Promoting Digital Wellbeing Through Real-Time State Classification of Psychophysiological Sensor Networks

Matthew Pike, Christopher Roadknight

The Second Neuroadaptive Technology Conference, July 2019

Digital wellbeing, Psychophysiological sensor networks, Deep learning, Real-time state classification, User state prediction, Neuroadaptive interfaces

Abstract

Boundaries between digital experiences and daily activities have blurred as technology becomes more pervasive. This study explores the application of psychophysiological sensor networks for real-time user state classification to promote digital wellbeing. Using deep learning (DL), we propose a novel pruning algorithm to enhance operational efficiency and classification accuracy in constrained mobile environments. By leveraging physiological data from sources such as ECG, fNIRS, EEG, and EDA, DL models can predict cognitive states like mental workload, fatigue, and task engagement. Neuroadaptive interfaces can then mediate user interactions with digital systems, improving focus, reducing interruptions (e.g., mobile notifications), and enhancing productivity. The study demonstrates the impact of correlated activity pruning (CAP) in refining multi-layer networks for efficient real-time classification of user states.