Integrating LLM in Agent-Based Social Simulation: Opportunities and Challenges

A research position paper examines the integration of Large Language Models (LLMs) in agent-based social simulations, highlighting both opportunities and limitations. The study proposes Hybrid Constitutional Architectures that combine classical agent-based models with small language models and LLMs to balance expressive flexibility with analytical transparency.

• →LLMs show promise in replicating human cognition aspects like Theory of Mind reasoning but suffer from cognitive biases and behavioral inconsistencies.
• →Projects like Generative Agents (Smallville) and AgentSociety demonstrate emerging applications but face challenges in behavioral fidelity and reproducibility.
• →Hybrid approaches integrating LLMs into established platforms like GAMA and NetLogo may offer better balance between flexibility and transparency.
• →LLM-based agents provide operational value in interactive simulations but raise epistemic concerns in explanatory or predictive modeling.
• →The proposed Hybrid Constitutional Architectures framework suggests a stratified integration of classical ABMs, SLMs, and LLMs for improved social simulation.