Jun Yuan is a postdoctoral research fellow at the Digital Technology for Democracy Lab at the University of Virginia. He holds a PhD in data science from the New Jersey Institute of Technology and an MS in mathematical sciences from Clemson University. His research promotes democratic values in digital technologies through interpretability-centered approaches. He developed a red teaming strategy that injects bias into machine learning models while evading current interpretation techniques like LIME and SHAP, enabling non-experts to audit models without training data access. He has contributed to projects examining the limitations of state-of-the-art large language models, focusing on reasoning gaps.
His interdisciplinary work includes collaborations with university administrative offices on admission and course evaluation analytics, and with social scientists and responsible AI experts on fair hiring systems. He also designed five interactive ranking tools (TRIVEA, CARE, Recourse, HILSA, TalkToRanker) that enhance transparency and fairness in rank-based decision-making.
