Emergent Communication through Negotiation

TL;DR

This paper investigates how communication emerges in multi-agent negotiation environments, showing that prosocial behavior enables effective use of ungrounded communication channels and highlighting the role of cooperation in language development.

Contribution

It introduces two communication protocols in negotiation settings and demonstrates that prosocial agents can develop effective communication strategies, unlike self-interested agents.

Findings

Prosocial agents learn to use cheap talk for negotiation.

Self-interested agents fail to utilize ungrounded communication.

Agent identifiability improves negotiation outcomes.

Abstract

Multi-agent reinforcement learning offers a way to study how communication could emerge in communities of agents needing to solve specific problems. In this paper, we study the emergence of communication in the negotiation environment, a semi-cooperative model of agent interaction. We introduce two communication protocols -- one grounded in the semantics of the game, and one which is \textit{a priori} ungrounded and is a form of cheap talk. We show that self-interested agents can use the pre-grounded communication channel to negotiate fairly, but are unable to effectively use the ungrounded channel. However, prosocial agents do learn to use cheap talk to find an optimal negotiating strategy, suggesting that cooperation is necessary for language to emerge. We also study communication behaviour in a setting where one agent interacts with agents in a community with different levels of prosociality and show how agent identifiability can aid negotiation.