"Induction and Simplicity" Gilbert Harman and Sanjeev Kulkarni, Princeton University We are concerned with the reliability of inductive methods in the light of certain foundational results in statistical learning theory. In this talk, we compare simple enumerative induction with methods that take into account some ordering of hypotheses, perhaps by simplicity. We compare different methods for balancing data-coverage against an ordering of hypotheses in terms of simplicity or some simplicity substitute. Then we consider how ideas from statistical learning theory might shed light on some philosophical issues. In particular, we distinguish two ways to respond to Goodman's (1965) "new riddle of induction", corresponding to these two kinds of inductive methods. We discuss some of Karl Popper's ideas about scientific method, trying to distinguish what is right and what is wrong about these ideas. We consider how the two sorts of induction discussed in this talk might apply to first language acquisition. And finally, we consider how an appeal to simplicity or some similar ordering might provide a principled way to prefer one hypothesis over another skeptical hypothesis that is empirically equivalent with it.