Source-linked AI summary
Artificial Intelligence and the Future of Psychiatry: Insights from a Global Physician Survey
P. Murali Doraiswamy, Charlotte Blease, Kaylee Bodner
TL;DR
Mental health faces a large burden, stigma, and shortages, while evidence about practicing physicians’ views of AI’s effects on psychiatry has been limited. This global survey asked psychiatrists about AI/ML fully replacing humans across 10 tasks. The clearest resistance concerned empathetic care, while documentation and diagnostic synthesis were more replaceable in respondents’ judgments.
Problem
Limited attention has been paid to practicing physicians’ views of AI’s impact on medical professions, particularly in mental health care.
Method
The study surveyed psychiatrists through Sermo, a global platform of verified and licensed physicians, about AI/ML’s potential to replace 10 psychiatric tasks.
Results
3.8% felt AI/ML would make their jobs obsolete, while 83% considered technology unlikely to provide empathic care as well as or better than the average psychiatrist.
Takeaways & Limitations
The findings support integrating intelligent technologies to enhance mental health care and reskilling doctors as needed, rather than replacing physicians.
Takeaways & Limitations
The survey cannot determine causality or predictive validity and may be affected by sampling and response biases.
Abstract
from arXiv · showhide
Futurists have predicted that new technologies, embedded with artificial intelligence (AI) and machine learning (ML), will lead to substantial job loss in many sectors disrupting many aspects of healthcare. Mental health appears ripe for such disruption given the global illness burden, stigma, and shortage of care providers. Using Sermo, a global networking platform open to verified and licensed physicians, we measured the opinions of psychiatrists about the likelihood that future autonomous technology (referred to as AI/ML) would be able to fully replace the average psychiatrist in performing 10 key tasks (e.g. mental status exam, suicidality assessment, treatment planning) carried out in mental health care. Survey respondents were 791 psychiatrists from 22 countries. Only 3.8% of respondents felt that AI/ML was likely to replace a human clinician for providing empathetic care. Documenting (e.g. updating medical records) and synthesizing information to reach a diagnosis were the two tasks where a majority predicted that future AI/ML would replace human doctors. About 1 in 2 doctors believed their jobs could be changed substantially by future AI/ML. However, female and US-based doctors were more uncertain that the possible benefits of AI would outweigh potential risks, versus their male and global counterparts. To our knowledge, this is the first global survey to seek the opinions of physicians on the impact of autonomous AI/ML on the future of psychiatry. Our findings provide compelling insights into how physicians think about intelligent technologies which may better help us integrate such tools and reskill doctors, as needed, to enhance mental health care.
Introduction
Mental health disorders impose a substantial global burden amid stigma and severe shortages of psychiatric care. Against predictions that AI may disrupt healthcare, the survey examines psychiatrists’ views of AI/ML’s effects on psychiatric work, risks, and benefits.
- 10-15% of the population is estimated to have a mental health disorder, among the leading causes of worldwide morbidity and mortality.
- Stigma, low funding, and acute shortages of mental health professionals are identified as key barriers to addressing global mental health needs.
- Psychiatrist shortages are markedly unequal globally, with rates in low-income countries estimated at 100 times lower than in high-income countries.
- 1.3 billion people in India are served by only about 9,000 psychiatrists, illustrating the scale of diagnostic and treatment gaps.
- Limited attention has been paid to practicing physicians’ views of AI’s impact on medical professions, particularly in mental health care’s long-term, empathetic relationships.
- The survey investigates psychiatrists’ opinions of autonomous AI/ML technologies, including their effects on psychiatric jobs and their potential risks and benefits.
Methods
The study used an exploratory cross-sectional global survey of psychiatrists registered on Sermo, an anonymous networking and survey platform for verified physicians. The sample was designed to represent psychiatrists across major world regions.
- The exploratory study targeted approximately 750 psychiatrist respondents through random sampling of Sermo’s registered psychiatrists.
- Sermo is an online platform for physician networking and anonymous survey research, exclusive to verified and licensed physicians.
- Sampling sought representation from the United States, Europe, and the rest of the world.
- The survey collected nationality, demographics, perceptions of the future of psychiatry, workforce perceptions, and practice characteristics.
- Survey results were de-linked from respondents’ personal identifiable information to create de-identified data.
Survey instrument
The survey asked psychiatrists whether future AI/ML could fully replace, rather than merely aid, human doctors across 10 routine psychiatric tasks. Neutral or non-opinion responses were excluded to elicit substantive judgments.
- 10 universal psychiatric tasks were assessed, spanning documentation, examination, interviewing, risk detection, diagnosis, treatment planning, referral, prognosis, and empathetic care.
- Task descriptions used neutral language and generic terms such as “machines” and “technology” to refer to AI innovations.
- Respondents rated whether future technology could fully replace and perform each task as well as or better than the average psychiatrist.
- The six response options ranged from “extremely unlikely” to “extremely likely,” without “don’t know,” neutral, or no-opinion choices.
- Participants predicting replacement as somewhat likely, likely, or extremely likely estimated when technology would acquire that capacity using five time ranges.
Data and Statistical methods
The final sample comprised 791 psychiatrists from 22 countries, and analyses summarized respondent characteristics and opinions about AI/ML replacement across 10 psychiatric tasks. Likelihood responses were also collapsed into positive and negative categories for selected contrasts.
- De-identified data were analyzed using summary statistics, 95% confidence intervals, and descriptive statistics for characteristics and replacement opinions.
- For some contrasts, responses were collapsed into positive replacement opinions versus negative replacement opinions, with non-overlapping 95% confidence intervals treated as qualitatively different.
- 791 psychiatrists from 22 countries constituted the final respondent sample.
- Participants worked in public clinics (52%), private practice (35%), and academia (13%).
Opinions about AI/technological replacement of physician jobs
Psychiatrists anticipated varied effects of AI/ML on their future work over the next 25 years, with substantial change more common than obsolescence.
- 48.7% expected AI/ML to have no or minimal influence on psychiatrists’ future work over the next 25 years.
- 3.8% felt future technology would make their jobs obsolete.
- 47% predicted that AI/ML would moderately change their jobs over the next 25 years.
Opinions about AI/technological replacement of specific psychiatric tasks
Psychiatrists viewed AI/ML as more capable of replacing documentation and diagnostic information synthesis than core interpersonal, evaluative, and treatment-planning tasks.
- 83% considered it unlikely that future technology could provide empathic care as well as or better than the average psychiatrist.
- Most psychiatrists considered replacement unlikely for mental status examinations (67%), acute homicidal-thought evaluation (58%), and medical-history interviews (58%).
- Replacement was also considered unlikely for inpatient-versus-outpatient referral decisions (55%), personalized treatment planning (53%), and suicidal-thought evaluation (52%).
- 83% judged future technology likely to replace human physicians for documentation, such as updating medical records.47% expected this capacity within four years, while another 37% estimated five to ten years.
- 54% believed future technology likely to fully replace physicians in synthesizing information to reach diagnoses.32% predicted this capacity within four years, with another 41% estimating five to ten years.
- Psychiatrists’ predicted timelines for AI/ML capacity to replace them varied across the specific psychiatric tasks.
Opinions on Potential Benefits and Risks of Future Technologies/AI
Most respondents did not clearly conclude that AI/ML’s potential benefits would outweigh its risks in psychiatry. Judgments varied by physician gender and practice location.
- Risk-benefit judgments varied by physician gender and practice location.
- 40% were uncertain that AI/ML benefits would outweigh possible risks or harms, while 25% said they would not.
- 36% felt that future AI/ML’s potential benefits would outweigh possible risks in psychiatry.
- 23% of women versus 41% of men predicted that AI benefits would outweigh possible risks.
Predictions about how AI/ML technologies could help or harm clinical care
Psychiatrists identified operational, informational, and access-related benefits of future AI/ML, alongside concerns about empathy, privacy, assessment, workload, and professional judgment.
- Respondents were invited to submit open-ended qualitative comments elaborating on their survey choices.
- Respondents identified potential benefits including eliminating human error, standardizing and personalizing care plans, integrating big data, and improving scalability where psychiatrists are scarce.
- Other proposed benefits included more truthful patient responses, training beginner psychiatrists, streamlining workflow, and clarifying currently opaque disease etiologies.
- Identified risks included lack of empathy, reduced personhood, job-displacement antipathy, less privacy, and greater fatalism.
- Respondents also worried that AI might assess mental status incomprehensively, increase workload and burnout, and cause physicians to forsake creative thinking.
Discussion
Psychiatrists generally expected AI/ML to transform rather than eliminate their work, while remaining skeptical that it could perform core relational and complex clinical tasks as well as humans. They also identified potential efficiency gains alongside substantial ethical and professional risks.
- About 1 in 2 psychiatrists believed AI/ML would substantially change their jobs, but only 3.8% felt it would make their jobs obsolete.
- 83% considered it unlikely that future technology could provide empathic care as well as or better than the average psychiatrist.
- Psychiatrists also viewed mental status examinations, dangerous-behavior evaluations, personalized treatment planning, and other complex psychiatric tasks as unlikely to be performed as well by AI/ML.
- Female psychiatrists and US-based psychiatrists were more uncertain about AI/ML benefits outweighing risks than male and non-US psychiatrists.
- Respondents saw AI/ML as potentially reducing administrative burden, supporting monitoring and individualized treatment, scaling care, and reducing errors, but also risking privacy, transparency, diagnosis, empathy, and physician control.
Strengths and Limitations
This first global survey drew a relatively large, diverse sample of practicing psychiatrists across 22 countries. Its findings are preliminary because of sampling limitations, possible confounding, and the inability to establish causality or predictive validity.
- Strengths: The survey was the first global investigation of psychiatrists’ opinions about AI/ML’s impact on psychiatry.
- Limitations: Sampling was limited by the relative absence of respondents from developing nations, platform-based sampling, response biases, and unmeasured confounding.
- Methods: The survey focused on broad psychiatric functions rather than granular subtasks to reduce anthropocentric bias.
- Limitations: The survey cannot determine causality or predictive validity, and the impact of machine learning on psychiatry may not be known for decades.Accordingly, the findings should be interpreted as preliminary and replicated or expanded through population surveys.
- Contribution: Despite these limitations, the survey provides foundational insights into how psychiatrists think about future technologies in mental health care.
Conclusions
Future research should include patients, people with mental illness, and informaticians or AI experts alongside psychiatrists. Combining these perspectives could support better development and validation of machine-learning technologies and preparation for their implementation in mental health care.
- Conclusions: Future surveys should examine patients’ and people with mental illness’s views on AI’s impact on psychiatry and mental health services.
- Conclusions: Forecasts from informaticians and AI experts could be combined with these perspectives to broaden understanding of mental-health technologies.
- Conclusions: Such combined insights could help develop and validate machine-learning technologies and prepare mental-health professionals and patients for implementation.