Computer may be able to identify depression from Instagram posts

Filter, aesthetic choice correlate with clinical depression diagnosis

Instagram filter choices help computers figure out which users suffer from depression, according to a new study from researchers at Harvard University and the University of Vermont.

For the study, which was not peer-reviewed but posted to online repository arXiv.org, researchers analyzed 166 Instagram users' posts and asked whether the participants had been diagnosed with clinical depression by a mental health professional.

Researchers started with 70% of the posts in their data set and focused on four areas of the photos: number of faces, Instagram filters, color and brightness, and metadata such as the number of comments or likes.

Researchers found that depressed people posted bluer, darker, and grayer photos; had fewer faces in individual photos—but more faces in their photo collections overall; and had fewer likes, but more comments on their posts. Overall, they used fewer filters—but when they did, they were to use the black-and-white Inkwell filter.

The findings align with previous research concluding that people with depression gravitate toward grays, blues, and darker colors. In general, depressed people also interact in smaller social groups.

Based on these initial findings, the researchers built an algorithm designed to screen photos and determine whether users were depressed. They tested the algorithm by running it on the final 30% of posts in their data set.

"We were right about people being healthy 84% of the time," says Andrew Reece, lead author and psychology and computational science graduate student at Harvard. 

In fact, the computer was correct 54% of the time it identified someone as depressed—compared with unassisted primary physicians' 42% accuracy rate. 

Related: Checklist for behavioral intervention teams

"It's not an A+, but it's a 25% improvement over those human rates," Reece says.

However, "you can't even really compare the two in a formal sense" because primary doctors examine the general population and don't sort people into binary categories of "depression" and "no depression" like the computer does, Reece says.

The program will likely not replace doctors as a way of screening individuals for depression, but might be able to help physicians screen large swaths of people more quickly.

"The goal lies in empowering the clinician with tools that would help people manage or cope with these challenges above and beyond what they're currently using," says Munmun De Choudhury, a social and computer scientist at the Georgia Institute of Technology who was not involved in this study (Chen, "Shots," NPR, 8/24). 


Next in Today's Briefing

Around the industry: BYU adjunct returns from competing in Rio Olympics

Next Briefing

  • Manage Your Events
  • Saved webpages and searches
  • Manage your subscriptions
  • Update personal information
  • Invite a colleague