Humanlike Emergent Language in Multi-Agent Systems

The study of emergent language examines how artificial agents autonomously develop communication strategies to achieve shared goals. These emergent languages hold promise for creating robust, adaptive, and context-aware communication systems. However, there are significant limitations in the field, including a lack of evaluation and benchmarking frameworks and limited experimental setups that mimic human language use. My PhD research seeks to address these limitations by developing a comprehensive framework that integrates existing metrics and introduces novel measures to capture critical linguistic features. Additionally, I will analyze how experimental parameters influence the characteristics of emergent languages. My focus will be on the exploration of complex and humanlike communication settings, such as population-based, semicooperative, targeted, and symmetric linguistic exchange, drawing inspiration from multi-task and continual learning. Through these advancements, my research aims to enhance our understanding of the interplay between language, meaning, and environment in artificial emergent communication systems.