Let us consider the following case, an adapted version of a case from Ross and Schroeder (2014, p. 261), to see how the kind of commitment which, according to Joyce, is involved in small-world decisions can help clarify the sense in which the reasoning disposition involved in belief is defeasible:
Peanut Allergy: Sarah made three sandwiches—one containing tuna, one containing almond butter and one containing peanut butter—that she placed in the fridge. The sandwich containing tuna was placed on the left, the one containing peanut butter in the middle and the one containing almond butter on the right. Sarah’s nephew Algernon is visiting for lunch, and Sarah is aware that he has a severe peanut allergy. Sarah also believes that the sandwich containing peanut butter is placed in the middle of the fridge and her credence for this is .8 (the sandwiches were placed in the fridge hours ago and Sarah is well aware that her memory is not infallible). In addition, while Sarah believes that Algernon likes almond butter, she is just slightly more confident that he does not like tuna than the contrary (her credence that he does not like tuna is .55). Algernon asks Sarah for a sandwich. When Sarah goes to the fridge, she can tell, by visual inspection, which is the tuna sandwich, but she cannot tell which is the peanut butter sandwich and which is the almond butter sandwich.
Sarah believes that the sandwich placed on the right does not contain peanut. By the Reasoning Disposition Account, this means that she is defeasibly disposed to treat the proposition (p2): ‘the sandwich placed on the right does not contain peanut’ as true in the course of her deliberation. The way Sarah is disposed to deliberate in order to determine what she prefers to do when her nephew asks her for a sandwich can thus be represented by Table 2, where ‘the almond butter sandwich’ is, given what Sarah believes, the one placed on the right.
As Sarah is willing to please her nephew and is more confident that he does not like tuna than the contrary, if she deliberated according to Table 2, she should decide to give Algernon the sandwich that is placed on the right in the fridge. The expected utility of this option would be the highest. However, given what is at stake in this case and given that Sarah is not absolutely certain that the sandwich placed on the right does not contain peanut, it seems that Sarah should not deliberate according to Table 2. More specifically, it seems that she should not treat p2 as true in the context of her deliberation, although she believes that p2 is true.
The kind of commitment that is involved in small-world decisions according to Joyce can help to account more precisely for the situation described in the Peanut Allergy case. Although Sarah believes that p2 is true, assuming that she is aware of what is at stake—her nephew’s life—there is a risk that taking her uncertainty as to p2 into account would change her preferences. That is, given her degree of uncertainty as to p2 and what is at stake, there is a risk that deliberating according to Table 2 would not result in the same decision as the decision she would take if she deliberated according to Table 3.
We suggest that Sarah’s confidence that taking her uncertainty as to p2’s truth would change her preferences is precisely what can override, if sufficiently high, her disposition to treat p2 as true in her reasoning. In other words, her belief that p2 defeasibly disposes her to treat p2 as true in her reasoning, in the sense that she is disposed to reason in such a way insofar as her confidence that taking into account her uncertainty as to p2’s truth in the course of her deliberation would change her preferences is not higher than a given threshold.Footnote 5 In addition, in case Sarah’s disposition is overridden, she is presented with a choice: either taking her uncertainty as to p2’s truth into account in her deliberation or not taking it into account. This is so because what overrides Sarah’s disposition to treat p2 as true in her reasoning is her confidence—but not her absolute certainty—that her preferences would not be the same if her uncertainty as to p2’s truth was taken into account. A context in which an agent’s disposition to treat p as true in her reasoning is overridden thus corresponds to a specific choice situation. In such a context, the agent has a choice between taking the risk of leaving some information critical to her decision out of her deliberation—that is, the possibility that ¬p—or taking this information into account in spite of her belief that p is true. In the next section, we will discuss in more details the factors that are relevant to determine an agent’s confidence that her preferences would be overturned if her uncertainty as to p’s truth was taken into account in the course of her deliberation. For now, let us focus on another question raised by this proposal, which concerns the justificatory status of the agent’s belief in contexts in which it is not rational for the agent to treat the believed proposition as true in the course of her deliberation.Footnote 6
According to Ross and Schroeder (2014, p. 271), an agent is not justified in occurrently believing that p in contexts in which it is not rational for her to treat p as true in the course of a deliberation, although she believes dispositionally that p. They (2014, p. 271) rely on the following principle to motivate this claim:
Justification principle: If having attitude a essentially involves being disposed to ϕ under circumstance C, then an agent A is justified to occurrently have attitude a in C only if it is rationally permissible for A to ϕ in C.
It follows straightforwardly from the Justification principle and the Reasoning Disposition Account that if an agent believes that p and is not rationally permitted to treat p as true in a given choice situation, then she is not justified, in that situation, to occurrently believe that p. One advantage of the conjunction of these principles is that it allows offering an explanation of the phenomenon of pragmatic encroachment on knowledge. However, as outlined by Locke (2014), the conjunction of the Justification principle and the Reasoning Disposition Account also leads to seriously counterintuitive consequences. Locke considers the following case:
Liz’s Bet: One day Liz is offered a bet on whether she was born in England. In fact, Liz was born in England, and her reasons for believing so are just like anyone else’s: her parents told her she was born in England, her aunts and uncles tell stories about going to see her in the hospital, she has never had a problem when dealing with the government, and so on. However, the terms of the bet are as follows: if Liz was born in England, Liz gains £1; if Liz was not born in England, Liz is tortured for the next 30 years. (Locke 2014, p. 39)
As pointed out by Locke, given the evidence Liz has in this case, it is very hard to deny that she is justified in occurrently believing that she was born in England—that is, in consciously believing that she was born in England in this particular context. On the other hand, given the bet that is offered to her, it seems that she should not act on her belief and that she should not treat the proposition ‘I was born in England’ as true when deciding whether or not to take the bet. According to Locke, the Reasoning Disposition Account goes wrong in requiring that Liz has any disposition to treat the proposition ‘I was born in England’ as true in this situation. By his own account, it is possible for Liz to believe this proposition and yet not to be disposed to treat it as true in this particular choice situation. But, as outlined in the previous section, there are serious reasons to be dissatisfied with Locke’s analysis of Liz’s Bet case.
We believe that the problem raised by Liz’s Bet case results rather from the Justification principle than from the Reasoning Disposition Account. Let us first note that the Justification principle is not as intuitive as it seems to be when we consider more carefully the distinction that can be made between dispositional belief and occurrent belief. Ross and Schroeder (2014, p. 272) note that given this principle, an agent can be justified in having the dispositional attitude a in C without being justified in occurrently having this attitude in C. But this situation appears somewhat odd. For when the reasons one has for having (dispositionally) the belief that p justify having this belief in C, why are they not also sufficient to justify occurrently believing that p? In addition to the Justification principle, a principled explanation of why a subject needs more reasons for being justified in having the occurrent belief that p than for being justified in having the dispositional belief that p is required.
Setting aside the question of the initial plausibility of the Justification principle, let us now focus on what we take to be the main problem of this principle. According to the Justification principle, the fact that it is not rationally permissible for an agent to treat p as true in the course of her deliberation means that the agent, in the choice situation she is in, lacks justification for occurrently believing that p. But the fact that it is not rationally permissible for an agent to treat p as true in the course of a deliberation can actually mean many different things. It could mean, for instance, that the agent should not suppose that p in the course of her deliberation. When one supposes that p, one treats p as true in the course of a deliberation. It could also mean that the agent should not acceptp in the course of her deliberation. Now, the fact that an agent should not suppose that p in a given context does not necessarily mean that she lacks justification for occurrently believing p in that context. This point is even clearer in the case of acceptance which is more closely related to the way one acts than the case of supposition. The attitude of acceptance is generally regarded as being under our direct voluntary control and as involving treating a proposition as true in the sense of being ready to act as if it was true. Consider the following definition from Cohen:Footnote 7
To accept that p is to have or adopt a policy of deeming, positing, or postulating that p—i.e. of including that proposition or rule among one’s premises for deciding what to do or think in a particular context, whether or not one feels it to be true that p. (Cohen 1992, p.4)
Cohen’s definition makes it clear that (i) acceptance is an attitude adopted in the context of a particular deliberation, (ii) that this attitude consists in treating a given proposition as true in one’s reasoning, and, finally, (iii) that one can decide voluntarily to accept or not to accept a given proposition, regardless of what one believes (or feels to be true in Cohen’s terminology).
With this characterization in mind, it is plausible that the fact that an agent should not decide to accept a given proposition in the course of a particular deliberation does not necessarily mean that she lacks justification for occurrently believing this proposition. Take the case of a lawyer discussed by Cohen (1992, p. 20) who has every reason to believe that her client is guilty of the crime she is accused of. As she has to defend her client, it is rational for the lawyer to deliberate as if her client was not guilty. It is rational for her not to accept the proposition ‘my client is guilty’ when she deliberates about the case and makes decisions in court in spite of the reasons she has for believing it. But the fact that the lawyer should choose not to accept the proposition ‘my client is guilty’ does not mean that she lacks justification for believing it occurrently or dispositionally.
If it is not rationally permissible for an agent to treat p as true in the course of a deliberation in the sense that she should not accept p in this context, one cannot thereby conclude that she lacks justification for occurrently believing that p. It might be the case that, in the particular choice situation the agent is in, she should not accept that p despite her disposition to treat p as true in this choice situation and despite the justification she has for believing—both dispositionally and occurrently—that p. The Justification principle is problematic because it abstracts from the various reasons for which, in a particular choice situation, it may not be permissible for an agent to treat p as true. Yet, in Liz’s Bet case for instance, it is quite plausible that Liz should not treat the proposition ‘I was born in England’ as true in the sense that she should decide not to accept this proposition in spite of her disposition to treat it as true, and in spite of the justification she has for believing it. The fact that Liz should not treat this proposition as true in that sense does not mean that she lacks justification for occurrently believing it. As Locke puts it, if anyone has justification for believing anything, then Liz has justification for believing that she was born in England. What the Justification principle overlooks is the fact that although the attitude of belief essentially involves a disposition to treat p as true, the fact that it is not rationally permissible for an agent to treat p as true might be related to facts that have nothing to do with the justification the agent has for believing that p, either dispositionally or occurrently. It might be related to facts that constitute reasons to decide not to treat p as true in spite of the justification the agent has for believing that p.