Skip to Content
MIT Technology Review

Are psychiatrists really ready for the AI revolution?

Machine learning can help manage a wide range of mental health disorders. But the psychiatric profession is worryingly unprepared for this change, according to a global survey.

Person sitting on couchPerson sitting on couch
Person sitting on couch
Kelly Sikkema | Unspalsh

The World Health Organization estimates that up to 15% of the population experiences mental health disorders. That has significant consequences. For example, suicide is the second- or third-leading cause of death for young people in most countries. And as the population ages, the rate of dementia is set to triple over the coming decades.

At the same time, access to mental health professionals is sorely lacking in many parts of the world, particularly in low-income countries. India, for example, has a population of 1.3 billion served by only 9,000 psychiatrists.

But technological advances can help. Smartphones and wearable sensors offer people the ability to monitor themselves and to benefit from the way deep learning can analyze the data. Indeed, these techniques are already being used to detect the changes in mood that indicate bipolar disorder or to detect people at risk of depression.  

So the scene is set for artificial intelligence to become a disruptive force in psychiatry. Indeed, that’s exactly what many observers predict.

But what of psychiatrists themselves? These professionals will have to play a key role in any change that artificial intelligence brings to the field. So their view ought to be a useful indicator of its potential.

Enter Murali Doraiswamy at Duke University School of Medicine in Durham, North Carolina, and couple of colleagues. This team has surveyed psychiatrists around the world to find out how they view machine intelligence and its likely impact on mental health care.

“To our knowledge, this is the first global survey to seek the opinions of physicians on the impact of autonomous artificial intelligence/machine learning on the future of psychiatry,” say the team. Curiously, the results appear to say more about psychiatrists than about the state of technological readiness or its potential.

The team’s method was straightforward. The researchers randomly chose a sample of 750 professional psychiatrists registered with an online database of over 800,000 health-care professionals around the world, including 22 countries in North and South America, Europe, and Asia; 30% were women and two-thirds were white.

The respondents clearly felt that machines could never learn some skills. “An overwhelming majority (83 per cent) of respondents felt it unlikely that future technology would ever be able to provide empathic care as well as or better than the average psychiatrist,” say Doraiswamy and colleagues. Interestingly, a survey of family physicians in the UK showed they had a similar view.

The group was also divided on the risks that artificial intelligence might pose.  “Only 23 per cent of women predicted that the benefits of AI would outweigh the possible risks compared to 41 per cent  of men,” say Doraiswamy and colleagues.

But they think they know why. “The gender differences in AI risk perception may be commensurate with a large body of findings that women are more risk averse than men,” they say.

The most interesting results are in the way respondents feel machine intelligence will change their jobs. Three-quarters of them thought that artificial intelligence will play an important role in managing data, such as medical records. And about half thought it would fully replace human physicians when it comes to synthesizing information to reach diagnoses.

Other areas of health care are already experiencing such benefits. Machine-learning techniques can outperform radiologists and pathologists in in certain circumstances. This ability to reach more accurate diagnoses has huge implications for patient treatment and safety.

And yet only half the psychiatrists felt that artificial intelligence would substantially change their jobs (presumably the same half who think AI can better diagnose conditions than humans).  Perhaps predictably, less than 4% thought AI could replace human psychiatrists completely.

Doraiswamy and colleagues have a potential explanation for this. “Doctors may be overvaluing their skills and/or underestimating the rapid pace of progress in intelligent technologies,” they say.

Either way, that has important implications for this profession. Earlier this year, the World Economic Forum published a report called “Empowering 8 Billion Minds,” which highlighted the growing burden of mental illness around the world.

The report pointed out that apps focused on mental health are among the fastest-growing sectors in the global digital health market. These kinds of apps could make a big difference. But if this survey is to be believed, psychiatrists around the world are broadly unprepared for the changes to come.

Ref: arxiv.org/abs/1907.12386 : Artificial Intelligence and the Future of Psychiatry: Insights from a Global Physician Survey