Structured AI-Supported Assessment (SAISA): A Novel Design Integrating ChatGPT to Support Clinical Reasoning and Reflective Engagement in Veterinary Students.

Abstract

The rapid emergence of generative artificial intelligence (AI) tools such as ChatGPT is reshaping higher education and challenging traditional approaches to assessment. In veterinary education, these developments raise critical questions about how clinical reasoning and professional judgement can be validly evaluated in an AI-enabled learning environment. Rather than excluding AI from assessment, there is growing interest in designing structured assessment models that integrate AI while preserving educational validity. This study describes and evaluates a Structured AI-Supported Assessment (SAISA) designed to integrate ChatGPT into veterinary assessment in a controlled, pedagogically grounded manner. SAISA was implemented within a fifth-year veterinary course (Equine Medicine and Surgery) and comprised three AI-integrated exercises: an AI-mediated role-playing consultation with ChatGPT functioning as a simulated client or instructor, an AI-supported stepwise clinical case resolution task, and an AI-based case creation and critical appraisal exercise. Student performance in a final non-AI-assisted case-based examination served as the primary outcome measure and was compared with a historical cohort from the same course and a concurrent cohort from a related course without AI-supported assessment. Students exposed to SAISA scored significantly higher on the final non-AI-assisted examination than both comparison cohorts. Exploratory survey data indicated high levels of reported engagement, reflective behavior, and confidence in clinical reasoning during AI-integrated assessment activities. These findings suggest that structured integration of generative AI within assessment is associated with improved performance on case-based tasks requiring clinical reasoning, supporting the redesign of assessment structures, rather than restriction of AI use alone, as a viable strategy for preserving assessment validity in an AI-enabled educational context.