About
I study why socially assistive robots fail to survive beyond research labs — and how to fix it. Through a 28-month co-design process and a 6-month in-the-wild deployment at a pediatric speech therapy center, I found that the core reason: creating and updating robot interaction requires researcher involvement, making real world adoption unsustainable.
My current work addresses this directly: I am building an LLM-powered authoring social robot system that enables non-technical users — therapists — to design, customize, and iterate on robot interactions without engineering support. The goal is a system where domain experts are authors of robot and AI behavior, not passive operators of researcher-defined scripts.
Prior to my Ph.D., I conducted research on contextual affect modeling at Korea University (ACM ICMI 2022), and worked as a research intern at KIST (2023) on multi-robot systems for hospital environments and human-robot trust.
Research Projects
An LLM-powered authoring system that gives non-technical clinicians direct control over AI-driven robot interactions — enabling them to create, customize, and iterate on therapy content without engineering support.
A 6-month real-world deployment with 9 children in an active speech therapy clinic — therapists operated the robot independently across repeated clinical sessions, surfacing what pre-scripted AI systems cannot sustain.
Developed and evaluated a multi-robot system for hospital isolation wards, incorporating telemedicine, emergency alerts, and delivery services.
Investigated whether users prefer robots that act proactively or those that follow specific verbal commands — with implications for designing AI systems where end users maintain meaningful control.
Explored how perceived hierarchies between robots in a team affect human trust and service evaluation.
Comparison study on human perception versus CNN-based contextual emotion recognition models, evaluating AI system perception against human baselines.
Publications
Contact