Simulating Judicial Decisions with LLMs: How Public Opinion on Social Media Shapes Labor Law Outcomes

ACL ARR 2025 February Submission7770 Authors

Abstract: This paper explores how social media discussions influence judicial decision-making in labor disputes. Using 309,642 comments on labor market conditions from Chinese social media platform Douyin and 10,000 representative labor case outcomes, we analyzed key labor issues and their sentiment patterns, revealing growing dissatisfaction with labor practices. Through a simulation experiment with Large Language Models (LLMs), we examined the impact of public opinion on judicial decisions. Our findings show that social sentiments significantly influence judicial outcomes, with a stronger effect on cases involving lower-skilled occupations. Additionally, different LLMs exhibit varying sensitivities to public opinion, with legal-specific models displaying the highest sensitivity, contrary to expectations. Notably, introducing public sentiment substantially alters the judicial decisions of certain LLMs, particularly in cases related to labor rights and lower-skilled workers. This study highlights the potential of social media discourse to shape judicial fairness, especially in labor disputes.

Paper Type: Long

Research Area: Ethics, Bias, and Fairness

Research Area Keywords: model bias/fairness evaluation, policy and governance

Contribution Types: NLP engineering experiment

Languages Studied: Chinese

Submission Number: 7770