With ubiquitous devices becoming more and more intelligent, they can robustly capture low-level behaviors such as daily activities and physical health metrics. However, current devices are still far from understanding and modeling users’ high-level behaviors such as mental wellbeing. A large body of research has emerged with the recent advances in mobile sensing and machine learning techniques. However, explainability and personalization are the two critical challenges before achieving the deployability of behavior models. This talk presents our new explainable and personalized behavior modeling techniques for depression detection. As applications of behavior models, this talk will also cover our recent progress of a novel intervention technique for smartphone overuse that is closely related to mental wellbeing, as well as a new interaction customization technique that fulfills individual needs.
Xuhai “Orson” Xu is a 4th-year PhD student at Information School from University of Washington, advised by Prof. Anind K. Dey, and Prof. Jennifer Mankoff. His research interests span across human-computer interaction, ubiquitous computing, and applied machine learning. His research leverages sensing data from everyday devices and develops explainable and personalized machine learning techniques to model daily behavior related to human wellbeing. Based on behavior models, he also designs new behavior change intervention methods and novel interaction techniques for wellbeing promotion. His research has won Best Paper Honorable Mentioned Award in CHI 2018 and CHI 2020, and has been covered at various media outlets such as ACM News, Hackster.IO, and UW News.