Raz on Standard and Non-Standard Reasons

In the new Oxford anthology Reasons for Action, Joseph Raz has a really interesting article on the difference between epistemic reasons and reasons for action (and it's actually comprehensible. I know. Shocking). Here's the broad claim: epistemic reasons diverge from reasons for actions at two points-- first, while values that may contribute or constitute reasons for action are plural, truth-directedness, not value, drives reasons to believe/epistemic reasons. Second, we can follow non-standard reasons for action but not epistemic reasons. That is, we can act in conformity with a reason even if that is not the reason we are acting; we cannot believe in conformity with a reason for belief without taking it to be evidence of the truth of the proposition we're believing.

I think this is an awesome project. But arguments will be hard to make. Apparently there's been a little skirmish between Setiya and Raz over whether epistemic reasons can be non-standard, Setiya falling on the side that argues that there can. The example he uses is this:

"Imagine that someone promises Jake a large sum of money if he believes that P. By all accounts this is a reason to have that belief. I claim that it is a non-standard reason for it cannot be followed [directly]. But suppose Jake does not think that. Suppose that he is philosophically minded ... and believes that the promise is an ordinary, standard, reason for believing that P, and that as a result he comes to believe that P." (Raz "On Adaptive Reasons," 18)

This strikes me as a weird and not very plausible example. Are we really capable of talking ourselves into beliefs and genuinely holding them without giving ourselves some (believable) evidence that the propositions are true? I think you have to either be talking about a truth that comes into being by the person's say-so that P (like maybe God's, if you think God's word is a truthmaker) or have a Quine-ish or disquotational theory of truth. Anyway, a case worth thinking about.