A machine learning algorithm is able to identify suicidal people by analyzing the things they say, researchers reveal in the journal Suicide and Life-Threatening Behavior.

“Death by suicide demonstrates profound personal suffering and societal failure,” explain the study authors, who hope to see their new approach used to help caregivers assist those who are at risk of developing suicidal tendencies.

“This methodology easily can be extended to schools, shelters, youth clubs, juvenile justice centers, and community centers, where earlier identification may help to reduce suicide attempts and deaths,” said study co-author John Pestian in a statement.

To test their algorithm, the team recruited 379 participants, some of whom had been diagnosed as suicidal, while others were mentally ill but not suicidal and some were neither. Each patient was then asked a series of questions such as “does it hurt emotionally?”, “do you have any fear?”, and “do you have hope?”

By registering what each person said in response to these questions, as well as their tone of voice, the computer was able to identify which of the three categories they fell into, with an accuracy of 85 percent.

It was particularly successful at recognizing those who were suicidal, achieving a 93 percent accuracy for this group.

Though the algorithm was designed to pick up on a range of verbal and non-verbal cues, the study authors reveal that certain tell-tale signs helped the computer to complete its task. For example, those with no mental illness or suicidal behaviors “tended to laugh more during interviews, sigh less, and express less anger, less emotional pain, and more hope.”

According to Pestian, the results of this study provide evidence of the value that technology can bring to suicide care. “When you look around health care facilities, you see tremendous support from technology, but not so much for those who care for mental illness,” he said. “Only now are our algorithms capable of supporting those caregivers.”