Researchers have experimented with artificial intelligence (AI) technology in analyzing voice and language patterns to screen for mental health issues. The process, they say, has shown promise so far.

person speaking on the phoneShare on Pinterest
Voice analysis facilitated by AI could help people receive the mental health support they need.

“Serious mental illnesses such as schizophrenia, bipolar disorder, and major depression affect nearly 10 million people in the United States and result in significant symptom burden, lower life expectancy, and cost to the healthcare system,” write Dr. Armen Arevian and colleagues.

They hail from the University of California, Los Angeles (LA) and the University of Southern California, also in LA. Their results appear in a new paper in the journal PLOS One.

The severity and symptoms of these mental health conditions can fluctuate greatly over time, which means that people may not always receive the appropriate care they need to maintain their quality of life.

However, according to the researchers behind the new study, AI technology — which has become increasingly interesting to researchers and medical professionals alike — may allow people to receive better care.

In their new study, the investigators wanted to see if they could successfully use interactive voice recognition technology to screen for changes in mental health. The results, they say, have been promising so far.

The researchers worked with 47 people receiving care at a community-based mental health clinic. All participants lived with a mental health condition, such as bipolar disorder, major depressive disorder, schizophrenia, or schizoaffective disorder.

Dr. Arevian and team monitored the mental health of the participants for up to 14 months using a specially designed application that the participants were able to access by calling a toll-free number.

The participants had to do this twice per week, and the process involved answering three open-ended questions asked by a computer-generated voice. The three questions were:

  • How have you been over the past few days?
  • What’s been troubling or challenging over the past few days?
  • What’s been particularly good or positive?

On the other end, the AI program was able to identify and analyze keywords and voice patterns for each person who “dialed in.” The program picked up on the participants’ choice of words for their answers and how these changed each time they called.

Following their experiment, the researchers concluded that the AI program was about as efficient at monitoring changes in the participants’ mental states as their own doctors.

“The way people answer questions and the way they change their answers over time is unique to each [person],” says Dr. Arevian, adding that the researchers “were looking at a person as a person, and not as a diagnosis.”

One benefit of using the toll-free number was that it was not costly for the individuals requiring support, and they did not need to own a sophisticated device such as a smartphone.

They could make their call from any cell phone, landline, or payphone.

“Technology doesn’t have to be complicated. In this study, [people] didn’t need a smartphone or any a phone at all. It could be simple and low tech on the patient end, and high tech on the backend,” says Dr. Arevian.

The researchers are hopeful that in the future, specially trained AI could help people get the assistance they need sooner — by, for instance, picking up immediately on negative changes in a person’s mental state.

“[AI] allowed us to illuminate the various clinically meaningful dimensions of language use and vocal patterns of the [participants] over time and personalized at each individual level.”

– Senior study author Prof. Shri Narayanan

The participants also reported good experiences. “They said speaking to a computer-generated voice allowed them to speak more freely,” notes Dr. Arevian.

“They also said it helped them feel less lonely because they knew that someone would be listening to it, and to them that meant that someone cared,” he adds.

The application the researchers used was called MyCoachConnect.