Abstract

This paper integrates two different ways to face the emergence of communication systems. First, information-theoretic models argue for the emergence of Zipfian properties based on the minimization of communicative energy, combining speaker and hearer efforts. Second, decentralized agent-based models focus on the emergence of a shared communication system in a population of individuals. The main aim here is to explore how a decentralized agent-based model of language formation can exhibit purely from local speaker-hearer interactions the minimization of communicative energy. Numerical simulations show an evolution towards the minimum of communicative energy for populations communicating with Zipfian languages. Our results suggest thus a new way to understand energy-based approaches to the formation of human language.