Regularities No Fluke

Abstract

Abstract Humean accounts of causality deny the existence of any causal factor that imparts regularity to empirical data. According to the cosmic fluke objection, this makes Humeanism very unlikely, since it must attribute the extensive regularities that we observe to mere chance. In defence of Humeanism, I argue that this objection equivocates between two senses of regularity. Both science and everyday cognition rely not on global, residue-free compressions of data (whole-string regularities), but on component patterns that appear together with nonzero noise (two-part regularities). Any data set, even one that is algorithmically random, can be decomposed into such patterns and noise. Consequently, the existence of such regularities in empirical data is not surprising, even in the absence of a mechanism devoted to making data regular.