Genetic Programming and Agent-Based Computational Economics: From Autonomous Agents to Product Innovation

Abstract

Despite their great development over the last decade, most ACE (agent-based computational economics) models have been generally weak in demonstrating discovery or novelty-generation processes. In this sense, they are not very distinct from their counterparts in neo-classical economics. One way to make progress is to enable autonomous agents to discover the modular structure of their surroundings, and hence they can adapt by using modules. This is almost equivalent to causing their “brain” or “mind” to be designed in a modular way. By this standard, simple genetic programming is not an adequate design for autonomous agents; however, augmenting it with automatic defined terminals (ADTs) may do the job. This paper provides initial research with evidence showing the results of using ADTs to design autonomous agents.