Title: MindMap: A Learning-Based Policy-Conditioned Review System for Mental Health Text Annotation

Speaker: Fan Weiguo, Henry B. Tippie Distinguished Chair Professor of Business Analytics, University of Iowa

Moderator: Chen Gang, Young Associate Research Fellow, Department of Information Management and Business Intelligence

Time: July 14, 2026, 10:00–11:30 a.m.

Venue: Guoshun Campus, Room 105, Li Dak-Sum Building

Abstract: Mental health text annotation is a highly specialized classification task in which label determination depends on how ambiguous evidence is selected, structured, and interpreted. Although multi-agent AI systems increasingly adopt parallel architectures, they typically differentiate agents through manually assigned role prompts, isolated tool access, or sequential workflows. We propose MindMap, a novel learning-based policy-conditioned review architecture that redefines multi-agent parallelism. Rather than relying on workflow-based isolation, MindMap deploys parallel Policy Instance Agents that operate simultaneously over a shared epistemic rationale graph. To construct these agents, we introduce a data-driven reasoning-chain modeling approach. Different judgment policies, such as supportive, conservative, and broad-recall policies, are not manually assigned virtual roles, but are empirically derived from patterns of evidence use and annotation rationales in the training data. Across six predictive backbone models, MindMap improves the average Macro-F1 from approximately 0.768 to 0.942. Controlled ablation experiments show that the dominant performance gains arise from the empirical modeling of reasoning chains, while the parallelization of multiple policies produces significant improvements for low-confidence samples.

Bio: Fan Weiguo is the Henry B. Tippie Distinguished Chair Professor of Business Analytics at the Henry B. Tippie College of Business, University of Iowa. He received his Ph.D. from the Ross School of Business at the University of Michigan in 2002. His research interests include artificial intelligence, information retrieval, data mining, text analytics and natural language processing, social media analytics, and business intelligence. His research has been published in MIS Quarterly, Information Systems Research, Management Science, Journal of Management Information Systems, Production and Operations Management, INFORMS Journal on Computing, IEEE Transactions on Knowledge and Data Engineering, Information Systems, Information Systems Journal, Communications of the ACM, Information & Management, Journal of the American Society for Information Science and Technology, Information Processing & Management, and Decision Support Systems, among other journals. He has published more than 300 peer-reviewed journal articles and conference papers, seven of which have been recognized as Web of Science ESI Highly Cited Papers. He was named a Clarivate Highly Cited Researcher in 2025. His work has received more than 23,000 citations on Google Scholar, with an h-index of 73 and an i10-index of 223. He has also provided consulting services to Fortune 500 companies over many years and has served as an expert reviewer for multiple government agencies and research funding organizations.