Time to Talk: LLM Agents for Asynchronous Group Communication in Mafia Games

2025-06-05

AI Summary

This work addresses the challenge of natural turn-taking timing for LLM-based agents in asynchronous, multi-agent social interactions—such as social deduction games—where no explicit turn constraints exist. We propose the first agent architecture that explicitly models speaking timing as a core cognitive capability, introducing a multi-stage decision framework jointly optimizing semantic coherence and temporal rhythm through integrated temporal modeling, dialogue state tracking, and social intent inference. Evaluated on a novel, manually curated asynchronous Mafia dataset, our agent achieves human-level win rates in real-time mixed-human–agent gameplay, attains a 72.3% deception success rate (i.e., being misclassified as human), and exhibits a Pearson correlation of 0.91 with human timing distributions—demonstrating significantly enhanced social integration. The code and dataset are publicly released.

Technology Category

• Multiagent Systems: Modeling other Agents
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

LLMs are used predominantly in synchronous communication, where a human user and a model communicate in alternating turns. In contrast, many real-world settings are inherently asynchronous. For example, in group chats, online team meetings, or social games, there is no inherent notion of turns; therefore, the decision of when to speak forms a crucial part of the participant's decision making. In this work, we develop an adaptive asynchronous LLM-agent which, in addition to determining what to say, also decides when to say it. To evaluate our agent, we collect a unique dataset of online Mafia games, including both human participants, as well as our asynchronous agent. Overall, our agent performs on par with human players, both in game performance, as well as in its ability to blend in with the other human players. Our analysis shows that the agent's behavior in deciding when to speak closely mirrors human patterns, although differences emerge in message content. We release all our data and code to support and encourage further research for more realistic asynchronous communication between LLM agents. This work paves the way for integration of LLMs into realistic human group settings, from assistance in team discussions to educational and professional environments where complex social dynamics must be navigated.

Problem

• Develops asynchronous LLM-agent for group communication timing
• Evaluates agent in Mafia games versus human players
• Enables realistic LLM integration in human social dynamics

Innovation

• LLM agent for asynchronous group communication
• Adaptive decision on when and what to say
• Mirrors human speaking patterns in games