Real patterns and indispensability

Abstract

While scientific inquiry crucially relies on the extraction of patterns from data, we still have a far from perfect understanding of the metaphysics of patterns—and, in particular, of what makes a pattern real. In this paper we derive a criterion of real-patternhood from the notion of conditional Kolmogorov complexity. The resulting account belongs to the philosophical tradition, initiated by Dennett (J Philos 88(1):27–51, 1991), that links real-patternhood to data compressibility, but is simpler and formally more perspicuous than other proposals previously defended in the literature. It also successfully enforces a non-redundancy principle, suggested by Ladyman and Ross (Every thing must go: metaphysics naturalized, Oxford University Press, Oxford, 2007), that aims to exclude from real-patternhood those patterns that can be ignored without loss of information about the target dataset, and which their own account fails to enforce.