Exploring the role of Emotion Recognition algorithms in relation to Burnout and in supporting Digital Care and Patient Empowerment

This doctoral thesis examines the role of emotion recognition algorithms, with a specific focus on Speech Emotion Recognition (SER), in relation to burnout syndrome, treated here as a focal but bounded construct, and to the broader objectives of digital care and patient empowerment. Burnout, officially recognised by the World Health Organization (WHO) as an occupational phenomenon, is increasingly prevalent across diverse professional sectors and is associated with severe psychological, organisational, and societal costs. Despite its impact, diagnostic practices remain largely dependent on self-report measures, which, while valuable, are limited by subjectivity, cultural biases, and their inability to capture dynamic changes in real-world contexts. The research adopts a multidisciplinary perspective that integrates psychology, occupational health, and artificial intelligence. It is guided by the premise that chronic stress and burnout manifest in subtle but measurable changes in vocal production, reflected in acoustic, prosodic, and temporal features. By developing a digital application for collecting structured voice samples and psychometric data, this work investigates the potential of voice biomarkers to complement established diagnostic instruments, such as the Maslach Burnout Inventory (MBI) and the Oldenburg Burnout Inventory (OLBI). At the core of the thesis is a conceptual model that situates SER within a layered framework, encompassing data collection, feature extraction, classification through machine learning, validation against psychometric benchmarks, and application within digital health platforms. This structure ensures that computational methods are not only technically accurate but also psychologically interpretable and clinically relevant. Empirical analyses demonstrate that AI-driven voice analysis can enhance the validity and sensitivity of burnout assessment, providing non-invasive, scalable, and continuous monitoring capabilities that traditional tools alone cannot achieve. The study also addresses the ethical and practical dimensions of implementing SER in healthcare and occupational settings. Considerations of privacy, transparency, algorithmic bias, and user trust are positioned as fundamental prerequisites for adoption, highlighting that technological progress must be accompanied by ethical safeguards and regulatory compliance. SER is therefore presented as a promising assessment direction rather than a clinically validated tool. By situating emotion recognition within the broader discourse of digital health, this research suggests that SER may enrich psychological assessment, inform occupational well-being strategies, and contribute to patient empowerment in carefully bounded ways. Ultimately, the thesis suggests that emotion recognition algorithms may serve not only as scientific tools for advancing affective computing but also as potentially useful instruments for informing healthcare delivery, supporting earlier and more personalised forms of care in appropriate contexts where appropriately validated and implemented.