Key Takeaways
- Medical students spend most of their clinical time on paperwork and computer work, leaving little room for direct patient interaction.
- AI tools such as the scribe “Heidi” and diagnostic aids like OpenEvidence can generate accurate histories and differential diagnoses instantly, reducing clerical burden.
- While medical schools are beginning to teach AI literacy, most prohibit students from using AI in clinical settings, creating a gap between training and real‑world practice.
- Learning to diagnose without AI remains essential for developing clinical reasoning, but integrating AI early can free time for empathy‑building and patient‑centered care.
- Empathy declines during medical training; off‑loading rote tasks to AI could help reverse this trend and strengthen the patient‑doctor relationship.
- Future curricula must teach not only how to use AI tools safely but also how to evaluate their limits, ethical implications, and impact on clinical workflow.
The Current Reality of Medical Student Workflows
In the hospital where I work, artificial intelligence is everywhere, yet medical students like me are barred from using it in patient interactions. I spend hours interviewing patients, scribbling notes, and then typing those details into the electronic medical record (EMR). “The majority of my day is spent in front of a computer screen,” I note, a reality that leaves little time for bedside conversation. This clerical overload mirrors a national trend: rising documentation demands and insurance requirements have pushed physicians toward burnout, with many reporting they feel pressed for time and dissatisfied with the amount of paperwork required.
AI Scribes and Diagnostic Assistants in Action
Beside me, supervising physicians activate “Heidi,” an AI scribe that records an entire patient encounter and instantly produces a polished, grammatically perfect history. In the same workroom, OpenEvidence—a tool often described as “ChatGPT for doctors”—is pulled up on at least one screen, offering rapid differential diagnoses and treatment suggestions. These technologies can generate paragraphs of detailed patient history in seconds, a task that would otherwise consume a significant portion of a student’s shift. As one physician remarked during a shift change, “Heidi cuts my note‑writing time by half, letting me focus on the patient rather than the keyboard.”
Why Students Are Still Practicing the Old Way
Despite the evident efficiency gains, medical students are prohibited from deploying AI during clinical encounters. The rationale, as educators explain, is twofold: first, students must master the art of diagnosis independently to cultivate clinical reasoning and critical thinking; second, there is concern that reliance on AI could cause learners to “offload their thinking” and fail to develop essential skills. I agree that learning to identify conditions such as heart disease, pneumonia, or diabetes without algorithmic help is fundamental. Yet the current split means I devote most of my energy to typing notes rather than honing the interpersonal skills that truly define physician competence.
The Educational Landscape: AI in Medical Schools
Medical schools are experimenting with AI, but the applications remain limited. Many institutions use AI as a personalized learning coach or a research aid, and 77 % of degree‑granting schools reported some AI coursework in 2024. At Harvard Medical School, for example, students can access AI models for experimentation, research projects, and schoolwork through brief modules and ad hoc workshops covering prompt crafting and hallucination detection. However, these sessions rarely translate into hands‑on clinical use. As one curriculum director confessed, “We are still uncomfortable letting students lean on AI in real patient care because we fear it will become a crutch rather than a catalyst for learning.”
The Empathy Crisis and AI’s Potential Role
A compelling reason to rethink AI restrictions looms large: empathy erodes as medical training progresses. Students learn implicit lessons to speed through interactions to accommodate administrative work, which diminishes the relational core of medicine. If AI assumes the rote tasks of note‑taking, data entry, and preliminary differential generation, students could reclaim minutes—if not hours—each day for genuine patient engagement. “AI is well positioned to take over the rote tasks and allow us to focus on patients,” I argue, noting that such a shift could directly counteract the empathy decline observed across cohorts.
A Patient Story Illustrating the Power of Presence
I recently experienced how those extra minutes can change outcomes. I met a patient bleeding from his gut who needed an urgent colonoscopy but initially refused. Over two weeks, I carved out a few extra minutes each afternoon to sit with him, learning about his job as a bartender, hearing childhood memories of his sister, and sneaking him peanut M&Ms from the cafeteria. I listened to his frustration about prolonged workups, his shame at being shuttled back and forth to the bathroom, and his fear of loneliness. One afternoon he said, “I feel like you see me as a person.” That affirmation paved the way for his eventual consent to the procedure. His words capture the essence of what physicians truly do: “Doctors don’t just diagnose disease. Doctors don’t even treat disease. Doctors treat people.”
Toward a Curriculum that Marries Clinical Reasoning with AI Fluency
To prepare future physicians for a world where AI is ubiquitous, medical education must evolve beyond basic AI literacy. The Association of American Medical Colleges plans to release a list of foundational AI competencies this fall, likely covering proficiency in using AI tools, understanding their social and ethical implications, and evaluating their safety and effectiveness in clinical settings. However, competence alone is insufficient; students must also learn to discern when AI augments judgment and when it risks undermining it. As one senior faculty member warned, “We need to teach students not just how to prompt an AI model, but how to critically appraise its output, recognize bias, and remain ultimately responsible for the patient’s care.”
Conclusion: Reclaiming the Human Core of Medicine
The tension between mastering traditional diagnostic skills and embracing AI assistance is real, but it need not be zero‑sum. By permitting supervised, reflective use of AI in clinical environments—starting with scribing and decision support—medical schools can alleviate the documentation burden that steals time from patient contact. That reclaimed time can then be invested in listening, empathizing, and building the therapeutic relationships that, as my bleeding‑gut patient reminded me, are the very heart of healing. In short, training students to wield AI wisely will not replace the clinician’s judgment; it will free the clinician to be the healer patients truly need.
https://www.bostonglobe.com/2026/09/19/opinion/artificial-intelligence-medical-education/