The Evolution of Language Groups among Cooperating Digital Predators
Many species of animals have evolved complex means for communicating with one another. Oftentimes, communication is essential for the execution of tasks that require cooperation between individuals, such as group hunting and mate selection. As a result, communication itself becomes essential for survival. While these facts are readily observed, the evolutionary processes underlying them are less understood, in large part because observational - much less controlled - studies of these processes are impossible. Both the timescales and population sizes required for such studies are simply too great.
To address these problems, this thesis uses simulated predators to study the evolution of language in animals. These digital predators evolve to perform two cooperative tasks: hunting and mate selection. After the populations of predators have evolved to perform both tasks successfully, the population is decomposed into both language groups and cooperative groups. Spectral clustering identifies predators that speak similar languages, while merge clustering is used to find those groups of predators that are the most successful when working together.
Analysis of the groups generated by these two different methods shows that the most successful pairings are not necessarily those in which the two individuals are speaking the same language. Rather, organisms can evolve to speak a different language than the one to which they respond. Moreover, even though one task -- mate selection -- evolves earlier in evolutionary history, the language diversity it produces counteracts any head-start provided for the evolution of the second task. Thus, not only is language important for the evolution of cooperative task success, but the appearance of language groups can also play a determinant role in the evolution of cooperation.