PRESENTED BY MMS MEMBER ADAM RODMAN, MD, MPH, FACP
In a compelling talk, “Towards a Medical Superintelligence,” at the Committee on Information Technology awards dinner and celebration on June 10th, Massachusetts Medical Society Member Adam Rodman, MD, MPH, FACP, assistant professor at Harvard Medical School and director of AI Programs at the Shapiro Center for Research and Education at BIDMC, explored the evolution of clinical reasoning and the emerging collaboration
between physicians and artificial intelligence.
Drawing on classic research from the 1970s, Dr. Rodman highlighted how expert clinicians often arrive at correct diagnoses more quickly than novices — not because they follow more steps, but because they possess rich, experience-based knowledge structures
known as "scripts." This dual-process model of thinking, popularized by Kahneman and Tversky, shows how physicians blend fast, intuitive heuristics (System 1) with slower, analytical processes (System 2).
Rodman emphasized that clinical decisions are also shaped by environmental and emotional context (“situated cognition”) and increasingly by other people’s inputs and technology (“distributed cognition”) — particularly in the digital era, where physicians
often integrate others’ documentation and insights via electronic health records.
Enter artificial intelligence. Large language models (LLMs) like GPT-4 have achieved remarkable results in simulated diagnostic tasks, even outperforming physicians in solving complex clinical cases like New England Journal Clinicopathologic Conferences.
When prompted to “think aloud” using structured reasoning (a method called “chain-of-thought prompting”), these models show marked improvements in diagnostic accuracy.
However, Dr. Rodman cautioned against assuming AI always enhances human performance. In multiple trials, clinicians using AI did not outperform AI alone — and in some cases, human input diluted the quality of the decision. Interestingly, AI did
help improve communication tasks, such as disclosing medical errors or coordinating care, where human strengths in empathy and nuance remain critical.
The presentation also introduced emerging autonomous AI systems, including Google’s AMIE (Articulate Medical Intelligence Explorer), capable of independently conducting patient interviews and formulating diagnostic impressions. Early results show AMIE
is often preferred by patients and performs comparably to primary care physicians. Live clinical trials are now underway.
As AI capabilities expand, a key challenge remains: designing systems that genuinely augment rather than replace physician judgment. Dr. Rodman closed by urging thoughtful exploration of this "strange new type of collaboration" — one that blends human
expertise with computational reasoning to improve care.
Learn more on AI from Dr. Rodman in our AI Learning Revolution: How to Apply AI in Continuing Education online course. For more AI developments in medicine, visit NEJM AI.
