8th August 2017 – Researchers have coded a computer program to look for signs of depression in photos that people post on social media.
According to research published in the journal EPJ Data Science, a program has been developed that can correctly identify depressed individuals from their social media photos 70% of the time. In comparison, previous research has shown that GPs can make a correct diagnosis of depression, without assistance from questionnaires, scales or other measurement instruments, 42% of the time.
The study involved 166 users of a popular social media app, including 71 people with a history of depression. The researchers used the program to analyse 43,950 photos. It searched for details in the photos that are associated with healthy and depressed people by using colour analysis, metadata components and algorithmic face detection. For example, the program can be coded to look for the following characteristics:
Are there people present?
Is the setting in nature or indoors?
Is it night or day?
Did the photo receive any comments?
How many 'likes' did it get?
The researchers report that depressed people tend to post photos that are bluer, darker and greyer. They also state that the more comments social media posts receive, the more likely depressed people posted them, but the opposite is true for likes received. Although depressed people are more likely to post photos with faces, they have a lower average face count per photograph than non-depressed participants. When filters are used, depressed people are more likely to prefer a filter that converts colour photos to black and white, while healthy people favour a filter that lightens the tint of photos.
According to the researchers, human ratings of photo attributes such as happy, sad, etc. were weaker predictors of depression and did not correlate with computationally generated features. These results suggest new avenues for early screening and detection of mental illness.
Limitations of the study
The study is limited by the generalisation of the term 'depression' in the data collection process. Depression is a complex, multifaceted illness and is frequently present with other health conditions. The researchers concluded that future research should seek to address the reliability of the computer model against more finely tuned definitions of depressive disorders.
For this study, a total of 509 participants were recruited, but 43% of them dropped out of the study because they did not want to consent to sharing their social media data. Concerns about data privacy will therefore need to be addressed if a similar model is developed for widespread clinical use.
To provide even greater transparency and choice, we are working on a number of other cookie-related enhancements. More information