Generative Artificial Intelligence in Medicinal Chemistry Education: Student Engagement, Confidence Calibration, and Safety-Related Uncertainty in Case-Based Learning

Abstract

The current study focuses on the function and use of generative artificial intelligence in teaching medicinal chemistry, with particular emphasis on learning process dynamics, decision-making power, and confidence estimation. A total of 101 undergraduate students in pharmacy schools were assigned to work on four case-type medicinal chemistry issues with optional artificial intelligence help. The variables in this research included task-specific AI use, confidence (pre/post), agreement with AI-generated outputs, and perceived certainty regarding chemical correctness and safety. Findings include that while there was considerable variation in usage patterns depending on the task type, there was considerable usage for explanatory tasks, with reduced reliance for higher-stakes decision-making tasks (structural modification/ADMET). Confidence recalibration was also found to vary: while most participants showed increased confidence after interacting with AI, many showed little or no confidence. Notably, there was considerable agreement with the output of AI, even in situations of doubt about its correctness or safety. The qualitative results showed that students mainly viewed AI as an entity for conceptual scaffolding, and their experiences within the disciplines influenced conditional trust and challenges in the process of verification. In general, the research indicates that generative AI is a high-fluency cognitive aid that alters traditionally assessed metacognitive control and monitoring and not just improves it. AI can help students understand medicinal chemistry concepts but also highlights the weaknesses of their capacity for the process of verifying the correctness and safety of the results.