Facebook has been rolling out the updates recently, both across their title platform and Instagram, could this algorithm be the next one? Scientists have been working on an algorithm that uses trigger words to detect undiagnosed or developing depression on Facebook.

There’s no denying that there is a tangible link between social media and mental health, for many different reasons. However, with this invention, it may be possible to pinpoint possible mental health issues in social media users before it becomes a clinical problem by using keywords.

Depression is a predominant mental health problem worldwide. In fact, according to the mental health charity ‘Mind’, who run a survey in England every 7 years, 3.3 in 100 people suffer from it, that’s over 3% of the population. This can manifest itself in mild, moderate or severe depression, with specific types like Seasonal Defective Disorder (SAD), postnatal depression and even prenatal depression possible. In its mildest form, depression can just be a feeling of low spirits and in its most severe form, it can be life-threatening; the figures below highlight this.

suicide and self harm rates uk in relation to depression on facebook

With modern living being a continuous pressure on mental health, with more financial worries and societal changes, like the pressures of social media (particularly for young people), than ever before - it’s not hard to see why mental health is becoming a more and more publicised issue.

This algorithm could be a great way of utilising something the majority of us use, sometimes to our detriment, and tailor it to be for our benefit. With over one billion users worldwide, Facebook is certainly not short of people it may be able to help.

"What people write in social media and online captures an aspect of life that's very hard in medicine and research to access otherwise,"

"It's a dimension that's relatively untapped compared to biophysical markers of disease. Considering conditions such as depression, anxiety, and PTSD, for example, you find more signals in the way people express themselves digitally."

- H. Andrew Schwartz, study author and computer scientist at Stony Brook University

How will it detect depression on Facebook?

The initial study collected data from consenting Facebook users over several months on the lead up to their depression diagnosis. The researchers studied the data and discovered that their algorithm was able to accurately predict future depression. Indicators which triggered the algorithm included mentions of hostility and loneliness. For example, words like “tears” and “feelings,” and the use of more first-person pronouns like “I” and “me” were telling.

The University of Pennsylvania’s Positive Psychology Center and Stony Brook’s Human Language Analysis Lab have been the base for the World Well-Being Project (WWBP), and they have been studying how people’s inner feelings and emotions are reflected through their words for the past six years. In 2014, WWBP’s founding research scientist, Johannes Eichstaedt, began looking into if it was possible for social media to predict mental health outcomes, with a focus on depression.

“Social media data contain markers akin to the genome. With surprisingly similar methods to those used in genomics, we can comb social media data to find these markers,”

“Depression appears to be something quite detectable in this way; it really changes people’s use of social media in a way that something like a skin disease or diabetes doesn’t.”

- Johannes Eichstaedt

person on laptop on Facebook

The algorithm was initially mentioned and published in a paper within Proceedings of the National Academy of Scientists, as opposed to using data from those with depression who volunteered it, they instead identified data from people consenting to share Facebook statuses and electronic medical record information.

It was then built using 524,292 Facebook updates, with a portion of the individuals participating in the study later being diagnosed with depression. WWBP focused in on the words and phrases that they believed were most frequently associated with statuses alluding to depression. They then categorised them into 200 topics, allowing them to identify what they called "depression-associated language markers". The language of the depressed group and the control group could then be compared to spot patterns between the two.

Around 1,200 people consented to provide both Facebook and medical data, 114 of these people diagnosis of depression in their medical records. Each of the people who had a depression diagnosis were then paired with five people who did not, offering a total control sample of 683 people.

The results...

The aim was to create as realistic a scenario as possible for the algorithm to learn, grow and adapt, making it more efficient at targeting depression on Facebook. It found that those likely to have, or with, depression were also likely to post longer posts more often, with their average annual word count 1,424 words higher than the control group. As well as an increased likelihood to use words associated with depressed moods (tears, cry, pain), loneliness (miss, much, baby), hostility (hate, ugh, f**kin), anxiety (scared, upset, worry), and rumination (mind, a lot).

The algorithm was able to predict future depression as early as three months before an official documentation of mental illness was made in a medical record by using these emotional, cognitive, and interpersonal markers.

How can it combat depression on Facebook?

“There’s a perception that using social media is not good for one’s mental health, but it may turn out to be an important tool for diagnosing, monitoring, and eventually treating it,”

“Here, we’ve shown that it can be used with clinical records, a step toward improving mental health with social media.”

- H. Andrew Schwartz

Eichstaedt says their wish is that ‘one day, these screening systems can be integrated into systems of care.’ with the hope that the algorithm ‘could directly funnel people it identifies into scalable treatment modalities.’ This would be beneficial for those who are less likely to seek help, three out of four suicides (76%) are by men and suicide is the biggest cause of death for men under 35. Men are also less likely to access psychological therapies than women. With only 36% of referrals to IAPT (Increasing Access to Psychological Therapies) being men. These “yellow flag” alerts that the algorithm will use may be particularly useful in offering men a route to take when they maybe don’t realise they should be looking for one.

Though the study currently has limitations (like the small sample area and relying on medical records as opposed to a structured interview), it does offer a promising look at what could come when it comes to mental health diagnosis and detecting indications of depression on Facebook.