Current status and future prospects of AI technologies for stress detection

This review explores the application of artificial intelligence (AI) for stress detection. Chronic stress poses substantial health risks and remains a persistent societal issue. While effective prevention requires accurate and timely assessment, current evaluation methods often rely on subjective indicators, which can be biased under high cognitive load and are difficult to deploy for real-time feedback. The integration of AI technology with physiological and physical signals holds promise for addressing the challenges associated with subjective indicators. Specifically, we discuss how deep learning has advanced the field by shifting from manual feature engineering to representation learning, enabling the extraction of complex patterns directly from raw data. Despite these advancements, major challenges remain, including performance degradation in real-world settings, the “black box” nature of models, and privacy concerns. In response, we examine emerging solutions such as multimodal sensing, explainable AI (XAI), and Federated Learning, alongside the potential role of large language models (LLMs). We conclude by discussing the future transition from experimental validation to practical application, highlighting just-in-time adaptive interventions (JITAIs) as a key framework.