Realism, Simplicity, and Topology

Kevin T. Kelly
Carnegie Mellon University



Abstract:  This is joint work with Hanti Lin, University of Californtia, Davis and Konstantin Genin, University of Toronto

Scientific realists assure us that simpler theories are better-confirmed by simple data and are, therefore, more worthy of belief.  Scientific anti-realists respond that the the data might look simple for eternity if the complex theory is true, in which case the realist's assurance would lead to eternal error.  We show that the realist's position follows from a learning-theoretic argument for Ockham's razor.  The argument is based on the interaction of two topologies on possible worlds: the realist topology of arbitrary similarity in reality, and the empirical topology of arbitrary empirical similarity.  The former pertains to the ends of inquiry, and the latter characterizes the means.