People can reason about the preferences of other agents, and predict their behavior based on these preferences. Surprisingly, the psychology of reasoning has long neglected this fact, and focused instead on disinterested inferences, of which preferences are neither an input nor an output. This exclusive focus is untenable, though, as there is mounting evidence that reasoners take into account the preferences of others, at the expense of logic when logic and preferences point to different conclusions. This article summarizes the most recent account of how reasoners predict the behavior and attitude of other agents based on conditional rules describing actions and their consequences, and reports new experimental data about which assumptions reasoners retract when their predictions based on preferences turn out to be false.