The focus of intelligent systems is on "making things easy" for people through automation. However, for many cognitive tasks--such as learning, creativity, or sensemaking--there is such a thing as too easy or too automated. In such cases, making decisions about whether, what, and how to automate is a challenging HAI design problem. In this talk, I will outline my approach to designing AI experiences (AIX) by considering analogous models of human cognition. To exemplify my approach, I will present an AI-assisted tool I developed, called texSketch, for supporting learners during science text comprehension through active reading, and diagramming. I will then offer an overview of my work on AIX design processes and tools. I will conclude with my research agenda for developing AI systems for reasoning and decision-making, including designing for AI uncertainties in critical human tasks.
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 HAI's cross-disciplinary problem. Hari holds a master's degree in Information from the University of Michigan and a bachelor's degree in Telecommunication Engineering from Visvesvaraya Technological University in India.