Epistemic rationality is about collecting accurate beliefs about the world. Instrumental rationality is about achieving your goals.
Certain sects, such as the rationalists, believe that epistemic rationality is always instrumentally rational: that is, that holding accurate beliefs can’t hinder you from achieving your goals relative to holding a contradictory inaccurate belief, and that holding inaccurate beliefs can’t help you achieve your goals relative to holding a contradictory accurate belief, and therefore that people should believe things that are true and not believe things that are false. (This of course can’t be true in principle, since there’s no reason in principle why an agent that terminally values believing false things couldn’t exist; but it’s assumed that humans don’t terminally value believing false things.)
Certain other sects, practiced in various Caribbean islands, believe that it’s possible for someone in a certain sort of religious rite to be possessed by spirits that fly over the ocean from Africa. This came up recently when I was talking to a rationalist, and he said that the practitioners of these sects should stop believing that, because it’s not true, because (for example) if you built a big net over the ocean, you wouldn’t catch the spirits in it. The effects of spirit possession could simply be due to hypnosis.
The interesting question that falls out of this is: is there anything analogous to the case where spirit possession is due to hypnosis or some similar natural mechanism, but where only people who believe in spirit possession are susceptible to the form of hypnosis that causes the effects? (I’m assuming that spirit possession is a positive state.)
If there are things that a human could want to do that can only be done by believing false things, the statement that epistemic rationality is always instrumentally rational seems much less defensible: its proponents would have to either give up or argue that the negative side-effects of holding false beliefs must always outweigh the benefits of being able to do the things they need the beliefs for.
It would, however, also be possible to distinguish between epistemic beliefs and instrumental beliefs. The rationalist account of belief seems incomplete: beliefs aren’t merely about correspondence with the world, but also about thedish signaling and social strategies (which most rationalists acknowledge), as well as myth, action orthogonal to the substance of the belief, aesthetic and conceptual associations, etc. That is: beliefs, like most social phenomena, are complex social technologies. For example: the rite is motivated by the belief in spirit possession, but presumably has some social functions, which of course rely on the belief; the spirits fly over from Africa because the culture that practices these rites was brought to the Caribbean from Africa; etc.
There’s surely a culture somewhere in the world that could be experimented on—a culture where people are socially required to believe something that’s demonstrably untrue…