Automated stress detection might not be the office panacea it appears to be

The writer is a science commentator

Ping! Up pops a message from your boss, chasing the report you should have delivered yesterday. Other distractions come thick and fast: a colleague texts to say he is ill; another interrupts your frantic typing of the executive summary to remind you that an online seminar starts in five minutes.

You can sense your stress levels rising — and so, it seems, can your computer. Researchers in Switzerland have found that people under stress use a keyboard and mouse differently from their carefree colleagues. The technology of automated stress detection is intended to help individuals privately manage their own wellbeing, but it is not hard to imagine unintended consequences. Stress-tracking can theoretically allow employers to pick out those who flourish under pressure — and those who struggle.

Mara Nägelin and colleagues at the Chair for Technology Marketing and Mobiliar Lab for Analytics at ETH Zurich put about 90 volunteers into a simulated group office environment. All had to perform standard tasks, such as scheduling appointments or analysing data. Some were constantly interrupted by chat messages; others had to undergo a mock job interview. During the experiment, researchers tracked levels of the stress hormone cortisol and heart rate variability; they also monitored volunteers’ keyboard and mouse use via an app. The volunteers self-reported their stress levels at regular intervals.

Stressed volunteers, it turned out, typed faster and moved their mouse more erratically than less anxious peers, with typing and clicking patterns a more accurate reflection of self-reported stress than heart rate variability. “In general, [when you are stressed] you are faster but less precise in your movements, so you type faster and make more errors,” Nägelin, a PhD student, explains. Typing was a fitful, stop-start affair among the tense; in comparison, the less stressed tended to tap for longer periods, with extended breaks. Stressed volunteers using a mouse were more likely to overshoot their target and covered longer distances on screen.

The results were published last month in the Journal of Biomedical Informatics. The researchers nod to “neuromotor noise theory”: the idea that stress causes our motor skills to take a hit. They are now collecting data “in the wild” to refine their machine learning stress measurement model: 40 employees at the insurance company Swiss Mobiliar have volunteered to use the app at work. This, Nägelin says, will illuminate how stress varies between individuals and across different tasks: “The technology has a lot of potential, but there are many research gaps to fill.”

One gap is understanding how workers will act on the data. Erika Meins, director of the Mobiliar Lab, explains that the technology, currently only an academic project, is intended to help individuals manage their own wellbeing; users can choose to get a readout or prompt at a time that suits them. The main goal, Meins says, is to stop acute stress, which can boost performance, from turning into chronic stress, which damages health: “There are very effective ways of dealing with stress, such as slow-paced breathing or yoga. But the challenge is to employ them at the right moment.” A recent survey by her lab suggests a majority of workers are willing to try the tech, though many justifiably harbour privacy concerns.

Won’t companies want to identify who is coping well in the office — and who isn’t? Meins is adamant the digital stress app is about self-management, not surveillance: “This is not a monitoring system and the data does not go to the company. This is very important to us.” Stress-tracking data could also conceivably be used by unhappy employees to boost claims of unreasonable work demands. Meins says the lab is exploring how to deploy the tech ethically and responsibly.

Sir Cary Cooper, a professor of organisational psychology and health at Manchester university’s Alliance Manchester Business School, points out that many workers already know they are stressed. “If you’re fatigued and the app is telling you to take a break, that could be useful,” Cooper says. “But there are often more deep-rooted causes, such as unrealistic deadlines, unmanageable workloads and bullying bosses.” He is also wary of the tech’s Big Brother potential.

Unforgiving taskmasters may indeed salivate at the prospect of identifying the most resilient subordinates. In that case, fairness demands that the technology cuts both ways — and is also used to weed out those bullying bosses.