The Illusion of Discretion

Discretion is the thesis that if one has sufficient evidence for p, it is epistemically permissible for one either to believe p or be agnostic on whether p. This paper argues that Discretion is not as obvious as many assume—indeed, that it is false. §1 observes that Discretion is not supported by the fact that we lack positive epistemic duties to take doxastic stances on propositions. §2 considers a different kind of example often taken to support Discretion and argues that it provides no more support than was provided by the case against positive epistemic duties. §3 gives a direct argument against Discretion. The upshot of these sections is that if agnosticism is an attitude in its own right rather than a mere form of non-belief (as recent writers argue), we should reject Discretion. With that conclusion in mind, §4 gives an error theory to explain away the appeal of Discretion. §5 explores the implications of rejecting Discretion, showing among other things how it undermines a major case for the view that practical reasons can be motivating reasons for belief. 1 What Discretion Is and Why It Is Not a Datum Traditional epistemologists have been interested in the epistemic norms governing three coarse-grained doxastic attitudes: belief, disbelief, and agnosticism. If we regard disbelief in p as belief in the negation of p,1 we can reduce the trichotomy to a dichotomy: belief and agnosticism. Focusing on this dichotomy, we can ask: (Q) Are there cases where it is epistemically permissible for one to adopt one or the other of these attitudes? Or is one always singled out as epistemically permitted? While rarely defended directly, the following answer finds considerable sympathy: Discretion: Even if one has sufficient epistemic reason to think that p, it can be epistemically permissible for one either to believe p or be agnostic on whether p.2 Some might insist that Discretion is an unassailable datum—an upshot of the apparent fact that there is no positive epistemic duty to take doxastic stances on proposition. The evidence supports an infinite number of uninteresting conclusions whenever it supports 1This is widely assumed, though Schroeder (2008: 102–104) provides compelling reasons for thinking that there is something worth calling “disbelief” that isn’t thus reducible. But his reasons don’t affect these introductory points. We will later see that they support my position. 2I borrow this term from McHugh (forthcoming), though he uses the term to describe the psychological claim that sufficient evidence “psychologically allows” belief and agnosticism. He also accepts the corresponding normative claim. That is not surprising. It is natural to think if both claims were true, the latter would help to explain the former: it would be because we understand that the evidence permits belief and agnosticism that we are psychologically able to respond to sufficient evidence with either belief or agnosticism. I will assume this connection later in discussing implications for doxastic voluntarism.