“We’re increasingly understanding that what people do online is a form of behavior we can read with machine learning algorithms, the same way we can read any other kind of data in the world,” lead author Johannes Eichstaedt, founding research scientist at the World Well-Being Project (WWBP) in Philadelphia, told Wired.
Eichstaedt’s team, co-led by H. Andrew Schwartz, a principal investigator of the WWBP, studied data from nearly 1,200 social media users who agreed to grant access to both their posts and their electronic medical records (EMR). Of those who participated, only 114 had dealt with depression in the past.
“For each of these 114 patients, we identified 5 random control patients without a diagnosis of depression in the EMR, examining only the Facebook data they created before the corresponding depressed patient’s first date of a recorded diagnosis of depression,” study authors wrote. “This allowed us to compare depressed and control patients’ data across the same time span and to model the prevalence of depression in the larger population.”
Researchers were then able to determine whether what they refer to as “depression-associated language markers” depicted “emotional and cognitive cues.” These included sadness, loneliness, hostility, rumination and increased self-reference.
The linguistic markers, according to researchers, could predict depression fairly accurately as soon as three months before the individual received a diagnosis.
Still, Eichstaedt says, there is a different method before turning to social media as a reliable tool to diagnose depression. “It would be irresponsible to take this tool and use it to say: You’re depressed, you’re not depressed,” he told Wired.
Eichstaedt also stated that the social media algorithm is comparable to a DNA analysis.
“Social media data contain markers akin to the genome,” Eichstaedt said, according to Medical News Today. “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 skin disease or diabetes doesn’t.”
Eichstaedt says he is hopeful one day that this type of information could prove helpful in making diagnoses and treatments.
“The hope is that one day, these screening systems can be integrated into systems of care,” he said. “This tool raises yellow flags; eventually the hope is that you could directly funnel people it identifies into scalable treatment modalities.”