Alternate Adaptive Agent Architectures and Behavioral Consequences

The Learning Classifier System (LCS) and its descendant, XCS, are promising paradigms for adaptive agent construction. Whereas LCS allows classifier payoff predictions to guide system performance, XCS focuses on payoff-prediction accuracy instead, allowing it to evolve “optimal” classifier sets in particular applications requiring rational thought. We examine LCS/XCS performance in artificial situations with broad social/commercial parallels, created using the non-Markov Iterated Prisoner’s Dilemma (IPD) game-playing scenario, where the setting is sometimes asymmetric and where irrationality sometimes pays. We systematically perturb a “conventional” IPD-playing LCS-based agent until it results in a full-fledged XCS-based agent, contrasting the simulated behavior of each LCS variant with the XCS agent in terms of a number of performance measures. Our intent is to examine the XCS paradigm to understand how it better copes with a given situation (if it does) than the LCS perturbations studied.