Scientists Invent Algorithm That Can Predict Depression Diagnosis From Your Facebook Updates

sittiphong/Shutterstock.com

There are plenty of articles and research papers out there showing a connection between social media use and mental illness – but what if those same social media sites can be used to diagnose cases of depression before an individual seeks clinical attention? 

It turns out your Instagram filter can be a surprisingly good indicator of whether or not you are depressed, and now computer scientists at Stony Brook University and the University of Pennsylvania have invented an algorithm that uses Facebook language to predict a user's diagnosis of depression.

"What people write in social media and online captures an aspect of life that's very hard in medicine and research to access otherwise," H. Andrew Schwartz, study author and computer scientist at Stony Brook University, said in a statement. "It's a dimension that's relatively untapped compared to biophysical markers of disease. Considering conditions such as depression, anxiety, and PTSD, for example, you find more signals in the way people express themselves digitally."

The algorithm, described in a paper published in Proceedings of the National Academy of Scientists, was built using 524,292 Facebook updates, some of which were from individuals who were later diagnosed with depression. Researchers singled out the words and phrases most frequently used and categorized them into 200 topics to identify so-called "depression-associated language markers". The language of the depressed group could then be compared to that of the control group to spot patterns between the two. 

To test their invention, the researchers analyzed the content and frequency of 683 city residents' Facebook posts. Of those, 114 Facebook users had a depression diagnosis in their medical records, and they found that the algorithm could identify them thanks to specific language markers. In fact, its ability to predict a depression diagnosis was on par with screening surveys benchmarked against medical records.

"Social media data contain markers akin to the genome," Johannes Eichstaedt, study author and University of Pennsylvania post-doc, explained. "With surprisingly similar methods to those used in genomics, we can comb social media data to find these markers. Depression appears to be something quite detectable in this way."

Language markers associated with emotional, cognitive, and interpersonal processes (including hostility, loneliness, rumination, and sadness) could all help predict depression up to three months before an official diagnosis. As previous research has shown, the algorithm found that people with depression were more likely to use first-person singular pronouns like Imy, and me. They were also more likely to use words associated with depressed moods (tearscrypain), loneliness (missmuchbaby), hostility (hateughfuckin), anxiety (scaredupsetworry), and rumination (mindalot).

People who were later diagnosed with depression were also likely to post longer posts more often, with their average annual word count 1,424 words higher than the control group. However, unlike older studies, the algorithm did not find that depressed users were more likely than others to post at specific times of the day. 

With roughly 6 percent of the US adult population (16 million) affected by depression in any given year and fewer than half of those receiving adequate treatment, the researchers hope language patterns on social media could serve as a red flag to encourage depressed users to seek treatment – of course, whether you want your social media account monitoring your mental health is another matter entirely.

"There's a perception that using social media is not good for one's mental health," Schwartz added. "But it may turn out to be an important tool for diagnosing, monitoring, and eventually treating it. Here, we've shown that it can be used with clinical records, a step toward improving mental health with social media."

Comments

If you liked this story, you'll love these

This website uses cookies

This website uses cookies to improve user experience. By continuing to use our website you consent to all cookies in accordance with our cookie policy.