[Abstract]

With the dramatic advancements in artificial intelligence (AI) technologies, many people desire to use AI in their everyday work. However, it is worth emphasizing that although people expect to interactively acquire the assistance they need from AI in a timely manner, satisfying their expectations may be challenging because in-depth analysis and understanding of user tasks and AI models are required. In addition, visual interfaces play an important role as a medium of communication and interaction to realize human–AI collaborations. In this talk, I present how we can foster collaborations between humans and AI with two case studies in different domains. In the first case study, I describe how we designed a visual interface that enables experts in the congestion control domain to effectively analyze traffic congestion and broadcast congested roads with the help of a deep-learning model. In the second study, I introduce a novel visual interface that allows novice users to design mobile user interfaces that elicit real-time assistance from AI models. The experiment results indicate that UI designs created with our visual interface received significantly higher scores than those without. Moreover, the users of our visual interface reported that using our visual interface is more fun, effective, efficient, and comfortable compared with other design tools without AI assistance.

 

 

[Biography]

Sungahn Ko is an associate professor in the School of Computer Science and Engineering at POSTECH, Pohang, Korea. His research interests include Vvisual analytics, HCI, and language models with a focus on Human-AI Collaboration. He received the doctoral degree in electrical and computer engineering from Purdue University in 2014.