The use of reputation as noise-resistant selection bias in a co-evolutionary multi-agent system

Little attention has been paid to the relationship between fitness evaluation in evolutionary algorithms and reputation mechanisms in multi-agent systems, but if these could be related it opens the way for implementation of distributed evolutionary systems via multi-agent architectures. In this paper we investigate the effectiveness with which reputation can replace direct fitness observation as the selection bias in an evolutionary multi-agent system. We do this by implementing a peer-to-peer, self-adaptive genetic algorithm, in which agents act as individual GAs that, in turn, evolve dynamically themselves in real-time. The evolution of the agents is implemented in two alternative ways: First, using the traditional approach of direct fitness observation (self-reported by each agent), and second, using a simple reputation model based on the collective past experiences of the agents. Our research shows that this simple model of distributed reputation can be successful as the evolutionary drive in such a system. Further, we discuss the effect of noise (in the form of "defective" agents) in both models. We show that, unlike the fitness-based model, the reputation-based model manages to identify the defective agents successfully, thus showing a level of resistance to noise.