Unobtrusive stress assessment in knowledge work using real-life environmental sensor data

Prolonged work stress has an extensive negative impact across modern society. In recent years, it has become an increasing issue, specifically in cognitively demanding knowledge-intensive professions. To address the global necessity of the timely detection and reducing of work stress, sensor-based automated methods for measuring stress are emerging. To date, most stress assessment studies have been conducted in laboratory environments investigating the short-term effects of artificially induced stress. However, in the real world, stress is caused by a wider variety of factors over time, implying a need for more information with long-duration experiments in actual daily work life. Moreover, individuals perceive stress differently, and adaptation to personal characteristics is required for the best results.

This dissertation proposes a scientifically novel way of detecting work stress using machine learning and unobtrusive sensors in different types of knowledge work environments. The applicability of indoor environmental quality and human motion sensor data for continuous work stress assessment of individual employees was examined by conducting three longitudinal real-life experiments and an online survey. The results suggested that the developed stress monitoring system can help in assessing perceived work stress on a daily basis. Moreover, the survey results revealed that the use of environmental sensors for continuous work stress assessment is acceptable, and knowledge workers are willing to share their stress-indicative data to promote well-being at work. The presented findings enable automated follow-up of employee stress, suitable for the future development of personalized well-being solutions to avoid excessive work stress at individual and organizational levels.