Get access

Learning Conditional Information


  • An earlier version of this paper was presented at KogWis 2010 (Potsdam) as part of the symposium ‘Neue Theorien der Rationalität.’ I thank the organizers Markus Knauff and Wolfgang Spohn for the invitation and the audience for stimulating questions and remarks. Thanks also to David Atkinson, David Etlin, David Over, Niki Pfeiffer, Wolfgang Spohn, and an anonymous referee for this journal for helpful comments and/or discussions. Special thanks to Katya Tentori for a number of long and inspiring discussions about the subject matter of this paper.

Faculty of Philosophy, Oude Boteringestraat 52, 9712 GL Groningen, The Netherlands.


Some of the information we receive comes to us in an explicitly conditional form. It is an open question how to model the accommodation of such information in a Bayesian framework. This paper presents data suggesting that there may be no strictly Bayesian account of updating on conditionals. Specifically, the data seem to indicate that such updating at least sometimes proceeds on the basis of explanatory considerations, which famously have no home in standard Bayesian epistemology. The paper also proposes a still broadly Bayesian model of updating on conditionals that explicitly factors in explanation. The model is shown to have clear empirical content and thus to be open to empirical testing.