Artificial Life VI

On Searching for Generic Properties of Non-Generic Phenomena: An Approach to Bioinformatic Theory Formation

P. Hogeweg
Utrecht University


In this paper we first shortly review the current view of the evolution of complexity and novelty in biotic evolution. Next we show that the basic processes thereof do happen automatically and are generic properties of systems including the basic mechanisms of Darwinian evolution {\it plus} {\bf local} as opposed to global interactions. Thus we show that the so generated multilevel evolution can be studied within the paradigm 'simple rules lead to complex phenomena'. We derive some results demonstrating the power of such multilevel evolutionary processes to integrate information at multiple space and time scales. Nevertheless we also point out shortcomings of such an approach which necessarily uses a priori chosen and preferentially relatively simple interaction schemes. However straightforward extensions towards more complex interaction schemes generally leads to ad hocness and over-determinedness, rather than fundamentally new behavior of the system, and often to less understanding of that behavior. Nevertheless biological theory formation needs a method to go beyond the generic behavior of simple interaction schemes. We propose to use evolutionary optimization of very trivial fitness functions which are obtainable in many different ways to push back the necessary a priori choices and to zoom in to interesting non generic phenomena and their general properties. . We thus derive insights in relationships between sets of derived properties at several scales. We discuss how these approaches can be used in biological theory formation. focusing on information accumulation and utilization in replicator systems and immune systems.

Back to ALife6 Table of Contents / Keyword Index / Author Index / Meeting Schedule