Naegelin, Mara
[Author];
id_orcid0 000-0002-0490-7510
[Author]
;
von Wangenheim, Florian
[Contributor];
Schinazi, Victor
[Contributor];
Ferrario, Andrea
[Contributor]
The Stress in Your Data: Towards the Automated Detection of Work-Related Stress
Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
Work-related stress has become a pervasive issue in modern-day working environments, precipitated by the digital and economic transformations of the last decades and compounded by the recent pandemic. Stress can have severe detrimental effects on employees' mental and physical health, leading to enormous costs for healthcare systems and economies. Consequently, solutions to effectively and efficiently prevent and manage work-related stress are needed. Monitoring stress in a continuous, unobtrusive manner can make employees more aware of their stress, improve self-regulation and allow them to take preventative, tailored action before it becomes chronic and negative health effects begin to manifest. The field of stress detection research thus investigates the development of automated stress detection systems by leveraging machine learning (ML) methods and various data sources connected to the physiological or behavioural stress response. However, this research faces a number of open challenges on the way from laboratory experiments to field studies to practical deployment in real-life work environments. This thesis aims to address some of these challenges along a series of empirical studies. The majority of previous works have focused on physiological signals such as heart rate variability (HRV), which can now be collected with increasingly accurate and comfortable consumer-grade wearables. For work environments, data on mouse movement characteristics and keystroke dynamics data resulting from computer interaction have been suggested as particularly suitable options for automated stress detection due to their non-invasiveness and straightforward collection. However, the number of lab and field studies investigating mouse and keyboard data for stress detection is limited, and a consensus on their predictive potential has yet to be reached. Furthermore, existing lab studies aiming to simulate work-related stress may not adequately reflect real-life conditions. Therefore, the first empirical article included in this ...