A computer-science professor at Dartmouth College is building a smartphone application that can detect users’ levels of happiness, stress, and loneliness, he says, with the hope of helping students monitor their mental health.
The app, called StudentLife, draws on sensor data from smartphones to “infer human behaviors,” says the professor, Andrew Campbell. It was inspired partly by the mental-health struggles that Mr. Campbell’s brother experienced while in college. The professor also wants to test his hypothesis, based on classroom observations, that students’ fluctuating stress levels correspond to their behaviors.
“I feel as if there’s a divide between faculty and students. We don’t want to delve too deeply in their lives, but we want to be as helpful as we can,” Mr. Campbell says. “Is there anything that could give us an objective view?”
To find out, he ran a 10-week experiment in the spring of 2013 that compared behavioral data gathered by the app with self-reported assessments of mental health by 48 student subjects. Doctors and psychologists at Dartmouth and the University of Texas at Austin, who helped design the study, recommended which health surveys to use.
Using the GPS, motion-sensor, and microphone features built into smartphones, as well as machine-learning algorithms that infer behaviors, StudentLife recorded participants’ movement, sleep, and number of face-to-face conversations each day. Users uploaded their class assignments to a shared Google calendar so Mr. Campbell could track their workloads.
“I was really interested, in days they had lots of deadlines, what were their stress levels?” he says.
The results, according to Mr. Campbell, showed that the behavioral data correlated significantly with participants’ mental health and academic performance. Findings suggested, for example, that students who sleep more or have more conversations are less likely to be depressed, and that students who have higher grade-point averages tend to be less physically active.
Victor Schwartz, medical director of the Jed Foundation, which promotes emotional well-being among college students, has reservations about the app’s diagnostic capabilities. “It’s not really telling you if you’re depressed. It measures activities that he’s trying to associate … with being more likely you’re depressed,” Dr. Schwartz says. “I think there could be really very valuable targeted utilities for this kind of thing, but I think we need to be aware that there could be pitfalls.”
Objective behavioral feedback may motivate some users, Dr. Schwartz believes, but it could also make others feel guilty. “For every positive way of this playing out, I could easily imagine a negative scenario where they’re getting pressure and they’re getting pissed off.”
Mr. Campbell is concerned about the security of data gathered by StudentLife. He thinks it should be up to students to decide with whom they share their data. “I err on the side of being incredibly conservative when it comes to private data,” he says. “You own this data, but if you want to share it with your counselor, that’s completely up to you.”
Using an app like StudentLife as part of a treatment plan designed by a mental-health professional could be beneficial, Dr. Schwartz acknowledges. One of the first signs of bipolar disorder, for instance, is irregular sleep patterns, and so a counselor who notices changes in a patient’s sleep data could make a diagnosis more quickly, he says.
“In those well-thought-out circumstances, it could be extremely helpful,” Dr. Schwartz says. “I could see where this would be a really nifty tool.”
Mr. Campbell’s write-up of the experiment was nominated as a “best paper” last month at the ACM International Joint Conference on Pervasive and Ubiquitous Computing. A second study, planned for the spring, will provide feedback to participants during the course of the experiment.
The finding Mr. Campbell found most surprising? “No correlation between grades and attendance in class,” he says.