While we may have epistemic reasons for a particular belief and a practical reason not to have it... The outcome of "conflict" between adaptive and practical reasons is not, as in genuine conflict between practical reasons or between epistemic ones, that the better reason prevails. They are not in competition, and reasons of neither kind can be better than reasons of the other. Rather, adaptive reasons, being the standard reasons for belief or for having emotions, prevail. Practical reasons, being non-standard, can 'win' only by stealth. (p.23)
He adds in a footnote: "Note though that standard and non-standard reasons for action, both being practical, do conflict in a straightforward way. And the same is true of reasons for intentions."
So how are we to assess a situation where your standard and non-standard reasons for action or intention conflict? Take, for example, Kavka's toxin puzzle. Clearly the thing to do is to bring it about that you have other (i.e. standard) reasons to intend to drink the toxin -- to keep a promise or to avoid some self-inflicted punishment, perhaps. But in the absence of those new reasons, i.e. reasons you can rationally follow, wouldn't it be irrational for you to form the intention? (It'd be fortunate, for sure, but that merely shows that sometimes it's fortunate to be irrational, right?) In what sense is this 'conflict' any different from the belief case?