For instance, one can doubtless stay crazy for ever, but if one could explain to crazy-you that taking the antipsychotic drugs at least most of the time will make it possible to have a job, a partner-of-the-appropriate-sex, a life that doesn't involve being locked up all the time, etc., then I suspect that even crazy-you might agree that the price is worth paying.

Don't be sure about that. One of the reasons schizophrenics are so problematic is that they can't be trusted to take their meds. Apparently in a subset of schizophrenics, the condition causes them to see the world as 'vivid', bright, popping out at them in great details; and the medications bring them down to the grey dull normal perceptions - which they don't like and sometimes will skip their medications to retrieve. Even if that does mean risking the trappings of a normal life they've managed to piece together.

So doing something that gratifies you-while-high but more severely harms you-while-not-high is likely to be bad for you-on-average.

So can we derive a general principle from this? Perhaps: 'we should do what benefits us over the long run, and not the short'. So to take the cheesecake example: the duration of me which suffers for having eaten the cheesecake (both directly and for having broken my diet) is much much longer than the duration of me enjoying the cheesecake, and so we shouldn't eat the cheesecake. If we leave out mention of peaks entirely, then we avoid issues like 'but the hallucinogen feels awesome!' Or to take the wireheading example: if the wireheading could be sustained over a long time, comparable to a normal lifespan, and we're not talking the Niven scenario of dying in a few weeks, why wouldn't we permit the wireheading? It is almost by definition not harmful. (And arguments about higher pleasures of life than direct neural stimulation, are, I think, addressed by Mills's old quip about 'better to be Socrates dissatisfied'.) Basing our decisions on duration seems to me to deal with silver chairs satisfactorily. We appeal to decisions made 'sitting quietly' because that's what the rest of our life, averaged out, is like, and per the foregoing we should privilege the long-term over the short.

(And this privileging of duration even has some historical precedent; in Indian philosophy, one point of contention between Sankara/Vedantin and the Yogacara darsana was over being & perception; the Yogacara argued that we had no way to say that dreams were less real than reality, since we perceived both. Sankara's arguments consigning dreams to falsity were that waking life is longer than dreams, and waking life endures and contradicts dreams, while the reverse did not occur. Our argument here is somewhat analogous: which desire, the long-term average of our many selves, or the short-term current desire, is more 'real' and thus the one we should follow?)