Researchers used machine learning to scan Instagram posts for mental health indicators, including color analysis and facial detection

It can be difficult for those struggling with mental illness to reach out for help or even recognize that they may be suffering, which is what lead two researchers to develop an algorithm which they claim can detect depressed individuals from what they post on Instagram.

In a study conducted by Harvard University’s Andrew Reece and Chris Danforth of the University of Vermont, machine learning tools were applied to the Instagram posts of 166 people. Extracting data from nearly 44,000 participant photos, the algorithm used color analysis, metadata components and algorithmic face detection to guess if an individual was depressed.

The researchers found that photos posted by depressed individuals were more likely to be bluer, grayer and darker. Those without depression, meanwhile, were more likely to have many photos with other people in them, and their photos (perhaps unsurprisingly) also tended to get more likes, although photos posted by those with depression received slightly more comments.

The machine learning algorithm was able to accurately identify those participants with depression 70% of the time compared to doctors, who were only able to correctly identify depressed patients 42% of the time without the use of a questionnaire or other screening tools.

Although the study was only carried out on a small group of individuals, the technology holds promise as a way to help medical health professionals in identifying mental illness. Most interestingly, the algorithm could even detect depression at the same rate of accuracy even when the analysis was restricted to posts made before depressed individuals were first diagnosed. The researchers hope that the algorithm could eventually be further developed to detect those at risk for suicide.

Reece/Danforth Study


Lead Image: Christopher Burns | Unsplash

It can be difficult for those struggling with mental illness to reach out for help or even recognize that they may be suffering, which is what lead two researchers to develop an algorithm which they claim can detect depressed individuals from what they post on Instagram.

In a study conducted by Harvard University’s Andrew Reece and Chris Danforth of the University of Vermont, machine learning tools were applied to the Instagram posts of 166 people. Extracting data from nearly 44,000 participant photos, the algorithm used color analysis, metadata components and algorithmic face detection to guess if an individual was depressed.