Just as people display their emotions through body language and behaviour, your emotional state can now be detected using your Instagram account.
Researchers at the University of Vermont and Harvard University have shown that machine learning algorithms can successfully detect depression from Instagram photos. Currently, the computerised method has a success rate of 70%, a vast improvement on the 42% achieved by general practice doctors diagnosing in-person.
The computerised method has a success rate of 70%
In order to collect the data, the scientists recruited 166 volunteers from Amazon’s Mechanical Turk, half of whom had reported having clinical depression in the last three years. Using well-established psychology research, their Instagram feeds were analysed for signs of depression, including brightness, colour and shading preferences. Pixel analysis of the dataset showed that depressed individuals tended to post bluer, darker and greyer photos than their healthier colleagues. From Instagram’s pre-set filters, it was found that the black-and-white ‘Inkwell’ was the most popular among depressed people. On the other hand, healthy individuals seemed to favour warmer filters such as ‘Valencia’.
Along with colours, the amount of faces shown in photos could also be used to indicate depression. Although depressed people were more likely to post a photo including a face, their photos had, on average, fewer people within the shot. “Fewer faces may be an oblique indicator that depressed users interact in smaller settings,” says Chris Danforth, a professor at the University of Vermont. This is corroborated by previous research linking depression to less social interaction. Or, it indicate that depressed people take more ‘selfies’. However, according to Danforth, this ‘sad-selfie’ hypothesis remains untested.
“So much is encoded in our digital footprint,”
“So much is encoded in our digital footprint,” Danforth explains. “Clever artificial intelligence will be able to find signals, especially for mental illnesses.” It is certainly true that this new technique has potential, especially in terms of early onset diagnoses, avoiding false diagnoses and reducing the cost of mental health screening. However, it is still in its early stages. “This study is not yet a diagnostic test, not by a long shot,” says Danforth. “But it is a proof of concept of a new way to help people.”
Image Credit: [http://www.sonypark360.net/wp-content/uploads/2016/11/2016-11-11-image-8.jpg]