February 13, 2026, Webb Hall 1100

We study A/B experiments that are designed to compare the performance of two recommendation algorithms. Prior work has observed that the stable unit treatment value assumption (SUTVA) often does not hold in large-scale recommendation systems, and hence the estimate for the global treatment effect (GTE) is biased. Specifically, units under the treatment and control algorithms contribute to a shared pool of data that subsequently train both algorithms, resulting in interference between the two groups. In this paper, we investigate when such interference may affect our decision making on which algorithm is better. We formalize this insight under a multi-armed bandit framework and theoretically characterize when the sign of the difference-in-means estimator of the GTE under data sharing aligns with or contradicts the sign of the true GTE. Our analysis identifies the level of exploration versus exploitation as a key determinant of how data sharing impacts algorithm selection.

Jingyan Wang is a Research Assistant Professor at the Toyota Technological Institute at Chicago. She was previously a postdoctoral fellow at Georgia Tech, affiliated with the School of Industrial and Systems Engineering and the Algorithm and Randomness Center. She received her PhD in Computer Science from Carnegie Mellon University and her B.S. in Electrical Engineering and Computer Sciences from UC Berkeley. She uses tools from statistics and machine learning to study the evaluation of people (in hiring, admissions, and peer review) and of algorithms (in recommendation systems). Her interdisciplinary research has been published in statistics, machine learning, artificial intelligence, human computation, and economics and computation. She was the recipient of the Best Student Paper Award at AAMAS 2019, and was selected as a Rising Star in EECS and in Data Science.