Implicit Behavioral Alignment of Language Agents in High-Stakes Crowd Simulations

2025-09-19

AI Summary

Language-driven generative agents exhibit a persistent behavior-realism gap in high-stakes social simulations, leading to distorted collective behavior and diminished credibility. To address this, we propose the Persona-Environment Behavioral Alignment (PEBA) framework—the first to formalize agent–environment interaction as an implicit distribution matching problem, grounded theoretically in Lewin’s behavior equation. PEBA introduces PersonaEvolve, a novel instruction-free iterative optimization algorithm that jointly leverages large language models and high-fidelity social simulation environments to achieve end-to-end behavioral alignment. Evaluated on active-shooter scenario simulations, PEBA reduces behavioral distribution divergence by 84% relative to unguided baselines and improves alignment by 34% over explicit-instruction methods. Crucially, optimized personas demonstrate strong cross-scenario generalization. This work bridges foundational behavioral theory with modern generative AI, advancing realistic, scalable, and trustworthy social simulation.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Language-driven generative agents have enabled large-scale social simulations with transformative uses, from interpersonal training to aiding global policy-making. However, recent studies indicate that generative agent behaviors often deviate from expert expectations and real-world data--a phenomenon we term the Behavior-Realism Gap. To address this, we introduce a theoretical framework called Persona-Environment Behavioral Alignment (PEBA), formulated as a distribution matching problem grounded in Lewin's behavior equation stating that behavior is a function of the person and their environment. Leveraging PEBA, we propose PersonaEvolve (PEvo), an LLM-based optimization algorithm that iteratively refines agent personas, implicitly aligning their collective behaviors with realistic expert benchmarks within a specified environmental context. We validate PEvo in an active shooter incident simulation we developed, achieving an 84% average reduction in distributional divergence compared to no steering and a 34% improvement over explicit instruction baselines. Results also show PEvo-refined personas generalize to novel, related simulation scenarios. Our method greatly enhances behavioral realism and reliability in high-stakes social simulations. More broadly, the PEBA-PEvo framework provides a principled approach to developing trustworthy LLM-driven social simulations.

Problem

• Addressing the Behavior-Realism Gap in generative agent simulations
• Aligning agent behaviors with expert expectations and real-world data
• Enhancing behavioral realism and reliability in high-stakes simulations

Innovation

• PEBA framework models behavior as person-environment function
• PEvo algorithm iteratively optimizes agent personas using LLMs
• Implicit persona refinement aligns behaviors with expert benchmarks