Advertisement
Tech by VICE

How Your Phone Can Tell If You're Depressed

The unseen and largely unsung components of our devices are often what most accurately reflect our personal lives in bits of data.

by Jordan Pearson
Sep 22 2014, 5:35pm

Our phones are full of data that says far more about us than we might imagine, and in all kinds of unexpected ways. Often overlooked components like accelerometers and light sensors can be used to identify you, for example. Now, researchers at Dartmouth College have figured out how to use smartphone sensor data to track users' moods, which they think could be used to identify symptoms of depression.

The aim of their experiment, described in a paper presented at last week's Ubicomp 2014 conference, was to see if autonomous data collection could help indicate depression, loneliness, and stress in college students. 

The researchers gave smartphones loaded with an app they designed, StudentLife, to 75 Dartmouth students and collected their accelerometer, microphone, light sensor, and GPS data for 10 weeks. The data collection was totally passive and offered no feedback to the students; the phones acted as silent black boxes, constantly sucking in data and calculating the results.

By tracking the amount of sleep subjects got, how many social conversations they had (measured by tracking microphone activity and filtering out lecture hours), and how much they moved around during the day, the Dartmouth researchers found a significant correlation between the results garnered from sensor data and self-reported surveys on mental health, as well as clinical definitions of depression.

"We identify strong correlation between automatic sensing data and a broad set of well-known mental well-being measures, specifically, PHQ-9 depression, perceived stress (PSS), flourishing, and loneliness scales," the authors write. "Results indicate that automatically sensed conversation, activity, mobility, and sleep have significant correlations with mental well-being outcomes."

How passively collected data can be used to analyze mood. Image: Dartmouth

A college campus made the perfect living lab for their experiment, the researchers write, because (and I'm paraphrasing here) college can be really stressful and shitty. 

Remember those first days of school every September? They were always so full of promise. Maybe you had a pristine binder you swore you'd actually take notes in this time around. Maybe you were really planning on buying some expensive textbooks instead of noodles and beer to forget that this is a choice you have to make in the first place. 

And then, a month later, maybe you were clinging to a bottle of cheap whisky and having stress-sex in the library. The binder, presumably, was lost to the sands of time and too many morning coffee spills.

The study's results support such anecdotal tales of collegial psychic degradation, according to the researchers. While most subjects reported few depressive symptoms at the beginning of the trial, by the end of the term, the students were getting less sleep, being less social, and staying put more often. Importantly, they reported feeling more stressed out, lonely, and depressed. As you might guess, these sharp dips in mood were highly concentrated around midterm and final exam periods.

While being a helpful reminder that mental health services on college campuses are really important, the study also drives home an increasingly salient point out our connected and quantified existence: Our devices are an extension of ourselves, for all intents and purposes. Our behaviours and their hardware create something utterly unique in a weird kind of organic-digital symbiosis, like a data fingerprint.

While hacking iPhone gyroscopes for personal information isn't really a common exploit right now, sensor data is an often overlooked area of data privacy and cryptography. It's important to remember that although locational and communicative information usually get all the attention in terms of privacy, the unseen and largely unsung components of our devices are often what most accurately reflect our personal lives in bits of data.