Stabilizing and Destabilizing Factors of Nurse Retention: A Comparative Analysis of Two Healthcare Settings in Austria.

“AI literacy” is a critical skill for healthcare professionals, essential for tasks that are becoming more common, such as interpreting diagnostic algorithms, working with clinical decision support, and communicating with AI developers. The Association of American Medical Colleges (AAMC) recently (October 2025) put together a list of “AI Competencies for Medical Educators”. The first category is “Understanding AI (What is AI and how does it work?)”, including topics such as understanding AI fundamentals, describing the statistical properties of AI and varied AI approaches to data, and describing the strengths and weaknesses of various AI systems. While these competencies define what learners should know, they do not prescribe how educators can effectively teach such abstract and technical material. Generative AI now offers a promising pathway to bridge this gap by enabling rapid creation of interactive, hands-on learning experiences that make these AI principles digestible and attainable by health profession students who typically do not have a computer science background. Recent literature has rapidly cataloged the expanding uses of generative AI in medical education, including curriculum development, personalized learning materials, simulated patients, assessment support, and educational resource generation. These reviews consistently emphasize both the promise of these tools and recurring risks, including hallucinated or inaccurate content, algorithmic bias, academic integrity concerns, privacy, inequitable access, overreliance, and the need for faculty oversight. In contrast to these prior works focusing on leveraging AI for , we discuss here the potential of leveraging AI to teach given the growing push for AI literacy in HPE. Teaching health AI concepts is distinct from much of traditional HPE: rather than the recall of clinical facts or the mastery of procedural skills, it requires quantitative, engineering-oriented reasoning about concepts such as probabilistic model outputs, statistical performance metrics, and how a model’s behavior shifts as its settings are changed. Concepts such as the way an interpretable model’s predicted risk changes as its inputs are varied and the movement of a model’s operating point along an ROC curve as its decision threshold is adjusted can be naturally conveyed through interactive web and mobile interfaces. With only brief natural language prompts, large language models (LLMs) can now generate fully functional web-based simulations, dynamic visualizations, and data-driven exploratory tools that bring complex AI concepts to life for learners. Modern commercial LLMs like ChatGPT and Claude can easily be leveraged to create manipulable website interfaces such as sliders, toggles, or simulated clinical environments that respond in real time to user input, allowing learners to experiment with clinically oriented AI concepts without requiring any programming expertise. New specialized no-code AI tools for generating such interfaces are increasingly available.