LLM Agents Empower Simulation for the Future of Work: Universal Basic Income, Employment, and Well-being

ACL ARR 2025 February Submission7518 Authors

Abstract: UBI policy remains one of the most widely studied topics in economics, drawing significant attention for its potential social and financial impacts. However, real-world UBI experiments are costly and constrained in scale, limiting their feasibility for large-scale analysis. The emergence of LLM-based society simulations offers a promising alternative, enabling detailed economic and social modeling at a lower cost. We propose an agent-based simulation where Large Language Models (LLMs) role-play individuals in a virtual economy to evaluate UBI policies. By integrating real-world data, our model captures complex human behaviors, including financial decisions and mental well-being. We successfully replicated outcomes from five real-world UBI trials across economic and mental metrics, with ablation studies confirming that LLM role-playing agents produce more realistic and insightful simulations. Our work demonstrates how LLM-powered simulations can advance UBI research and inform policy design. Codes are available here: https://anonymous.4open.science/r/LLM-UBI-7837

Paper Type: Long

Research Area: Computational Social Science and Cultural Analytics

Research Area Keywords: Large Language Models, Simulation, Universal Basic Income, Agents, Agent-Based Modeling, Employment, Well-being

Contribution Types: NLP engineering experiment

Languages Studied: English

Submission Number: 7518