Fitness as Informational Fit: An Information Theoretic Approach to Multilevel Requisite Variety

Abstract

The article starts with the familiar notion that evolution can be thought of as handling uncertainty with regard to variety in population structures. The resolution of uncertainty communicates information through negative entropy, which increases fitness. This setup can be generalized. Information turns out to be the link between diversity of types in populations and unfolding environmental patterns in time. Conditioning of information over multiple levels reduces uncertainty, which increases fitness. While previous research has applied this logic to temporal patterns through Kelly’s bet-hedging criteria (stochastic switching), this article shows how the same logic can also be applied to multilevel population structures. Examples include taxonomies and geographic distributions. The result is a reformulation of fitness as a multilevel communication process between the evolving population and its environment. The multilevel recursion provides a single setup to investigate diversity relations between any population hierarchy and environmental patterns. Information theoretic optimization through bet-hedging reveals that fitness can be optimized for the case in which there is one specialized type per environmental state. This is reminiscent of the notion of requisite variety, and shows how fitness optimization can be understood in terms of a multilevel informational fit between the evolving population and environmental patterns.