A picture may be worth a thousand words, but with machine learning it's also worth one useful diagnosis. A new algorithm tested with volunteer Instagram users accurately detected depression markers 70 percent of the time, reports Technology Review.
This is one heck of a filter.
RELATED: Putting the Art in Artificial Intelligence: Photos
The eagle-eyed algorithm was created by University of Vermont associate professor Chris Danforth and Harvard University doctoral candidate Andrew Reece. It works using color analysis, metadata components and face detection, according to their paper published on ArXiv.
In order to test the algorithm, the researchers used the Amazon Mechanical Turk crowdsourcing marketplace platform to recruit volunteers. They narrowed the group down to 166 participants who were active on Instagram, screened them for depression, and asked them to self-report diagnoses. The participants also consented to a one-time collection of their Instagram photo histories for analysis -- totaling 43,950 images.
Feeling blue is real. Depressed participants posted photos that were bluer, grayer, and darker than ones uploaded by healthy users. Plus, when the depressed users did use a filter they tended to pick Inkwell, which transforms color photos into high-contrast black-and-white ones, the researchers reported. "Conversely, healthy participants most disproportionately favored the Valencia filter, which lightens the tint of photos," they wrote.
RELATED: Tool Knows Your Age, Sex From Social Media Updates
Danforth and Reece also asked a different group of Mechanical Turk workers to rate a random selection of 20 photos from the full set, determining how interesting, likable, happy and sad each one seemed on a scale of 0 to 5. The algorithm ended up being much better than the human crew at detecting depression markers, correctly identifying 70 percent of depressed cases.
Even though the algorithm could tell how many faces were in the photos, the "sad selfie" hypothesis has yet to be tested. If a search for #sad on Instagram is any indication, there are a lot of people who seem like they need help. Maybe one day a therapeutic filter will help give them a different outlook.
WATCH VIDEO: The Science of Depression