Belief, Desire, and Revision

According to Humean theory, conduct is motivated by desire and merely guided by belief. An anti-Humean view maintains against this that belief does the whole job, since desire is just a certain species of belief: namely, belief about what would be good.1 This Desire-as-Belief Thesis, at least in its simplest form, cannot be correct, for it is at odds with the way in which we think it rational to change our minds. The purpose of this paper is to make that argument precise. I start from an account of rational belief revision in which doxastic states are represented as sets of worlds and show how this account can be extended to include a representation of an agent's desires. The notion of a belief-desire system introduced in section 3 leads to a non-quantitative decision theory that bears the same relation to standard decision theory as the belief-set scheme bears to the theory of subjective probability. The argument against the Desire-as-Belief Thesis is presented in section 4.