Validating LLM-generated political networks with survey-reported partisan exposure
Abstract
Large language models (LLMs) are increasingly used to simulate human behavior in social science research, including decisions, interactions, and social relationships. However, in network formation tasks, existing evaluations often focus on aggregate network properties, while networks with plausible structural properties may still fail to reproduce individuals' social and political environments. Here, we generate LLM-based political networks using personas sampled from the nationally representative ANES 2020--2022 Social Media Study and use respondents' survey-reported partisan exposure as an empirical benchmark for the political environments generated for the corresponding personas. We first characterize the structural properties and political assortativity of the generated networks, and then compare ego-level partisan exposure with the survey. We find that the structure and political assortativity of the generated networks vary across experimental configurations. More importantly, when comparing generated partisan exposure with the survey, some models generate stronger separation between Democratic and Republican egos than reported in the survey, with higher same-party and lower cross-party exposure, while others mainly underestimate Republican exposure across ego groups. Moreover, networks with similar political assortativity can still generate different partisan environments around the same personas, including among networks with plausible structural properties. Finally, generated networks show group-specific errors, under- or overestimating exposure to either Democrats or Republicans for specific social groups. These findings show that evaluating LLM-based social simulations only through aggregate outcomes can hide errors in the local environments through which those outcomes emerge, highlighting the importance of validation at both aggregate and individual levels.
Related
- Partisan voter model on complex networks: Dynamics of local ordering
- Parliamentary mention networks reflect the organisation and dynamics of Finnish political discourse
- Experienced partisan segregation across patterns of mobility behavior in US cities
- Validating a BDT-based Electron-Positron Identification Algorithm at CLAS12 with Experimental Data
- A Phase-Space Inclusive Figure of Merit Based in Optimal Transport for Validating Monte Carlo Reweightings
BibTeX
@misc{tonin2026validating,
title = {{Validating LLM-generated political networks with survey-reported partisan exposure}},
author = {Marco Tonin and Michele Tizzoni and Bruno Lepri and Luca Maria Aiello},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.23199},
eprint = {2610.23199},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.23199},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Validating LLM-generated political networks with survey-reported partisan exposure
AU - Marco Tonin
AU - Michele Tizzoni
AU - Bruno Lepri
AU - Luca Maria Aiello
PY - 2026
DA - 2026/10/16
PB - arXiv
DO - 10.48550/arxiv.2610.23199
UR - https://arxiv.org/abs/2610.23199
AB - Large language models (LLMs) are increasingly used to simulate human behavior in social science research, including decisions, interactions, and social relationships. However, in network formation tasks, existing evaluations often focus on aggregate network properties, while networks with plausible structural properties may still fail to reproduce individuals' social and political environments. Here, we generate LLM-based political networks using personas sampled from the nationally representative ANES 2020--2022 Social Media Study and use respondents' survey-reported partisan exposure as an empirical benchmark for the political environments generated for the corresponding personas. We first characterize the structural properties and political assortativity of the generated networks, and then compare ego-level partisan exposure with the survey. We find that the structure and political assortativity of the generated networks vary across experimental configurations. More importantly, when comparing generated partisan exposure with the survey, some models generate stronger separation between Democratic and Republican egos than reported in the survey, with higher same-party and lower cross-party exposure, while others mainly underestimate Republican exposure across ego groups. Moreover, networks with similar political assortativity can still generate different partisan environments around the same personas, including among networks with plausible structural properties. Finally, generated networks show group-specific errors, under- or overestimating exposure to either Democrats or Republicans for specific social groups. These findings show that evaluating LLM-based social simulations only through aggregate outcomes can hide errors in the local environments through which those outcomes emerge, highlighting the importance of validation at both aggregate and individual levels.
ER -