AI Chatbots’ Medical Hallucination: Innovation of References Hallucination Score and Comparison of Six Large Language Models (Preprint)

Abstract

Artificial intelligence (AI) chatbots have gained use recently in medical practice by healthcare practitioners. Interestingly, their output was found to have varying degrees of hallucination in content and references. Such hallucinations generate doubts about their output and their implementation.

We propose a reference hallucination score (RHS) to evaluate AI chatbots’ citation authenticity.

Six AI chatbots were challenged with the same ten medical prompts, requesting ten references per prompt. The Reference Hallucination Score (RHS) is composed of six bibliographic items and the reference’s relevance to prompts’ keywords. RHS was calculated for each reference, prompt, and type of prompt (basic versus complex). The average RHS was calculated for each AI chatbot and compared across the different types of prompts and AI chatbots.

Bard failed to generate any references. ChatGPT 3.5 and Bing generated the highest RHS (11), while Elicit and SciSpace generated the lowest RHS, and Perplexity was in the middle. The highest degree of hallucination was observed for reference relevancy to the prompt keywords (61.6%), while the lowest was reference titles (33.8%). AI chatbots generally had significantly higher RHS when prompted with scenarios or complex format prompts.

The variation in RHS underscores the necessity for a robust reference evaluation tool to improve the authenticity of AI chatbots. Also, it highlights the importance of verifying their output and citations. Elicit and SciSpace had negligible hallucination, while ChatGPT and Bing had critical levels. The proposed AI chatbots’ RHS could contribute to ongoing efforts to enhance AI’s general reliability in medical research.