AI trainees: Never-skilled future doctors?

The increasing use of AI tools by medical trainees raises concerns about 'never-skilling,' where individuals may never develop fundamental clinical reasoning abilities. Tools like OpenEvidence, an AI chatbot, provide rapid answers to complex medical queries, allowing trainees to bypass the crucial learning process of independent thought and problem-solving.

Historically, medical training has been an apprenticeship, shaping physicians through failure and uncertainty. However, AI tools insert themselves into this cognitive development, potentially hindering the formation of independent judgment. This risks creating a generation of supervisors of reasoning rather than actual reasoners.

A study in Nature Medicine highlights that AI tools, even those citing current literature, can be unreliable. Trainees recognize AI as a potential crutch but feel compelled to use it to keep pace with peers.

The solution requires structural changes from medical institutions, not just individual restraint. A proposed approach is 'reason first, consult AI second,' where trainees must demonstrate unaided reasoning before using AI. This mirrors aviation's approach to automation, emphasizing the preservation of manual competence.

Furthermore, trainees should be taught to critically interrogate AI outputs, discerning accurate information from subtle flaws. This involves simulated drills and assessments focused on disciplined judgment rather than reflexive skepticism. The goal is not to make training harder but to cultivate core competencies. Patients need doctors who can critically evaluate AI, not be subordinated by it, developing rich bedside judgment that AI alone cannot replicate.