COM Research Seminar: The New Data, the Old Problem: Validation in the Age of Synthetic Social Science

20 Jul 2026 (Mon)

On 26 June 2026, Prof. Winson Peng from Michigan State University delivered a seminar titled “The New Data, the Old Problem: Validation in the Age of Synthetic Social Science.”

In the seminar, Prof. Peng examined the methodological evolution of social science from the age of made data, produced through researcher-designed surveys, experiments, and interviews, to the age of found data, captured through platforms, devices, and digital traces. He argued that the field is now entering an age of synthetic data, where large language models and AI agents can generate responses, interactions, and simulated social environments. The presentation contended that despite these methodological shifts, the central problem remains the same: validation. The key question, he suggested, is not simply whether new forms of data are larger, more naturalistic, or more flexible, but whether they can validly support the claims social scientists want to make. Using LLM-driven social simulation as an empirical illustration, Prof. Peng demonstrated why behavioral plausibility is only a starting point, and proposed a layered view of validation that considers the data-generating process, construct representation, behavioral correspondence, inference, and generalization. The talk emphasized that the larger task is to develop claim-aware validation strategies that clarify what synthetic data can, and cannot, tell us about human communication and social life.

Faculty members and doctoral students from the Department of Media and Communication attended the session and engaged in a lively and intellectually stimulating discussion. In responding to questions from the audience, Prof. Peng shared valuable insights on computational social science and the methodological challenges posed by synthetic data in communication research.

Dr. Winson Peng (Ph.D. in Communication, City University of Hong Kong) is a Professor in the Department of Communication at Michigan State University and Editor-in-Chief of Human Communication Research. He previously served as Associate Editor of the Journal of Communication and as Guest Editor for Communication Methods and Measures, Asian Journal of Communication, and the Journal of Communication. His research interests include computational social science, health communication, political communication, and mobile analytics. He has published more than eighty articles in highly ranked SSCI and SCI journals.