Abstract: As AI, and especially large language models, moves into clinical workflows and everyday well-being, health care is shifting from clinician-led treatment toward AI-mediated, self-directed care. These systems promise accessibility and emotional support, but they also introduce quieter risks. Clinicians over-trust routine AI outputs, patients cede judgment about their own experiences, and models trained on normative data frame different ways of thinking and communicating as problems to be fixed. I approach these questions as a design researcher who has spent more than fifteen years building technology with, rather than for, autistic people and their families. That work has taught me to ask whose "normal" a system encodes and who bears the cost when it fails. Drawing on studies in emergency care, mental health self-reflection, and family communication with minimally verbal autistic children, as well as work evaluating AI with the communities most exposed to its harms, this talk examines what it takes to design health AI that preserves agency and respects neurodiversity. I argue for two shifts. To preserve agency, we must move from seamless automation to intentional cognitive friction that keeps people's own judgment in play. To respect neurodiversity, we must move from normative correction to scaffolding that supports people's own goals and ways of communicating. I close with design principles that center lived experience and keep technology in the periphery of the human connection at the heart of care.
Biography: Hwajung Hong is a Tenured Associate Professor of Industrial Design at KAIST, where she directs the Data, Interaction, and Design (DxD) Lab and is affiliate faculty at the KAIST College of AI. She creates conceptual and computational tools for inclusive AI experiences. Grounded in participatory design, her lab builds generative AI systems for mental health and well-being, designs assistive technologies with neurodivergent people and people with disabilities, and audits algorithm-driven platforms for accessibility and fairness. It also develops benchmarks and sociotechnical audits to evaluate how AI models perform and affect people in the real world. With Google's support, it explores human–agent collectives in which people orchestrate teams of AI agents. Her research appears at ACM CHI, CSCW, FAccT, and NeurIPS, and has received Best Paper Awards at CHI 2025 and CHI 2026. She is a recipient of multiple Google research awards and was an invited speaker at an ICML 2026 workshop, presenting "Beyond the Average: Orchestrating Human–AI Co-Creativity in Industrial Design." She is currently a Visiting Scholar in Computer Science at Stevens Institute of Technology, hosted by Prof. Jina Huh-Yoo.
Hosted By: Jina Huh-Yoo, SoC Associate Professor
Upcoming Seminars:
