Deliberators must be imperfect

This paper argues that, with certain provisos, predicting one's future actions is incompatible with rationally deliberating about whether to perform those actions. It follows that fully rational omniscient agents are impossible, since an omniscient being could never rationally deliberate about what to do (omniscient beings, the paper argues, will always meet the relevant provisos). Consequently, theories that explain practical reasons in terms of the choices of a perfectly rational omniscient agent must fail. The paper considers several ways of defending the possibility of an omniscient agent, and concludes that while some of these may work, they are inconsistent with the aim of explaining practical normativity by appeal to such an agent.