Today’s approaches in data-driven interface design translate observations in user behavior into interface features. However, little consideration is given to data models that are the computational foundations of these interactions. I will introduce end-to-end techniques that a) builds computational representations that capture the diverse and nuanced task context and b) use those representations as building blocks for designing interfaces. I will illustrate three different techniques with examples from understanding the landscape of using web-scale cooking instructions, understanding multiple users’ step-by-step demonstrations of a 3d modeling task, and designing voice interactions for tutorial videos.
Minsuk Chang is a Ph.D. student in the School of Computing at KAIST. His research in HCI focuses on techniques for discovering, capturing, and structuring task context from user interaction data to create novel learning opportunities in the wild. He has previously interned at Adobe Research, Autodesk Research, and Microsoft Research.