Studies In Bayesian Confirmation Theory

The major research object of this thesis is the (subjective) Bayesian approach to induction and confirmation, involving (1) some of its theoretic backgrounds, (2) some of its advantages and (3) some challenges to it. There are five chapters in the thesis. I shall first introduce the central Bayesian ideas and some of its theoretic and historical backgrounds.Then in Chapter 2, I present one of the successful stories of Bayesian confirmation theory: This theory resolved the Duhem-Quine problem, because it successfully depicted the subjective judgments of scientists.In Chapter 3, I present successful story II of this theory: Bayesian confirmation theory resolved the Paradox of the Raven. There is a standard resolution of the Paradox of the Raven within a Bayesian framework, and a newly improved Bayesian resolution that Fitelson proposed.In Chapter 4, I discuss the Old Evidence Problem. Subjective Bayesian has such a criterion: e confirms h if and only if P(h/e) > P(h). However, we have some evidence e that is already known when the question of confirmation is raised, i.e., P(e) = 1. In this case, it can be proved that no matter what h is, P(h/e) = P(h). It follows from the criterion that e cannot confirm h at that time. But this contradicts our intuitive belief that E can provide confirmation for h.Finally, in Chapter 5 I discuss the problem of Measure Sensitivity. The validity of Bayesian confirmation theory depends on choice of measure, that is, on which of measure functions is taken to be measure confirmation, and different measure function will produce different and even conflicting results.