BiAffect app: can typos give insight into your mental health?
A team of researchers at the Center on Depression and Resilience at the University of Illinois at Chicago is working on technology that could monitor users’ mood and cognition—important indicators of mental-health stress—by tracking their typing patterns with an iPhone app called BiAffect. Initial research has found it is possible to predict episodes of mania and depression among users with bipolar and major depressive disorder based on changes in their typing habits.
For instance, a manic episode may be preceded by rising numbers of typos, faster typing, more frequent use of the “delete” key or tremors detected by the phone’s accelerometer, which measures the device’s tilting and orientation. During depressive episodes, users withdraw from their personal technology and tend to send short, infrequent messages.
“It doesn’t track what you type, but how you type it,” says Dr. Alex Leow, an associate professor from the university’s College of Medicine and lead researcher on the project.
‘ Rather than waiting until the patient shows up in somebody’s office for intervention, you could do the intervention in real time through the same device that’s monitoring their symptoms. ’
—Dr. Olusola Ajilore, an associate professor of psychiatry on the BiAffect team
The BiAffect study is part of a larger trend in psychology of attempting to measure how the brain functions using “digital exhaust” from users’ daily interactions with technology. There are over 10,000 mental-health-related apps on the App Store, according to the National Institute of Mental Health’s most recent count.
The trouble is that evidence-backed apps like BiAffect that can substantiate their claims of effectiveness make up just 3% of that total, says Dr. Adam Haim, an NIMH expert on mental-health apps. “There’s a lot of promise,” he says, “but there’s a lot of false promise out there as well.”
The NIMH considers the development of digitally delivered interventions a priority because they have the potential to make treatment available to more people, decrease medication switch rates, reduce hospitalizations and perhaps lower medical costs in the long run.
However, apps like BiAffect that passively collect data come with a host of ethical concerns, mostly related to privacy. False positives can have costly consequences for users if an app fails to properly de-identify data or is not compliant with federal health-data protection standards, says Dr. Haim. People “need to understand the potential implications of opening themselves to the release of data to a third party,” he says.
BiAffect’s aim is to be “as unobtrusive as possible,” says Dr. Leow. After downloading the app, users opt into the study and allow the specially coded BiAffect keyboard to replace their iPhone’s default version. The app then operates behind the scenes whenever a person uses their phone, compiling a trove of objective data.
The goal with this approach, the researchers say, is to make an unbiased real-time assessment of users’ mental state in their natural environments. Typical mental-health assessments are made by polling patients on their state of mind through questionnaires or in-person meetings in clinical settings, and they are prone to self-reporting bias and other collection flaws.
Researchers made BiAffect available to the general public on the App Store in March and it has since yielded over 8,150 hours of data from more than 1,300 users. The app was created as part of the Mood Challenge, a contest funded by the Robert Wood Johnson Foundation that challenged researchers to find new ways to use Apple Inc.’s open-source app development platform, ResearchKit, to study mood disorders.
BiAffect’s creators intentionally didn’t limit the app for use in clinical settings, though they believe it could be most effective when used by mental-health professionals to monitor their patients. Such use is probably several years off, said Dr. Olusola Ajilore, an associate professor of psychiatry on the BiAffect team.
But researchers hope the app will eventually allow clinicians to offer just-in-time interventions. “Rather than waiting until the patient shows up in somebody’s office for intervention, you could do the intervention in real time through the same device that’s monitoring their symptoms,” he says.