Resiliency, propensities, and causal necessity

Abstract

This article proposes a concept of ‘resiliency’ as a degree to which a probabilistic hypothesis resists change in the face of new information. It argues that resiliency is a valuable tool for understanding propensities – hypothetical probabilities associated with repeatable processes, and the difference between genuine statistical laws and spurious correlations. It suggests that resiliency can be used to characterize different types of statistical randomness, such as exchangeability and independence, and connects these ideas to concepts of causal ordering, shielding-off, and the confirmation of simple nonstatistical laws – AI-generated abstract.