Preference conditional on circumstances and past preference satisfaction

I’ve men­tioned con­di­tional prefer­ences be­fore. Th­ese are prefer­ences that are de­pen­dent on facts about the world, for ex­am­ple “I’d want to be­lieve X if there are strong ar­gu­ment for X”.

But there is an­other type of prefer­ence that is con­di­tional: my tastes can vary de­pend­ing on cir­cum­stances and on my past ex­pe­rience. For ex­am­ple, I might pre­fer to eat ap­ples dur­ing the week and or­anges on week­ends. Or, be­cause of the mir­a­cle of bore­dom, I might pre­fer or­anges if (but only if) I’ve been eat­ing ap­ples all week so far.

What if I cur­rently want ap­ples, would want or­anges to­mor­row, but falsely be­lieve (to­day) that I would want ap­ples to­mor­row? This is a known prob­lem with “one-step hy­po­thet­i­cals”, and a strong ar­gu­ment in prac­tice for as­sess­ing prefer­ences over time rather than at a sin­gle mo­ment .

In the­ory, there are meta-prefer­ences that al­low one to get this even at a sin­gle mo­ment , such as “I want to be able to fol­low my differ­ent tastes at differ­ent times” or a more for­mal­ised de­sire for va­ri­ety and ex­plo­ra­tion.