The Epistemic Value of Novel Predictive Success in Scientific and Criminal Investigations: A Bayesian Explanation

TL;DR

This paper uses Bayesian analysis to compare the epistemic value of new versus old evidence in criminal and scientific contexts, clarifying when novel predictive success strengthens theories.

Contribution

It provides a Bayesian framework to evaluate the epistemic significance of new evidence relative to old evidence in theory assessment.

Findings

New evidence can sometimes provide stronger support than old evidence.

Bayesian analysis clarifies conditions under which novel predictions are epistemically valuable.

Insights are applicable to both criminal investigations and scientific theory evaluation.

Abstract

Because there are similarities between the evaluation of alternative stories in criminal trials and the evaluation of scientific theories, scholars have looked to literature in epistemology and the philosophy of science for insights on the evaluation of evidence in criminal trials. The philosophical literature is divided, however, on a key point -- the epistemic value of novel predictive success. This article uses a Bayesian network analysis to explore, in the context of a criminal case, the circumstances in which "new evidence" discovered after a theory is propounded, can provide stronger (or weaker) support for the theory than "old evidence" that was accommodated by the theory. It argues that insights from analysis of the strength of "new" and "old" evidence in the criminal case can be applied more generally when assessing the relative merits of prediction and accommodation in scientific theory development, and are thus helpful in addressing the longstanding philosophic controversy over this issue.