Proceedings of the Annual Meeting of the Cognitive Science Society

In the process of collectively inventing new words for new con-cepts in a population, conflicts can quickly become numerous,in the form of synonymy and homonymy. Remembering all ofthem could cost too much memory, and remembering too fewmay slow down the overall process. Is there an efficient be-havior that could help balance the two? The Naming Game isa multi-agent computational model for the emergence of lan-guage, focusing on the negotiation of new lexical conventions,where a common lexicon self-organizes but going through aphase of high complexity. Previous work has been done onthe control of complexity growth in this particular model, byallowing agents to actively choose what they talk about. How-ever, those strategies were relying on ad hoc heuristics highlydependent on fine-tuning of parameters. We define here a newprincipled measure and a new strategy, based on the beliefsof each agent on the global state of the population. The mea-sure does not rely on heavy computation, and is cognitivelyplausible. The new strategy yields an efficient control of com-plexity growth, along with a faster agreement process. Also,we show that short-term memory is enough to build relevantbeliefs about the global lexicon.