Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits

2026-05-16

AI Summary

This study demonstrates that minor implementation differences in large language model (LLM)-based social simulations can trigger substantial macro-level outcome variations—up to a 76-percentage-point shift in cooperation rates in Prisoner’s Dilemma scenarios—thereby confounding scientific conclusions with implementation artifacts rather than genuine social mechanisms. Through case studies on the Prisoner’s Dilemma and social media echo chambers, the authors systematically uncover such “butterfly effects” and reveal significant heterogeneity in model sensitivity. To address this, they introduce TRAILS, a tiered robustness auditing framework spanning agent, interaction, and system levels, which underscores that robustness validation must be tailored to specific models and claims. This work establishes robustness as an essential prerequisite for using LLM-based social simulations in scientific inference.

Abstract

The scientific claims drawn from LLM social simulations should be no stronger than the robustness audits that support them. Generative agents bring new expressive power to agent-based modeling, enabling simulations of collective social processes like cooperation, polarization, and norm formation. Yet they also introduce complexity through additional architectural choices, such as agent specification, memory representation, interaction protocols, and environment design. Small perturbations that appear minor to researchers can cascade into macro-level outcomes through repeated interaction, creating a "butterfly effect." Consequently, scientific claims drawn from LLM social simulations may reflect implementation artifacts rather than the social mechanisms being modeled. We support this position with two case studies: a repeated Prisoner's Dilemma and a social media echo chamber simulation. Across multiple models, minor perturbations in persona format and game-instruction framing shift cooperation rates by up to 76 percentage points, while network homophily and hub assignment produce significant and consistent shifts in polarization metrics. We also find that sensitivity is unevenly distributed across both architectural choices and model families: the same perturbation that produces the 76 pp shift in one frontier model only shifts another by 1 pp. Robustness is therefore a property that should be measured per claim and per model, not assumed. To address this validation gap, we introduce TRAILS (Taxonomy for Robustness Audits In LLM Simulations), a robustness-audit taxonomy spanning three levels of simulation design: agent (micro-level), interaction (meso-level), and system (macro-level). We call for robustness to become a first-order validation requirement before LLM social simulations are used to explain mechanisms, evaluate interventions, or inform decisions.

Problem

• LLM social simulations
• robustness audits
• scientific claims
• implementation artifacts
• agent-based modeling

Innovation

• robustness audit
• LLM social simulation
• TRAILS
• generative agents
• butterfly effect