At Miami’s Miller School of Medicine, AI Curriculum begins with art

In addition to attending lectures and anatomy labs, first-year students at the University of Miami’s Miller School of Medicine are asked to gather around works of art. Their interpretations are layered, imaginative, sometimes contradictory. Later, instructors show them how an AI system parsed the same canvas: detailed, logical, but flat.

“You see immediately the difference in human perspective,” said Dr. Gauri Agarwal, an associate professor of clinical medicine and associate dean for curriculum at the University of Miami Miller School of Medicine. She also spearheads the medical humanities class that is a requirement for first-year medical students.

“A large language model analyzing the same painting will often come up with a few interpretations for which it can provide abundant detail, but not with the same degree of variety and depth as our students. This illustrates to them the value of their connection to the human experience. In that same vein, humans are essential in ensuring that AI tools are used responsibly in practice with thoughtful human judgment and insight.”

That juxtaposition captures the tension now running through medical education. Artificial intelligence is creeping into patient notes, diagnosis software, and even medical exams. The question for schools is no longer whether to teach AI but how to do so without hollowing out the humanity at the center of medicine.

The rush to teach AI in healthcare

In the past two years, medical schools have moved with unusual speed to embed AI into their curricula. Stanford was first, tying its medical training to the university’s Institute for Human-Centered AI through the RAISE-Health initiative. Harvard followed, introducing a pair of hands-on courses and a new PhD track in AI in Medicine, which drew 400 applicants for just seven spots. Mount Sinai went further still, offering every student access to ChatGPT Edu — a locked-down, higher-ed version of OpenAI’s flagship model — and weaving AI into its MD and graduate programs.

Miami, though, is charting a slightly different path. In early August, the Miller School of Medicine opened the Office of AI in Medical Education, the first of its kind in Florida. Its mission is not just to add classes but to reorganize how students — and faculty — think about technology.

Dr. Latha Chandran, the school’s executive dean, framed the move as part of a larger project called NextGenMD, a sweeping curriculum overhaul that has already produced two graduating classes. 

“Artificial intelligence will undoubtedly have a seismic effect on medical education and clinical practice, the likes of which we have never experienced before,” she said. “We wanted to be at the forefront of it as a national leader and thinker in this space. This prompted the creation of the Office of AI in Medical Education, one that can capitalize on the tremendous resources at our university and build programs to train students, residents, and faculty in the technical, ethical, and clinical possibilities within an AI-integrated healthcare infrastructure.”

The Miller School’s flagship offering is an elective: Introduction to Artificial Intelligence in Medicine and Public Health. In it, students learn the mechanics of machine learning, neural networks, and natural language processing. They also build simple AI agents — projects that force them to translate abstract algorithms into tools that might educate patients or deliver clinical knowledge.

For Dr. Shirin Shafazand, faculty director of the new office and professor of medicine, the most important part of the course is not the coding but the reckoning with AI’s limits. “Students responded favorably [to the course], mentioning that they gained considerable foundational knowledge and appreciated learning how to create useful AI agents,” she said. “But just as important are the small-group sessions where they discuss ethical and regulatory aspects with faculty and industry experts. They learn to see AI as a companion to human ingenuity, not a replacement.”

The next generation of medical scholarship

A 2025 study in BMC Medical Education found that more than 80 percent of medical students expressed positive attitudes toward AI, with most calling it an effective and credible learning tool that could help optimize study time. Another international review of curricula, published last year, concluded that while AI is appearing more frequently in medical education, programs are still weighted heavily toward ethics and evaluation, with fewer opportunities for technical training in coding or data science. 

“AI education is emerging but not yet universal,” said Lisa Howley, PhD, the AAMC’s Senior Director for Transforming Medical Education. “According to the AAMC’s Curriculum SCOPE Survey, approximately 77% of medical schools report teaching about AI—an over 20% increase in one year. The momentum is impressive and the rate of increased attention is something we rarely see in such a short time.”

That balance is intentional, Howley explained.

“We don’t expect every medical student to become a coder. The priority is understanding AI’s capabilities and limits, integrating it into clinical reasoning, and addressing ethics, equity, and patient safety.”

Still, the field is uneven. “Some schools are moving quickly while others are just beginning,” she noted. “We view one of our roles as helping close those gaps by providing frameworks, faculty development, and shared resources so that all students, regardless of institution, gain foundational preparation in AI.”

Faculty themselves acknowledge the barriers: limited infrastructure, crowded schedules, and a shortage of instructors with AI expertise. Howley underscored these points: “Key challenges include faculty capacity, the rapidly evolving nature of AI, and ensuring equity of access across schools. The AAMC is addressing these through competency development, convenings, and partnerships that make AI education sustainable and accessible.”

Taken together, these findings paint a nuanced picture: medical students are broadly receptive to AI and recognize its immediate value, but deeper engagement through formal curricula can breed more measured enthusiasm. At the same time, institutions struggle to balance ethics-driven instruction with practical, technical fluency—and to clear logistical and infrastructure roadblocks that stand in the way of more meaningful AI integration.

Measuring the impact

Miami is also trying to answer a harder question: not whether students like AI, but whether it makes them better doctors. 

“As we introduce new AI tools that help students practice clinical decision-making and provide feedback on performance, we will do so in the setting of education trial design,” Shafazand said. “We’ll measure student performance before and after the introduction of an AI learning tool. The goal is to train physicians who can improve patient care and outcomes with compassion and innovation.”

That ambition comes with challenges. Faculty need training just as much as students. Infrastructure lags. And the regulatory environment for medical AI is unsettled at best. 

“We have to be extremely nimble and focused,” Chandran admitted. “The ideal balance between the inputs of the man and the machine in providing optimal patient care remains to be determined carefully. At this point, there are lots of uncertainties — but lots of interesting and exciting possibilities as well.”

Dr. Agarwal, along with Miller student Isha Harshe and fellow professor Kenneth W. Goodman, published a paper in June that captures the tension at the heart of this new curriculum — algorithms may generate answers, but only humans can navigate uncertainty with empathy. For Miami, training students to see that difference is the point.