Policy, ethical, and practical considerations of artificial intelligence use in the internal medicine clerkship: results of a survey of clerkship directors.
Abstract
Purpose
The integration of artificial intelligence (AI) into medical education is accelerating, yet the policies guiding its use remain unclear. This study aimed to characterize the landscape of AI-related policies in U.S. internal medicine (IM) clerkships and examine clerkship directors' (CDs) perspectives on ethical and practical considerations for student AI use.
Method
An annual, nationally representative survey of IM CDs at 141 U.S. allopathic medical schools with membership in a medical education professional association was conducted from September through December 2024. One section included 21 questions on AI use in clerkships, policies about its use, and CDs' perspectives towards students' AI use. Descriptive statistics summarized policy prevalence and perspectives, and qualitative content analysis described free-text responses about policies.
Results
The response rate for the survey section on AI was 80.9% (114/141 schools). Most CDs reported no formal policy on AI use (45/114, 39.5%) or were uncertain about whether one existed (39/114, 34.2%). When policies existed, they required students to obtain explicit permission or supervision for AI use and mandated disclosure when using AI-generated content. CD perceptions of hypothetical student usage of AI varied: 49/114 (42.9%) described AI use for a reflection essay as "cheating" compared with 6/114 (5.2%) for clinical diagnosis (p < 0.001). Most CDs (84/114, 73.7%) reported that they would encourage AI use for at least one clinical skill domain-including differential diagnosis and patient education-but few (7/114, 6.1%) reported that their institution provides firewall-protected AI tools for clinical use.
Conclusions
Policies guiding AI use in IM clerkships do not yet exist or are not well known by CDs. Policy gaps and varying perspectives on appropriate AI use may create ethical and practical ambiguity for students and their clinical supervisors. Institutions should prioritize developing clear policies and secure infrastructures to support responsible, supervised AI use in medical education.