Modeling technology evolution using generalized genotype-phenotype maps

Abstract

Modeling open-ended technological evolution is notoriously challenging. The most successful models to date have been grounded in specific domains such as electronic circuit design. This paper presents an alternative approach based on a generalization of Kauffman's NK model. In this approach, boundedly rational agents combine components into products and systems whose value is determined by a random fitness landscape in which components may vary in their pleiotropy, or the number of genotypic functions they enable. The authors are developing a family of agent-based models using this framework, the first of which explores the evolution of platform architectures. Preliminary results from this model show that platforms emerge most strongly under conditions of frequent but moderate environmental change or a moderate number of correlated market niches. Copyright 2012 ACM.