Date: Tuesday, May 12; 12:00-1:00 PM, Arlington Campus
This Brown Bag session builds on the March 23 conference organized by CONTRA. The talk introduces a multi-agent simulation framework for modeling policy debates using persona-driven retrieval-augmented generation (RAG). Unlike standard large language model (LLM) outputs, the framework grounds autonomous agents in the specific corpus of each speaker (e.g., speeches, publications, and interviews) to ensure that arguments remain traceable to documented positions rather than generic training data.
To validate the framework, the recorded conference debate will be used as a benchmark to compare the synthetic output against this “ground truth.” The key question is whether a document-grounded multi-agent simulation can reproduce similar lines of argument, points of disagreement, and areas of consensus.
If successful, this approach could provide policymakers with a useful tool for anticipating critiques, identifying potential coalitions, and understanding likely areas of policy convergence before proposals enter public debate.
As part of this effort, the team plans to develop an AI model that simulates policy debates based on speakers’ documented positions. The simulated debates will then be compared with the actual conference discussions and analyzed after the event. This Brown Bag will present the results of this comparison and assess the reliability of AI for simulating policy discussions.
Speakers
Naoru Koizumi – Schar School of Policy and Government, George Mason University. Koizumi is Professor of Public Policy, Associate Dean for Research and Grants at the Schar School of Policy and Government, and co-director of CONTRA (Corruption, Networks and Transnational Crime Research Center). Her research focuses on health and medical policy, particularly kidney transplantation systems and global kidney trade and trafficking. Her work uses quantitative and computational approaches, including social network analysis and modeling, to examine organ transplantation systems and illicit organ trade.
Meng-Hao Li – Schar School of Policy and Government, George Mason University. He serves as Managing Director of the Center for Biomedical Science and Policy, Research Fellow at CONTRA (Corruption, Networks and Transnational Crime Research Center), and Adjunct Faculty at the Schar School of Policy and Government. He is a public policy scholar and data scientist whose research combines computational methods—such as network analysis, machine learning, and large language models—with policy analysis to address complex challenges in health and governance.
Yang Yu – Schar School of Policy and Government, George Mason University. Yu is a Ph.D. student in Public Policy at the Schar School of Policy and Government. His research focuses on data-driven public policy analysis and interdisciplinary studies related to health policy and biomedical systems. His work has contributed to projects examining kidney transplantation outcomes, organ trafficking networks, and health system analytics.