New Study Shows That Your Instagram Posts Reveal A Lot More About You Than You Think

Someone’s Instagram account says a lot about that person. What they eat, what they wear, where they go, and maybe even what their mental health is like. In a recent study, researchers say that Instagram photos can help predict who might have depression. Interestingly, the signs may not be what you think.

The study involved 166 Instagram users and over 40,000 photos. The researchers tested a machine learning program that analysed the photos, looking for signs of depression. The algorithms used a range of factors to determine if a user had depression. These included facial expression, composition and colour hue, and the program was surprisingly effective in predicting who had depression. The success rate even held before a user received a diagnosis.

Perhaps heralding a time yet to come, the study also found that the program was much more effective than humans who analysed the photos. More than half of the diagnoses from general practitioners who looked at the same accounts and photos were false positives, while the majority of machine generated diagnoses were correct. The takeaway here is that while it might not come as a huge shock that a purpose-built machine is better at that purpose than an educated human, the real markers of depression are often more complex than we may assume.

instagram-1372870_1920

So what are some signs? Well, the model isn’t perfect but there are some standout features. Depressed users are more likely to filter out colours from photos, and to include more blues and greys. This is consistent with past findings that people with depression are less drawn to colour than others. Other markers include isolation, again already correlated, a higher likelihood of including a face in a photo (yet with less faces on average) and self-focused language.

More than half of the diagnoses from general practitioners who looked at the same accounts and photos were false positives, while the majority of machine generated diagnoses were correct.

The authors write that “[The] present work may serve as a blueprint for effective mental health screening in an increasingly digitalized society. More generally, these findings support the notion that major changes in individual psychology are transmitted in social media use, and can be identified via computational methods.”

Hopefully, in the near future, we see doctors working with these sorts of programs to identify and provide help to people who may need it. It seems that doctors don’t necessarily know what to look for, and any help is good help with this sort of thing.