The rise of wearable devices such as the Apple Watch and Fitbit offer us the ability to turn our daily lives into an never-ending catalogue of interpretable data: how many steps we take, the number of calories we consume, our REM sleep cycles, and even the health of our hearts.
The researchers say their mobile-sensing system, which consists of fitness bracelets, sensors and a custom app, can measure employee performance with about 80 percent accuracy.
The system monitors physical and emotional signals that employees produce during the day and uses that data to create a performance profile over time that is designed to eliminate bias from evaluations. The technology may be in its infancy, researchers say, but it could signal the beginning of a new era of virtual assistants that will redefine our relationships with intelligent machines.
Someday, they hope their technology might function like “Samantha,” the artificially intelligent assistant in the 2013 movie “Her” (minus the unhealthy romance), providing someone with valuable insights about their productivity, stress levels during meetings or lifestyle habits that impact their ability to perform their job.
“I know when I’ve had a bad week, but in the moment I might not know the facts that have contributed to me having a poor week,” said Andrew Campbell, a professor of computer science at Dartmouth College. “We set out to discover whether there was a way to move the needle from an almost backward way of assessing people’s workplace performance to using more objective measures.”
“We want to use that information to empower workers to tell them whether they’re being influenced by levels of stress or sleep or other factors that may not be immediately obvious to them.”
Campbell said he came up with the idea for a mobile sensing system after spending a year as a visiting faculty member at Google, where he was surprised that one of the world’s premier technology companies still relied upon a traditional performance review involving subjective write-ups from employees and their supervisors. Around the same time, he said, he read about researchers predicting depression using mobile sensing data. If it was possible to predict someone’s mental health by analyzing their social media feeds and smartphone data, Campbell wondered, could similar data be leveraged to improve employee performance evaluations?
To find out, he assembled a team of researchers to pore over “passive sensing data” from 750 workers (both supervisors and non-supervisors) inside a high-tech company and a management consulting firm over a one-year period ending in April. The workers were fitted with a wearable fitness tracker that monitored heart functions, sleep, stress, and measurements such as weight and calorie consumption, as well as a smartphone app that tracked their physical activity, location, phone usage and ambient light.
Location beacons placed in the home and office measured participants time at work and breaks from their desk, giving researchers a comprehensive window into their day from one hour to the next.
The information was processed by cloud-based machine-learning algorithms that classified performance using factors such as the amount of time spent at the workplace, quality of sleep, physical activity and phone usage.
Campbell’s partner, Pino Audia, a professor of management and organizations at the Tuck School of Business at Dartmouth, said research shows that conscientious people, who are often more detailed-oriented and disciplined, tend to be more productive. What the research does not explain, he said, is what habits make someone conscientious in the first place, leaving a gap in knowledge that researchers hoped to fill.
“Very often when people try to detect what drives performance, they rely on personality, which actually reveals little about someone’s ability to do their job well,” he said. “Evaluations can be biased because they are infused with stereotyping of people or political influences inside an office. But when you can extract a pattern over weeks and months, we can be more certain that assessment is objective and neutral."
What did researchers find?
Despite studying populations with two different working styles — one largely in an office or lab and the other group often working from home or traveling — the results showed, perhaps not surprisingly, that high performers tended to have lower rates of phone usage. They also experience deeper periods of sustained sleep and are more physically active than their lower performing colleagues.
Researchers discovered that high-performing supervisors tended to be more mobile during the day, but they visited a smaller number of distinct places during their working hours. High-performing non-supervisors, meanwhile, tend to spend more time at work during the weekends, researchers said.
The team’s findings were published in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies.
Both Campbell and Audia — who have committed to keeping their subject’s data completely private — say that a more refined version of their assessment technology could be used to empower workers. Future versions, they said, could be tailored to individual jobs and provide workers with meaningful information about changes in their mental well-being during meetings or suggestions for reducing stress each week.
But they also acknowledge that the valuable private data could prove volatile if it falls into a company’s hands without employee consent. Campbell suggested there might be a middle ground, such as companies offering incentives to employees who opt into a program that treats precise assessment data as one tool among several for evaluating performance.
Barring that, he said, the technology might best be left in workers’ hands alone.
“If there was any point down the road where I could have an application on my phone that could provide an objective assessment of my performance, that might be an incentive for workers to use it," he said. “Imagine being able to say, ‘Here’s the evidence that I deserve to be promoted or that my boss is standing in my way.’"
“I can’t really look into a crystal ball, but I’m hopeful this passive sensing technology will be used to empower the workforce rather than used against them," he added.
Dig Deeper: Personal tech + Privacy
Want to learn about how to keep your personal information private? Check out our curated list of stories below.
Americans receive more than 5.2 billion automated calls in a month. But there are new apps to help stem the deluge.
As tech columnist Geoffrey Fowler slept, a dozen marketing companies used his iPhone to learn his number, email, location and IP address.
Changing privacy default settings means you’ll get less personalization from some services, but it can slow down the number of eerie on-the-nose ads driven by data siphoned by major companies.