The epistemologist John Pollock has implemented computationally an architecture for a rational agent which he calls OSCAR. OSCAR models both practical and theoretical (or epistemic) reasoning. I argue that Pollock's model of practical reasoning, which has seven components, is superior not only to the two-component belief-desire model stemming from Aristotle, but also to the three-component belief-desire-intention model developed especially by Michael Bratman. Despite its advantages, Pollock's model of practical reasoning is incomplete in at least three respects: it is solipsistic, it is egoistic and it is unsocial.