According to a new study, Instagram posts might actually be useful for the health as a team of scientists used them to look for signs of depression in the users. The researchers conducted their research with help from a team of volunteers and a machine learning-based software that analyzed their photos.
Instagram Posts Deemed Helpful in Keeping Track of the Mental Health
The study involved over 160 volunteers recruited from an online crowdsourcing platform, Amazon’s Mechanical Turk. These were asked to provide information regarding their past depression diagnoses. They also had to fill in a questionnaire which looked to assess the respective responder’s level of depression.
Then, the research team took a look at their Instagram posts, which, combined, included almost 44,000 photos. These were analyzed using a special software based on artificial intelligence.
This noted that people that showed signs of depression were more likely to post grayer, bluer, generally darker pictures. They also appeared to be less likely to use filters, and when using them, mostly chose black and white ones.
People with depression were also noted to be more liable to post pictures with people in them, but with fewer individuals when compared to the other group. Participants that did not exhibit signs of depression appeared more likely to use filters, especially ones that livened up their photos.
Machine Learning Results Compared to Volunteer Ones
A follow-up test then had another group of volunteers look at the Instagram posts of the study participants. These were asked to rate the photos, which showed that they were generally capable of spotting users with depression but to an extent.
Still, their results were less impressive than those of the machine learning software, which managed to accurately detect people with depression in 70 percent of the cases.
“This study is not yet a diagnostic test, not by a long shot. But it is a proof of concept of a new way to help people. This points toward a new method for early screening of depression,” stated Christopher Danforth.
He is a study co-author and professor of mathematical, natural, and technical sciences part of the University of Vermont.
Study results are available in the journal EPJ Data Science.
Image Source: Wikimedia