Considering Humans in Acquiring Data, Building Machine Learning Models, and Creating Data-sensemaking Support Systems

Tuesday, February 18, 2020 - 11:00
Blakeslee Room
Ray Hong

Although machine learning (ML) models leverage large-scale data to drive major innovations in nearly every domain, cognitive efforts required for humans to draw meaningful insights and informed decisions have grown significantly.  My research focuses on building a human-centered, cognitively efficient, and interactive environment where humans can better leverage ML models and data to draw insights and make decisions. My research, Machine-aided Data Sensemaking (MaD), aims at exploring the following three questions: (1) How can we design a Data labeling environment to acquire better quality datasets using fewer human resources? (2) How can we support data scientists to build more interpretable and trustworthy ML Models? Finally, (3) How can we create data-sensemaking support Systems to facilitate a user or a group of users’ reasoning about the data and/or models?  In this talk, I will introduce my approaches that show how applying theoretical and technical constructs in Human-Computer Interaction (HCI), Information Visualization (InfoVis), and Computer-Supported Collaborative Work (CSCW) can improve the way people interact with data, ML models, and data-sensemaking systems. Then I will explain how I would extend my future research agenda with the faculty and students in CS@SFSU.



Ray Hong is a Visualization for Machine Learning Postdoctoral Fellow at New York University. He received his Ph.D. in the Department of Human-Centered Design & Engineering (HCDE) at the University of Washington. Before he joins HCDE, he worked at Samsung Research as a researcher for 6 years, working on building user interfaces for consumer electronics devices. His research contributes to the field of Human-Computer Interaction (HCI), Computer Supported Collaborative Work (CSCW), and Information Visualization (InfoVis). He has published 19 papers in top-tier conferences closely related to his research. Seven of his papers are published at CHI and two are published at CSCW, one of which was recognized with the Best of CHI honorable mention award and another received the best of CSCW honorable mention award. Aside from these, he holds 9 granted US patents. He contributes to the research of Machine-aided Data Sensemaking (MaD) by (1) conceptualizing and building novel designs that can improve upon our current ways of interaction with data and ML models, and (2) extending our empirical understanding of how a user or a group of users interact with data and models to accomplish their targeted tasks in varying scenarios.