Ph.D. Dissertation Chapter 1 Abstract

We review several systems of probability logic along with the uses of each. Probability logic is any system of probability theory formulated within a logical framework for the purpose of using logical techniques to analyze probability theory or its applications. We first examine the probability logic of Haim Gaifman. His analysis uses techniques from Boolean algebra and recursion theory to clarify how the complexity of probability distributions affects uses of probability as an inductive logic.

Dana Scott and Peter Krauss generalize Gaifman's logical formulation of probability theory. Within it they prove important analogues of theorems of ordinary logic, along with extensions particular to probability models, by combining model-theoretic techniques with techniques from functional analysis.

Jerome Keisler et al have developed a novel model theory of probability accompanied by axioms complete under associated proof rules. They use the logical system to establish the importance of non-standard techniques within probability theory and also to prove powerful theorems in and techniques for stochastic processes.

Ph. D Dissertation Chapter 2 Abstract

We investigate Stephen Simpson's philosophical motivations for the program of classifying theorems of applicable ("ordinary") mathematics by using axioms of second-order arithmetic; the program he dubs reverse mathematics. In particular we scrutinize his claim that the particular codings within a branch of mathematics do not affect the classifications of the theorems within the branch. We examine Xioakang Yu's coding of finite measure theory into second-order arithmetic both as an example of coding issues and also in order to develop her system further. Finally, we outline some desiderata for proper codings of mathematics.

Ph.D. Dissertation Chapter 3 Abstract

We characterize the logical requirements simple versions of de Finetti's exchangeability theorem within second-order arithmetic by using the measure theory coding developed by Yu. We find that the theorem is provable within the subsystem $RCA_0$, thereby contributing to Simpson's reverse mathematics program. In ordinary mathematics, de Finetti's theorem is equivalent to a simple version of the Riesz Representation Theorem, which in a weak sense is true in Yu's system because her coding of measures as functionals relies on Riesz's theorem. These facts call into question the classification that de Finetti's theorem (and Riesz's theorem) have been assigned through Yu's coding of probability theory. Finally, we discuss some assumptions of Yu's coding of measure theory which indicate that more general de Finetti theorems are unlikely to be proven within the framework.