A cultural evolutionary model for artifact capabilities

The use of tools or artifacts is essential to the human race and has been the subject of recent research in Artificial Intelligence. How individual agents acquire these capabilities and how they evolve can be considered vital steps towards understanding complex group capabilities. In a previous study, we designed and implemented an extended version of a theoretical model for artifact capability that accommodated biological evolution and learning via exploratory methods. Historical knowledge and genetic algorithms were combined with learning techniques to build agents that could learn either individually from observations of their own behaviour or socially by observation from a distance. In this study, we incorporate a collaborative form of cultural learning into the model in an effort to enhance the artifact capability-learning agents. This is accomplished via the design of a cultural evolutionary model that utilizes genetic and cultural algorithms to complement the cognitive abilities of the agents. Learning agents belonging to a social network cooperate with and benefit from each other by sharing individual experiences. Results obtained from the multi-agent simulation implementation confirm the efficiency of social learning over individual learning and demonstrate the benefits of cultural over biological evolution. They also suggest that as artifacts get more complex, social agents learning via cultural influence outperform those learning by observation from a distance.