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

The paper presents a novel approach that integrates Quality Diversity (QD) optimization with Large Language Model (LLM)-powered agents to automatically generate diverse prompts for simulating human-like team behaviors in collaborative environments. By conducting human-subject experiments, the authors demonstrate that their method effectively captures a wide range of coordination and communication behaviors, which are typically challenging to observe without extensive data collection. This research is significant for practitioners as it provides a systematic way to enhance AI-assisted decision-making and improve human-agent collaboration by leveraging synthetic models of diverse human interactions.