US researchers analyzed Facebook posts of around 700 users and found that it was possible to predict depression with great precision long before the official diagnosis. In 114 people diagnosed with depression, the first symptoms were detected three months before the diagnosis was decided: the accuracy of the prediction was more than 70%. The article has been published in the journal Proceedings of the National Academy of Sciences.
In general, the diagnosis of mental disorders can be deduced by analyzing how a person talks about himself: what words he uses to describe his emotions, as well as what he says about his own well-being. Unfortunately, these symptoms may not be very pronounced and in the worst case, a depressed person could simply deny that there is a problem.
Therefore, in recent decades, psychologists have developed different scales and tests that can more accurately identify a person in depression or anxiety disorder. Speech analysis is also useful, but especially for automated methods. Therefore, MIT researchers have recently suggested to diagnose depression from the language of a patient using a neural network.
Scientists under the leadership of Johannes C. Eichstaedt of the University of Pennsylvania have suggested that depression can be diagnosed long before going to a psychiatrist by analyzing Facebook posts. To do this, they collected open data in the publications of 683 users who requested medical assistance: 114 of them were diagnosed with depression and the rest served as a control group. Data were collected in two and a half years. The researchers analyzed the users' vocabulary, the duration of the publications, their frequency, as well as the demographic data (geotag).
Accuracy of the forecast based on the time of the post. / Eichstaedt et al. / PNAS 2018
With the help of these data, scientists were able to diagnose depression through patient publications with great precision: from publications made six months before diagnosis, depression could be predicted with an accuracy of 62% and 72% accuracy in places published three months earlier. The vocabulary of depressed patients, according to the scientists, contained many emotional words and words associated with well-being and status.
Word clouds are more common in publications of depressed patients. /Eichstaedt et al. / PNAS 2018
It is not known if the developed method will be used by the social network itself. Last year, Mark Zuckerberg announced that the company would analyze user publications to identify potential victims of suicide.
Depression can be expected not only from publications on social networks, but also from photos. The American researchers came to this conclusion after analyzing 166 Instagram profiles, of which 71 belonging to people with depressive symptoms. A statistical model created by researchers has identified people with depression with a 70% chance (more specifically psychiatrists). The study was published in the journal EPJ Data Science.
This news was originally published in N + 1, science that adds.
About N + 1: It is the first online scientific and technological publication that allows the total or partial reproduction of its contents through communication, bloggers and influencers, making mention of the text and the link to the web: "This news has it was originally published in the journal N + 1, science that adds: Www.nmas1.org ".