Should Irrational Preferences Count?
Consider, as one example, the endowment effect, the observation that individuals value items that belong to them more than items that do not even if, as in the classic Cornell coffee cup experiment, who owns what is the result of random chance. I explain this as a commitment strategy designed to enforce property rights in a world without police and courts, the human elaboration of the territorial behavior observed in many animal species.
The usual rule in economics is to take values as we find them expressed in behavior. In deciding whether one situation is more or less economically efficient than another we are judging whether it does a better or worse job of giving people what they are observed to want, not going behind preferences to judge whether it does a better or worse job of giving them what they ought to want.
Suppose you accept my explanation for the endowment effect, or some similar explanation for some similar, apparently irrational, pattern of behavior—that it exists not because it serves the present interest of the individual but because it served the (reproductive) interest of other individuals long ago in a very different environment. Should you still take it as a given in evaluating economic institutions?
Before answering "obviously yes," which I am tempted to do, you might want to consider a simpler question of the same sort. You observe A add some cyanide in B's wine glass, while B is looking the other direction. You ask B if he wants to drink what is in the glass. B replies that he does. Do you conclude that his drinking it is, on the principle of revealed preference, a good thing?