The truth-convergence of open-minded Bayesianism
Tom F. Sterkenburg and Rianne de Heide
The truth-convergence of open-minded Bayesianism
Tom F. Sterkenburg and Rianne de Heide
The Review of Symbolic Logic 15(1):64-100, 2022, doi:10.1017/S1755020321000022 philsci archive proc
Accepted to the Formal Epistemology Workshop 2019, and the biennial conference of the European Philosophy of Science Association.
What is this paper about?
This paper studies an open-minded form of Bayesian learning in which new hypotheses may be introduced during the learning process. It shows that a natural existing proposal can lose the usual truth-convergence guarantee and develops a modification that restores a version of it.
Summary
Wenmackers and Romeijn formalize an open-minded Bayesian inductive logic that can dynamically incorporate statistical hypotheses proposed during the learning process. We show that their proposal does not preserve the classical Bayesian consistency guarantee of merger with the true hypothesis. We diagnose the problem and offer a forward-looking open-minded Bayesian approach that preserves a version of this guarantee.
Topics
Bayesian learning and generalized Bayes · Foundations of statistics, probability and learning