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

Area of Science:

• Health Sciences Librarianship
• Artificial Intelligence in Information Science

Background:

• Collection development is crucial for health sciences libraries.
• Generative AI offers potential tools for library professionals.

Purpose of the Study:

• To evaluate the efficacy of generative AI models in health sciences library collection development.
• To assess AI model performance in identifying subject gaps and suggesting new resources.

Main Methods:

• Four generative AI models (ChatGPT 4.0, Google Gemini, Perplexity, Microsoft Copilot) were tested over six months.
• Two distinct prompts were used: one for generating eBook titles and another for identifying collection gaps.

Main Results:

• AI models showed inconsistencies and inaccuracies when generating eBook titles.
• All tested models provided useful analysis and accurate Library of Congress call numbers for identifying subject gaps.
• Inaccuracies and hallucinations were noted in some AI model outputs.

Conclusions:

• Large Language Models (LLMs) are currently unreliable as primary tools for health sciences collection development.
• LLMs can function as supplementary tools for analyzing collection coverage and identifying subject gaps.