1. Introduction
You are a juror in adjudicating whether Jack is guilty of murder. Together with other jurors, you all carefully go through the relevant evidence. But back in the jury chamber where you are asked to express your opinion, you find that half of the jurors agree with you that Jack is guilty whereas the other half disagree. How should this finding affect your belief?
The answer depends – among other things – on whether a single body of evidence allows multiple doxastic attitudes towards a proposition.Footnote 1 This is the debate between the ‘Uniquer’ and the ‘Permissivist.’ The former holds
Uniqueness
Given a body of evidence and a proposition, there is a unique doxastic attitude that's rational to take towards the proposition.Footnote 2
Permissivism is the negation of Uniqueness. It says that, in some cases, multiple attitudes towards a proposition are rational given the same evidence.Footnote 3 Permissivists typically think that the same evidence can allow different attitudes given differences in factors such as epistemic standards, background beliefs, practical stakes, ways of weighing theoretical virtues, and epistemic goals.Footnote 4
An important motivation for Uniqueness is the so-called ‘arbitrariness argument.’ It says that learning that one is in a permissive case would result in some unacceptable arbitrariness with regard to one's doxastic attitude (White Reference White2005).
Currently, Permissivists’ responses to the argument focus on explaining how Permissivism doesn't lead to arbitrariness (Kelly Reference Kelly, Steup, Turri and Sosa2014; Meacham Reference Meacham2014; Schoenfield Reference Schoenfield2014; Simpson Reference Simpson2017). A prominent attempt along this line is made by what I call ‘Standard-Permissivists,’ who claim that our beliefs are not arbitrary since they are uniquely singled out by our epistemic standards. In reply, Uniquers claim that appealing to standards merely pushes the worry of arbitrariness from the level of beliefs to the level of standards (White Reference White2005: 452; Feldman Reference Feldman and Antony2007: 205−6). So, the challenge for Standard-Permissivists becomes explaining how we can avoid the arbitrariness in our standards.
I have two goals in this paper. First, I argue that Standard-Permissivists’ attempts to meet the above challenge are unlikely to succeed and thus Permissivists shouldn't try to avoid the arbitrariness in our beliefs by appealing to standards. Second, I defend a new response to the arbitrariness argument by arguing that, contrary to common conception, the arbitrariness resulted by Permissivism is acceptable.
The paper proceeds in the following way. In Section 2, I explain the arbitrariness argument and why the Standard-Permissivist's response would relocate the worry from beliefs to standards. In Section 3 and Section 4, I criticize two attempts to avoid the arbitrariness in standards. Section 3 criticizes the claim that we can avoid arbitrariness because we have a standard-dependent reason to favor our own standards, and Section 4 criticizes the claim that we can avoid arbitrariness because we have a standard-independent reason to favor our own standards. Then in Section 5, I argue that the arbitrariness resulted by Permissivism is acceptable. My core idea is that the arbitrariness resulted by Permissivism is analogous to the arbitrariness in permissive actions and the latter arbitrariness is intuitively acceptable. Then three possible objections against this analogy, which are all motivated by the thought that beliefs aim at the truth, are discussed and rejected. In addressing the last objection, I draw inspiration from the recent debate on transformative experience. Section 6 concludes the paper.
2. The arbitrariness argument and Standard-Permissivism
Here is White's (Reference White2005) argument that Permissivism leads to arbitrariness. Suppose that you are a juror mentioned in the beginning of this paper. After rational deliberation, you believe that Jack is guilty. However, suppose that you learn that you are in a case that instantiates Permissivism (hereafter ‘a permissive case’) – given your evidence, it's rational to believe that Jack is guilty and it's both rational to believe that Jack is not guilty. In this case, do you have any good reason to keep your belief instead of switching to the other belief? It seems not. Clearly, you cannot choose to keep your belief by claiming that it's more rational. So, if you do choose to keep your belief, you can only do so on other grounds. For instance, you might keep your belief on the grounds that you are more used to it, or that it makes you sleep better, or that this is the choice you've made by flipping a coin (‘if the coin lands heads up, I will keep my belief; otherwise I will switch; and I just see the coin landing heads up.’) But these are all arbitrary (or ‘irrelevant’) considerations, and it seems irrational for one to keep a belief on the basis of these arbitrary considerations. So, you don't have good reason to keep your belief, and thus it's no longer rational for you to hold the belief. But this means that Permissivism is unstable: once you learn that two beliefs are both rational, it would no longer be the case that the two beliefs are both rational; instead, it seems that suspending judgment becomes the uniquely rational attitude.
Here is a clearer formulation of the above argument:
The Arbitrariness Argument (AA)Footnote 5
P1. If one learns that belief not-p is both rational as one's own belief p, and if one chooses to keep one's belief, then one can only make the choice arbitrarily.
P2. For any situation, it's irrational to hold a belief in the situation if one can only arbitrarily choose the belief.Footnote 6
Therefore,
C1. When one learns that belief not-p is both rational as one's own belief p, it's irrational to believe p.
Therefore,
C2. Permissivism is unstable in some sense: one's belief cannot survive learning that one is in a permissive case.Footnote 7
Of course, one response to AA is to simply bite the bullet and accept that one's belief can never survive learning that one is in a permissive case. But for Permissivists, it is unattractive to admit that their view is unstable in this way – it's odd that one's rational belief cannot survive learning a proposition that correctly states what's rational in one's situation (White Reference White2005: 450 − 1). Biting the bullet is especially unattractive for those Permissivists who think that their view has important implications for the debate on peer disagreement – if our beliefs can never survive learning that we are in a permissive case, then we couldn't cite ‘we are in a permissive case’ as a reason to be steadfast when encountering peer disagreement. So, the Permissivists in the literature have understandably looked for another way out.
One influential way out is to distinguish Intra-personal Permissivism from Inter-personal Permissivism (Kelly Reference Kelly, Steup, Turri and Sosa2014). The former says that a body of evidence can allow a single person to take different attitudes towards a proposition; the latter says that it can allow different persons to take different attitudes. According to Inter-personal Permissivists, AA threatens only Intra-personal Permissivism, for what leads to arbitrariness is your learning that it's rational for you to believe p and rational for you to believe not-p.
However, it's natural to doubt whether Inter-personal Permissivism can avoid arbitrariness: for any factors that allegedly allow different persons to respond differently to the same evidence, the arbitrariness plaguing Intra-personal Permissivism would just resurface with respect to those factors (White Reference White2005: 452; Feldman Reference Feldman and Antony2007: 205−6).
To see this point, consider a prominent version of Inter-personal Permissivism, which I call ‘Standard-Permissivism.’ It says that different persons can respond differently to the same evidence because they can adopt different epistemic standards (which are also called ‘epistemic rules’ or ‘methods of reasoning’; see Schoenfield Reference Schoenfield2014: 199; Meacham Reference Meacham2016: 472−3; Titelbaum and Kopec Reference Titelbaum, Kopec, Balcerak-Jackson and Balcerak-JacksonForthcoming). More exactly, Standard-Permissivism says:
Standard-Permissivism:
(A) A belief is rational just when (a1) it conforms to one's standard and (a2) one's standard is a rational one;
(B) There are multiple rational standards, and different persons could adopt different rational standards.
Here are four clarifications of Standard-Permissivism. First, following Schoenfield (Reference Schoenfield2014: 199), we can understand a standard as a function from evidence to doxastic attitudes. It's global: for any evidential input, it outputs an entire doxastic system. A global standard can be understood as consisting of a system of local standards, which concern attitudes about a particular proposition. For instance, the following are all local standards: <if e1, then believe p1>; <if e1, then believe −p2>; <if e2, then believe −p1>; <if e2, then believe p2>…; Second, also following Schoenfield (Reference Schoenfield2014: 200−2), we can understand ‘one's standard’ in this doxastic way: a standard of the form ‘if E, believe p’ is one's standard just in case one believes that following this standard is most truth-conducive.Footnote 8 Third, a charitable reading of (a2) in claim (A) should say that, for a belief to be rational in an evidential situation, it's enough that the relevant local standard – the one recommending the belief in question as a response to the evidence in question – is rational, so that one's belief needs not be corrupted by the irrationality of some irrelevant local standards within one's global standards system. For this reason, ‘standards’ will mean ‘relevant local standards’ unless otherwise stated. Last, Standard-Permissivists can divide on what makes a standard a rational one. Some are radical, claiming that any standard is rational as long as the global standards system containing it is coherent; most are more modest, claiming that there are also substantial constraints (e.g., Immodesty and the Principal Principle). This paper will focus on the modest version.
Standard-Permissivism is popular among Permissivists because it subsumes many versions of Permissivism. A paradigm example is Subjective Bayesianism, according to which rational credence depends both on one's evidence and on one's priors. It's a version of Standard-Permissivism because one's priors encode one's epistemic standards.Footnote 9 Besides, if we take one's epistemic standards to be also reflected in one's favorite way of weighing evidence, epistemic values, or theoretical virtues, then Standard-Permissivism can also subsume any version of Permissivism that takes these factors to be the extra-evidential determinant of rationality. For this reason, I will use Standard-Permissivism as a representative Inter-personal Permissivist position in this paper.
With the above clarifications, let's examine Standard-Permissivists’ response to AA. The response is that, in a known permissive case, you can keep your belief non-arbitrarily by claiming that your belief is uniquely singled out by your standard (Meacham Reference Meacham2014; Schoenfield Reference Schoenfield2014; Simpson Reference Simpson2017; Titelbaum and Kopec Reference Titelbaum, Kopec, Balcerak-Jackson and Balcerak-JacksonForthcoming).
However, this response still faces the arbitrariness worry (White Reference White2005: 452; Feldman Reference Feldman and Antony2007: 205 − 6). If Standard-Permissivism is true, then presumably you could learn that you are in a standard-permissive case, a case where a competing standard is as rational as yours. In such cases, it seems that you can only arbitrarily choose to keep your standard. Thus, it's irrational for you to keep your standard. So, we have:
The Standard Version of the Arbitrariness Argument (SAA)
P1*. When one learns that one's own standard and a competing standard are both rational, one can only arbitrarily choose one's own standard instead of the competing one.Footnote 10
P2*. For any situation, it's irrational to hold a standard in that situation if one can only arbitrarily choose the standard.
Therefore,
C1*. When one learns that one's own standard and a competing standard are both rational, it's irrational to hold one's standard.
Therefore,
C2*. Standard-Permissivism is unstable in some sense: one's standard cannot survive learning that one is in a standard-permissive case.
If the original AA is worrisome for Intra-personal Permissivists, then SAA should also be worrisome for Inter-personal Permissivists like Standard-Permissivists. Indeed, some Standard-Permissivists have anticipated this worry and have provided some responses (Schoenfield Reference Schoenfield2014: 201; Simpson Reference Simpson2017). Before we get into the responses, however, I want to make two clarifications.
First, proponents of AA apply it both to Permissivism about credence and to Permissivism about belief (including the radical kind that allows both believing p and believing not-p and the moderate kind that allows believing p and suspending judgment.) In this paper, I will primarily focus on the radical kind of the belief version. I will assume that my arguments – both the one about why attempts to avoid arbitrariness are unsuccessful and the one about why Permissivists can accept arbitrariness – can be generalized to the credence version. And if the radical kind of the belief version can withstand the arbitrariness challenge, presumably the moderate kind can withstand the challenge as well.
Second, what exactly is it to choose a standard, or a belief, arbitrarily? There is little precise characterization in the current literature.Footnote 11 Given White's (Reference White2005) frequent talk of how Permissivism leads to choosing a belief by coin-flipping, we might understand choosing a belief arbitrarily as choosing randomly. But this understanding would be too restrictive: if I choose a belief because I am most used to it, then I am not choosing randomly but intuitively still arbitrarily. Besides, it doesn't explain why premise P1 seems plausible: just because two beliefs are both rational, it doesn't follow that one can only choose one randomly.
So, I will use a broader characterization: a person chooses a belief or a standard arbitrarily just in case he chooses it according to an epistemically irrelevant property of the belief or the standard. Paradigmatic examples of epistemically relevant properties are those that epistemologists often work with: rational, true (or truth-conducive), justified, being knowledge (or being knowledge-conducive), sensitive, safe, reasonable, manifesting good epistemic dispositions, etc. And paradigmatic examples of epistemically irrelevant properties include: popular, familiar, sleep-conducive, and being the result of a random-selection process like coin-flipping.
Admittedly, the above characterization is not precise: some properties are neither clearly epistemically relevant nor clearly epistemically irrelevant. However, it accords with our intuition that choosing a belief arbitrarily is choosing according to some ‘irrelevant’ considerations, and it can explain why P1 and P2 in AA at least seem plausible: P2 seems plausible because, apparently, favoring a belief on epistemically irrelevant grounds seems problematic; P1 seems plausible because, although we can separate rationality from other epistemically relevant properties in evaluating other people's beliefs or standards, it's hard to do so from a first-personal deliberative point of view. For instance, for a person who is consciously deliberating on what to believe, it's odd to think ‘although it's rational for me to believe q, belief q is not true (or not reasonable, not sensitive, etc.).’ So, once you admit that two competing beliefs are both rational for you, it's odd for you to favor one of them by claiming that they are not equally good when it comes to other epistemically relevant properties; that is, if you do favor one of them, you can only do so by invoking its epistemically irrelevant properties. (Later, we will see that some Permissivists try to block arbitrariness by rejecting this intuitive thought.)
With the above clarifications, let's see Standard-Permissivists’ response to SAA.
3. Attempt One: standard-dependent reasons to favor our own standards
3.1. Truth-conduciveness as a standard-dependent reason
For convenience, I will use ‘a known standard-permissive case’ to refer to a case where one learns that one's own standard and a competing standard are both rational.
Suppose that there are two standards, R1 and R2, about what to think about certain proposition p in response to certain body of evidence E. And
R1 says: ‘if E, believe p’;
R2 says: ‘if E, believe not-p.’
You adopt R1, but then you learn that R2 is also rational. According to premise P1*, you can only favor your standard R1 by invoking its epistemically irrelevant properties. And we have explained why this premise is intuitively plausible: once you admit that R2 is also rational, it's odd for you to claim that R1 is better when it comes to other epistemically relevant properties.
Now, it's exactly this idea – that rationality necessarily goes together with other epistemically relevant properties from a first-personal deliberative point of view – that is denied by an influential response to SAA. According to Schoenfield (Reference Schoenfield2014: 200−2), even if you learn that R2 is also rational, you should still think that your own standard R1 is more truth-conducive, that is, following R1 is more likely to lead you to the truth. The reason is that you are committed to thinking that R1 is more truth-conducive, because thinking so is constitutive of the fact that you have R1 as your standard according to our definition of ‘possessing a standard.’ If so, you can appeal to the following belief as a non-arbitrary reason to favor your standard:
T = ‘Standard R1 is more truth conducive.’
It's important to note that your belief T is standard-dependent on R1 – Since believing T is constitutive of the fact that you have R1 as your standard, you can rationally believe T only if you actually hold R1 as your standard. If you were to hold R2, you would be committed constitutively to thinking that R2 is more truth-conducive.Footnote 12 More generally, we have:
A standard-dependent belief One's belief p is dependent on a standard S if and only if: one can rationally believe p only if one is actually holding S instead of a competing one as her own standard.
In the following sub-section, I argue that the standard-dependence of T on R1 makes the above response to SAA viciously circular. In fact, Schoenfield (Reference Schoenfield2014: 201) also admits that some circularity is involved, although she quickly argues that the circularity is not vicious, for reasons I will discuss below. So, in what follows, my primary task is to explain why the circularity is indeed vicious.
3.2. Why we cannot appeal to standard-dependent beliefs
First, note that for you to invoke T as a reason to favor your standard R1, it must be rational for you to believe T – after all, irrational beliefs cannot serve as reasons. For Standard-Permissivists, this means that the relevant local standard – the one that determines the rationality of your belief T – must be a rational standard, since a core claim of Standard-Permissivism is that the rationality of any belief is determined by the rationality of the relevant local standard. But what's the relevant local standard that determines the rationality of T? Answer: Your standard R1. This is because, although R1 doesn't directly tell you to believe T, it does so indirectly given that your belief T is standard-dependent on R1. So, if it were irrational for you to hold R1 as your standard, it would be irrational for you to believe T.
Now, is it rational for you to hold R1 as your standard? Admittedly, we can grant that, before you enter the known standard-permissive case – that is, before you learn that R2 is also rational – it's rational for you to hold R1. After all, we have stipulated that you learn that you are in a standard-permissive case; more exactly, suppose your learning happens at time t, our stipulation is that you learn that, at the moment right before t, it was rational to hold R1 and also rational to hold R2. Those advocates of SAA merely want to ask whether your standard survives as rational when you learn at t that you are in a standard-permissive case. So, it's unproblematic for the Standard-Permissivist to say that, before you enter the known standard-permissive case, it's rational for you to hold R1 as your standard and thus it's rational for you to believe T.
However, in order for you to invoke T as a reason to favor your standard when you enter the known standard-permissive case, it must still be rational for you to believe T. This in turn requires that R1 is still a rational standard for you to hold as you enter the case. But remember that the Standard-Permissivist's task in addressing SAA is exactly to explain why R1 is still a rational standard for you to hold when you enter the known standard-permissive case (i.e., to explain why R1 can survive as a rational standard in such a case). So, if you are a Standard-Permissivist who invokes T as a reason to favor your standard, your response to SAA would essentially look like this:
It's still rational for me to hold R1 as my standard when I am in the known standard-permissive case; therefore, my belief T is still rational when I am in the case; therefore, I have a non-arbitrary reason to favor my standard R1 in the case; therefore, it's still rational for me to hold R1 as my standard when I am in the case.
This reasoning is circular and the circularity is vicious – the conclusion is directly used as a premise. (And note that it's a kind of belief-circularity, not what is sometimes called ‘benign rule-circularity,’ the kind of circularity involved in implicitly using a rule to argue for the reliability of the rule. For classic discussion on the distinction, see Alston Reference Alston1986 and Bergmann Reference Bergmann2004). So, I conclude that Standard-Permissivists should not invoke standard-dependent beliefs like T in response to SAA.
Now, why does Schoenfield think that invoking standard-dependent beliefs wouldn't lead to viciously circular reasoning? She gives an anti-skepticism reason: Even when we are not in a known standard-permissive case, we can only invoke standard-dependent beliefs as reasons to favor our standards; if we are not allowed to do so, it would be irrational for us to hold our standards, no matter whether we are in a known standard-permissive case, and this will lead us to broad skepticism (Schoenfield Reference Schoenfield2014: 202).
To this way of defending the circularity, my response is two-fold. First, it's not true that we can only invoke standard-dependent beliefs to favor our own standards – in fact, in the next section, I will discuss possible standard-independent reasons to favor our own standards. To see the possibility, we need to recall that standards like R1 and R2 are merely local standards. In some cases, if I merely want to justify a particular local standard of mine, I can appeal to beliefs that are independent of the local standard in question, even if they might still be dependent on other local standards of mine, standards that are not called into question.
Second, even if we lack independent reason to favor a standard, we are not forced to accept skepticism. This is because, according to a general anti-skeptical strategy (Wright Reference Wright and McManus2004; Vavova Reference Vavova2018), the need to justify favoring our own standards arises only when we learn that a competing standard is equally good or better in some important respects, respects like rationality or reliability (which happens in known standard-permissive cases). We are entitled to regard our own standards as the uniquely best unless we have reason to think otherwise. So, for a standard like counter-induction, if we haven't learned that it's also rational, we can continue to use our inductive standard without having a positive reason to favor it.
In conclusion, on pain of viciously circular reasoning, Standard-Permissivists should not attempt to avoid arbitrariness by claiming that we have non-arbitrary standard-dependent reasons to favor our own standards. In light of this lesson, you might think that Standard-Permissivists should attempt to find a non-arbitrary standard-independent reason to favor our own standards. I turn to such attempts in the following section.
4. Attempt Two: standard-independent reasons to favor our own standards
Again, let's suppose that you hold standard R1 and then you learn that a competing standard, R2, is also rational. To say that you have a non-arbitrary standard-independent reason to favor your standard R1 is to say that there is certain property F that meets the following conditions:
C1: You can think that ‘only R1 has F (or that it has F to a greater degree than R2 does – this complication is omitted hereafter for simplicity).’
C2: ‘Only R has F’ is a non-arbitrary reason to favor R1; that is, F is an epistemically relevant property (and it's relevant in a positive way so that it's epistemically desirable).
C3: The belief ‘Only R1 has F’ is standard-independent of R1; that is, you could hold this belief even if your standard were R2.Footnote 13
Now, I won't argue for the ambitious claim that Standard-Permissivists’ attempt to search for such property F is doomed. Rather, I will only lay out three general difficulties faced by such an attempt.
The first difficulty is this: C1 and C3 would make C2 hard to meet. Given C1, F cannot be the property ‘rational’ or any epistemically desirable property that is closely connected to rationality (e.g., ‘reasonable’) – it's odd to claim ‘although both standards are rational, only mine is reasonable’; Given C3, F cannot be the property ‘truth-conducive’ or any epistemically desirable property that is closely connected to truth-conduciveness (e.g., ‘reliable’) – the belief ‘Only R1 is reliable’ would not be standard-independent of R1. But given the central place of truth and rationality in epistemology, it's hard to see how ‘only my standard has F’ can be a non-arbitrary reason to favor one's standard. Admittedly, in this paper, I haven't given a rigorous definition of ‘an arbitrary reason’ in a choice of beliefs or standards; I've only said that a reason is arbitrary just when it's epistemically irrelevant. But this rough characterization is enough to make us worry whether one can find a non-arbitrary reason to favor one's own standard – presumably, all paradigmatically epistemically relevant properties are closely connected to truth-conduciveness or rationality in one way or another. (For instance, ‘reliability,’ ‘sensitivity,’ and ‘safety’ are arguably all closely connected to truth-conduciveness; ‘reasonableness’ and ‘manifesting good epistemic dispositions’ are arguably closely connected to rationality.)
Here is a second difficulty: C2 and C3 taken together seems inconsistent with your being in a known standard-permissive case. More exactly, if C2 and C3 are satisfied, it's hard to see how you can ever learn that standard R2 is also rational. C3 says that you can believe ‘only R1 has F’ even if your standard were R2. But if F is an epistemically desirable property as C2 says, then, even if your standard were R 2, you can believe ‘R 1 is epistemically better than R2 given that only it has F.’ That is, R2 is self-undermining in some sense: a person holding R2 can believe that a particular competing standard is epistemically better. This casts doubt on whether R2 is really a rational standard.Footnote 14
Here is a third difficulty: even if we have found a property F that meets the three conditions, so that you can keep your standard non-arbitrarily by claiming that only it has F, it's difficult to generalize this way of avoiding arbitrariness to other known standard-permissive cases. For there is no guarantee that a condition similar to C1 will hold for every rational person in a known standard-permissive case. That is, there is no guarantee that any rational person in such a case can believe that it's his own standard, rather than the other standard, that has this epistemically desirable property F.
To see the above difficulty more clearly, consider a candidate of standard-independent reason that has been offered in the literature. It has been argued that one can favor one's own standard because it's more ‘catered to one's own cognitive abilities’ (Simpson Reference Simpson2017). For instance, suppose that you and your friend are two detectives working on the same criminal case. Your common evidence consists in a group of witness's testimonies and a group of car traces. In response to the evidence, you and your friend reach different conclusions as to whether the suspect is guilty. This is because you two follow different standards: your standard gives much less weight to witness's testimonies and more weight to car traces than your friend's standard does. You prefer your standard because, in general, you are not good at analyzing testimonies but very good at analyzing car traces; your friend's cognitive capacities might be the opposite. Given your cognitive limit, it's better for you not to give testimonial evidence substantial weight.
A problem with this way of avoiding arbitrariness is that it's not generalizable, even granting that ‘my standard is most catered to my cognitive abilities’ is a non-arbitrary reason for you to favor your own standard. This is because it's not the case that any person who possesses the standard ‘if E, believe p’ is committed (rationally or constitutively) to thinking that this standard is most catered to his cognitive abilities. (In contrast, as explained above, any person who possesses the standard ‘if E, believe p’ is constitutively committed to thinking that following ‘if E, believe p’ is most truth-conducive.) For example, we can imagine a person whose cognitive abilities are perfect in all respects, such that he wouldn't find one rational standard more catered to his abilities than another one. Given this possibility, ‘my standard is most catered to my cognitive abilities’ is not a reason available to all rational persons in a known standard-permissive case. It's especially worrisome that the reason is not available to cognitively perfect agents so that such agents couldn't avoid arbitrariness while ordinary agents ironically can.
In conclusion, the Permissivist's attempt to block arbitrariness in beliefs by invoking standards is unlikely to succeed, because it's difficult to block the arbitrariness in standards. If the Permissivist claims that we have a standard-dependent reason to favor our own standards, then he must accept certain viciously circular reasoning. If he claims that we have a standard-independent reason to favor our own standards, he will face the three difficulties discussed above. Since Standard-Permissivism is a representative Inter-personal Permissivist position, I conclude that to distinguish Inter-personal Permissivism from Intra-personal Permissivism cannot answer AA.
In the following section, I argue that such a distinction is not necessary to answer AA. This is because even the Intra-Personal Permissivist has a good response to AA: he doesn't need to block arbitrariness in beliefs, because the arbitrariness is acceptable in a known permissive case. That is, the Intra-Personal Permissivist can reject premise P2 in AA. (For simplicity, ‘Permissivism’ hereafter will refer to ‘Intra-Personal Permissivism’ – I will not be concerned with Inter-Personal Permissivism because, presumably, if P2 in AA is false then P2* in SAA is also false.)
5. A more promising response: denying P2
5.1. Denying P2: arbitrary beliefs compared to arbitrary actions
So far, P2 in AA has received little attention in the literature.Footnote 15 This is surprising, given that P2 is very strong: it says that, in any situation, a belief is irrational if it can only be chosen arbitrarily. Admittedly, it's intuitive that choosing a belief arbitrarily is normally problematic. Normally, we should only consider epistemically relevant properties. But it's much less intuitive that it's never rational to choose a belief arbitrarily, not even when we learn that the two attitudes we choose from are both rational.
In what follows, I argue that we should allow arbitrariness in beliefs in those known belief-permissive cases. My argument consists of two steps. First, in a known action-permissive case, it seems acceptable to choose an action arbitrarily; second, arbitrarily choosing a belief in a known belief-permissive case is analogous to arbitrarily choosing an action in a known action-permissive case, which means that if the latter is allowed then the former should also be allowed. In what follows, I will defend these two claims. Then I will answer some objections against this comparison, objections motivated by the idea that beliefs aim at the truths.
For starters, we need to clarify what it is to choose an action arbitrarily. We have defined choosing a belief arbitrarily as choosing according to some epistemically irrelevant properties of the belief. Similarly, we can define choosing an action arbitrarily as choosing according to some practically irrelevant properties of the action, where practically relevant properties are those that practical-rationality theorists typically work with: promoting desires-satisfaction, conforming to practical norms, being recommended by an ideal advisor, etc. Again, a typical example of choosing an action arbitrarily is to choose randomly by flipping a coin.
Can you sometimes choose an action arbitrarily? The answer is positive. It's true that, normally, choosing an action arbitrarily is problematic. Normally, we should choose actions according to their practically relevant properties. But it's much less intuitive that it's never acceptable to choose an action arbitrarily, not even when you learn that the two competing actions you are choosing from are both rational (hereafter, I call these cases ‘known action-permissive cases.’) For instance, when you learn that driving home and walking home are both rational because they are equally good means to get home, it seems that you can choose arbitrarily. Or, when you are sitting in a restaurant, knowing that ordering steak and ordering pizza are both rational because you like them equally well, it seems that you can choose arbitrarily.
So, it seems that you can choose an action arbitrarily in a known action-permissive case. But this gives us a reason to think that you can choose a belief arbitrarily in a known belief-permissive case. Suppose that I am a Permissivist because I think that belief p is rational just in case p best explains one's evidence and that there is no uniquely rational way to balance virtues of an explanation – e.g., I think that prioritizing simplicity over adequacy to data is as rational as prioritizing adequacy to data over simplicity. Then suppose that I encounter two competing theories that can explain my data equally well, one of them slightly simpler and the other slightly more adequate to data. In this case, why cannot I choose arbitrarily between the two theories? Arbitrarily choosing a theory is normally problematic because normally I don't know that the two theories I choose from explain my data equally well. But once I know that two theories explain my data equally well, it doesn't seem problematic to arbitrarily choose one.
Now, a natural way for the Uniquer to resist the above analogy between arbitrary belief and arbitrary action is to point out a crucial difference between belief and action: beliefs aim at the truth – when one forms a belief about p, one must aim that one believes that p if and only if p is true. And between two opposing attitudes (believing p and believing not-p), only one can satisfy this aim. In contrast, two opposing actions can both satisfy the aim of action. For example, driving home and walking home can satisfy my desire of getting home equally well.
This is a serious worry for the analogy. However, granting the truth-aim hypothesis about belief, it's not immediately clear how this hypothesis explains that beliefs can never be arbitrary even though actions sometimes can. So, in the next sub-section, I discuss three possible ways in which a Uniquer might use the truth-aim hypothesis to mount an objection against arbitrarily choosing a belief in a known permissive case, and I argue that the Permissivist can answer all of them satisfactorily.
5.2. Three objections
Objection 1: An arbitrary belief-forming process is not reliable. The first objection to arbitrarily choosing a belief is simple: Insofar as beliefs aim at the truths, rationality requires reliability. That is, a rational belief-forming process must be reliable. But arbitrarily choosing a belief is not a reliable belief-forming process. For instance, if you choose your belief by flipping a fair coin, there is only 50% chance that you would end up with a true belief: if p is true, you have 50% chance to get the truth because you have 50% chance to believe p; the same holds if not-p is true. Since arbitrarily choosing a belief is not a rational process, a belief chosen arbitrarily is irrational.
Response: This objection assumes that a rational belief-forming process must be reliable. This assumption is problematic because it begs the question against Permissivism. Here is why.
Suppose you have evidence E and you are deliberating whether to believe p or to believe not-p. If you believe p, it's natural to interpret your belief-forming process as ‘believing p-like propositions on the basis of E-like evidence’; and if you believe not-p, it's natural to interpret your belief-forming process as ‘believing not-p-like propositions on the basis of E-like evidence’ (where ‘not-p-like propositions’ are propositions of the same type with not-p). But these two processes cannot both be reliable. If the former process is reliable, it means that, in most cases in which E-like evidence obtains, a p-like proposition is true. If so, it cannot also be the case that, in most cases in which E-like evidence obtains, a not-p-like proposition is true. That is, it cannot be the case that the latter process is also reliable. Therefore, two competing processes with the same evidence as inputs cannot both be reliable. Thus, assuming that a rational belief-forming process must be reliable will imply that two competing beliefs in response to the same evidence couldn't be both rationally produced. If so, Permissivism should be rejected directly, even if there is no fancy argument from arbitrariness.Footnote 16
It's worth noting that the assumption that reliability is necessary for rationality presumably underlies White's ‘quick-and-dirty’ argument against Permissivism (Reference White2005: 447). The argument says that Permissivism must be false because, first, one's belief p is rational given one's total evidence E just in case E supports p, and second, it cannot be the case that E both supports p and supports not-p. White doesn't say much about the second premise. A reasonable guess is that this premise is motivated by the assumption that E supports p only if the objective probability of p given E is high – it's natural to think that the same evidence cannot make two competing beliefs both objectively likely to be true. If this guess is correct, then it's no wonder that Permissivists have quickly resisted this ‘quick-and-dirty’ argument, claiming that it's question-begging in assuming a notion of evidential support that the Permissivists won't accept (Kelly Reference Kelly, Steup, Turri and Sosa2014: 308). That is, when the Permissivist claims that the same evidence can support a pair of competing beliefs, what he is saying is not that the same evidence can make two competing beliefs both objectively likely to be true, but that the same evidence can make one belief likely to be true from some rational subjective perspective and make the other belief likely to be true from another rational subjective perspective.
Objection 2: You often have good reason to think that an arbitrary belief-forming process is unreliable. We have seen that posing an externalist criterion such as reliability on rationality would beg the question against Permissivism. Motivated by this thought, a Uniquer might revise his objection into a more internalist-friendly one. He can say that, an arbitrary belief-forming process is problematic not because it's in fact unreliable but because you have reason to think it's unreliable – after all, it's very easy to know that a belief-forming process like coin-flipping is unreliable. So, the objection is that, for a belief-forming process to be rational, you must lack reason to think that it's unreliable, and this necessary condition is not satisfied by an arbitrary belief-forming process.Footnote 17
Response: This objection is better than the last one, since posing ‘lacking a reason to think that the process is unreliable’ as a necessary condition of a rational process wouldn't beg the question against Permissivism. Presumably, two competing processes can both satisfy this necessary condition – you might lack reason to think that a process that produces belief p given E is unreliable and also lack reason to think that a process that produces belief not-p given E is unreliable, even though you know that one of the two processes is unreliable (compare: you might lack reason to think that belief p is false and also lack reason to think that belief not-p is false, even though you know that one of the two beliefs is false). However, the Permissivist has a good answer to this objection.
For starters, is ‘lacking reason to think that the process is unreliable’ a plausible necessary condition of a rational process? It's tempting to answer ‘yes,’ and it's natural to explain this answer by appealing to defeat: reason to think that a process is unreliable defeats the rationality of beliefs produced by the process. However, this explanation should be qualified: reason to regard a process as unreliable is not in itself a defeater; it's a defeater only coupled with a reason to think that an alternative process is more reliable. Suppose you originally rationally believe p. Then an epistemological oracle tells you ‘your belief-forming process is unreliable, but alternative belief-forming processes – including those that would produce suspending judgment in your situation – are no better.’ That is, the oracle merely tells you that your process is unreliable, without telling you that an alternative process is more reliable. It seems that, in this case, you can rationally keep your belief p – after all, you have no reason think that changing it to believing not-p or to suspending judgment would better serve your truth goal.
So, for reason to regard a process as unreliable to make the process irrational, you must have reason to think that an alternative process is more reliable. This point can explain why arbitrarily choosing a belief is normally problematic. For normally, arbitrarily choosing a belief means choosing a belief without examining one's evidence, and we do have reason to think that an alternative process – for example, forming beliefs by carefully examining evidence – is more reliable than arbitrarily choosing a belief without examining evidence. But the question is whether this is also true for arbitrary choices in a known permissive case. Suppose that, after carefully examining your evidence about p, you find that the evidence both makes belief p rational and makes belief not-p rational (because, say, the two competing theories can explain your data equally well). Do you have reason to think that there is a belief-forming process that is more reliable than the process of arbitrarily choosing a belief? Since the only non-arbitrary belief-forming process in this case would be one that produces suspending judgment, our question becomes this: Do you have reason to think that the process that produces suspending judgment in a known permissive case is more reliable than arbitrarily choosing a belief in the case?
Answering this question would be tricky, because it's unclear what makes a process of the form ‘if C(p), suspend judgment on p’ reliable. We cannot say that its reliability is determined by the truth ratio, given that its outputs don't have truth-values. However, since suspending judgment is a way of balancing the benefit of a true belief with the risk of a false belief, a natural way of defining the reliability of such a process is to appeal to epistemic decision theory – although suspending judgment doesn't have truth-values, it can figure in epistemic-utility ranking in which it's worse than a true belief but better than a false belief. So, we can say that this process is more reliable than a competing process if it's more epistemic-utility conducive, that is, if it tends to produce attitudes with greater epistemic utility. If so, then our question in the last paragraph becomes this: Do you have reason to think that the process that produces suspending judgment in a known permissive case is more epistemic-utility conducive than an arbitrary process is?
The Permissivist can answer this question negatively. He can say that, there are plausible measures of epistemic utility that imply that processes producing suspending judgment in a known permissive case are not more epistemic-utility conducive than an arbitrary process. Just consider a measure that values truths and falsehoods symmetrically – that is, a measure that says that a true belief is better than suspending judgment to the same degree in which suspending judgment is better than a false belief. For example, consider a measure U that has U(True Belief) = 1, U(Suspension) = 0, and U(False Belief) = −1. Then a process of the form ‘If C(p), suspend judgment on p’ would produce 0 utility on average in C-like situations, while a process of the form ‘if C(p), arbitrarily pick a belief about p’ would also produce 0 utility on average in C-like situations (for 50% times it produces 1 and for another 50% times it produces −1). If this symmetric measure is plausible,Footnote 18 the Permissivist can conclude that there is no reason to think that the process that produces suspending judgment in a known permissive case is more epistemic-utility conducive.
In conclusion, it's true that arbitrarily choosing a belief is normally problematic, because we normally have reason to think that an alternative process such as forming beliefs by examining evidence is more reliable. But in a known permissive case, you know that examining evidence can only take you so far, so that if you don't choose a belief arbitrarily, you will have to suspend judgment. And since you have no reason to think that suspending judgment better serves your epistemic goal (at least according to some plausible epistemic-utility measures), it's not irrational for you to choose a belief arbitrarily in a known permissive case.
Objection 3: An arbitrarily chosen belief allows you to make an instrumentally irrational plan about your future beliefs. So, the Uniquer's first two objections to arbitrarily choosing a belief in a known permissive case can be answered. I now turn to the third possible objection, which I think is the most serious one. A distinctive feature of this objection is that it specifically aims to explain why it's irrational to arbitrarily choose a belief in a known permissive case. Here is the basic idea.Footnote 19
First, given the truth-aim hypothesis, whenever you form a belief, you aim to form a true belief. But this implies that, once you believe that p, you should ‘bind’ yourself to continue to hold this belief at those future moments with no new evidence about p. That is, if you believe that p, then you shouldn't plan to change this belief at those future moments when you don't gain new evidence, on pain of instrumental irrationality. Given that you aim at the truth, it would be instrumentally irrational of you to plan to change your belief without new evidence: from your current point of view according to which p is true, you would be changing your belief into a false one.Footnote 20
But the problem is that, if arbitrarily choosing a belief in a known belief-permissive case were (epistemically) rational, you would be allowed to be in the above instrumentally irrational mental state. To see why, consider this example. Suppose that, at t0, I learn that it's both rational to believe that (r) it's going to rain and rational to believe not-r, and I arbitrarily choose the former. Now, suppose that, at a later moment t1, I face a choice about whether to keep belief r, and I haven't gained any new evidence about the weather. As I am deliberating at t1, I remember that I arbitrarily chose at t0 to believe r. Assuming that my goal at t1 is to hold a true belief about the weather, am I required to keep believing r? It seems not. After all, according to my memory, my earlier choice was made quite arbitrarily (say, by flipping a coin), so that remembering what this arbitrary choice was doesn't give me any evidence that the chosen belief is more likely to be true. If so, why let this earlier arbitrary choice constrain my current choice on whether to believe r? But this means that, at t0, I can foresee that my future belief-formation at t1 is not constrained by my current arbitrary belief-choice at t0. Then, at t0, I can plan to change my belief at t1 – If you foresee that an action is rational at a future moment with no new evidence, then it seems that you are allowed to plan to do it at that future moment.
In conclusion, here is the objection:
(Q1) Given the truth-aim hypothesis, it's instrumentally irrational for you to [believe p while planning to change this belief at a future moment with no new evidence].
(Q2) If arbitrarily choosing a belief in a known belief-permissive case were epistemically rational, you would be allowed to [believe p while planning to change this belief at a future moment with no new evidence].
Therefore,
(Q3) Arbitrarily choosing a belief in a known belief-permissive case is epistemically irrational – presumably, being epistemically rational won't lead one to being instrumentally irrational, when one's aim is the truth in question.Footnote 21
This argument can explain how actions are dissimilar with beliefs in a way such that the former can sometimes be arbitrary but the latter cannot – premise Q1 cannot be extended to action. To illustrate, suppose that I am deliberating on whether to order pizza or to order steak for today's lunch, knowing that I like the two kinds of meals equally well. And suppose that I arbitrarily pick pizza. Given my choice this time, is it instrumentally irrational if I plan to order steak next time when I face the same decision problem in the same situation? The answer is clearly negative. Knowing that I like steak equally well, it's instrumentally rational to order pizza this time while planning to order steak next time. So, unlike the case of belief, arbitrarily choosing an action in a known action-permissive case won't lead one to instrumental irrationality.
Response: First, in principle, a Permissivist could resist Q2 of the above argument by citing norms of diachronic coherence (such as conditionalization) – planning to change belief without new evidence is planning to violate such norms and thus planning to be irrational. However, I won't take this line of response here, because I agree with a lot of scholars that norms of diachronic coherence are not intrinsically normative; so, I don't think that such norms are binding in every case. Particularly, I find it hard to explain why such norms are binding in those cases where one knows that one's earlier beliefs are arbitrary.Footnote 22 But if there are cases where one can rationally be diachronically incoherent, it's hard to explain what's wrong with planning to be diachronically incoherent in those cases. One might think that a plausible explanation involves instrumental rationality – it seems instrumentally irrational to plan to change belief without new evidence. And this is exactly what premise Q1 says. Therefore, my following discussion primarily focuses on resisting Q1.Footnote 23
Although Q1 looks plausible, it should be rejected. It's not always instrumentally irrational to plan to change belief without new evidence, even though you now think that this plan would lead you to a false belief. More specifically, it's not instrumentally irrational to plan for a belief change in known permissive cases.
My reason for making the above exception to Q1 is inspired by the recent debate on ‘transformative experience.’ (See Paul Reference Paul2014, Reference Paul2015; Pettigrew Reference Pettigrew2015.) A crucial feature of a piece of transformative experience is that undergoing this experience will ‘transform’ you in the sense that it will change your values.Footnote 24 A typical example discussed in the literature is the experience of having a child. When you don't have a child, you might not like having a child because you value career more than family, but when you become a parent, you probably will value family more than career.
Due to this feature, it's an interesting question as to whether it's instrumentally rational to plan to undergo a transformative experience. Suppose that you don't have a child now, and currently you hate having a child, but you also know that you will value the opposite once you have a child. Is it rational for you to plan to have a child? At first thought, this plan seems irrational: given your current value that ‘career is more important than family,’ carrying out the plan would lead you to a ‘bad’ value. But this plan would not be irrational if it's viewed from the valuative point of view that you would adopt once you carry out the plan. From that future point of view, the current plan to have a child is a very wise one. And as Pettigrew (Reference Pettigrew2015) has argued, it's unclear why you should assess actions from your current point of view as opposed to that future point of view; in other words, it's unclear why instrumental rationality must always focus more on your current value-satisfaction than on your future value-satisfaction. So, if we want to accommodate our intuition that a decision to undergo transformative experience is sometimes instrumentally rational, or at least sometimes not instrumentally irrational, we should say that it's sometimes not instrumentally irrational to adopt a plan that would lead to a ‘bad’ value from your current valuative point of view.
Now, we can see why Q1 doesn't apply to known belief-permissive cases. When you have arbitrarily chosen a belief from two both rational ones, the decision about whether to keep this belief in the future is similar to a decision about whether to undergo a certain transformative experience; The only difference is that what's being transformed is not your value but your belief. So, just like it's not instrumentally irrational to plan to undergo a transformative experience, it's not instrumentally irrational to plan to change your belief. It's true that, if you rationally believe p, you would rationally believe that the best way to hold a true belief about p in the future is to keep belief p unless you get new evidence. But you hold this means-end belief only because you are viewing the world from your current point of view according to which p is true. Once you carry out the plan to change your belief, you will take an opposite point of view according to which not-p is true. And once you take that point of view (which you now think is a both rational point of view), you will rationally regard your original means-end belief as false: you would think that the best way to hold a true belief about p at the moment is to believe not-p. More importantly, just like instrumental rationality shouldn't always give priority to our current values in those cases involving a transformative experience, it shouldn't always give priority to our current beliefs in those known permissive cases. Especially, in those cases, you cannot decide to favor your current point of view over your future one by citing the reason ‘of course I should view the world and make plans according to my current belief and my current value – my future opposite belief is false and my future opposite value is bad!’ If you think so, you are already favoring your current point of view over your future one.
So, here is the lesson: Q1 doesn't apply in known permissive cases. Q1 assumes that we should always favor our current point of view in order to be instrumentally rational, but it's unclear why we should do so in known permissive cases.
This lesson also allows us to see what goes wrong in the following objection against my resistance of Q1. According to the objection, it's inappropriate to compare change of belief to change of value, because correct values can be ‘objectively pluralistic.’ That is, although relative to a certain valuative point of view, an opposite value would be bad, there is no independent, or objective, criterion by which we can say which value is ‘objectively better’; In contrast, two opposite beliefs can never be both ‘objectively correct’ in the sense of satisfying the independent criterion of truth, even if they can be equally good in respects like rationality.
This objection is misguided. It's true that two opposite beliefs can never be both objectively correct, so that if my future belief conflicts with my current one, one of the beliefs must be ‘incorrect.’ But this doesn't give me a reason to favor my current belief. After all, from an objective, independent, point of view – a perspective abstracting away from my current belief – I don't have reason to think that my current belief is more likely to be true than my future belief. And since the purpose of the above comparison between change of belief and change of value is merely to show that we don't need to always favor our current point of view, the comparison should not be resisted on the grounds that correct values are sometimes pluralistic but correct beliefs are not.
To conclude, arbitrarily choosing a belief in a known belief-permissive case doesn't lead to instrumental irrationality relative to the truth aim. So, the third objection against arbitrarily choosing a belief in a known permissive case also fails. Absent other possible objections, I conclude that the analogy between arbitrary beliefs and arbitrary actions is a good one: both are normally problematic, but both are not problematic when one knows that the two options one chooses from are both rational.
6. Conclusion
In this paper, I have argued that current attempts to avoid arbitrariness are unlikely to succeed. I have used Standard-Permissivism as a representative to explain why this is the case. Appealing to standard-dependent reasons to avoid arbitrariness would engage in viciously circular reasoning, and appealing to standard-independent reasons to avoid arbitrariness would face several difficulties. Then I have argued that a better response to AA is to simply accept the arbitrariness in known permissive cases. Once we find that arbitrariness is not problematic in the case of actions and that there is no good reason to reject the analogy on the grounds that beliefs aim at the truth, we can think that the arbitrariness resulted by Permissivism is actually not problematic. In conclusion, we should not dislike Permissivism on the grounds that it leads to arbitrariness.Footnote 25
