Either by nature or education, we tend to identify bright and light colors with good spirits, and dark colors with depression. This intuition has been confirmed by numerous investigations, and could have an unexpected application in the age of social networks. Would it be possible to find out if a person suffers from depression by analyzing the photos they upload to Instagram ?
This is what Andrew Reece (Harvard University) and Chris Danforth (University of Vermont) believe, who have developed an algorithm that determines if a person is depressed from the analysis of the photos that are posted on that website.
These two American researchers have found a relationship between the colors of the images uploaded to Instagram and the mental health of the person who posted them there. That link is so strong that Reece and Danforth suggest it could be used for early detection of mental problems.
The researchers asked study participants (166 people, 70 of them with clinical depression ) to fill out a series of questionnaires, including some used to diagnose depression. Later, they asked them to share their publications with them on this social network. With the help of a group of turkers, the scientists analyzed more than 40,000 photos and collected data on their color saturation, contrast and tones, values that make an image appear vivid, washed out, with a tendency to gray …
They also used facial detection software to count the number of faces in each image, since that is an indicator of the subject’s social life, and they checked the number of “likes” and comments each photo garnered.
The algorithm kicks in
Armed with that data, Reece and Danforth used an algorithm to look for a relationship between depression and the type of photos. They found it quickly: depressed people tended to post bluer, gray, and darker pictures than healthy people, and they also garnered fewer “likes.” Interestingly, they also tended to post more photos with faces, but when they did, the number of faces that appeared was less than in photos of people without mental problems.
In addition, they found that the depressed used more of the Instagram filter that changes the color to black and white. According to the researchers, their algorithm correctly identified 70% of previously studied subjects diagnosed with depression.
Reece and Danforth argue that their work demonstrates that “important changes in the psychology of the individual are reflected in their use of social networks, and can be identified with computational methods.” If they’re right, these tools could one day help early detection of certain mental disorders, making them easier to cure.
