Complex Energy Landscapes in Populations

15/05/2008, 14:15, D5-153

Juan G. Diaz Ochoa, Institut für Theoretische Physik, Bremen

We conducted an investigation of the effect that variability of the agent's environment has in the agent's adaptability and, in general, in the co-evolution of a population. First we assume that the agents are a kind of perceptrons, with a given memory size, that adapt to their environment; however, these agents have a kind of decentralized mind, i.e. they are not only able to process information but also have preferences in the selection of the mutual interaction according to more or less risk aversion. This second feature introduces strong variability of the population's environment. Second, we consider co-evolution not only of the agent's actions but also of its memory size, depending on the agent's fitness. We show that changes in the intensity of the environmental fluctuations introduce a phase transition in the frequency of cooperators and the diversity of the population. We also find out that extreme variability of the environment induces changes in the critical exponents and that this variability can promote more diversity in the population.