Abstract
We propose a linguistic prediction game with competitive andcooperative variants, and a model of game players based onfinite state automata. We present a complexity metric forthese automata, and study the coevolutionary dynamics ofcomplexity growth in a variety of multi-species simulations.We present quantitative results using this complexity metricand analyze the causes of varying rates of complexity growthacross different types of interactions. We find that whileboth purely competitive and purely cooperative coevolutionare able to drive complexity growth above the rate of geneticdrift, mixed systems with both competitive and cooperativeinteractions achieve significantly higher evolved complexity.