Finnish researchers measure stress in knowledge workers with AI

VTT's tool utilises an artificial intelligence-based algorithm to analyse the movements, detect changes in movement patterns and convert the data into easily understandable metrics. It envisions that organisations can use the tool to measure stress levels among staff, identify causes of stress and take action early, thus reducing stress-related sick-leaves.

While previous studies have shown that employees who are stressed tend to move the mouse pointer quicker, for longer distances and less precisely, the research centre reminded that it is important to calibrate the measuring tool for each employee to account for individual traits in movement patterns.

Atte Kinnula, senior scientist at VTT, said the research team opted to perform the calibration by asking employees to self-report whether or not they felt stressed on a particular day. Currently, he revealed, the tool can learn individual traits from roughly 30 reports, roughly a third of the number of reports required by comparable studies.

“For the solution to be practical, this learning phase must still be improved,” he conceded.

Taught with data collected from real workplaces in the past four years, the artificial intelligence tool can currently detect stress at an accuracy of 71 per cent on a daily and 84 per cent on a quarterly basis. Scientists at VTT believe the accuracy could be improved by adding other computer-related behavioural data, such as typing tempo.

They have also developed an organisation barometer that presents the results in an easy-to-read, positive and engaging way – all the while preserving the privacy of employees.

VTT stated that one key premise for ensuring transparency and preventing the misuse of data is making sure employees have ownership of the data they produce and the ability to choose whether or not the results or shared. Another is that anonymisation should guarantee not only that individual employees cannot be identified from the shared data, but also that employees who chose not to share their data remain anonymous.

As part of the development project, the scientists conducted an international survey of more than 1 200 knowledge workers. It found that 80 per cent of workers would be willing to share their data anonymously with their employers if it was used to support their own and colleagues’ wellbeing.

The research centre is testing solutions also in other work environments, including factories, with a view to introducing tools to monitor and manage stress in different occupational groups.