Synthetic Social Data: Trials and Tribulations
Published in arXiv preprint arXiv:2510.19952, 2025
Large Language Models are increasingly used to simulate human participants and generate synthetic data for computational social science. This work critically assesses validity, representativeness, bias amplification, and privacy implications, offering guidelines for responsible research with synthetic persona data.