ISyE Seminar Series: Shixiang (Woody) Zhu
"Human-Centered Decision Making: From Curating Recommendations to Auditing Decisions"
Shixiang (Woody) Zhu
Assistant Professor at Heinz College of Information Systems and Public Policy, and a faculty affiliate
Wilton E. Scott Institute for Energy Innovation
Carnegie Mellon University
About the Seminar:
In many high-stakes applications — from infrastructure planning to healthcare and public policy — decisions are not made by algorithms alone but are mediated by human judgment. This talk presents a human-centered framework for decision support that emphasizes two complementary stages of human-algorithm interaction: recommendation and assessment. First, I introduce Generative Curation, a novel paradigm for designing recommendation systems that generate a set of diverse, high-quality candidate actions. Rather than optimizing a single best solution, generative curation learns a distribution over actions that maximizes the likelihood that the best item in a small, manageable portfolio aligns with the decision-maker’s unobserved preferences. This approach balances quantitative optimality with qualitative diversity, enabling algorithms to better support human judgment in contexts where objectives are partially specified or inherently subjective. Second, I present CREDO (Conformalized Risk Estimation for Decision Optimization), a new framework for auditing user-specified decisions under uncertainty. CREDO provides a distribution-free upper bound on the probability that a given decision is suboptimal by leveraging generative modeling, conformal prediction, and inverse optimization geometry. This enables decision-makers to assess the robustness of their chosen actions without requiring access to the full objective function or decision distribution. These two approaches highlight the shift from automation to collaboration in AI-driven decision-making — enabling systems that generate better choices and evaluate existing ones, while keeping humans firmly in the loop.
Affilitated Papers:
- "Conformalized Decision Risk Assessment"
- "Balancing Optimality and Diversity: Human-Centered Decision Making through Generative Curation"
About the Speaker:
Dr. Shixiang (Woody) Zhu is an Assistant Professor at Carnegie Mellon University’s Heinz College of Information Systems and Public Policy, and a faculty affiliate at the Wilton E. Scott Institute for Energy Innovation. His research lies at the intersection of machine learning, operations research, and statistics, with a focus on human-centered decision-making under uncertainty. Woody develops methodologies that bridge predictive modeling and optimization with real-world human judgment, particularly in high-stakes domains such as energy systems, public policy, and critical infrastructure. He has collaborated with utility companies, national laboratories, and public-sector partners to translate data-driven insights into practical decision-making tools.
If you wish to be added to the ISyE Graduate Seminar Series emailing list, please email Event Coordinator Emily Rice at [email protected].