about the case being believed. The (indirect) epistemic reason that's in effect in this case applies to all the agents, just as a new item of relevant evidence would be a (direct) epistemic reason for all agents to alter their views. The interpersonal force of both kinds of reasons owes to the fact that all agents have epistemic reasons to aspire to truth, regardless of their personal aims and interests. In short, an epistemic reason that tells in favour of believing P has its (epistemic) normative force for different agents irrespective of their contingent aims and interests. And this is the case for indirect epistemic reasons, we maintain, in much the same way as it's the case for direct epistemic reasons.

Indirect epistemic benefits and religious belief

As indicated in the introduction, we now want to ask how the consideration of indirect epistemic reasons for belief can affect the rational standing of beliefs about matters of religion. We will argue that there are at least some situations in which indirect epistemic reasons support some religious belief, subject to caveats which reflect the concerns we discussed earlier, relating to self-deception and self-contradiction.

We'll begin by discussing several examples that are close to home for philosophers. Consider the major contributions to knowledge that have been made, in post-war Anglophone philosophy, by philosophers whose work is partly motivated by the aim of substantiating or clarifying some aspect of their prior religious convictions. Some of that research has been done by authors who self-consciously situate their work in the sub-discipline of philosophy of religion. For instance, in attempting to defend the view that a certain kind of religious experience justifies theistic belief, William Alston (Reference Alston1991) formulates an influential defence of a view that has general and wide-reaching epistemological ramifications, namely, that there is no non-circular way to defend the justificatory import of sensory perception. In a nearby region of the philosophical map, Alvin Plantinga's work, which broadly aims at defending the warranted status of Christian theism, builds in novel contributions to prior understandings of (i) externalist epistemologies founded on a conception of a proper functioning perceptual apparatus (Plantinga (Reference Plantinga1993)), and (ii) the concept of beliefs being properly basic, i.e. justified on non-inferential grounds (Plantinga (Reference Plantinga2000)). Outside of research that's self-consciously situated in philosophy of religion, Elizabeth Anscombe's Catholic faith was a discernible guiding force in her ground-breaking work on the nature of intention (Anscombe (Reference Anscombe1957)), and in her (related) influential critiques of mid-century ethical theory (Anscombe (Reference Anscombe1958)), both of which germinated major research programmes that have expanded our breadth of understanding in ethics and philosophy of action. Obviously we shouldn't pretend to have any definitive biographical knowledge of what motivated these particular individuals in their work. Still, there are reasons – plausible inferences, and textual evidenceFootnote ¹⁹  – in light of which it's reasonable to think that the ideas and arguments these authors gave us, and the further knowledge their research agendas generated, owe a large debt to these authors’ prior religious convictions, and to the (entirely natural) impulse that most thoughtful people have to want to show that their deepest convictions are underwritten by a coherent and credible set of ideas about the world.

To the extent that having prior religious convictions has led to some philosophers producing novel ideas and arguments, and thereby contributing to good epistemic outcomes, this is partly an upshot of mundane facts about the sociology of the philosophical academy, namely, that at a particular time it was populated nearly entirely by people who regarded Catholic moral doctrine and Dutch Reformed theology as views that didn't merit any serious intellectual consideration.Footnote ²⁰ In short, the indirect epistemic benefits of religious belief in this domain owe to the ostensible absence of religious belief within the domain, and to the familiar Millian principle that in an intellectually homogeneous group, adding people with unorthodox views is epistemically salutary, since the ensuing disputes tend to reduce the rate of unthinking acceptance of false orthodoxies by people in that group. But the general upshot of the above is not specific to religious belief. Where there's no-one who believes P in a group or social environment, then having someone around who believes P is liable to have some benefit on some of the epistemic goals – the acquisition of knowledge, the transmission of true beliefs – for that group or environment. So, where everyone's an atheist, the introduction of a religious believer is epistemically salutary. But equally, where everyone's a religious believer, the introduction of an atheist is epistemically salutary. Similar points can be made along different axes. Where everyone's a moral realist, the introduction of a nihilist is salutary. Where everyone thinks alternative medicine is a sham, the introduction of a thoughtful defender of alternative medicine is salutary. Are there any reasons to think that there are indirect epistemic benefits that result from religious belief, beyond those reasons that obtain under particular demographic conditions, as a result of the general point noted here?

One reason why indirect epistemic benefits might be generated by people's adherence to religious belief-systems, in particular, is the fact that religious belief-systems tend to be, as the term suggests, systematic in their linking together of propositions in metaphysics, epistemology, ethics, etc. The intellectual effort that's involved in trying to get a complicated set of views across these different domains to ‘hang together’ will tend to elicit intellectual creativity – in argumentation, and in the elaboration of conceptual frameworks – that may result in novel contributions to knowledge, even if it's just knowledge of the logical cartography of the terrain being explored by the proponent of a particular religious belief-system. In a manner that's analogous to how the metrical constraints of a sonnet or a haiku elicit aesthetic ingenuity, the constraints that come with trying to reconcile the elements of the religious belief-systems that one identifies with may elicit intellectual ingenuity. In general, where one sees that one's beliefs are unorthodox, one may recognize that there are indirect epistemic benefits that stand to be gained by one's believing as one does. But where one recognizes that one's views are unorthodox and that they have a complex systematicity that needs working out, one may see a double indirect epistemic benefit in the offing: not just in challenging the easy acceptance of debateable orthodoxies among one's opponents, but also in the work that one will have to do in order to give an account, to an antecedently unsympathetic audience, of how the different parts of one's belief-system coalesce and interact.

None of this is to suggest that it would be rational for a person to try to nurture belief in a religious belief-system where she judges that the relevant evidence and arguments manifestly fail to rationally support that belief-system. As we said in discussing Atheist Scientist, and contrasting it with the Permissive Scientist scenario, the kind of case in which indirect epistemic reasons may legitimately get a grip on some agent, A, is one where A antecedently judges that the probative force of the relevant evidence makes belief in P rationally permissible, and where this leads A to favour this belief-state over some other belief state – such as agnosticism towards P – which she also regards as a rationally permissible response to the evidence. If recognition of the potential indirect epistemic benefits of holding some religious belief leads someone to delusively nurture a belief which, by her own lights, represents a rationally impermissible assessment of the evidence, this certainly would be irrational. But if it leads someone to nurture a belief that by her lights represents a rationally permissible assessment of the evidence, there's nothing amiss about that. So in at least some situations, indirect epistemic reasons can rationally favour the nurturing of religious beliefs.

Taking account of indirect epistemic reasons as reasons for belief

We concede that the reflective mindset involved in trying to take account of indirect epistemic reasons is a peculiar one. Someone might even deny that there's any coherent notion of ‘taking account of’ or ‘being influenced by’ such reasons to be provided. According to Jonathan Adler (Reference Adler2002), there is no psychologically workable way to approach belief-formation apart from the one in which we form beliefs based solely on our ingenuous assessment of the evidence. Indeed, this is why epistemic norms operate in a distinctive – he insists, wholly evidentialist – normative domain. Beliefs are not acts; they are representations of how we take the world to be. And we cannot take the world to be any way other than how we in fact take it to be. Brian Talbot, in defending a form of epistemic consequentialism that's complementary to our position here, rejects Adler's view on this matter. For Talbot, reasons generated by indirect epistemic considerations and reasons grounded in routine appraisals of evidence can exert their normative force on people's beliefs in harmony with one another, insofar as both kinds of reasons ultimately derive from, and operate in the service of, the same epistemic ends (Talbot (Reference Talbot2014), 615–616).Footnote ²¹ Although we're sympathetic to this suggestion, we also think it makes matters sound implausibly smooth and simple. Attempting to take account of indirect epistemic reasons while remaining faithful to one's assessment of one's first-order evidence is liable to be confusing. In this respect, paying attention to indirect epistemic reasons for belief does seem to have something important in common with paying attention to pragmatic reasons for belief. In both kinds of cases, the agent's doxastic attitudes become sensitive to her consciously apprehensible motivations, and so in both kinds of cases trying to take these reasons into account involves a jarring interruption of the unmotivated (or not-consciously motivated) observations and appraisals that are the aetiology of normal, evidence-based belief. One might even suggest that to categorize indirect epistemic reasons as bona fide epistemic reasons, as we do, is to overlook this point of similarity in how agents consciously take account of both kinds of reasons as reasons.

On the other hand, though, to classify indirect epistemic reasons as merely pragmatic reasons is to suggest that they are – like other pragmatic reasons for belief, such as what's comforting to believe; or what's ego-boosting to believe – a type of consideration that ought to be ignored by intellectually serious truth-seekers. And that, we have argued, is a mistake. Once we start thinking about the roles we occupy in communities of inquiry, and about how our beliefs can contribute to good and bad social epistemic dynamics – in other words, once we start doing (a certain kind of) social epistemology – we are already, whether we like it or not, contemplating the indirect epistemic consequences of our beliefs, and conducting our intellectual business in a way that's liable to be influenced by this consideration. When someone tries to more self-consciously take these indirect epistemic reasons into account, she shouldn't be thought of as weakening in her commitment to epistemic rationality, and letting her beliefs be guided by mere pragmatic concerns. Attempting to include indirect epistemic reasons in one's reflective considerations should be thought of as a commendable effort to pay attention to all the epistemic reasons for belief that apply in relation to the issue at hand.

Still, the question of precisely how an epistemically rational agent can take indirect epistemic reasons into account, and go on forming beliefs that are sensitive to the evidence, but with this extra consideration in mind, is a difficult one. We conclude with some tentative remarks on this, linking back to the non-evidential arguments for religious belief from the introduction, Pascal's wager, and James's will to believe.

Both Pascal and James advert to pragmatic considerations which, so they claim, favour belief in God. At the same time, both recognize that really to believe in God is to ingenuously think it's true that God exists, and not just to acknowledge the pragmatic reasons that favour thinking as much. And in different ways, both authors try to bridge this gap between pragmatic motive and ingenuous belief by recommending some form of enacted commitment to the beliefs one has (alleged) non-evidential reasons to adopt. Where we've postulated a distinction between having a direct epistemic reason to believe P, and having an indirect epistemic reason to nurture a belief that P,Footnote ²² this notion of enacting a commitment to P – of being a believer in P, as opposed to just having a belief that P – narrows the gap between these two things. To think of beliefs as enacted commitments is to reject a familiar understanding of how epistemic rationality and practical rationality interact. We don't always start by settling our ideas about what we believe, and then, following that settlement, choose how to conduct our lives. In our efforts to make our beliefs and practices cohere, sometimes the former adjust to fit the latter. Put simply, we work out our beliefs and practices in tandem. Indeed, according to James at least, there are cases in which it's only by undertaking a commitment to certain practices that we'll even encounter all of the evidential reasons that support the beliefs which would make sense of those practices. The very idea of religious belief hinges on action, James says, and if we insist that decisive evidential reasons for belief always have to precede action, à la Clifford,Footnote ²³ then – in matters of religion, and in other areas of life as well – we cut ourselves off from truths of potentially great significance (James (Reference James1912), 29–30). The remedy to this is to allow that people may decide to live in the light of beliefs about whose truth they're uncertain.

Our question, remember, is how an epistemically rational agent can take indirect epistemic reasons into account. In the domain of religion, enacting a commitment to a particular belief may mean spending time with a faith community, or engaging in devotional practices like prayer and meditation, or identifying oneself as a member of the faith and promoting it to others. We don't think that taking account of indirect epistemic reasons will, in general, require agents to regularly perform analogues for all of these. But the third element above – identifying oneself as an adherent of a view and promoting it to others – does seem like a simple and palatable way of enacting a commitment to a belief which one takes to be supported by indirect epistemic reasons. In acting as an advocate for a certain view and signalling one's pro-attitude towards it, one can thereby find oneself becoming more cognisant of the view's rational merits. These things are all to the good, on a Jamesian or Pascalian view, since the point of enacting a commitment to a belief, on such views, is to engage in practices which betoken one's pro-attitude towards that belief, and which simultaneously can make the belief's merits, as a picture of the world, more vivid than they were in the abstract. Much more could be said about all this. Our contention, for now, is simply that there is some sense to be made of the notion of responding to indirect epistemic reasons as reasons, and that a better sense of what this involves might be unearthed by looking at those who favour ‘enacted commitment’ as an approach to religious belief.Footnote ²⁴