I suppose I am a limited doxastic voluntarist: though I haven't thought about this question in much depth my tendency is to say that there are some beliefs over the formation of which I have direct voluntary control. That is, there are some believable contents — call them propositions — that I can bring myself to believe at will, others that I can bring myself to disbelieve at will, and still others about which I can suspend judgment, thereby enacting something like the epoché (ἐποχή) of such ancient skeptics as Sextus Empiricus.
Note that the issue concerns the formation of beliefs, not their maintenance, and note the contrast between direct and indirect formation of beliefs. Roughly, I form a belief directly by just forming it, not by doing something else as a means to forming it. Suppose the year is 1950 and you are a young person, sincere and idealistic, eager to consecrate your life to some cause higher than a bourgeois existence of consumption in suburbia. You have vibrant stimulating friends who are members of the CPUSA. They tell you that the Revolution is right around the corner. You don't believe it, but you want to believe it. So you go to their meetings, accept Party discipline, toe the Party line, and soon you too believe that the Revolution is right around the corner. In this example, the formation of belief is indirect. You do various things (go to the meetings, repeat the formulas, toe the line, etc.) in order to acquire the belief. But then in 1956 you learn of Krushchev's denunciation of Stalin and your belief in the glorious Revolution and its imminence suddenly collapses to be replaced by an opposing belief. The formation of the opposing belief is direct.
A correspondent supplies an example of the third case, that of suspending belief:
Suspending belief. Sometimes in the face of good or strong evidence that p, I refuse to believe that p or again that not-p. I suspend any opinion on p.
This has always been my attitude on OJ and the murders he was charged with. Recently I talked with someone who had been teaching OJ knife-fighting in conjunction with a Commando-style TV show that never got launched. His evidence was excluded from the trial. Even in the face of this new evidence that OJ was competent with a knife, I do not form a opinion as to whether or not OJ killed his wife. (This is close to the classical skeptical epoche, except I do not bother to inquire and try to build a counterbalancing case for the opposite belief. Pyrrhonian skepticism says I always can build such a case and the result will be spontaneous cessation of belief.)
In both these cases [I omitted my correspondent's first example] I think it's clear that what I believe (or don't believe ) is a function of what I will or wish to believe, trumping the evidence on hand and any reasonable induction therefrom. Hopefully, in both case it is also a principled refusal on my part to buy into beliefs that condemn other people. To believe my wife a poisoner or OJ a killer, I require evidence several parsecs beyond a reasonable doubt. You can say that that standard is too liberal [too stringent?], but I can choose to live (and die) by it and it is for several grounds an attractive ethics of belief.
This addresses my concern about the possibility of an ethics of belief. My correspondent suspends judgment, holds no opinion, on the guilt or innocence (as charged) of O.J. Simpson. By suspending judgment, he deliberately impedes or rather prevents the formation of two beliefs, the belief that O.J is guilty and the belief that O.J. is not guilty. I find that I have the same power of doxastic abstention, except that in this particular case I assent to the proposition that O. J. is guilty since I judge the evidence that he is guilty as charged to be overwhelming, and the notion that 'racism' played any part in this case utterly absurd. (My ethics of belief is perhaps less stringent than my correspondent; but we both have an ethics of belief.)
Our question does not concern the content of an ethics of belief, but the very possibility of one whatever its content. Since 'ought' implies 'can,' if I ought to withhold judgment in some cases — and surely there are some cases in which I ought to withhold judgment — then I can withhold judgment. I have the power to withhold judgement; hence my epoche is voluntary. So here seems to be a case in which believing/disbelieving is under the direct control of the will: I decide to neither believe nor disbelieve. And from this it follows that the application of deontological categories is legitimate. "You ought not believe that your neighbor Jones is a homosexual on the basis of such flimsy evidence as that he is unmarried and has a Martha Stewart-like interest in home furnishings! You ought to suspend judgment!"
My provisional conclusion is that our manifest ability to suspend judgment in some case shows that we do have direct voluntary control over some of our believings. I have no control over my believing that a naked woman is standing in front of me if in fact such is the case (in good light, etc.) And I have no control over my believing that the Imperium Romanum no longer rules Western Europe (to adapt an example from Alston, Beyond 'Justification', p. 63). But I do have direct voluntary control over my believing that neighbor Jones is a homosexual.