Noisy Population Dynamics Lead to Efficiently Compressed Semantic Systems

Nathaniel Imel, Richard Futrell, Michael Franke, Noga Zaslavsky

Keywords: emergent communication, efficient compression, semantic typology, game theory

TL;DR: Evolutionary game dynamics of signaling with noisy perception leads to semantic category systems achieving efficient compression.

Abstract: Converging cross-linguistic evidence suggests that that human vocabularies are shaped for efficient communication, but we know little about the agent-based dynamics that could explain their evolution. In this paper, we show that very general population dynamics of signaling games lead to the emergence of information-theoretically efficient meaning systems. In numerical simulations, we observe that noisy perception of meaning can result in evolved systems with higher efficiency.

Submission Number: 23