Stress Monitoring in the Era of AI Boom - Where Should We Begin?

Abstract

This Systematic Literature Review (SLR) explores the growing convergence of social sciences, medical sciences, and computer engineering in the development of Artificial Intelligence (AI)-enabled multimodal stress monitoring systems. We aim to provide a comprehensive overview of the current state of AI-enabled stress monitoring and identify key areas requiring further research and development. Our analyses reveal a fundamental gap in current literature: many studies lack a clear theoretical definition of stress and a standardized methodological design to measure the multifaceted nature of stress. Although our findings highlight a popular method of integrating multimodal data including self-evaluations, physiological signals, behavioral reactions, and responses to environmental stressors, the data sources are often inconsistent and only partially integrated due to personal, contextual, and technical constrains. Moving beyond current focus on maximizing the technical challenges, this review also identifies other key areas, i.e., improving model’s accuracy, realizing real-world usability, enhancing data quality, safeguarding ethical use of data, and establishing user trust that require future research to develop more rigorously.