Bayesianism, Ravens, and Evidential Relevance

Abstract

Bayesian confirmation theorists have proposed a variety of solutions to Hempel's paradox of the ravens. I examine those of Suppes and Horwich and argue that they do not completely avoid counter-intuitive results about the relevance of data. The Bayesian explication of evidential relevance is also susceptible to the same relevance problems that infect Hypothetico-Deductivism. I explore a possible escape to the problem of old evidence, but conclude that it only leads to problems of the same sort-any datum can be relevant to any hypothesis in any circumstance. I argue that the Bayesian evidence relation is not sufficient or necessary to determine what counts as evidence. Such difficulties warrant pursuit of alternative explications of evidential relevance. I show how the raven's paradox may be avoided by bringing in causal considerations.