Seminar: Centering People in the Design of AI-Powered Applications
PhD Candidate, University of Michigan
Tuesday, February 16, 2021
9:30am - 11:00am
AI is now prevalent in both everyday and high-stakes software applications. However, end-users regularly encounter undesirable experiences from AI. A fundamental challenge is that existing software practices fall short when creating AI for diverse human needs. AI development prioritizes automation and task efficiency over other human values. User experience (UX) design methods fail to support the designer's role in shaping key AI components such as training-data, labels, and learning models. That means an "AI-first" approach will fail to effectively integrate AI features within diverse human experiences.
In this talk, I provide an overview of the systems and techniques I have developed for centering people in the design of AI-powered applications. I will first discuss how designing AI experiences requires considering analogous human cognitive models. I will then talk about the role of end-user data as a design probe for UX and AI professionals to co-design AI experiences. Finally, I will offer solutions for UX designers to incorporate diverse human needs within AI application design.
Hari Subramonyam is a PhD candidate in the School of Information at the University of Michigan. His research focuses on ways to operationalize the vision of human-centered AI. By combining technical HCI work with qualitative studies of AI software development in practice, he brings multiple perspectives to bear on the cross-disciplinary problem of HAI. He has interned at Microsoft Research, Adobe Research, and Xerox PARC. His work has received two Best Paper awards at CHI. Hari holds a bachelor's degree in Telecommunication Engineering from Visvesvaraya Technological University in India, and a master's degree in Information from the University of Michigan.