New ACM Study Improves Everyday Stress Monitoring While Reducing User Burden
August 6, 2026 A new study from the Health SciTech Group introduces a context-aware approach to everyday stress monitoring that combines smartwatch photoplethysmography (PPG), smartphone context, active learning, and reinforcement learning.
The system determines when it is most useful to ask a participant for a brief ecological momentary assessment, helping improve model performance while reducing unnecessary interruptions. In the study, online adaptation increased the Random Forest F1 score from 0.21 to 0.32, while adding contextual information increased it to 0.36. Personalized models also improved area under the curve by as much as 0.10.
The paper, “Enhancing Performance and User Engagement in Everyday Stress Monitoring: A Context-Aware Active Reinforcement Learning Approach,” was authored by Seyed Amir Hossein Aqajari, Ziyu Wang, Ali Tazarv, Sina Labbaf, Salar Jafarlou, Brenda Nguyen, Nikil Dutt, Marco Levorato, and Amir M. Rahmani.