Scientists Created an Algorithm That Diagnoses Depression by Instagram Pictures

Harvard and Vermont university researchers developed an algorithm that is able to diagnose depression by photographs that the users post on Instagram, The Next Web writes.

The scientists have studied about 44,000 photographs taken by 170 people, 70 of them diagnosed with clinical depression, and analyzed their accounts by activity level, number of likes, number of faces in the photographs and color characteristics of the images. This data was used to develop a program that can detect level of depression of a certain user.

It turned out people with depression tend to post photographs with low contrast, and the prevailing colors are blue and gray. Their photographs have much fewer likes and comments than the photographs of people who don’t suffer from depression, and one of the filters that they use most often is Inkwell that turns color images to black-and-white.

After teaching the algorithm, the authors tested its ability to distinguish between depressed and nondepressed users: the program was right in 70% of cases. To compare, doctors could correctly diagnose depression in patients only in 42% of cases.