AI tool can predict psychosis risk from your speech

A hidden clue in people's language can help predict if they would develop psychosis

AI tool can predict psychosis risk from your speech Representational Image | Pixabay

Scientists using artificial intelligence have discovered a hidden clue in people's language that can accurately predict whether they are likely to develop psychosis in future.

The machine-learning method more precisely quantifies the semantic richness of people's conversational language, a known indicator for psychosis.

The research, published in the journal npj Schizophrenia, shows that automated analysis of the two language variables―more frequent use of words associated with sound and speaking with low semantic density, or vagueness―can predict whether an at-risk person will later develop psychosis with 93 per cent accuracy.

Even trained clinicians had not noticed how people at risk for psychosis use more words associated with sound than the average, although abnormal auditory perception is a pre-clinical symptom.

“Trying to hear these subtleties in conversations with people is like trying to see microscopic germs with your eyes,” said Neguine Rezaii, who conducted the research at Emory University in the US.

“The automated technique we've developed is a really sensitive tool to detect these hidden patterns. It's like a microscope for warning signs of psychosis,” said Rezaii, who is now at Harvard University in the US.

“It was previously known that subtle features of future psychosis are present in people's language, but we've used machine learning to actually uncover hidden details about those features,” said Phillip Wolff, a professor at Emory University.

The findings add to the evidence showing the potential for using machine learning to identify linguistic abnormalities associated with mental illness, said Elaine Walker, an Emory professor.

The onset of schizophrenia and other psychotic disorders typically occurs in the early 20s, with warning signs―known as prodromal syndrome―beginning around age 17.

About 25 to 30 per cent of youth who meet criteria for a prodromal syndrome will develop schizophrenia or another psychotic disorder.

Using structured interviews and cognitive tests, trained clinicians can predict psychosis with about 80 per cent accuracy in those with a prodromal syndrome.

Machine-learning research is among the many ongoing efforts to streamline diagnostic methods, identify new variables, and improve the accuracy of predictions. Currently, there is no cure for psychosis.

“If we can identify individuals who are at risk earlier and use preventive interventions, we might be able to reverse the deficits,” Walker said.

“There are good data showing that treatments like cognitive-behavioral therapy can delay onset, and perhaps even reduce the occurrence of psychosis,” she said.