Friday, 3 October 2003
The Tracking View of Scientific Evidence
Sherri Roush, Rice University
12:05 pm, 817R Cathedral of Learning

Abstract: I present a concept of evidence (confirmation) that contains some old and some new ideas. Robert Nozick's ‘tracking’ conception of evidence is best understood, I argue, in terms of the likelihood ratio P(e/h)/P(e/-h). A further condition is required if we want evidence to give a good reason to believe a hypothesis, I argue, and I make this the condition that P(e) be high, rather than that P(h/e) be high. This is in contrast to the typical Bayesian recommendation that the “expectedness” of the evidence be low.