【SDS Colloquium Series】Diversified Bayesian Learning: Optimal Control with Multiple Biased Information Sources
November 10, 2025 SDS Colloquium Series
SDS Colloquium Series | |
| Topic | Diversified Bayesian Learning: Optimal Control with Multiple Biased Information Sources |
| Speaker | Jussi KEPPO, Provost's Chair Professor and Head, Department of Analytics & Operations, NUS Business School Research Director, Institute of Operations Research and Analytics, National University of Singapore |
| Host | Zizhuo WANG, Professor & Associate Dean (Education), School of Data Science, CUHK-Shenzhen |
| Date | 10 November (Monday), 2025 |
| Time | 11:00 AM - 12:00 PM, Beijing Time |
| Format | Hybrid |
| Venue | Room 401, Dao Yuan Building |
| Live on WeChat Channels |  |
| Language | English |
Abstract | |
| We consider a decision-maker (DM) who can acquire signals from multiple biased information sources to learn about a hidden state before making an earning decision. Unbiased signals are also available but come at a higher acquisition cost. The DM jointly optimizes both learning (information acquisition) and earning decisions to minimize expected loss. This problem is motivated by applications such as medical diagnostics, revenue management and financial investments, where decisions often rely on multiple biased information sources. In contrast to existing literature that primarily focuses on stopping problems with unbiased information, we develop a Bayesian decision framework that accommodates general earning decisions, and models multi-source learning using a hierarchical Bayesian network, explicitly capturing intrinsic biases. We fully solve the model and explicitly characterize the optimal acquisition policy: For small budgets, cost-effectiveness dominates, prioritizing cheap biased sources. As budgets grow, diversification across all biased sources becomes optimal as it mitigates risks from biases. With large budgets, acquisition shifts to costly unbiased sources as biased signals offer diminishing value due to their limited accuracy. We also illustrate the model through two use cases: a disease nowcasting example using real data that shows how budget allocation shifts with bias, cost, and precision, and a demand forecasting example that demonstrates the performance advantage of the proposed policy (learning algorithm) over benchmarks in simulation. | |
Biography | |
| Professor Jussi Keppo teaches risk management and analytics courses, and directs analytics executive education programs at NUS Business School. He is also the Head of the Department of Analytics & Operations at NUS Business School and Research Director of the Institute of Operations Research and Analytics at NUS. Previously, he taught at the University of Michigan. He has several publications in the top-tier journals such as Journal of Economic Theory, Review of Economic Studies, Management Science, Operations Research, and Journal of Business on topics such as investment analysis, banking regulation, learning, and strategic incentives. His research has been featured also in numerous business and popular publications, including the Wall Street Journal and Fortune. Professor Keppo’s research has been supported by several Asian, European, and US agencies such as the National Science Foundation. He serves on the editorial boards of Management Science, Operations Research, and Journal of Risk. He has consulted several startups, Fortune 100 companies, and financial institutions. | |
