Contents
Will he “take over the world” or will he serve the people? While writers and filmmakers are exploiting artificial intelligence horror stories, scientists are getting practical results by developing apps to help psychiatrists and their patients.
Researchers have developed an AI system – artificial intelligence – that can detect daily changes in speech that indicate a deterioration in a person’s mental health.
“We are not trying to replace doctors…”
Thanks to advances in artificial intelligence, computers can now help doctors diagnose diseases and monitor the vital signs of patients hundreds of miles away. University of Colorado Boulder researchers are working on the application of machine learning to psychiatry. They are designing a mobile app that, based on a patient’s speech, can classify their mental health status just as well as another person.
“We are by no means trying to replace doctors,” says Peter Foltz, a professor at the Institute for Cognitive Sciences. He is also co-author of a new article in the Bulletin of Schizophrenia outlining the promise and potential pitfalls of using artificial intelligence in psychiatry. “But we believe we can create tools that will allow psychiatrists to better manage their patients.”
In search of a reliable diagnostic method
Nearly one in five adults is living with a mental illness. Many of these people live in remote areas where access to psychiatrists or psychologists is very limited. Others cannot afford to see a doctor often, and do not have the time or money to pay for frequent visits. Even if the patient is regularly shown to a psychotherapist, he uses a conversation with the patient to diagnose and draw up a treatment plan. It’s an age-old method that can be subjective and not reliable enough, says paper co-author Brita Elvevog, a cognitive neuroscientist at the University of Tromsø in Norway.
“People are imperfect. They can become distracted and sometimes miss subtle speech cues and warning signs, says Dr. Elwevog. “Unfortunately, there is no objective blood test for mental health in medicine.” Scientists set out to find a more objective way to define the problem.
Using mobile devices and artificial intelligence, we can monitor patients daily
Seeking an “AI version” of such a blood test, Elwewog and Foltz teamed up to develop machine learning technology capable of detecting day-to-day changes in speech that could indicate worsening mental health. For example, in schizophrenia, the critical symptom may be sentences that do not follow the usual logical pattern. Changes in the tone or rate of speech may indicate mania or depression. And memory loss can be a sign of both psychological and mental problems.
“Language is an important factor in identifying the mental states of patients,” Foltz says. “Using mobile devices and artificial intelligence, we can monitor patients on a daily basis and capture the subtlest changes in their condition.”
How does it work?
The new mobile app prompts the user to answer a 5-10 minute series of questions over the phone. Among other tasks, the person is asked about their emotional state, asked to tell a short story, then in turn listen to the story and repeat it, and complete a series of motor skills tests using touch and swipe on a smartphone screen.
In collaboration with Chelsea Chandler, a graduate student on the faculty at the University of Colorado at Boulder, and other colleagues, the authors of the project developed an artificial intelligence system that can evaluate these speech patterns, compare them with previous responses from the same patient and a wider control group, and as a result evaluate the mental state person.
Accuracy and Reliability
In one recent study, a team of scientists asked clinicians to listen to and evaluate speech patterns from 225 participants. Of these, half had previously been diagnosed with serious psychiatric problems, and half were healthy volunteers from rural Louisiana and Northern Norway. The researchers then compared the results of the survey of doctors with the results of the artificial intelligence program.
Our task is not to shift decision making to machines, but to use them in what they do really well.
“We found that computer AI models can be at least as accurate as doctors,” says Peter Foltz with confidence. He and his colleagues are convinced that the day will come when the AI systems they develop for psychiatry will be in the office at the meeting of the therapist and the patient in order to help collect data or serve as a remote monitoring system for severe mental patients who need in lookout.
Control system
By detecting disturbing changes, the application can notify the doctor to pay attention and take control of the patient. “In order to avoid costly emergency care and unpleasant events, patients should proactively undergo regular clinical interviews with qualified professionals,” Foltz says. “But sometimes there just aren’t enough doctors for that.”
His previous development in the field of artificial intelligence is now widely used. Foltz is confident that the new project will also prove the effectiveness of machine learning technologies. In their article, the scientists urged colleagues to conduct even larger studies to prove the effectiveness and gain public confidence. This is important in order for artificial intelligence technology to be widely introduced into clinical psychiatric practice.
“The halo of mystery around AI does not help build trust, which is essential in the application of medical technologies,” they write. “Our task is not to shift decision making to machines, but to use them in what they do really well.” Thus, it is possible that psychiatry and medicine in general are on the verge of a new era in which artificial intelligence will become an important assistant to doctors in caring for the health of patients.