I've changed my mind -- there is a substantive difference between believing vs. getting yourself to believe. I had mistakenly thought of it as being like trying to distinguish jumping from getting yourself to jump. But that's the wrong analogy. When you take a magic pill to get yourself to believe, you are treating yourself as an alien object. This believing is not something that you do; it is something that is done to you. No intentional act of believing occurs. Instead, there is an intentional act of getting oneself to believe (by taking the magic pill), which then causes one to have the belief. This effect is a consequence of your action, rather than an action itself.
So, because you are acting upon yourself as an alien object, the appropriate analogy is thus NOT jumping vs. getting to jump (both of which are intentional actions), but rather, bleeding vs. getting yourself to bleed. Bleeding is not an intentional action. It's something that happens to you, rather than something you do. This is to describe it from the object's perspective. How about the actor? Here we should say that bleeding is a state of affairs: it's something that can be brought about, but not something that can be done. You cannot intend to bleed. That would be a category error, as you can only intend to perform actions, and bleeding is not an action. Instead, all you can do is bring it about that you (or someone else) bleed. So this 'bringing it about' is all that you can intend, too.
Can there be "reasons for bleeding"? Well, not if by 'reason' we mean 'reason for action', since bleeding is not an action. Instead, there might be reasons for getting yourself to bleed (i.e. bringing it about that you bleed). One might try to say there can be reasons for a state of affairs, defined as value-indicators. If R indicates that a state of affairs would be valuable, then it counts in favour of the state of affairs, and thus is (by definition) a 'reason' for it. But this seems a pretty clumsy definition. What it's really pointing to are reasons for bringing about the state of affairs. So they're really reasons for action, deep down. [Update: this isn't quite right. It might be impossible, or at least costly, to bring about a state of affairs that is nonetheless good in itself. I should have spoken instead of reasons for desiring the good state of affairs.]
And all this applies just as well to self-imposed beliefs. Suppose you decide that a belief P would be good to have, but you lack evidence for it and so cannot believe voluntarily. You must take a magic pill that will cause you to gain the belief. Here you never engage in any intentional action that could be described as "believing P". When you perform an intentional action, you perform it for a reason. But it is impossible to believe for practical reasons. So when you act on practical reasons, whatever your action is, it is not an intentional act of believing.
None of this is to deny the obvious fact that practical reasons can count in favour of beliefs in some sense. But we can clarify just what this 'sense' is. Practical considerations can count in favour of a belief qua state of affairs. That is, you can have practical reasons to bring this state of affairs about; to bring it about that you have this belief. But there can be no practical reasons for the intentional act of believing.
Well, maybe. This isn't so clear, actually. It is impossible to believe for practical reasons. But maybe there can be inaccessible reasons, i.e. reasons that cannot possibly be acted upon. Perhaps practical considerations are a reason for _you to adopt the belief for other reasons_. This is a little convoluted though, and I guess many would want to deny the existence of inaccessible reasons. But there's a possible argument against evidentialism here.
Though the whole debate still strikes me as fairly insubstantial. Presumably everyone agrees that moral considerations etc. could mean that you ought to take the pill and obtain a false belief no matter the evidence. But if evidentialists grant this much, then their claims about the "ethics of belief" can't be very strong. The disagreement isn't about what ought to be done. It's merely about how we should describe it. (The guy should swallow the pill. But maybe we should say this is an act of "getting to believe" rather than "believing", or whatever.) Meh.
Still, as far as classifications go, there does seem one interesting upshot from all of this: Musgrave's distinction doesn't go far enough. Not only should we distinguish "believing" vs. the "thing believed", but on the other side we need to distinguish the act of believing vs. the state of affairs wherein one has the belief.
So now we have three types of reasons relating to beliefs. We have truth-indicative reasons, or reasons for the thing believed. We have other (non-truth-indicative) epistemic reasons for belief. And we have practical reasons to bring a belief about.
Things are clear at last. And just three days before my essay is due, too. Heh, could've been worse, I suppose...