I present hitherto unexplored and unaccounted for uses of 'wants'. I call them advisory uses, on which information inaccessible to the desirer herself helps determine what it's true to say she wants. I show that extant theories by Stalnaker, Heim, and Levinson fail to predict it. I also show that they fail to predict true indicative conditionals with 'wants' in the consequent. These problems are related: intuitively valid reasoning with modus ponens on the basis of the conditionals in question results in unembedded advisory uses.
I consider two fixes, and end up endorsing a relativist semantics, according to which desire attributions express information-neutral propositions. The truth of a desire attribution depends on the state of information at the context of assessment. On this view, 'wants' functions as a precisification of 'ought', which exhibits similar unembedded and compositional behavior. I conclude by sketching a pragmatic account of the purpose of desire attributions, one that explains why it made sense for them to evolve in this way.
Posts
No comments:
Post a Comment