New Algorithm Can Detect Depression by Instagram Photos

98
instagram-and-depression
Valencia vs Inkwell filters, via The Next Web.

It’s no secret that colour has a strong link with our emotions and out mental well-being. Bright colours are strongly linked with positive emotions, while neutral or darker tones go hand in hand with negative ones. Andrew Reece and Chris Danforth from Harvard have decided to take this information and see if they could discern people’s well-being and emotional status by looking at their Instagram photos.

They then asked 500 employees from Amazon’s Mechanical Turk, with 170 of them agreeing to have their Instagram photos analysed. Out of the bunch, 70 were diagnosed as clinically depressed. By selecting the last 100 images uploaded by each volunteer, the team analysed a total of 40,000 images. Those diagnosed with depression also had 100 images taken before they had the diagnosis.

The results were as expected. Colours, hues, contrast, as well as the number of people present in the photos, have all been quantified in a “happiness” meter ranging from 0 to 5. With Instagram, people can add all sorts of filters to adjust their photos in accordance with their mood. This happens, more often than not, at a subconscious level, wit people feeling more comfortable wit filters which more closely resemble their state of mind.

So, the “Inkwell” filter which turns ordinary photos into black and white one, is something which depressed people do more often. They were also found to favour black and blue as a predominant colour in their photos, with an overall preference for darker, matte colours.

On the other end of the depression spectrum, happy volunteers mostly preferred the “Valencia” filter which lightens up images. However, the algorithm devised by the two researchers isn’t by any means perfect, showing an average of 70% accuracy. This new development can be seen as a double-edged sword. On the one hand, people suffering from depression can be detected as such via their Instagram photos. But on the other hand, total strangers can find things about you that maybe not even you don’t realise. And what’s more, this can be applied to basically everyone with a social network account.

“These findings support the notion that major changes in individual psychology are transmitted in social-media use, and can be identified via computational methods,” say Reece and Danforth.