Stephan Hartmann

London School of Economics

"Unification and Coherence"

Abstract:

Myrvold (2003) proposed a Bayesian explication of the concept of epistemic unification. This proposal entails that a hypothesis h1 unifies two pieces of evidence e1 and e2 better than a hypothesis h2 if and only if C(e1, e2; h1) > C(e1, e2; h2) with C(e1, e2; h) = Pr(e1, e2|h)/[Pr(e1|h) Pr(e2|h)]. Interestingly this explication is closely related to ShogenjiÕs measure for the coherence of an information set in epistemology. I will address this resemblance and argue for four theses: (i) The proper explication of epistemic unification is in terms of coherence. (ii) MyrvoldÕs measure should be rejected as it suffers from analogous problems as ShogenjiÕs coherence measure. (iii) The coherence measure proposed by Bovens and Hartmann in Bayesian Epistemology (OUP 2003) deals well with counter examples against MyrvoldÕs (and ShogenjiÕs) measure and (iv) helps us to make sense out of different types of unification in science. These types will be characterized by appropriate Bayesian Network models.

Back to Talks Page