5 reasons AI should never spell ‘game over’ for physicians

Generative AI chatbots have been outscoring human clinicians on perceived empathy for some time now. 

The operative word is, of course, perceived. However, to many young people, the bots may as well be more authentically caring than any flesh-and-blood persons in any line of consumer service and, indeed, every sphere involving one-to-one interaction. 

Those perceptions may be a problem for medical schools recruiting tomorrow’s doctors. After all, young people have long cited the desire to help people as a key to their choosing medicine as a career. 

Meanwhile certain corners of the healthcare AI industry continue hyping the notion that their technology will render physicians unneeded. 

Researchers in the U.S. and Italy give medical-career doomers reasons to stay the course in a paper published by the Journal of the American Medical Association April 29.

“Both claims—that AI has become compassionate and that physicians are becoming obsolete—arise from the same evidence, but both are wrong,” write Canio Martinelli, MD, and colleagues. “The algorithm did not become compassionate; however, the medical profession has drifted so far from the bedside that an AI language model can now outperform physicians on the very quality that medicine assumed would remain uniquely human.”

“Something,” the authors add, “has gone terribly wrong in medicine.”

With that dark acknowledgment registered, Martinelli and co-researchers break down the nature of the challenge and the opportunity to recover—or avoid losing—a sunny outlook.  

Among their salient points are these five. 
 

1. No AI chatbot has examined a patient, been able to interpret a patient’s facial expressions or carried out a plan of care through patient uncertainty. 

In experimental settings, the highest empathy ratings have been notched when AI-generated text was deliberately misattributed to a physician, Martinelli and colleagues point out. “People wanted to believe their physician wrote something so caring,” they remark. “The algorithm met their expectation.”
 

2. AI cannot replace a physician, but it can remove the burden of what replaced a physician’s clinical time.

Documentation, coding, triage and pattern recognition are “essential but substitutable,” the authors state. “A physical examination, a hand on the shoulder and the willingness to stay with patients and listen to their fears are irreducible and nondelegable.”
 

3. AI does not replace physician judgment; it preserves it.

“Large language models already match or exceed resident physicians in diagnostic accuracy under time pressure, which is precisely the condition in which human reasoning degrades,” Martinelli and co-authors write. “A physician whose cognitive reserves are spent on documentation has less capacity for the unpredictable moment in patient care.”
 

4. The physician at the bedside—together with the rest of the care team—must remain accountable for governing the AI system. 

“An absent physician cannot audit an algorithm, but a present one can,” the authors maintain. “The risk is not AI itself but AI integrated without clinician governance or redesign of the work around it.”
 

5. AI creates a fork in the road. 

“One path leads back to the bedside and the other leads further from the bedside,” Martinelli et al. write, adding that faster documentation and higher throughput makes the absent clinician “more efficient at being absent.”

“The technology does not choose, but the profession does,” the authors note. 

“The bedside was always the foundation, but in building everything above it we simply forgot that we were standing on it,” they conclude. “AI is not the end of the physician—it is the return of every reason to be one.”