Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits
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.
Related
- DeboRheo v2.0: Improved Robustness for Geometrical VOF Simulations of Viscoelastic Two-Phase Flows
- vibe-sketch for Chemical Structure Drawing and Calculation Preparation
- Carrington's Last Sunspot Drawing in 1870
- Quasi-Bessel Stokes skyrmions with enhanced topological robustness in free-space and turbulent propagation
- Radial encoding determines deformation robustness in far-field multimode fiber imaging
BibTeX
@misc{ye2026stop,
title = {{Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits}},
author = {Jinyi Ye and Lei Cao and Ding Chen and Emilio Ferrara},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2605.18890},
eprint = {2605.18890},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.18890},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits
AU - Jinyi Ye
AU - Lei Cao
AU - Ding Chen
AU - Emilio Ferrara
PY - 2026
DA - 2026/10/22
PB - arXiv
DO - 10.48550/arxiv.2605.18890
UR - https://arxiv.org/abs/2605.18890
AB - 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.
ER -