UnStressMe: Explainable Stress Analytics and Self-tracking Data Visualizations

Abstract

Self-tracking technology for behavior monitoring is prevalent in various aspects of human life. It enables users' activities tracking with data produced "in the wild", namely capturing real-world physical activity, sleep patterns, and stress levels, among others. Advanced new sensors integrated into commercial self-tracking devices have empowered a new era of sensing data exploration and self-improvement recommendations, aiming to enhance physical and mental well-being. However, the collected data and related inferred knowledge are not always well-explained or well-presented and discourage users' commitment leading to sensing devices' abandonment. To sustain user engagement with self-tracking technology for well-being, this paper introduces a comprehensive framework and respective fullstack web service called "UnStressMe" for the analysis of diverse data modalities tracked in the wild, the prediction of future stress behavior and the production and provision of personalized, model-agnostic explanations and interactive visualizations. We showcase the utility of our framework through a mental health use case, paving the way for explainable, transparent, and human-centric self-tracking technology.