The goal: a digital early warning system
At the Mobiliar Lab for Analytics at ETH Zurich, an interdisciplinary team is working to pre-empt such states of exhaustion by developing a digital early warning system that uses machine learning to detect stress in the workplace in real time. “Our first step was to find out how to measure the effects of social pressure and interruptions – two of the most common causes of stress in the workplace,” says psychologist Jasmine Kerr. Kerr is driving the project forward together with mathematician Mara Nägelin and computer scientist Raphael Weibel.
The three doctoral students are all lead authors on a recent study, details of which appeared in the journal Psychoneuroendocrinology. They used a university platform to recruit 90 participants, who agreed to take part in an experiment lasting just under two hours. To conduct their experiment, Kerr, Nägelin and Weibel transformed the Decision Science Laboratory at ETH Zurich into three group office environments. Each workstation was equipped with a chair, a computer with monitor and kits for collecting samples of saliva.
Playing the parts of employees at a fictional insurance company, the participants were asked to perform typical office tasks, such as typing up information from hand-written forms and arranging appointments with clients. While they did so, the researchers observed their psychobiological responses. At a total of six points during the experiment, the participants rated their mood on questionnaires, while a portable ECG device continuously measured their heartbeat. The researchers used the saliva samples to measure the concentration of the stress hormone cortisol.
In line for a promotion
For their experiment, the researchers divided the participants into three groups and exposed each group to a different level of stress. All groups were given the same workload. In the middle of the experiment, all participants were visited by two actors masquerading as representatives of the insurance company’s HR department. For participants in the control group, the actors staged a sales pitch dialogue, while in the two stress groups they pretended to be looking for the most suitable candidates for a promotion.