Jian Kang

Bio: Jian Kang is a final-year Ph.D. candidate in the Department of Computer Science at the University of Illinois at Urbana-Champaign. He received his M.CS. degree from the University of Virginia in 2016 and B.Eng. degree from Beijing University of Posts and Telecommunications in 2014. His research interests lie in trustworthy learning and mining on graphs. His research works on related topics have been published at several major conferences and journals in data mining and artificial intelligence (e.g., KDD, WWW, CIKM, TKDE, TVCG). He is the recipient of Mavis Future Faculty Fellowship and three reviewer awards (ICML’20, ICLR’21, CIKM’21).

Talk Title: Algorithmic Foundation of Fair Graph Mining

Talk Abstract: Graphs are ubiquitous in many real-world applications. To date, researchers have developed a plethora of theories, algorithms and systems that are successful in answering what and who questions on graphs. Despite the remarkable progress, unfairness often occurs in many graph mining algorithms, hindering the deployment of graph mining algorithms in high-stake applications. To make graph mining process and its results fair, it is crucial to propose a paradigm shift, from answering what and who to understanding how and why. Four desired properties (i.e., utility, fairness, transparency and robustness) are called for in order to build an algorithmic foundation of fair graph mining. In this talk, I will present our efforts in building the algorithmic foundation of fair graph mining by addressing the tensions among the desired properties. First, we present a generic algorithmic framework to understand how the graph mining results relate to the input graph. Second, we systematically study the individual fairness on graph mining, including the measurement, mitigation strategies and cost. Finally, we analyze the root cause of degree-related unfairness in graph neural networks and present a family of debiasing algorithms, followed by my thoughts about the future work.