Mark Bedau

We have studied the adaptation of mutation rates in a simple model of evolution. The model consists of a two-dimensional world with a periodically replenished resource and a uctuating population of evolving agents whose survival and reproduction are an implicit a function of their success at nding resources and their internal metabolism. Earlier work suggested that mutation rate is a control parameter that governs a transition between two qualitatively di erent kinds of complex adaptive systems, and that the power of adaptive evolution is maximized when the mutation rate is around this transition. This paper provides evidence that evolving mutation rates adapt to values around this transition. Furthermore, the mutation rates adapt up (or down) as the evolutionary demands for novelty (or memory) increase.

This paper investigates how environmental structure, given the innate properties of a population, affects the degree to which this population can adapt to the environment. The model we explore involves simple agents in a 2-d world which can sense a local food distribution and, as specified by their genomes, move to a new location and ingest the food there. Adaptation in this model consists of improving the genomic sensorimotor mapping so as to maximally exploit the environmental resources. We vary environmental structure to see its specific effect on adaptive success. In our investigation, two properties of environmental structure, conditioned by the sensorimotor capacities of the agents, have emerged as significant factors in determining adaptive success: (1) the information content of the environment which quantifies the diversity of conditions sensed, and (2) the expected utility for optimal action. These correspond to the syntactic and pragmatic aspects of environmental information, respectively. We find that the ratio of expected utility to information content predicts adaptive success measured by population gain and information content alone predicts the fraction of ideal utility achieved. These quantitative methods and specific conclusions should aid in understanding the effects of environmental structure on evolutionary adaptation in a wide range of evolving systems, both artificial and natural.