Yoon (Kyung) Shon

ElectroCalm: Investigating Physiological Effects and User Experience of Multi-Node Transcutaneous Electrical Stimulation for Stress Modulation

Yoon Kyung Shon, Pablo Paredes.

This project explores how electrical stimulation can be designed as a wearable calming interface beyond traditional clinical use.

To be submitted to the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT) 2026.

EMSeption: Exploring Electrical Muscle Stimulation for Reshaping Attitudes Toward Unhealthy Foods

Yoon Kyung Shon, Roksana Khanom, Nhi Tran, Jun Nishida, Pablo Paredes.

This study examines how electrical muscle stimulation (EMS) can influence food-related behaviors and attitudes, expanding the role of embodied feedback in persuasive HCI design.

To be submitted to the Journal of Medical Internet Research (JMIR) Formative Research.

When Awareness Isn’t Enough: Exploring Physical Interventions for Nail-Biting

Yoon Kyung Shon, Kyungyeon Lee, Jun Nishida, Pablo E. Paredes.

Nail-biting is an automatic behavior that some people struggle to stop voluntarily, even when they are aware of it and want to quit. We explored four wearable haptic probes with active nail-biters and found they favored motor interference over perceptual feedback, accepting reduced momentary motor agency to interrupt the behavior.

In Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA ’26). https://doi.org/10.1145/3772363.3798489

De-Stress Me: A Personal Digital Assistant

This system integrates physiological sensing and AI-driven feedback to deliver real-time stress monitoring and personalized interventions for daily well-being.

Wearable Stress Tracker for Blind and Low-Vision People

This glove-based wearable tracks stress through GSR sensing and haptic feedback on the inner wrist, enabling hands-free, non-visual monitoring for blind and low-vision users.

Federated Learning for Adaptive Client Weighting

This project proposes a federated learning approach that improves client selection and reweighting by tuning α and β parameters based on data distribution. Using CIFAR-10 and CIFAR-100 datasets, the method enhances both global and personalized models, outperforming FedAvg and FedProx through random search and Bayesian optimization.