Title: Achieving causal fairness in bandit based recommendation 

Abstract 

In online recommendation, customers arrive in a sequential and stochastic manner from an underlying distribution and the online decision model recommends a chosen item for each arriving individual conditional on the state of the environment. It is imperative to develop online recommendation algorithms to maximize the expected reward while achieving user-side fairness for customers, i.e., customers who share similar profiles will receive a similar reward regardless of their sensitive attributes and items being recommended. We study how to leverage offline data, incorporate causal inference, and adopt soft intervention to model the item selection strategy in contextual bandits. We present the d-separation based UCB algorithm (D-UCB) which can reduce the amount of exploration needed to achieve low cumulative regret, and the fair causal bandit (F-UCB) for achieving the counterfactual individual fairness. As the offline data often contain confounding and selection biases, ignoring these biases in causal bandits could negatively affect the performance of online recommendation. We present approaches of estimating conditional causal effects and deriving their bounds in the presence of compound biases. We further study how the derived causal bounds affect regret analysis in contextual bandits.   

Biography 

Dr. Xintao Wu is a professor and the Charles D. Morgan/Acxiom Endowed Graduate Research Chair and leads Social Awareness and Intelligent Learning (SAIL) Lab in the Electrical Engineering and Computer Science Department at the University of Arkansas.  He was a faculty member in College of Computing and Informatics at the University of North Carolina at Charlotte from 2001 to 2014. He got his BS degree in Information Science from the University of Science and Technology of China in 1994, ME degree in Computer Engineering from the Chinese Academy of Space Technology in 1997, and Ph.D. in Information Technology from George Mason University in 2001. Dr. Wu's major research interests include data mining, privacy and security, fair machine learning, and recently causal bandits and causal representation learning. Dr. Wu has published over 180 scholarly papers and received several paper awards including PAKDD'13 Best Application Paper Award, BIBM'13 Best Paper Award, CNS'19 Best Paper Award, and PAKDD'19 Most Influential Paper Award. He has also served on editorial boards of several international journals and many conference program committees of data mining and AI.