Absolutely! Depression changes one’s priors and one’s perception of evidence, making a depressed agent even further from rational than non-depressed humans (who are even so pretty far from purely rational).
That said, all agents must make choices—that’s why we use the term “agent”. And even depressed agents can analyze their options using the tools of rationality, and (I hope) make better choices by doing so. It does require more care and use of the outside view to somewhat correct for depression’s compromised perceptions.
Also, I’m very unsure what the threshold is where an agent would be better off abandoning attempts to rationally calculate and just accept their group’s deontological rules. It’s conceivable that if you don’t have strong outside evidence that you’re top-few-percent of consequentialist predictors of action, you should follow rules rather than making decisions based on expected results. I don’t personally like that idea, and it doesn’t stop me ignoring rules I don’t like, but I acknowledge that I’m probably wrong in that.
Specifically, “Suicide: don’t do it” seems like a rule to give a lot of weight to, as the times you’re most likely tempted are the times you’re estimating the future badly, and those are the times you should give the most weight to rules rather than rationalist calculations.
Absolutely! Depression changes one’s priors and one’s perception of evidence, making a depressed agent even further from rational than non-depressed humans (who are even so pretty far from purely rational).
That said, all agents must make choices—that’s why we use the term “agent”. And even depressed agents can analyze their options using the tools of rationality, and (I hope) make better choices by doing so. It does require more care and use of the outside view to somewhat correct for depression’s compromised perceptions.
Also, I’m very unsure what the threshold is where an agent would be better off abandoning attempts to rationally calculate and just accept their group’s deontological rules. It’s conceivable that if you don’t have strong outside evidence that you’re top-few-percent of consequentialist predictors of action, you should follow rules rather than making decisions based on expected results. I don’t personally like that idea, and it doesn’t stop me ignoring rules I don’t like, but I acknowledge that I’m probably wrong in that.
Specifically, “Suicide: don’t do it” seems like a rule to give a lot of weight to, as the times you’re most likely tempted are the times you’re estimating the future badly, and those are the times you should give the most weight to rules rather than rationalist calculations.