Researchers use AI to detect signs of depression on Instagram

Published August 22nd, 2016 - 04:48 GMT
The AI program was 70 percent accurate in detecting signs of depression, which was better than previous studies looking at the 42 percent success rate of GPs diagnosing patients. (Shutterstock/tanuha2001)
The AI program was 70 percent accurate in detecting signs of depression, which was better than previous studies looking at the 42 percent success rate of GPs diagnosing patients. (Shutterstock/tanuha2001)
Researchers developed a program that can recognize signs of depression in pictures on the social network

The study, carried out by researchers from Harvard University and the University of Vermont, used machine learning tools to identify markers of depression. It was found that the program was 70 per cent accurate in detecting signs of depression, which was better than previous studies looking at the success rate of GPs diagnosing patients – normally around 42 per cent accurate.

43,950 posts on the social media site were analyzed for the study using color analysis, metadata components and algorithmic face detection. Other factors were also taken into consideration such as how often users posted, as it is said depression can often be linked to reduced social activity, how many people were in the photo and the likes and comments received on each image.

The study, published on Arxiv, found photos posted by individuals diagnosed with clinical depression were more likely to be bluer, grayer and darker. They were more likely to not use an Instagram filter, but if they did it was often the Inkwell feature. Also, posts were less likely to receive likes but more likely to get comments.

The results also held when analyzing photos made before the users were first diagnosed with depression.

The study stated: “These findings demonstrate how visual social media can be harnessed to make accurate inferences about mental health.” And as Instagram has over 300 million people using the app every day it would be interesting to see how the study on the social network will be used to suggest new ways for early screening and detection of mental illness.

However, the researchers noted that the program has shortcomings but should be used “as a foundation upon which to build and refine subsequent models”. They also said issues such as the way social media changes change over time will mean the machine learning program would need to be continually updated to maintain effectiveness.

Early this year it was revealed that one in four people in the UK have been diagnosed with mental illness, with depression being the most common diagnosis.


©2019 Fars News Agency. All rights reserved

You may also like