Ten Tips for AI‑Assisted Key Feature Problems: A Validity‑Informed Guide for Medical Education

<ns3:p>Generative artificial intelligence (AI) can augment educators’ capacity to design high-quality Key Feature Problems (KFPs) for valid assessment of clinical reasoning and decision-making. This practice-oriented guide presents ten evidence-informed tips for using AI to develop KFPs that are aligned with learning outcomes, cognitively demanding, and contextually authentic. Drawing on the KFP literature and contemporary validity frameworks (content, cognitive and response processes, internal structure, and consequences), we synthesize practical strategies for translating outcomes into key features, constructing realistic vignettes, creating parallel case variants, targeting higher-order thinking, ensuring curricular alignment and learner-level appropriateness, diversifying complementary item formats, validating AI-assisted items through a stepwise workflow, delivering decision-specific feedback, iterating from learner performance data, and safeguarding equity, ethics, and governance. We illustrate these recommendations with concise examples and an adapted validation workflow that supports both formative and summative applications. Although AI can accelerate scenario construction and feedback drafting, human expertise remains essential to verify clinical accuracy, prevent bias and hallucinations, calibrate difficulty, and preserve assessment security. With transparent processes and expert review, AI can serve as a collaborative assistant rather than a replacement, helping medical educators build rigorous KFPs that enhance the assessment of clinical decision-making.</ns3:p>