Algorithmic Prompt Generation for Diverse Human-like Teaming and Communication with Large Language Models

Understanding human teaming and communication is essential for improving human-agent teaming. However, relying solely on data from large-scale user studies is impractical due to logistical, ethical, and practical constraints, necessitating synthetic models of diverse human behaviors. Recently, agents powered by Large Language Models (LLMs), with expert-designed prompts, have been shown to emulate human-like behavior in social settings. On the other hand, Quality Diversity (QD) optimization has been shown to be capable of generating diverse agent behavior, albeit without natural language communication. In this work, we combine QD optimization with LLM-powered agents to iteratively search for prompts that generate diverse team behavior in a long-horizon, multi-step collaborative environment. We first show, through a user study, that humans exhibit diverse coordination and communication behavior. We then show that our approach captures behaviors that are difficult to observe without large-scale data collection, and conduct a follow-up user study to demonstrate that the generated behaviors are human-like. Our findings highlight the combination of QD and LLM-powered agents as an effective tool for studying teaming and communication strategies in multi-agent collaboration.