14 Pages Posted: 6 Jul 2014
Date Written: February 25, 2014
Evidence is sometimes misleading; it can mislead researchers who use it to update their beliefs, even if they do so in an epistemically responsible way. But what exactly does it take for evidence to be misleading, and what is it for a researcher to be misled? We investigate three approaches to this question - based on significance tests, based on likelihood ratios, and based on credence shifts - and argue in favor of the latter approach. In particular, we argue that a function of credences known as the spherical rule provides the best criterion for deciding whether evidence is misleading.
Keywords: Misleading Evidence, Likelihood Ratios, Significance Tests, Credences, Epistemic Utility, Proper Scoring Rules, Brier Score
Suggested Citation: Suggested Citation