2.1 In What Way is Permissivism Arbitrary?

Permissive epistemologies deny that there is always one unique, rationally permissible doxastic attitude that an agent can hold towards proposition P on the basis of a body of evidence E. On its face, there's nothing arbitrary about a Permissivist stance, thus understood. Permissivism just expresses a – prima facie plausible – view about what's rational when it comes to judgments about how evidence probatively bears upon propositions. Almost any example of evidence bearing on a proposition can be used to illustrate this. Say we're trying to figure out who left the dirty dishes on the counter, and we gather together a few scraps of evidence which point to Anne as the likely culprit, though without being sufficient to confirm her guilt beyond all doubt. In such a case, it seems much more plausible to say something in the spirit of Permissivism than something in the spirit of Uniqueness. It seems much more plausible to say

Than it is to say

When the rival views are portrayed this way, Uniqueness appears to be extraordinarily strict. In Kelly's words (Reference Kelly, Steup, Turri and Sosa2014: 298), proponents of Uniqueness seem to be saying that there's never any ‘slack’, when it comes to the relationship between the evidence and what it's rational to believe. Given the same evidence, if A believes it's 50–70% likely that Anne left the dishes out, and B believes this is 70–90% likely, then Uniqueness says that it's necessary that one of these beliefs is irrational. And because that seems absurd, Uniqueness is false.Footnote ⁵ A number of authors have criticised Uniqueness along these lines.Footnote ⁶

These bizarre implications are one reason why we should be loath to endorse Uniqueness. As I said in §1, though, even if Uniqueness looks implausible, we still have to see whether Permissivism has untoward costs, and so we still have to examine TAO. The worry TAO raises for Permissivism is about the way an agent's attitudes ‘hang together’ once she consciously affirms a permissive set of judgments, for any matter on which she holds a determinate doxastic attitude, DA₁. Here's a rough formulation of the problem, in second-personal terms. If you say DA₁ and DA₂ are both rationally permissible doxastic attitudes to hold towards P given E, it's hard to see what reasons you could have – other than arbitrary reasons, like what's more fun to believe – for singling out either of these as your belief. Once you've registered the notion that both DA₁ and DA₂ are, by your own lights, rationally permissible, it's arbitrary to accept one in preference to the other. In this situation, you can figure out which out of DA₁ or DA₂ ultimately seems rationally superior, thus abandoning the judgment that they're equally rationally eligible. Or, if you're staunchly committed to the judgment about their parity, you can adopt a doxastic attitude that expresses this – like a partial endorsement of both DA₁ and DA₂ – and affirm this as the rationally appropriate doxastic attitude to hold towards P given E. But in either case, ridding your attitudes of the arbitrariness will involve you revising your beliefs away from the kind of permissive stance which affirms the rational permissibility of more than one doxastic attitude towards P in light of E.Footnote ⁷

What TAO is homing in on, to put it another way, is the quasi-akratic stanceFootnote ⁸ of the agent who, in addition to saying ‘there's a plurality of belief states which it would be rational to hold towards P on the basis of E’, also wants to say ‘and as far as I'm concerned, my belief about P is a credence of 0.6’. For the proponent of TAO, it is this pair of attitudes that spells trouble for Permissivism. I cannot hold fast to just one of the doxastic attitudes, DA₁, that can be rationally formed based on E, while at the same time thinking that this attitude lacks any general rational better-ness relative to another attitude, DA₂, which I could have easily arrived at instead. And to be clear, the problem with this pair of attitudes is not just that the agent has a particular credence, while being uncertain about the rationality of having that credence; Elga (Reference Elga2013) shows that these kinds of mismatches between first-order and second-order doxastic attitudes needn't be irrational. The problem in our case is about the particular nature of the mismatch between the agent's first-order and second-order attitudes. The problem is that the agent holds a particular credence while thinking that it would be rational to accept both that credence and one or more others.

2.2 White's Magical Pill Argument

White sums up his version of TAO in a rhetorical question: if my current beliefs are, by my own lights, not rationally obligatory for me, why should I take propositions that I actually believe as a basis for action and reasoning, instead of others which I don't believe but which, by my own lights, I would be rational in believing? (White Reference White2005: 455). Of course White's insinuation is that there is no satisfactory answer to this question. If I really do think DA₁ and DA₂ are both rational doxastic attitudes to hold towards P given E, then any reason I may have to privilege DA₁ as my belief would be epistemically arbitrary.

The main argument White uses to formulate TAO appeals to a thought experiment.Footnote ⁹ Take a case in which, by the Permissivist's lights, two different doxastic attitudes about P – let's say, a credence of 0.2 and a credence of 0.6 – are both permissible given evidence E. White then asks us to imagine that agent A is invited to take one of two magical pills. Taking either pill will cause A to have a determinate credence in P. One will give her a credence of 0.2 in P, the other will give her a credence of 0.6 in P. If A accepts the offer, she will take one of the two pills, although which one she takes will be settled by random selection. Now, according to White, doxastic attitudes formed via the pill-taking process cannot, upon reflection, be endorsed as rational. A can't stand behind her belief about P if she's aware that the belief was acquired via a chancy process, which may have just as easily resulted in her having a different belief about P. And this is true even if, by A's own lights, the doxastic attitude she acquires via pill-popping falls within her range of permissible doxastic attitudes. The problem with pill-popping is that it makes A arbitrarily fix on one belief-state, DA₁, which, while it is within the range of belief-states that A regards as rationally acceptable, nevertheless isn't worthy of being elected to a privileged doxastic status just because it was selected via a flukey procedure. But here lies the problem for the Permissivist. If, instead of pill-popping, A acquires her doxastic attitude, DA₁, in the normal way – i.e. by making a judgment about E's probative force relative to P – this seems vulnerable to the same sort of complaint as the pill-popping method, given a permissive framework. As in the pill-popping case, A arrives at a determinate doxastic attitude, DA₁, which is within her range of permissible doxastic attitudes. So far so good. But she arrives at it via a process – judging the force of the evidence – which, by her own lights, might have just as easily (and in a phenomenologically indistinguishable way) led her to have an entirely different doxastic attitude, DA₂. From the Permissivist's perspective, pill-popping and evidence-appraisal are both ways to acquire one among some plurality of rationally permissible doxastic attitudes, while leaving it to chance which particular doxastic attitude one acquires. As White says in a more recent discussion of these issues, the only way for you to endorse, on reflection, a belief that's the output of such a process, is to suppose that you're good at lucking into rational beliefs (White Reference White, Steup, Turri and Sosa2014: 316). Ergo, since Permissivism entails something absurd – namely, that it isn't rationally mandatory to favor (i) forming beliefs by assessing the probative force of evidence, over (ii) forming beliefs through a random-selection, pill-popping method – it must be false.

The general problem underlying the intuitions that White is pressing in the pill example can be stated as follows. If you're committed to the view that there is a plurality of doxastic attitudes it would be permissible to hold towards P given E, having this commitment obliges you to deny the accuracy of any belief-forming process that picks out a single doxastic attitude towards P (among the set of permissible doxastic attitudes) and elects it to the privileged position of being your belief about P. The Permissivist must, after all, be appealing to some standards to ground her judgment that (e.g.) both 0.2 and 0.6 are among the set of credences that it's rational to have towards P given E. But whatever those standards are, they will – if they're being treated as fixed bases for further judgment – impugn the rationality of any process which says that it is only a credence 0.2, and not a credence of 0.6, that you should hold towards P given E. If you adhere to the standards that underwrite the initial permissive judgment about P, then you ought to reject any method of evidence-assessment that tells you to adopt one narrower doxastic attitude towards P. And if you think that what the evidence shows is that there is a 20–60% likelihood that P is true, then your doxastic attitude concerning P should reflect precisely that judgment about what the evidence shows (e.g. by expressing a concomitant credence interval). But in that case, you will have to reject – as unacceptably arbitrary – any doxastic attitude which fixes upon a single point within that interval of probabilities, in preference to all the other candidate points within that interval.

2.3 Problems for Permissivists in Practical Decision-making

As Sophie Horowitz observes in her work on epistemic akrasia (Reference Horowitz2014: 727), “the irrationality of an agent's beliefs is often brought into sharper focus if we consider what happens when she puts her money where her mouth is”. White acknowledges this point with a discussion of how permissive sets of attitudes are liable to undermine legal verdicts. He sketches a case in which “the range of rationally permissible degrees of confidence in Smith's guilt is just broad enough to include my own conviction [that Smith is guilty] as well as a degree of doubt that would make a vote of Not Guilty appropriate” (White Reference White2005: 453). The permissive set of attitudes thus described seems highly unstable as a foundation for decision-making. I can issue the verdict that Smith is not guilty, thereby effectively disavowing my judgments about the probative force of the evidence. Or I can issue the verdict that he is guilty, contrary to my own acknowledgement that others would be fully rationally within their rights to issue the rival verdict. “Holding this moderately permissive view”, White says, “either leads to the absurd consequence that it is appropriate to vote against one's convictions, or it is self-undermining” (White Reference White2005: 454). The point seems to me well-made.

Here I want to present two new examples to further clarify what is rationally untenable about the kind of arbitrariness that arises in a permissive suite of attitudes. Like the above case, my examples do this by bringing in practical stakes. In my examples, however, the stakes are quantifiable. Consider:

Should A take a 2:1 wager on P's truth, given the permissive set of attitudes that I have ascribed to her in the description of the case? The answer is unclear. On one hand, A judges that P is 60% likely to be true, and given that judgment, the expected payoff of taking a 2:1 wager on P's truth is 20¢ for every $1.00 bet. So A should accept the wager. On the other hand, A judges that it is rationally permissible to think that P is 20% likely to be true, and given that probability estimate, a 2:1 bet on P's truth carries with it an expected loss of 60¢ for every $1.00 bet. So, by A's own lights, accepting the wager means acting in a way such that it would be rational to expect that loss. So A shouldn't accept the wager. I don't want to insist that there is perfect parity in the strength of the reasoning here. My point is just that it's hard to see how A can practically decide between the recommendations, since they both arise, quite straightforwardly, out of her own judgments about the epistemic rationality of different doxastic attitudes towards P given E. So far as A’s permissive set of attitudes in WAGER leads to conflicting recommendations, which it seems difficult to resolve, we have reason to see that set of attitudes as untenable in some sense or other.

Here is a structurally similar scenario with a different set of stakes.

Should A invest B's money in S? Again, it's difficult to say how A should act in this scenario. On one hand, A believes S will deliver a 10% return. On the other hand, A believes that it's rational to expect S to incur a 30% loss. Again, I don't want to insist that there is perfect parity in the strength of the reasoning in either direction. The point is that it is hard to see how A can rationally arbitrate between the recommendations, since they both stem from her own beliefs. So again, insofar as A's permissive set of attitudes in PORTFOLIO leads to irreconcilable recommendations about whether A should invest in S, we have reason to regard that set of attitudes as rationally untenable.Footnote ¹⁰

2.4 A Satisficing Resolution?

The lesson from both cases, and also from White's verdict example, is that A cannot have beliefs about whether P given evidence E, which are one hundred percent independent of her P-and-E-related higher-order beliefs, that is, her beliefs about what it is rationally permissible to believe with respect to P given evidence E. If A is willing to declare that the doxastic attitude that she in fact affirms, DA₁, is the only permissible doxastic attitude, then there's no problem. But if A really insists that other doxastic attitudes – DA₂, DA₃, etc. – are rationally permissible, then her partial allegiance to these other attitudes, even if it's just a matter of granting their rational permissibility, must in some sense undermine or ‘infect’ her purported allegiance to DA₁. An agent can either endorse a permissive view about the probative force of E in relation to P, or she can affirm some determinate doxastic attitude about P in light of E. But doing both generates problems.

Perhaps there's a relatively simple way for the Permissivist to wriggle out of the bind that I'm trying to impose. David Enoch sketches a view on which “there is one degree of belief that is maximally rational, but some other degrees of belief – though less than maximally rational – are still rationally permissible” (Enoch Reference Enoch2010: 957). This proposal resembles satisficing forms of consequentialism in ethics, distinguishing responses to the evidence that are rational in the sense of being ‘best’, from those that are rational in the sense of being ‘good enough’.Footnote ¹¹ And as Enoch says, there's something natural about this way of cashing out a Permissivist framework. The Permissivist can say “given the evidence, my doxastic attitude towards P is rationally permissible (and it's the best attitude, that's why I hold it), your attitude towards P is also rationally permissible (only in the sense of being good enough), and that other person's attitude towards P isn't rationally permissible at all”.

I don't think this solves the problem, though. Any account of Permissivism modelled on satisficing faces a dilemma. The satisficer has to divide the set of doxastic attitudes that she regards as rationally permissible into two categories: (i) Worthy beliefs, i.e. those that she thinks merit her actual acceptance, and (ii) Okay beliefs, i.e. those she thinks don't merit her acceptance, but whose rational permissibility she's willing to grant. The dilemma arises when the satisficer tries to explain what distinguishes the Worthy belief from the Okay beliefs, for a particular proposition P and body of evidence E. Suppose A grants that the distinguishing consideration is something epistemically arbitrary. In that case TAO returns with full force, and the practical problems as in WAGER can't be resolved by A treating her belief as decisive, since, by A's own lights, her belief is based upon epistemically arbitrary factors. By contrast, suppose A thinks there is an epistemically relevant factor that renders the Worthy belief about P superior to the Okay belief about P, given E. In that case, it's hard to see how A can maintain that others are (or would be) free from any epistemic blame for holding an Okay belief. To hold an Okay belief is, by A's lights, to evince an apprehensible rational failure in one's assessment of the relevant epistemic factors. And such failures are characteristically blameworthy, in a way that seems incompatible with calling them permissible – assuming the highly plausible principle that if φ-ing is permissible, then one isn't blameworthy for φ-ing. The dilemmas in cases like WAGER can now be resolved, if this is how the satisficer distinguishes worthy beliefs from Okay beliefs. But the resolution is enabled because the Permissivist in question is no longer in a position to credibly maintain that the beliefs other than her own are rationally permissible. In other words, if the satisficing approach resolves the problems generated by TAO, it does so only to the extent that it dissolves the commitments that distinguish Permissivism in the first place.

3.1 Schoenfield's Response

To rebut TAO, the Permissivist has to accomplish both of the following tasks. Task 1: she must identify something epistemically non-arbitrary about DA₁ – that is, her doxastic attitude towards P given E – in light of which it's rational for her to hold DA₁ in preference to other prima facie eligible doxastic attitudes. However – Task 2 – she has to answer to the demands of Task 1 in a way that doesn't lead to the non-Permissivist view, that DA₁ is in general rationally superior to these rival doxastic attitudes.

Schoenfield presents an account of standards-based Permissivism that's intended to satisfy these requirements. According to Schoenfield, in cases where you judge that DA₁ and DA₂ are both rationally permissible doxastic attitudes to hold towards P given E, it needn't be arbitrary for you to affirm DA₁ as your belief state, and you can have an explanation of why forming your belief by appraising the evidence is a better method than randomly selecting one option among the set of belief-states that are permissible by your lights. Schoenfield thinks you can say something like the following. ‘I hold DA₁ because it is the doxastic attitude that's enjoined by what I take to be the most truth-conducive epistemic standards, ES₁. However, there is an alternative set of rationally acceptable (although, by my lights, less truth-conducive) epistemic standards, ES₂, which recommends a different doxastic attitude towards P given E, DA₂. And it's rationally permissible for an agent using ES₂ to hold DA₂.’ This move seemingly addresses Task 1, because the consideration that's appealed to in support of believing DA₁ – namely, the fact that DA₁ is enjoined by the epistemic standards you regard as optimally truth-conducive – looks like an epistemically non-arbitrary reason to favor DA₁ over DA₂. And this move seemingly addresses Task 2 as well. You're not obliged to say that DA₂ is rationally impermissible compared to DA₁, because the rational status of any doxastic attitude is assessed relative to the agent's epistemic standards, and if the epistemic standards that enjoin DA₂ are rationally permissible, then DA₂ itself is rationally permissible as well.

What are epistemic standards, exactly, and what does it mean to employ a given set of epistemic standards? Epistemic standards, for Schoenfield's purposes (Reference Schoenfield2014: 199), are simply methods for forming beliefs based on evidence; ‘functions from bodies of evidence to doxastic states’. And to adopt standards ES₁ as your own, means to use ES₁ to form beliefs from your evidence, and to do so because you judge that this will reliably lead you to form doxastic attitudes that express a high degree of confidence in truths and a low degree of confidence in falsehoods. Now, as Schoenfield concedes, you make a peculiar set of judgments when you adopt one set of epistemic standards, ES₁, while accepting the rationality of some alternative set, ES₂. To adopt ES₁ is to regard these standards as optimally truth-conducive, and thus better than ES₂ in some sense.Footnote ¹² At the same time, though, there's also a sense in which you can regard ES₂ as just as good as ES₁. Different sets of epistemic standards weight different epistemic considerations differently – for instance, they can give more or less priority to believing the true relative to disbelieving the falseFootnote ¹³ – and for at least some such considerations, there isn't any viewpoint-independent way to evaluate the comparative merits of alternative weightings. Rational capacities are, as Schoenfield stresses, general-purpose, insofar as they aid agents in making judgments, conducting inferences, and appraising evidence across a wide range of circumstances and in response to many different forms of information. Different epistemic standards can serve different agents sufficiently well, with respect to their truth-conduciveness, across these different contexts. And thus, while an agent will naturally regard her own epistemic standards as optimally truth-conducive, she can also maintain that it's rationally permissible to employ some rival set of epistemic standards. That is the type of picture that Schoenfield proposes, at any rate.

3.2 Why TAO Applies to Schoenfield's Standards-based Permissivism

One doesn't have to dig far beneath the surface to unearth the difficulties that this picture brings with it, insofar as it's supposed to underpin a response to TAO. Schoenfield states the worry as follows.

Where before our problem was arbitrarily favoring one among two rationally permissible doxastic attitudes, now our problem is arbitrarily favoring one among two rationally permissible sets of epistemic standards. The problems of TAO recur for Schoenfield at the level of epistemic standards.

Or so it seems. An adequate response to this problem is in the offing, I believe, but the response that Schoenfield provides doesn't quite achieve what's required of it. Her response begins by noting that the Permissivist, A, who regards both ES₁ and ES₂ as rationally permissible epistemic standards, can favor ES₁ – and the doxastic attitudes it endorses – over ES₂, if and insofar as she judges that ES₁ is optimally truth-conducive, whereas ES₂ is merely adequately truth-conducive. There's nothing unacceptably dogmatic about these kinds of judgments, Schoenfield insists, because it is simply ‘an inescapable fact of epistemic life’ that we judge our own epistemic standards to be the best ones (by the lights of those same standards) and that we're unable to adduce compelling reasons in support of such judgments that are independent of our own standards (Schoenfield Reference Schoenfield2014: 202). Call this the No Independent Justification constraint, or NIJ.Footnote ¹⁵

Let's assume for the sake of argument that NIJ does indeed hold with the degree of generality that Schoenfield suggests. The core of the problem with appealing to NIJ, in this context, is that it leaves A in a position from which it is very hard for her to substantiate her professed view regarding DA₂, namely, that it is rationally permissible to accept DA₂, given E, even though A herself rejects it.

To see why, note firstly that there's no sense in the Permissivist appealing to the idea of epistemic standards to try to resolve TAO, unless she's going to be invoking rival epistemic standards, ES₂, as part of her explanation of what makes it rationally permissible to adopt DA₂. If A's own epistemic standards, ES₁, affirm the rational permissibility of both DA₂ and DA₁ – the latter being the doxastic attitude that A actually holds – without even any reference to some rival set of epistemic standards, ES₂, then we're back to square one with the arbitrariness problem. If this is the situation, then A is contemplating two eligible doxastic attitudes towards P given E, both of which are rationally permissible according to ES₁ – i.e. A's own epistemic standards, from which all of A's judgments derive – and neither of which can be privileged as A's belief regarding P unless an epistemically arbitrary factor is brought in to adjudicate.

So, in order for A to vindicate her professed position, that DA₂ is rationally permissible, A needs to offer an explanation of DA₂’s rationally permissibility which is built around the premise that DA₂ is the doxastic attitude which is endorsed by some rival set of epistemic standards, ES₂. And A must be able to endorse the rationality of ES₂ while she carries on in applying her own standards, ES₁.

The problem with this is that once A has cited the NIJ constraint as her basis for favoring ES₁ over ES₂, it seems that she shouldn't accord any weight to ES₂’s recommendations. After all, NIJ says that ES₁ is the reference point from which all of A's judgments about the rationality of epistemic standards are derived; if we assume NIJ, then ES₂’s merits, from A's perspective, are entirely determined by what ES₁ says about ES₂. And those determining standards, ES₁, say at least both of the following things about ES₂: (i) that the doxastic attitude endorsed by ES₂ in the case at hand is not the most accurate one, and (ii) that using ES₂ in general will result in an agent adopting less accurate doxastic attitudes than using ES₁ would. In short, from the perspective of ES₁, ES₂ has apprehensible deficiencies with respect to the accuracy of the doxastic attitudes it endorses. And in light of those deficiencies, it is natural to regard the adoption of ES₂ as liable to some kind of epistemic blame, and therefore – assuming the highly plausible principle that if φ-ing is permissible, then one isn't blameworthy for φ-ing – as not rationally permissible.

To be clear, Schoenfield will reject this inference. She explicitly denies that epistemic standards have to be optimally accurate or truth-conducive in their recommendation of doxastic attitudes in order to qualify as rationally permissible. And stated in the abstract her view seems plausible, particularly for anyone who comes at the issue with permissive intuitions. On Schoenfield's view, principles of rationality say what considerations count in favor of what beliefs, and it is at least possible for two agents to both be reliably paying attention to the same (correct) considerations in forming their views, while weighting those considerations differently, such that they sometimes arrive at different views, even while both doing what rationality demands of them (Schoenfield Reference Schoenfield2014: 202). Still, however much it seems reasonable, in the abstract, to prise apart accuracy and rational permissibility in evaluating epistemic standards, there seems to be something decidedly odd about A doing this while she's in the midst of a concrete dispute in which she's reflectively affirming (i) the superior accuracy of her standards, ES₁, compared to ES₂, as well as (ii) the apprehensible deficiencies of ES₂, vis-à-vis accuracy, by the lights of ES₁. (Recall also that, given the NIJ constraint, A cannot but arrive at these kinds of judgments, (i) and (ii), about ES₂’s deficiencies relative to ES₁.)

Of course it is open to Schoenfield to insist that when it comes to epistemic standards, accuracy and rationality can come apart in a way that doesn't disallow these kinds of combinations of judgments. There will be further problems to negotiate in maintaining this view. Presumably a great enough degree of inaccuracy in epistemic standards will have to disqualify those standards from being counted as rationally permissible, and it's not clear whether any principled explanation can be given to distinguish between the inaccurate epistemic standards that are rational and those that aren't. In any case, I haven't said enough to decisively show that Schoenfield's position is untenable. It's certainly not self-contradictory – or a denial of an analytic truth – to say that one can grant the rationality of epistemic standards even if they aren't optimally truth-conducive. But there are costs that come along with this position and which militate against it. It requires us to have a conception of epistemic standards on which an agent can endorse the rational permissibility of epistemic standards even while recognising deficiencies in those standards which, to all intents and purposes, seem like the sort of deficiencies that normally merit epistemic blame.

3.3 The Interaction between Epistemic Standards and Cognitive Abilities

In light of the above, the challenge for the Permissivist who aims to resolve TAO by appealing to epistemic standards is as follows. First, she must explain her acceptance of DA₂’s rational permissibility by appeal to the fact that DA₂ is endorsed by ES₂ – that is, by epistemic standards other than her own – because if her own epistemic standards endorse DA₂’s rational permissibility, TAO immediately recurs. Second, in order to defend her judgment about DA₂’s rational permissibility, in light of its endorsement by ES₂, she must find a way to vindicate the claim that ES₂ really is rationally acceptable, like her own epistemic standards, ES₁. But then, third, her explanation of ES₂’s rational acceptability must still leave her with the space to identify some epistemically non-arbitrary factor which justifies her in favoring ES₁ over ES₂.

How can the Permissivist square this circle? The answer lies in an appeal to the idea that there is some kind of interaction between the epistemic standards we adopt, i.e. the methods via which we form beliefs based on the evidence, and our cognitive abilities, i.e. the abilities that we put into practice in the application of our epistemic standards. Analogically, the idea is that the truth-conduciveness of an agent's epistemic software depends on what kind of cognitive hardware that agent is working with. Consider agent A, and her particular set of cognitive abilities, CA₁. It may be rationally preferable for A to use a particular set of epistemic standards, ES₁, insofar as these are – for an agent operating with CA₁ – the most truth-conducive standards via which to form beliefs from the evidence. But where that's the consideration that A appeals to in order to defend her adoption of ES₁, A can reconcile this appeal with her judgment that certain rival standards, ES₂, are no less truth-conducive per se than ES₁, since the truth-conduciveness of both ES₁ and ES₂ depends on the cognitive abilities of the agents employing those standards.

Here's an example to illustrate, which I'll call DETECTIVES. Veronica Mars and Nancy Drew are investigating a crime for which X and Y are the prime suspects. After thoroughly reviewing the same body of relevant evidence, Mars believes it's 70% likely X is the culprit and 30% likely Y is the culprit, whereas Drew believes it's 30% likely X is the culprit and 70% likely Y is the culprit. Can both detectives regard the other's views as rationally permissible, while also asserting the rationality of their own doxastic attitudes about X and Y's guilt? Can they be permissive? Yes. Suppose firstly that Mars and Drew employ different epistemic standards, i.e. they use different methods for assessing and interpreting their evidence in order to form their beliefs. Suppose, for instance, that they have different methods for evaluating the probative force of testimonial evidence, a type of evidence that's often an important part of the data they examine as detectives. Mars is excellent at recognizing insincere testimony – she's very sensitive to various subtle verbal and non-verbal tell-tale markers of insincerity – and thus she interprets different instances of testimony differently, ascribing a different probative force to insincere testimonial evidence. Drew is not very good at identifying insincere testimony, but she is excellent at making holistic, gestalt assessments of the combined probative implications of large bodies of disparate evidence. So unlike Mars, Drew doesn't pre-sort the testimonial evidence, based on its sincerity. Where Mars reviews the testimonial evidence in sequence, identifying each item's probative force in its own right, Drew reviews all testimonial evidence in the same light, and assesses the probative force of the testimonial data as a set. Mars applies her standards because she's cognitively well-equipped to apply them. She has the kind of perceptual and attentional abilities that enable her to accurately judge the sincerity of testimony. In a similar way, Drew is cognitively well-equipped to apply her standards. She has a particular imaginative facility that enables her to formulate credible explanatory narratives, based on gestalt interpretations of diverse bodies of evidence. Now here's the final, crucial detail. Although Mars and Drew are working together for the first time, they have a long track-record of working independently on other cases. And in light of their track-records, both of them have good reason to believe that when they apply their own epistemic standards, in forming beliefs about cases relevantly similar to the case at hand, they are very likely to get the answer right. In short, each of them is in the position of (i) having a combination of cognitive abilities and epistemic standards which reliably generates accurate beliefs, and (ii) having good reason to believe that this is the case.

Under these conditions it seems rational for each party to apply her own epistemic standards to the evidence and adopt the beliefs endorsed by those standards. Each detective uses epistemic standards that are optimally truth-conducive for her, given her cognitive abilities, and given the felicitous interaction between her epistemic standards and those cognitive abilities. At the same time, each of them would be able – if all the above facts were made available to them – to see that the other is in the same position, of using epistemic standards that are optimally truth-conducive for her, given her cognitive abilities, and therefore that the other is reasoning well in adopting the doxastic attitudes endorsed by her standards. So, despite having different doxastic attitudes in the case at hand, Mars and Drew could at least in principle regard each other's judgments as fully epistemically rational. And they can recognize that the other party's epistemic standards and the beliefs formed in light of them are rational, in a way that need not undermine their faithfulness to their own epistemic standards and the beliefs enjoined by them. This is because each of them can appreciate that the truth-conduciveness of the epistemic standards in question depends upon the cognitive abilities of the agent by whom they are employed. They can both appreciate, then, that in a pair-wise comparison, neither Drew's standards nor Mars's standards are more truth-conducive per se, or rationally superior per se. And therefore, although there is an epistemically non-arbitrary consideration that justifies each agent in applying her own epistemic standards and standing by the doxastic attitudes that they enjoin, it is one that still allows both agents to affirm the rationality of each other's beliefs.Footnote ¹⁶

4.1 Conciliationism and the Practical Problems Revisited

Here are two worries one might have about how the kind of standards-based Permissivism I've sketched above would operate in practice. First, if making sense of Permissivism in any particular case requires us to recognize a host of further considerations about the different (but equally good) cognitive abilities and epistemic standards of the two agents, as in DETECTIVES, these further considerations may change the case into something like a paradigmatic example of peer disagreement. In cases where there is a seeming parity in the epistemic credentials of two agents disagreeing about some proposition, the Conciliationist will say that the two agents should revise their doxastic attitudes towards a middle ground view (see §1). And any rational obligation to make such revisions seems to alter the case such that it's no longer permissive. The second problem is that it's not obvious how my proposed form of standards-based Permissivism can resolve pragmatic dilemmas like those discussed in §2.3. If Mars thinks it's 70% likely X is the culprit and 30% likely that Y is, but also thinks that Drew is rational in believing it's 30% likely that X is the culprit and 70% likely that Y is, should Mars accept a 2:1 wager on X being the culprit? The gamble seems vexed much like the cases of WAGER and PORTFOLIO in §2.3. Mars thinks X is 70% likely to be the culprit, and given that likelihood the expected payoff of taking the wager is 40¢ for every $1.00 bet. But Mars also thinks Drew is rational in believing that X is 30% likely to be the culprit, and if that probability judgment is granted the wager has an expected loss of 40¢ for every $1.00 bet. So the problem remains.

One way that a standards-based Permissivist could respond to such concerns would be to say more about the indirect epistemic benefits that result from people ‘sticking to their guns’ in the face of disagreement. Plenty of authors have suggested that in certain contexts there are epistemic goods that are better served if there is some diversity in people's beliefs.Footnote ¹⁷ Jonathan Matheson (Reference Matheson2015: 145 ff.) calls the type of considerations that arise from this ‘instrumental epistemic reasons’; Brian Talbot (Reference Talbot2014) calls them ‘truth-promoting non-evidential reasons for belief’; Kirk Lougheed and Robert Simpson (Reference Lougheed and SimpsonForthcoming) call them ‘indirect epistemic reasons’. It seems plausible that such considerations can at least sometimes be regarded as reasons for belief of some sort, and so perhaps in a case like DETECTIVES an agent could appeal to them to resolve the pragmatic deadlocks I've adverted to. Still, even in cases where such reasons could be invoked, the prevalent view would be that they cannot properly be characterized as epistemic reasons for belief; that they are, rather, merely pragmatic reasons (see Berker Reference Berker2013). It would be too great a detour here for me to try to comprehensively address this claim, but suffice it to say, if our strategy for defending Permissivism necessitates a defence of a position that's antecedently less plausible and less likely to be granted than Permissivism itself, then that strategy seems fairly unpromising.

The right way for a standards-based Permissivist to handle these kinds of problems, I think, is by distinguishing two types of circumstances that agents may find themselves in while trying to reflectively maintain a permissive perspective on a disagreement. Assume that in both types of circumstances the key conditions I outlined in relation to DETECTIVES hold: each of the agents is using epistemic standards that are optimally truth-conducive for her, given (i) her cognitive abilities, and (ii) the felicitous interaction between her standards and her cognitive abilities. The difference between the two types of circumstances is that in one the agents only have detailed knowledge about their own epistemic standards and cognitive abilities, i.e. what those epistemic standards consist in, how they're used to assess the probative force of evidence, how they interact with different kinds of cognitive abilities, and when these interactions do and don't lead to accuracy. Whereas, let's stipulate, in the other type of circumstances the agents have such knowledge about their own – and each other's – epistemic standards and cognitive abilities.

In the second type of circumstances the rational pressure on the agents to revise their beliefs in a conciliatory fashion will be stronger, and may ultimately oblige them to converge on one shared doxastic attitude. It is beyond the scope of this paper to provide a full account of the conditions under which such conciliatory belief-revision is rationally required.Footnote ¹⁸ One way or the other, though, it seems plausible that in at least some cases in which an agent knows of her counterpart's track-record of accuracy – and has good reason in light of this to think her counterpart is as likely as herself to be accurately judging the probative force of the evidence in the case at hand – conciliatory belief revision will be rationally required, in a way that unravels the initial permissive set of attitudes that the Permissivist was trying to defend.

But the same is not true in the first type of circumstances, in which the agents only have detailed knowledge about their own epistemic standards and cognitive abilities. Under such circumstances, agent A can have good reason to believe that in holding DA₁ towards P given E, she is in fact applying epistemic standards that are optimally truth-conducive for her given her cognitive abilities. Granted, she might also have good reason to believe that another agent, B, who holds DA₂ towards P given E could, in principle, be doing so as a result of her applying different epistemic standards, ES₂, which might be optimally truth-conducive for her (B), given the different set of cognitive abilities she has to work with. But without A herself having the cognitive abilities she would need to felicitously apply ES₂, and without her being in a position to confidently appraise the accuracy of ES₂ when applied by an agent who has those cognitive abilities, A seems justified in remaining steadfast in DA₁, and in treating that attitude as the basis for any practical decision-making linked to the relevant proposition. And all this is consistent with A accepting, in principle, that B's situation may ostensibly be the same from B's perspective, and also that if she (A) did learn that B – with her different cognitive abilities – had a track-record of accuracy in applying ES₂ that was equal to A's track-record of accuracy in applying ES₁, this would give both of them some kind of reason to revise their beliefs in a conciliatory manner. One can be a standards-based Permissivist, of the kind I've described in §3.3, while still conceding that, under particular circumstances, the disagreement of other agents with a known track-record of accuracy can generate conciliatory pressures.

When it comes time for the Permissivist to act under the guidance of her beliefs, there may still be some residual unease in holding DA₁ while thinking that others could be rational in holding DA₂. But this disquiet may be partly alleviated by reflecting again on the relation between epistemic standards and cognitive abilities. Consider how a Permissivist who isn't accustomed to thinking of things in this way – let's say, a run-of-the-mill Bayesian, B – is going to try to make sense of the uneasy feeling of arbitrariness when practical dilemmas arise. If B judges that DA₁ and DA₂ are both rational doxastic attitudes to hold towards P given E, she will regard the discrepancy – between her acceptance of DA₁, and someone else's acceptance of DA₂ – as a result of them having different prior probability distributions. Now, naturally B will see herself as rationally entitled to her priors. But still, when the practical dilemmas arise, and make these worries about arbitrariness more salient, it's difficult for B not to regard it as some kind of accident that she should have those priors, rather than the other ones which, after conditionalizing on E, would have resulted in her holding DA₂ instead of DA₁. There is no deeper fact about herself that B can look to in order to sustain a sense of proprietary connection to her priors – they're simply the ones she happened to have. It should be welcome news for the Permissivist, then, to be given explanatory resources that help her to have a sense of proprietary connection with her doxastic attitudes under such circumstances. The standards-based Permissivist needn't see it as just an accident that she holds her beliefs. This is because (i) her cognitive abilities are a part of her identity, and (ii) her acceptance of DA₁ owes to her having those cognitive abilities, since DA₁ is endorsed by epistemic standards whose fittingness for her is due to her cognitive abilities. The standards-based Permissivist has an answer to the critic who says that in sticking to her own judgment she's doing something entirely arbitrary. And that's some kind of progress.

4.2 A Less Permissive Permissivism?

Several prominent epistemological views – like Bayesianism, Coherentism, and Conservatism – are tacitly committed to Permissivism (see note 2). All these views, in their typical forms, tell us that the probative force of evidence E in relation to proposition P depends on factors that differ between different agents. And they all resist the notion of there being some uniquely-rationally-privileged agential standpoint, from which objective evidential support-relations for any values of P and E can be judged. The problem that TAO raises for such views is about how to understand the combination of attitudes that a self-conscious, reflective Permissivist needs to maintain. And this is where a standards-based approach – one that links epistemic standards to agent-specific cognitive abilities – offers some help. It helps explain why it needn't be epistemically arbitrary for A to hold DA₁ in preference to DA₂, even while A resolutely maintains that DA₂ is no less rationally permissible than DA₁. Notice, however, that the moves we are introducing to counter TAO will generate more stringent caveats for the doxastic license that Permissivism issues. It's not enough for the agent to say: ‘you seem to have arrived at your belief via rational methods, and I seem to have arrived at my belief via rational methods, so we're both entitled to believe as we do, and also to affirm the rationality of each other's beliefs’. If I'm a Permissivist I now have to bite at least two further bullets. I need some reason to think (i) that our different views are due to us having different cognitive abilities and applying different epistemic standards, and (ii) that the use of my epistemic standards is optimally truth-conducive for me, given the cognitive abilities that I have to work with in the application of my epistemic standards. And if it's rare for these conditions to obtain, then we'll only rarely be able to defend permissive sets of attitudes from the arbitrariness worries we've been considering.

Or so it appears. Whether a standards-based form of Permissivism ends up being painted into this corner depends upon what kind of epistemic modifiers are (or ought to be) applied to the qualifying conditions (i) and (ii), in order for an agent to activate this defence of a permissive stance. In short, the question is whether the agent needs to know – or be highly confident – that conditions (i) and (ii) obtain, or whether some weaker epistemic criterion is sufficient, like having a pro tanto good reason to believe that the conditions obtain. If the stringent standard is the right one, then permissive sets of attitudes will rarely be justifiable in real-world cases. As Nathan King notes, in a discussion of how these sorts of judgments bear on the epistemic significance of disagreement, it's difficult enough to see how our doxastic attitudes arise out of our evidence and our methods of evidence assessment (King Reference King2012: 267–8). It's presumably going to be even more difficult to make confident assessments about what kind of cognitive abilities agents are working with, and about how the truth-conduciveness of a particular set of epistemic standards is affected by the interplay between those epistemic standards and a certain suite of cognitive abilities.

Having said all that, I think a less stringent criterion – like having a pro tanto reason to believe that the conditions obtain – is more appropriate. All the Permissivist ever really needed was some kind of epistemically non-arbitrary factor – some kind of tie-breaker, so to speak – to justify her commitment to DA₁ alongside her affirmation of DA₂’s rational permissibility. And it seems like she has such a tie-breaker, as long as she has some pro tanto good reason to believe that the qualifying conditions (i) and (ii) obtain in the case under consideration. In effect, what the reflective Permissivist has to do to respond to TAO, once she has consciously affirmed the rational permissibility of person B's view, DA₂, is ask herself the following questions: (i) do I have any reason to think that mine and B's different doxastic attitudes vis-à-vis P owe to our applying different epistemic standards to the relevant evidence? And (ii) do I have any reason to believe that my epistemic standards are the more truth-conducive ones for me to employ, given my particular suite of cognitive abilities? If both of these questions can be answered in the affirmative, then the Permissivist has the tie-breaker she requires in order to avoid arbitrariness in holding DA₁ while also granting DA₂’s rational permissibility.

Still, affirmative answers to these questions aren't too cheaply available. The kind of standards-based version of Permissivism on offer here is permissive, but it's far from indulgent. And this is, I'd suggest, the kind of Permissivism that it makes sense to want. It's a familiar mindset to think something like ‘your appraisal of the evidence seems reasonable, but so does mine, so let's just say that both of us have rational beliefs’. The worry is what you will say when someone asks why you're favoring your view over the other one which – so you say! – is no less rational? What's disallowed by the kind of Permissivism that I'm espousing is the apathetic shrug. It's not enough to say ‘I can't pin down any good epistemic reason for favoring my belief over the other, but it's what I think, so that's that’. We may feel sympathy for the notion that, as Richard Fumerton puts it, “in the final analysis, there really is no alternative to the egocentric perspective” (Reference Fumerton, Feldman and Warfield2010: 106). But it doesn't follow from this that there's no middle-ground view between an indulgent Permissivism that says ‘I'm rational and so are you, because rationality comes fairly cheap’, and an irresponsible egocentrism in which the agent sees herself as the bearer of some ineffable, truth-conducive source of insight. A standards-based version of Permissivism – one which addresses TAO by positing an interaction between epistemic standards and cognitive abilities – shows how an agent can affirm the rationality of other people's doxastic attitudes, while being non-arbitrarily steadfast in her own doxastic attitudes. But it forces her to work a bit harder for this, and that's no bad thing.Footnote ¹⁹