Support more videos like this at patreon.com/rebecca!
A study has just been published in EPJ Data Science suggesting that doctors may be able to help diagnose depression in patients based on the patient’s Instagram account. As someone with crippling depression who mostly uses Instagram to post pictures of a happy puppy, I was intrigued to know more. It turns out, puppies are the #1 symptom of depression in Instagram users.
Researchers at Harvard and University of Vermont teamed up to first analyze the Instagram photos of people who had depression and create an algorithm based on the trends they found. They then applied that algorithm to a new set of Instagram users and found that it was pretty good at sorting healthy people from depressed people.
I say “pretty good” while you may notice other media outlets acting like it was great. It wasn’t. But it was pretty good.
The algorithm was only able to identify ? of the people with depression. But, when it did identify someone with depression, it was right 54% of the time. Doctors are currently estimated to be correct 42% of the time when they say a patient has depression, so the algorithm was slightly better at that, meaning that there are fewer “false positives,” or people who are told they have depression when they don’t.
When the algorithm identified someone who was healthy, it was right 84% of the time. So really, the headlines should more accurately state that your Instagram can show if you’re healthy, not necessarily if you’re depressed. Confusing, I know. Statistics are weird sometimes.
So what is it about depressed people’s Instagrams that stand out? Well, the researchers found that depressed people were more likely to favor darker pictures, blue and grey colors, and low saturation. They were also less likely to use filters, and if they did use a filter it was most likely to be the black and white Inkwell filter. They also found that depressed people were more likely to have faces in their photos, but less likely to have a large number of faces in each photo. That could suggest that depressed people favor selfies, but the researchers didn’t look at selfies in particular so that remains to be seen in a future study.
Remember that the whole point of Instagram and other social media networks is to pretend like your life is way better and more interesting than it really is, which makes this a particularly interesting study. The algorithm wasn’t looking at the photo subjects to see who was posting pictures that humans would think were “sad,” which means that it may be able to detect depression in someone who is usually good at hiding it. An algorithm like this will never be perfect, but it is interesting that something so simple can, in one respect at least, be slightly better than a human doctor, especially when it comes to an illness that a patient may even lie to a doctor about to avoid being detected. That doesn’t mean that machines are ready to replace doctors, or that they ever will, but it’s worth noting as a potential new tool to help doctors diagnose an illness that can be very difficult to recognize. As of now, there’s no blood test or x-ray that will tell you someone is depressed, but maybe this is the first step toward a social media test that can at least detect some warning signs.