A SCRIPTIONS OF IRRATIONALITY typically constitute a form of criticism, while ascriptions of rationality are a form of praise. (1) More specifically charges of irrationality involve personal criticism, in which the agent is negatively evaluated for having responded in certain ways. (2) It is often thought that being criticizable is evidence that one has done something one ought not to do-something one had decisive normative reasons not to do. (3) Assuming that this is so, the fact that charges of irrationality constitute a form of criticism suggests that there is a close connection between rationality and what we have reasons to do. This provides motivation for a normative, reasons-based account of rationality. According to this type of account, being rational is a matter of doing what one has most normative reasons to do. (5)
Reasons-based accounts of rationality face, however, an immediate problem. It seems that an agent can behave rationally despite relying on false considerations (for example, in situations in which the agent is guided by convincing but ultimately misleading evidence). Yet, plausibly, only facts, and not false beliefs, can be objective normative reasons--this sort of view finds widespread support in the literature. (6) So, on the face of it, there are cases in which rationality does not amount to responding to factual normative reasons, but rather to merely apparent reasons. In this way, several authors have developed theories of rationality as sensitivity to those considerations that appear as normative reasons to the agent. (7) Given that merely apparent reasons have no normative force, these theories fail to make rationality a genuinely normative notion. (8)
Recent proposals seem to offer a way out for factualist reasons-based accounts of rationality, in the face of the challenge posed by rational actions relying on false beliefs. (9) The way out would be to argue that there are always factual, normative reasons to which the behavior of a rational agent is sensitive and that rationalize such behavior. In particular, in cases in which a rational agent seems to be guided by false beliefs, the reasons that rationalize her behavior would be constituted by facts about how things appear to her--that is, by reasons about appearances, rather than by apparent reasons.
This paper has two main goals, one critical and one constructive. The critical goal is to show that these sophisticated reasons-based accounts of rationality are still unsatisfactory, at least when combined with a factualist conception of normative reasons. The constructive goal of the paper is to argue that theories of rationality as responding to apparent reasons manage to do justice to the main intuitions seemingly supporting reasons-based accounts of rationality, in particular the connection between ascriptions of rationality and criticism and praise. The key move will be to abandon the view that being criticizable requires acting against one's reasons, while being praiseworthy involves doing what one's reasons actually recommend.
THE REASONS-BASED ACCOUNT OF RATIONALITY
An intuitively attractive idea is that being rational is a matter of being properly sensitive to the reasons one has. (10) This idea leads to an account of rationality along the following lines:
Reasons-Based Account (RB) : A response [phi] by an agent S is rational if and only if [phi] is, on balance, sufficiently supported by the reasons possessed by S. (11) The response [phi] may be the performance of an intentional action or the adoption of some reason-sensitive attitude, such as believing or intending. Moreover, the reasons relevant for RB are normative reasons--that is, considerations that count in favor of the response in question. I will take it that such normative reasons are facts. These assumptions are shared by the advocates of RB I will be concerned with. (12)
I will further assume that reason possession is a perspective-sensitive relation involving some form of epistemic access to the fact constituting the reason, although I will not specify whether this access amounts to knowing, being in a position to know, believing, or something else. All the defenders of RB I am going to discuss agree that only possessed reasons determine what is rational for an agent to do (and also what she ought to do). One way of motivating this restriction is by offering an attractive account of three envelope cases, and of structurally analogous examples such as Parfit's mine-shaft scenario or Jackson's doctor cases. (13) In these examples, it is rational for the agent to [phi] despite her being aware that there exist facts beyond her epistemic ken that constitute decisive reasons to do something else instead. The type of view I am considering here deals with these cases easily, insofar as the relevant facts beyond the agent's epistemic ken do not figure among her possessed reasons, and thereby do not contribute to fixing what is rational for her to do. (14)
I will take it that the reasons possessed by an agent sufficiently support a certain response when the combined possessed reasons in favor of that response are at least as strong as the combined possessed reasons against it. Finally, if one wanted to specify what it is for an agent S to be rational in producing a response [phi], it would be necessary to add the further condition that S's response is properly based on the sufficiently supporting possessed reasons. This condition sees to it that S produces her response for the reasons that rationalize producing it.
On the face of it, RB plus the standard assumption that normative reasons are facts leads to bad results in cases in which seemingly rational agents are guided by false considerations. (15) There is the strong intuition that an agent can be rational despite relying on false beliefs, as long as from her point of view there appeared to be good reasons to behave as she did. (16) If one has extremely convincing evidence that p, then it seems rational to take p as a reason to produce a certain response, even if, unbeknownst to one, the evidence happens to be misleading and the belief that p is false. Think, for instance, of the agent who drinks from a glass of petrol, thinking that it contains gin and tonic. (17) More radically, it is very plausible that an agent deceived by a Cartesian demon, or a brain in a vat, can be as rational as her lucky counterpart in an ordinary environment. (18) In general, it seems that an agent will be rational if she takes proper heed of the considerations that, from her perspective, appear as good reasons, regardless of whether such considerations turn out not to be normative reasons after all. A deceived agent that is internally identical to her rational counterpart should also be regarded as rational (the deceived agent seems to be as blameless and as competent as her non-deceived counterpart).
Cases of deceived rational agents (I will call them "bad cases") put pressure on RB, since in these cases the agents' responses do not seem to be supported by normative, factual reasons and, nonetheless, they count as rational. Recent proposals by Kiesewetter and Lord suggest a reply to this problem on behalf of RB. (19) I will call this reply the appearance reasons view of rationality. According to this view, whenever an agent behaves rationally, there will be factual, normative reasons that rationalize her behavior. In the bad cases in which the agent relies on false beliefs, the facts that make her response rational will be, in general, facts about appearances--the fact that things appear to be to the agent as she believes them to be, or more broadly, facts that appear to support the agent's belief. (20) On this view, ifp is a reason for a response in a good case, then the fact that it appears to the agent that p is a reason in the corresponding bad case. Thus, the central thesis of the appearance reasons view is that appearance facts can be normative reasons for the agent's response: such facts count in favor of the response; they recommend it. (21)
Appearance Reasons View. If p is a reason to [phi] possessed by an agent S, then the fact that it appears that p is a reason to [phi] possessed by an agent S* in a situation subjectively indistinguishable from that of S. So, in the case of the glass with petrol, the fact that the glass appears to contain gin and tonic (the agent is in a bar and has ordered a gin and tonic) is a normative reason for the agent to drink from it, given that she is thirsty and fancies a gin and tonic. (22)
At this juncture, one may ask whether appearance reasons in the bad case are as strong as the reasons given by the appeared fact p in the good case. Kiesewetter thinks that they are, whereas Lord takes them to be weaker. These answers lead to different versions of the appearance reasons view of rationality. I will argue that both versions of the view are problematic.
DO WEAKER APPEARANCE REASONS RATIONALIZE?
Let us examine first views that assume that in bad cases the reasons favoring the response (call them pro-reasons) are weaker than in good cases. This assumption seems plausible. After all, these reasons are constituted by facts about appearances that p (where p is a reason in favor of the response), rather than by the fact that p, and, on the face of it, if these appearance facts give reasons for the response, it is only to the extent that they reliably indicate that p is the case. (23 )I further discuss this issue in section 3; for the time being I will just grant that the assumption is right.
Lord concedes that appearance reasons are weaker than the reasons that would be given by the appeared facts. However, he suggests that, on balance, the reasons possessed by the agent in a bad case support her response as strongly as in the corresponding good case, at least in an important sense. (24) This is so because it is not only the pro-reasons that are weaker in the bad case. According...