Opinion: AMA CEO: AI won’t replace doctors — it will work alongside them

Ezekiel J. Emanuel and Abe Butler-Baker have been debating the proper place for AI in medicine with American Medical Association CEO John Whyte. Now, they are taking their discussion to STAT’s First Opinion. Read Whyte’s essay below and read Emanuel and Butler‘s essay here. Would you want artificial intelligence to tell you that you have cancer? Not help a physician identify the cancer. Not analyze the scan, synthesize your medical history, or recommend the treatment most likely to work. Would you want AI, on its own, to tell you that you have cancer? Or guide you through what comes next? These questions get at something missing from much of the debate about AI and medicine. As AI becomes more capable, we will have to decide not only which parts of medicine can be automated, but which parts we want to automate. Those are not the same question. I recently debated this issue with oncologist and bioethicist Ezekiel Emanuel, who has argued that AI will eventually perform many of the functions physicians perform today, and in some cases do them better. He may well be right about many of those functions. AI can already interpret images, summarize medical records, suggest diagnoses, and answer health questions with a fluency that seemed impossible just a few years ago. But there is a leap in logic from “AI can do many things physicians do” to “AI can replace physicians.” That leap rests on an imprecise view of what practicing medicine actually is. I am optimistic about what augmented intelligence can do for medicine. I believe physicians should play a leading role in shaping and adopting these technologies. But I profoundly disagree with the conclusion that increasingly capable AI makes the physician expendable. That conclusion reduces medicine to a collection of tasks: reading an X-ray, identifying a diagnosis, renewing a prescription, selecting a treatment. If AI can perform each task as well as, or better than, a physician, the logic goes, why do we need the physician? Because medicine has never simply been the completion of tasks. Medicine is recognizing when the diagnosis does not quite fit. It is understanding why the treatment supported by the evidence may not be right for the person sitting in front of you. It is helping a patient choose among imperfect options. It is delivering life-changing news. It is knowing when to push and when to listen. It is the art and the science of medicine. Too much of the debate over AI in health care asks a deceptively simple question: Can AI outperform physicians? Studies comparing AI and physicians on diagnostic questions, image interpretation, and other clinical tasks are important. We need rigorous benchmarks to understand where AI performs well and where it does not. But passing a benchmark is not the same as practicing medicine. Patients do not arrive as carefully constructed clinical vignettes. They describe symptoms imperfectly. Their medical records contain gaps and contradictions. The decisions they make are influenced by their preferences, fears, family circumstances, and tolerance for risk. And medicine requires something benchmarks rarely measure: responsibility. Physicians are licensed. They operate within professional and ethical standards. Their decisions can be scrutinized, challenged, and reviewed. Responsibility can ultimately be assigned. As AI assumes more consequential roles in care, those questions become more important. Who is responsible when an autonomous system makes the wrong diagnosis? What happens when it confidently reassures a patient who needed immediate care? Who recognizes when an algorithm’s recommendation may be statistically sound but wrong for the person in front of it? As AI advances, the companies that create it must be willing to bear the responsibility, and take on higher levels of liability, when decisions are wrong. These are not arguments against AI. They are arguments against confusing technical capability with clinical responsibility. AI tools and advanced digital devices are already changing medicine. Patients increasingly begin the medical encounter before they enter an exam room. They search symptoms, review laboratory results, and ask chatbots what a diagnosis might mean. Used well, these tools can help people understand complicated information and formulate better questions. For physicians, the opportunities are even greater. AI can help gastroenterologists identify polyps they might otherwise miss. It can help radiologists detect cancers earlier. It can synthesize years of medical records and reduce administrative work that consumes time physicians would rather spend with patients. Most exciting, AI could democratize access to medical expertise. Consider cancer care. A patient’s options can still depend partly on whether they live near a leading academic medical center or hours away from one. Imagine instead that every person diagnosed with cancer could receive an AI-assisted second opinion incorporating the latest evidence, their complete clinical history, and the characteristics of their individual cancer — with a physician interpreting those recommendations and helping the patient decide what comes next. That is not a lesser vision for AI because a physician remains involved. It is a more ambitious one. The goal should not be to reproduce today’s health care system with fewer humans. It should be to use AI to give every physician and every patient capabilities that today are available only in the best-resourced settings. There is another reason technological capability should not determine the future of medicine: Patients must have a say in what that future looks like. Some moments in medicine require something other than information. When a biopsy comes back positive, the patient’s first question may not be, “What does the evidence recommend?” It may be, “Am I going to die?” When a patient must choose between a treatment that may extend life and one that may preserve quality of life, there may be no algorithmically correct answer. The decision depends on values, relationships, uncertainty, and an understanding of what matters most to that person. AI can inform those conversations. It may make them better. But it should not conduct them on its own. The role of physicians will change. It should. Some tasks physicians perform today will be automated. AI may allow doctors to spend less time searching for information, documenting encounters, and completing administrative work, and more time exercising judgment and caring for patients. That would be progress. But we should not confuse the automation of medical tasks with the automation of medicine itself. Someday, an AI system may be perfectly capable of telling you that you have cancer. That doesn’t mean it should be the one sitting across from you when you hear the news. John Whyte, M.D., M.P.H., is CEO of the American Medical Association.