Making the most of Artificial Intelligence and Large Language Models to support collection development in health sciences libraries

Abstract

This project investigated the potential of generative Al models in aiding health sciences librarians with collection development. Researchers at Chapman University's Harry and Diane Rinker Health Science campus evaluated four generative Al models-ChatGPT 4.0, Google Gemini, Perplexity, and Microsoft Copilot-over six months starting in March 2024. Two prompts were used: one to generate recent eBook titles in specific health sciences fields and another to identify subject gaps in the existing collection. The first prompt revealed inconsistencies across models, with Copilot and Perplexity providing sources but also inaccuracies. The second prompt yielded more useful results, with all models offering helpful analysis and accurate Library of Congress call numbers. The findings suggest that Large Language Models (LLMs) are not yet reliable as primary tools for collection development due to inaccuracies and hallucinations. However, they can serve as supplementary tools for analyzing subject coverage and identifying gaps in health sciences collections.