Sign In

Communications of the ACM

ACM News

Machines can Spot Mental Health Issues—If you Hand Over your Personal Data

View as: Print Mobile App Share:
Neguine Rezali.

Neguine Rezaii, a neuropsychiatry fellow at Massachusetts General Hospital, found that computer-based language analysis could predict which psychiatric patients were likely to develop schizophrenia with more than 90% accuracy, before any typical symptoms

Credit: Jake Belcher

When Neguine Rezaii first moved to the United States a decade ago, she hesitated to tell people she was Iranian. Instead, she would use Persian. "I figured that people probably wouldn't know what that was," she says. 

The linguistic ambiguity was useful: she could conceal her embarrassment at the regime of Mahmoud Ahmadinejad while still being true to herself. "They just used to smile and go away," she says. These days she's happy to say Iranian again. 

We don't all choose to use language as consciously as Rezaii did—but the words we use matter. Poets, detectives, and lawyers have long sifted through people's language for clues to look for their motives and inner truths. Psychiatrists, too: perhaps psychiatrists especially. After all, while medicine now has a battery of tests and technical tools for diagnosing physical ailments, the chief tool of psychiatry is the same one employed centuries ago: the question "So how do you feel today?" Simple to ask, maybe—but not to answer.  

"In psychiatry we don't even have a stethoscope," says Rezaii, who is now a neuropsychiatry fellow at Massachusetts General Hospital. "It's 45 minutes of talking with a patient and then making a diagnosis on the basis of that conversation. There are no objective measures. No numbers."


From MIT Technology Review
View Full Article



No entries found