Performance of AI Chatbots on Head and Neck Pathology Board-Style Exam Questions and Guidelines for Responsible Use

Abstract

The promising integration of artificial intelligence (AI), particularly large language models (LLMs) or AI chatbots, into medical education and practice necessitates rigorous evaluation of their capabilities. While chatbot performance has been assessed against standardized exams like the bar, medical and dental boards, and in diagnostic specialties such as radiology and ophthalmology, a critical gap exists within the complex discipline of pathology, specifically concerning subspecialty areas like head and neck pathology. Conducting a pilot study evaluating chatbot responses to head and neck pathology board-style questions is therefore essential and novel. Crucially for education, this pilot study evaluated six AI chatbots on their performance in answering multiple-choice questions (MCQs) retrieved from authorities in the field of head and neck pathology. Twenty MCQs relevant to head and neck pathology were identified from the public domain. A total of 120 responses from six chatbots were evaluated for response and citation accuracy. Although AI chatbots answered head and neck pathology board-style questions with 85-100% accuracy, citation accuracy and performance on image-based questions were poor. Considering the pedagogical principle that actively challenging exam questions enhances memory retention and learning outcomes compared to passive text review, AI-generated multiple-choice questions specifically designed for head and neck pathology study revision purposes were generated. The generated MCQs were determined to be at a more fundamental level and suitable for predoctoral dental students or first-year pathology residents. This pilot study probed the limitations of these tools in accurately addressing diagnostic and interpretive challenges. Furthermore, the recommendations and guidelines informed from the results of this pilot study represent a vital first step in the addressing the responsible use of AI for head and neck pathology education and practice.