Donghyun Lee AI Group
Trustworthy AI for High-Stakes Forecasting and Simulation
Social Science & AI at HUFS · Seoul, South Korea
Welcome to the Donghyun Lee AI Group. Established in 2017, we are based at Hankuk University of Foreign Studies (HUFS) in Seoul.
We study when AI predictions and simulations can be trusted — and when they should be withheld — in high-stakes physical and social systems. Our core program advances forecasting and Physics-AI for spatiotemporal systems under distribution shift, rare events, and imperfect observations. We emphasize calibration, external validation, failure detection, abstention, and decision utility.
🔬 Research Areas
- Methodological Core — Trustworthy Forecasting & Physics-AI: Uncertainty quantification, interpretable modeling, regime-shift evaluation, spatial and temporal holdouts, selective prediction, and reusable reliability-audit tools.
- Domain I — Environmental & One Health Systems: Early warning for freshwater CyanoHABs, water-quality risks, and infectious diseases under climate shifts, hydrological change, surveillance bias, and imperfect observations.
- Domain II — AI-Based Social & Behavioral Simulation: Evaluating the validity and reliability of LLM-based synthetic agents through human benchmarks, subgroup fidelity, cross-model sensitivity, and explicit non-use rules.
We welcome motivated Master’s and Ph.D. students and interdisciplinary collaborations. Read our alumnus interview with Sangwon Chae of NH Bank.
Contact Prof. Lee at donghyun.lee [at] hufs.ac.kr.