Bayesian Confirmation Theory: Inductive Logic or Mere Inductive Framework?

Published: Synthese, 141:365–379. 2004.

Abstract: To what extent is Bayesianism a positive confirmation theory, delivering particular judgements as to how evidence bears on scientific theories, and to what extent is it, as Colin Howson (Hume's Problems) has recently claimed, more like a framework for confirmation theory capable of accommodating any kind of substantive inductive assumption about the proper relation between evidence and theory? I ask these questions of what is perhaps the most popular version of Bayesian confirmation theory, that presented recently in Howson and Urbach's Scientific Reasoning and Earman's Bayes or Bust?, which I call, after Earman, modern Bayesianism. I conclude that

  1. Modern Bayesianism is more of a positive confirmation theory than Howson suggests, but that
  2. The principal source of modern Bayesianism's positive judgments of inductive relevance is not the Bayesian machinery itself, but rather what David Lewis calls the Principal Principle.

I then explain why the Principal Principle has, incorrectly, not traditionally been regarded as an inductive assumption.

See a PDF version of the paper.