时   间：2025年12月16日（周二） 13:30-15:00

地   点：管理学院思源楼524室

主  题：Feature Selection for Promotion: Learning to Avoid Information Overload

主讲人：李骋怀 阿尔伯塔大学商学院助理教授

主持人：田林 复旦大学管理学院教授

Abstract：Motivated by practical applications in e-commerce, we consider a product promotion problem where a product showcase webpage is personalized for each arriving customer during a campaign. The product has various characteristics that may affect the customers' willingness to purchase, and each customer may weigh these characteristics differently. However, showing all the product characteristics to customers might lead to information overload, which means customers can be less interested in the product when facing too much information on the product. Based on the evidence from a field experiment, we build a feature-based logit model to capture the information overload effect in customer choice behavior, such as clicking. Faced with this probabilistic choice model with unknown parameters, we design a statistical learning algorithm that carefully perturbs the showing characteristics of the product and then exploits the estimated model to optimize the product display page for each arriving customer. By dedicating  √T periods to exploration, the algorithm leveraging the maximum likelihood estimation achieves the expected cumulative regret of order √T logT. Finally, we conduct a field experiment in collaboration with a medical agent selling medical equipment. In this field experiment, we randomly display different versions of a product webpage to arriving customers, where these versions differ in the amount of displayed product information. Observing the information overload phenomenon from the experimental data, we calibrate our model with the data and numerically test the performance of our proposed algorithm.

Bio: Dr. Chenghuai Li is an assistant professor in the Alberta School of Business at the University of Alberta. His work examines how emerging technologies — from blockchain to AI — can help address challenges in supply chains, retailing, food systems, and clean energy transition. He also develops innovative data-driven solutions to operational problems, such as pricing and promotion. He received his Ph.D. from the Fuqua School of Business at Duke University in 2025.