In a paper published while this one was under review at Erkenntnis, Christian Piller (2009) independently develops a version of the Wishful Thinking Puzzle and offers an alternative solution to it. Piller applies the Wishful Thinking Argument to the common view that our rational interest in truth amounts to desiring, of each proposition P that interests us, both P → B(P) and B(P) → P. He uses the Wishful Thinking Argument as a reductio ad absurdum of the claim that we desire B(P) → P—i.e., as a reductio of Epistemic Responsibility. On Piller’s proposal, our rational interest in truth does not include the desire to believe propositions that interest us only if they are true. Instead, it includes only the desire to believe such propositions if they are true, and the desire not to believe any contradictory pairs of propositions.
A full discussion of Piller’s proposal and his arguments would not be possible in this space. The purpose of this appendix is to point to important points of agreement between Piller and me, and to explain why, in my view, contrastivism about desire provides a better resolution of the Wishful Thinking Puzzle than Piller’s proposal does.
Piller agrees with me that problems arise if we suppose Conative Consequence and Epistemic Responsibility characterize our rational ideal or our rational interest in truth. He also agrees that the Wishful Thinking Argument shows what the problems are: the principles would commit us to desiring whatever we believe to be the case or, as I put it, to never believing a proposition while desiring its denial.
I disagree with Piller’s diagnosis of the problem. He thinks Epistemic Responsibility alone is to blame, and thus that our rational interest in truth does not include the desire to believe only truths. As I explain above (in n. 4 and n. 5), Piller is wrong to think the Wishful Thinking Argument unequivocally undermines Epistemic Responsibility rather than the combination of Epistemic Responsibility with the other principles, especially Conative Consequence. Additionally, Piller seems to think that the only way to formulate the desire to believe only truths is as the desire that B(P) → P (i.e., as Epistemic Responsibility). Contextual Epistemic Responsibility provides an alternative formulation, not in terms of B(P) → P: To desire to believe only truths is to be such that, relative to contexts fixed with respect to P but variable with respect to B(~P), one prefers the way the world would be if ~B(~P) to the way it would be if B(~P). That is, taking it for granted that P is true, one would prefer not to believe the denial of P.
To solve the problem the Wishful Thinking Argument poses, Piller proposes a revision of the traditional view that our interest in truth is a matter of desiring of each proposition P that interests us, (a) that we believe P if P is true (i.e., P → B(P)) and (b) that we believe P only if P is true (i.e., B(P) → P). On Piller’s proposal, we do desire, of each proposition that interests us, that we believe it if it is true. But, instead of desiring to believe each proposition only if it is true, we simply desire not to believe any contradictory pairs of propositions (pp. 206–208). Rather than desiring B(P) → P, Piller thinks we desire ~(B(P) & B(~P)).
Piller’s proposal also requires an adjustment to the way we express the claim that we want to believe propositions if they are true (pp. 207–208). As he points out, expressing it as the desire that P → B(P) will not work, because we can plausibly derive Epistemic Responsibility from D(P → B(P)) and D ~(B(P) & B(~P)), in which case the Wishful Thinking Argument would go through after all. The derivation exploits the fact that → allows for contraposition, though, so Piller stipulates that the conditional involved in our desire to believe propositions if they are true does not contrapose. He suggests Robert Stalnaker’s indicative conditional as a suitable, non-contraposing conditional for the purpose. Note that, without a suitable, non-contraposing conditional, Piller’s proposal is unable to solve the Wishful Thinking Puzzle at all.
On Piller’s proposal, when one believes a falsehood, something has gone wrong, but the trouble is not simply that one has an erroneous belief. Suppose one believes ~P, but P is the case. In such a case, either one does or one does not also believe that P. If one does not believe that P, then one fails to satisfy the desire to believe all truths. And if one does believe that P, then one fails to satisfy the desire not to believe any contradictions (p. 206).
In my view, there are at least three ways in which Piller’s proposal is inferior to the contrastivist approach discussed in this paper.
First, the contrastivist approach provides a unified solution to the Wishful Thinking Puzzle for both full desire all things considered and the decision theoretic conception of desire. Piller’s proposal does not address the decision theoretic version of the puzzle at all.
Second, Piller’s proposal counterintuitively treats false belief as only derivatively undesirable, rather than undesirable as such. Cases of false belief are bad either because they are cases of ignorance or because they are cases of contradiction, on his proposal. But ignorance and error are, intuitively, very different sorts of failings, and they have different consequences. For example, it is worse to make a losing bet than to make no bet at all. The contrastivist proposal, however, accommodates the intuition that being mistaken is undesirable as such, because it accommodates the idea that we want to avoid false belief as wellas the idea that we want to acquire true beliefs.
When we consider cases of false belief that are also cases of contradictory belief, we must consider why we think it is bad to believe contradictions. There are three likely answers. First, it might be bad to believe contradictions because one cannot believe a contradiction without believing something false. Second, it might be bad because of explosion: everything follows from a contradiction, in classical logic. Third, it might be bad because believing contradictory propositions involves some sort of incoherence in one’s beliefs.
The first answer would not help Piller. He wants to explain the badness of believing falsehoods by appeal to the badness of believing contradictions. The explanation is circular if we understand the badness of believing contradictions by appeal to the badness of believing falsehoods.
The second explanation might look promising, but it does not help Piller either. We must ask why explosion is bad. Why is it bad to be committed to believing absolutely everything? The most plausible answer, again, is that then we would be committed to believing a great deal that is false. If the badness of believing a contradiction depends on the badness of being committed to believing everything whatsoever, then it ultimately appears to depend on the badness of false belief. But in that case too, Piller’s explanation of why false beliefs are bad turns out to be circular.
The third explanation is unhelpful for similar reasons. What is wrong with incoherence in one’s beliefs? The most obvious answer is that having an incoherent set of beliefs involves one in having a number of false beliefs, and false beliefs are bad. There is another point to be made here, though. It is irrational to believe contradictory propositions, but neither ignorance nor error is necessarily irrational. A rational agent might want to believe all truths or not to believe anything false. When he fails, he simply fails to get what he wants, but his failure is not a failure of rationality. When an agent believes P and also ~P, though, it is a failure of rationality, not just a failure of the agent to satisfy a desire for consistency. Piller’s proposal seems to treat not believing contradictions as something we rationally desire, rather than as something that is part and parcel of rationality itself.
The contrastivist proposal avoids these problems. Neither contrastivism nor Piller’s proposal treats our interest in truth as including the desire that B(P) → P. Unlike Piller’s proposal, though, contrastivism maintains the idea that we want to believe only what is true, not what is false. On contrastivism, the desire to believe only truths is not a matter of desiring B(P) → P, but of preferring ~B(~P) to B(~P) relative to contexts fixed with respect to P but variable with respect to B(~P).
Third, Piller’s proposal fails to solve the problem unless it incorporates a contrastivist component at the outset. His proposal requires understanding our desire to believe a proposition if it is true in terms of a non-contraposing conditional such as Stalnaker’s indicative conditional, which I will symbolize with ‘>’. A desire expressed with such a conditional, I contend, is best construed along contrastivist lines.
On Stalnaker’s semantics, ‘P > Q’ is true if and only if the closest possible world where P is true is a world where Q is true. Worlds are “closer” or “further away” depending on how similar they are to the actual world, on the contextually relevant metric of similarity. When P is true, the “closest” P-world is the actual world. But when P is actually false, which P-world is closest depends on the contextually relevant metric of similarity, and different contexts have different metrics. As the context varies, the truth conditions of ‘P > Q’, and thus what proposition it expresses, also vary (Stalnaker 1999). Consequently, the state of desiring that P > Q is not a propositional attitude at all. It is a collection of attitudes toward different propositions in different contexts.
The trouble now is that there are contexts relative to which someone who cares about truth would not desire P > B(P) for every proposition that interests her. Let P be some proposition that is actually true although the available evidence overwhelmingly favors ~P, and let C be a context such that worlds that match the actual world in terms of what evidence is available always count as “closer” than worlds that do not. In the closest P-world, the evidence still overwhelmingly favors ~P, and so believing P would require going against the overwhelming evidence. The agent who cares about truth, though, would not want to go against overwhelming evidence. After all, following the evidence is the best way we have to approach our goal of believing all the truths. Additionally, it is far from clear that we can believe something, in the face of overwhelming evidence to the contrary, without the mediation of some false beliefs, such as the false belief that the evidence is not really overwhelmingly against what one believes after all. At the very least, we could construct a context such that either worlds where one follows the evidence count as closer than worlds where one does not, or worlds where one’s deviations from the evidence depend on false beliefs are closer than worlds in which they do not. Relative to such a context, someone who cares about truth would not desire P > B(P), for every P that interests her.
Contrastivism gives Piller a way around this problem. Rather than treating the desire to believe all truths as a desire whose content is a conditional, contrastivism can treat it as the preference for B(P) over ~B(P), relative to contexts that are fixed with respect to P but variable with respect to B(P). Such a move would keep the conjunction of the desire to believe all truths and the desire not to believe any contradictions from being susceptible to the Wishful Thinking Argument. If we are willing to make that move, though, then there is no longer any need to follow Piller in denying that our interest in truth includes an interest in believing only what is true. Contrastivism allows us to treat the desire to believe only what is true along the lines of Contextual Epistemic Responsibility, as a preference for ~B(~P) over B(~P) relative to contexts fixed with respect to P but variable with respect to B(~P). This is not the same as desiring B(P) → P, it is not susceptible to the Wishful Thinking Argument, and it honors the intuition that we desire not merely to avoid contradiction, but to believe only what is true.
For Piller’s proposal to work, we apparently need to gloss the desire to believe all truths (that interest us) along contrastivist lines. If we are free to do that, then we are also free to gloss the desire to believe only truths along contrastivist lines. When we do that, though, the Wishful Thinking Argument does not go through, and Piller’s criticisms of the desire to believe only truths (which depend on the Wishful Thinking Argument) do not apply. So there is no need to follow Piller in denying that our interest in truth includes the desire to believe only what is true.
None of the above three points constitutes a knock-down refutation of Piller’s proposal. Nevertheless, I think they do show that the contrastivist approach outlined here has significant advantages over Piller’s response to the Wishful Thinking Puzzle.