Finding out that your phone has been monitoring you more closely than your closest friends can cause a certain kind of anxiety. Not in a dramatic way. Just the typical aftermath of a typical day, such as how quickly you type a text, how frequently you backspace, and whether or not you answered three calls. These seemingly insignificant behaviors can subtly indicate depression, sometimes even before the person experiencing it is aware of it, according to research on a phenomenon known as “digital phenotyping.”
The concept is no longer science fiction. Based on data that most of us never consider, it’s more akin to a slow-motion mood reading. It is simply described as continuous, moment-by-moment tracking of human behavior using personal devices by a Harvard-based research team. In actuality, this means that the sensors on your phone—such as the accelerometer, GPS, and screen taps—become a type of behavioral fingerprint. Furthermore, it turns out that depression leaves its own mark.
Think about mobility. During a depressive episode, people frequently stop wandering. They spend more time at home. There are fewer locations, shorter trips, and less variation in daily movement in GPS logs. Researchers at Harvard’s Onnela Lab based part of their entire monitoring platform, Beiwe, on this seemingly insignificant detail. It’s difficult to ignore how unglamorous the signals themselves are as you watch this develop in the research literature—they’re simply absence rather than dramatic confessions. fewer texts were sent. shorter responses. An unused phone for an extended period of time.
There are hints in typing itself. Longer pauses between words, more corrections, and slower keystrokes have all been connected to psychomotor slowing, a known characteristic of depressive episodes. This was the focus of a now-defunct company called Mindstrong, which developed clinical tools that treated typing rhythm in a manner akin to a digital neurological exam. The rhythm of your thumbs on a screen acting as an early warning system, long before someone answers the phone to inquire about how you’re doing, is subtly unsettling.
Additionally, voice is powerful. According to a Wall Street Journal article on behavioral biomarkers, patients with depression typically pronounce vowels less clearly, and their camera-captured smiles are smaller and nearly undetectable. Researchers studying acoustic patterns have found that suicidal people who speak in a breathy rather than tense voice have higher rates of reattempt. All of this is not precise. It’s not all fate. However, when considered collectively, it implies that our habits and bodies transmit something that our conscious minds haven’t yet caught up to.
It’s tempting to describe this as exciting, and to be fair, it is—early detection may result in earlier intervention and support before a crisis fully develops. However, underneath it all is a tension. Who is the owner of this data? What would happen if an advertiser, an employer, or an insurer had access to a pattern that appeared to be depressing? Mood-tracking sensors were not considered when current privacy laws were created, and ethicists are still working to close that gap.
Overreach is another risk. Not all quiet weekends are indicative of depression. Not all slow typists are having difficulty. Particularly when it comes to different cultures or lifestyles, algorithms trained on behavioral noise may misinterpret normal human variation as pathology. At its best, technology should enhance rather than take the place of a clinician’s judgment.
Will your phone be aware of this before you are? Maybe. Its ability to guess is improving. The degree to which you trust the individuals in possession of the data will likely determine whether that is comforting or slightly concerning. This is the true unanswered question, more so than the algorithms themselves.
