SIGCHI Korea
Local Chapter Summer Event

Don't Forget the Engagement: Coping with the Threat to Human Expressivity in the Age of AI

Sang Won Lee | Virginia Tech, USA & NAVER AI Lab

In this talk, I explore how AI can negatively affect human engagement in creative practices and diminish our capacity for expression by devaluing the act of creation itself. The current trajectory of AI development prioritizes efficiency, effectiveness, and convenience. While these goals are beneficial in many contexts, they can conflict with the nonlinear, seemingly inefficient engagement that characterizes creative practice. Drawing on examples from writing education and the design of intelligent personal assistants, I illustrate how narrowly focused AI systems can undermine opportunities for meaningful human expression. I advocate for a shift in perspective: rather than using AI solely to optimize tasks, we should design it to unlock new expressive dimensions and actively foster creative participation.

multi-modal learning and fairness in healthcare ai

Seungbae kim | University of South Florida, USA

Seungbae Kim, Ph.D. is an Assistant Professor at the Bellini College of Artificial Intelligence, Cybersecurity and Computing at the University of South Florida. He leads the Connected Social Artificial Intelligence Lab (CSAIL). Dr. Kim's research focuses on AI for social good, with applications in mental health, communication science, fairness in machine learning, and social network analysis. His work integrates multimodal machine learning, graph neural networks, and computer vision to address societal challenges across diverse platforms. He earned his Ph.D. in Computer Science from UCLA and previously held research roles at UCLA, KAIST, and Seoul National University.

In this talk, I will present how integrating diverse data streams—such as audio, video, gestures, physiological signals, and textual transcripts—enables the development of richer and more accurate models for the healthcare domain, while also addressing challenges related to data scarcity and bias. Specifically, I will introduce novel architectures that fuse co-speech gestures with acoustic features to enhance speech recognition for patients with communication disorders, as well as methods for combining facial and acoustic features to improve depression detection. To ensure these powerful models remain equitable under strict privacy constraints inherent to healthcare, I will also discuss reinforcement learning–driven synthetic data generation and fairness-aware pruning techniques designed to reduce performance gaps across subgroups.

visual analytics for responsible ai

Minsuk Kahng | Yonsei University, Korea | Website: https://minsuk.com

Minsuk Kahng began his position as an Assistant Professor in the Department of Computer Science and Engineering at Yonsei University in March 2025, after working as a Senior Research Scientist at Google DeepMind in the People + AI Research (PAIR) team. His research focuses on building interactive data visualization tools that help people responsibly develop AI systems. His work has led to deployed technologies (e.g., LLM Comparator for Google, ActiVis for Facebook) and recognized with prestigious fellowships from NSF and Google, as well as a dissertation award from Georgia Tech. He also worked as an Assistant Professor at Oregon State University and received his Ph.D. from Georgia Tech. 

While large tech companies invest heavily in training LLMs such as GPT and Gemini, these models are inherently imperfect. They can fail unexpectedly, reflect biases, or produce unsafe outputs. How can we help practitioners address these issues when developing AI systems? In this talk, I present my research on building novel visual analytics tools that help people interactively analyze model behavior. I draw upon my experience developing tools like LLM Comparator, designed to support model evaluation at Google. I also discuss my vision for human-centered workflows for curating high-quality datasets, making AI not only more powerful, but also more responsible.