Hanover — Over the past few years, a computer scientist at Dartmouth College has helped lead a wave of mobile sensing research across the country with applications in psychiatric treatment, workplace performance, education and the military.
Andrew Campbell’s work sparked national headlines in 2014 when he and a team of collaborators announced StudentLife, an app that uses the built-in microphone, GPS and other sensors built into smartphones to measure students’ health.
Back then, a study of 48 students who ran the app for 10 weeks revealed connections between their sleep and social habits and their health and academic performance.
“Big Brother Is Feeling You,” a story in Forbes quipped at the time.
In the following years, the mobile sensing field has practically exploded, pointing toward a not-so-distant future when people will depend on their phones not just for driving directions and restaurant reviews but also for help with intimate details of their mental and physical health.
The work of Campbell’s team has spawned the CampusLife consortium, a group of researchers at universities in the United States and Europe who use StudentLife and other apps to study student health in ways far beyond the Dartmouth professor’s original intention.
Separate studies by tech companies, other academics and U.S. military and intelligence agencies now are plumbing the possibilities of mobile sensing, as well.
“It’s really cool,” Campbell said during a recent interview in his Sudikoff Hall office, “because we really almost started the field here in 2013.”
Among many new endeavors outside of university campuses, StudentLife, an open-source app available for anyone to use for free, has seen use in psychiatric hospitals in New Hampshire and New York.
In a study at New Hampshire Hospital published in October, Campbell and a team of researchers found correlations between violent episodes from patients and behavioral trends captured by smartphones that gathered sensing data and self-reported information .
At Long Island Jewish Medical Center, in Queens, N.Y., many of the same scientists recently gathered information with smartphones on the sleep schedules, conversation habits and movements of patients with schizophrenia. Researchers are still analyzing data from that year-long, 100-person experiment in hopes of finding a way to predict schizophrenic relapse.
Dror Ben-Zeev, a University of Washington professor who collaborated on both studies, said he had been particularly proud of the results at New Hampshire Hospital, the state psychiatric hospital in Concord, given the condition of its subjects.
People there are often “quite ill,” he said, and “you’re approaching them at their worst state.”
Nevertheless, the researchers succeeded in engaging the patients in a week of data collection through cellphone apps that gathered sensing data and administered surveys.
The results didn’t show a correlation between the sensing results and violent ideation, but the surveys did, and in some cases helped alert clinicians to a developing problem, Ben-Zeev said.
Although researchers are still parsing data from the Long Island hospital study, preliminary results are promising.
The group, which included nearly a dozen collaborators, including Ben-Zeev, Campbell and some of his graduate students, published a paper this fall that follows the smartphone habits of a few specific patients before they experienced psychotic episodes.
One patient began reporting increased symptoms of psychosis to the smartphone’s in-app survey. Another patient’s sleep patterns shifted, showing increased device use between midnight and 6 a.m.
Another paper from the group published this year showed that the Long Island hospital study’s data could reliably predict patients’ performance on psychiatric exams that clinicians normally have to perform. “That was pretty cool,” Ben-Zeev said, noting that the researchers had been able to make these health assessments just by having patients carry around phones with the app installed.
Ben-Zeev also noted that, in analyzing sensing information, researchers have to be careful about the assumptions they make when searching for connections between data and behavior.
If, for instance, a participant’s phone shows more activity at night, it might be happening because a roommate or partner has started using the device, rather than as a result of changing sleep patterns. Or, in another case, the phone’s GPS might determine that a person is now staying home for most of the day.
“Who’s to say that spending more time at home is a bad thing?” Ben-Zeev said. “We sort of project what we anticipate this would mean onto the individual. But what if the person got a job and is now working from home as a programmer, and is actually functioning better than they used to?”
As they handle the challenges inherent to mobile data interpretation, researchers are looking forward to their next step: technology that uses its measurements to take direct action. “I think the objective is to see (if) not only can we monitor but can we intervene?” Ben-Zeev said.
Significant barriers remain, however, before consumers will be able to download apps that not only monitor their health, job performance and social interactions but also offer suggestions for improvement.
Campbell said he expected that large and thorough studies would be necessary to iron out any glitches before computer programs can begin making direct interventions into people’s behavior.
“Until you have a study of a hundred thousand people,” he said, “you’re not going to have the confidence to say, ‘This phone can make an intervention.’ ”
Campbell predicted that it would take the resources of a large company, such as Google or Apple, to “scale up” sensing technology to the point where it can reliably serve the general public.
And as it happens, the computer science professor recently completed a research fellowship with Google, where, among other projects, he researched smartphone sensing at Verily Life Sciences, a research entity owned by the internet giant.
Campbell and his collaborators are far from the only scientists exploring mobile sensing technology.
The Defense Advanced Research Projects Agency, the research arm of the U.S. military that created early versions of the modern internet, is funding the development of a mobile app that will passively assess soldiers’ readiness for battle.
“Warfighter Analytics using Smartphones for Health,” as the project is known, will use mobile devices already carried by U.S. personnel to provide up-to-date health information to soldiers and their doctors, according to the agency’s website.
The U.S. government is also collaborating with Lockheed Martin, the University of Memphis, the University of Notre Dame, the University of Southern California and Campbell on a project called Mosaic that will use mobile data to monitor workplace performance.
The 42-month endeavor will fund several different research teams investigating how stress affects workers’ decision-making and productivity, according to an announcement from the Intelligence Advanced Research Projects Activity, the government research agency that is heading the project.
In addition to “task performance,” the project summary says, researchers are hoping to identify “organizational citizenship behaviors” that promote an effective workplace culture, along with “counterproductive work behaviors” that don’t.
Many of Campbell’s collaborators are taking sensing technology in different directions on their own.
Those include John Kane, a psychiatrist and professor at Hofstra University School of Medicine who worked with Campbell and Ben-Zeev on the study in Queens.
Just last week, Kane and another group of researchers working with the Japan-based company Otsuka Pharmaceutical announced that the U.S. Food and Drug Administration had approved its first digital pill — a product called Abilify MyCite that combines aripiprazole, a known antipsychotic medication, with a sensor that tracks whether patients are taking it.
“It gives the clinical team, and potentially the caregiver as well, information about how the patient is doing,” Kane said of the sensing pill, which sends information to a smartphone app. “We have a hard time knowing for sure when people are taking their medicine, so this is a very helpful tool.”
Back in his Dartmouth office, Campbell applauded the spread of sensing technology and the medical advances it offered.
The potential benefits, he said, were evident all around him. Dartmouth students, according to a recent campus climate survey, report having been diagnosed with depression at a rate of 19 percent — a figure that represents an increase from the past.
Roughly a quarter of students said depressive symptoms had harmed their academic performance.
The 58-year-old Campbell, a native of England who ran his 13th marathon this year, has his own personal connection to the issue. Decades ago, his brother dropped out of university while battling with depression. Although the brother later re-enrolled, the experience inspired Campbell to apply his field to the circuitry (such as it is) of the human mind.
Sitting back in his chair, Campbell took a sip of Darjeeling tea and gazed ahead at his work table, which was strewn with cellphone components. “How can we help these people?” he asked.
Rob Wolfe can be reached at rwolfe@vnews.com or at 603-727-3242.