Using Generative AI to Support Data-Driven Decisions in STEM Library Collections
Library collection management can benefit from data-driven decision-making. However, it can be challenging to evaluate print materials against several retention criteria simultaneously, such as historical usage, recency of use, rarity, and electronic access. Here, we present a workflow that uses a generative AI chatbot to operationalize a collection development policy by translating policy criteria into rules-based decisions. We describe our experiences applying the workflow during two large-scale weeding projects for science, engineering, and computer science print collections at an academic library. This human-in-the-loop approach combines flexibility with systematic decision-making, enabling consistent application of retention criteria while preserving opportunities for librarian review and judgment when needed. In addition, it supports deeper analysis of borrowing trends that can inform future collection development policies. The workflow had a significant practical impact. Across the two weeding projects involving approximately 30,000 items, the approach enabled thousands of collection decisions to be applied consistently using policy-defined criteria while reserving librarian effort for review and exception handling. This approach offers an efficient, scalable, and accessible model for data-driven collection management. This talk was presented at the AI and Academic Libraries Online Conference on November 16, 2026.
Authors
- Charles D. Griego (ORCID: https://orcid.org/0000-0002-2051-7491)
- Melanie Gainey
Publication Details
- Journal
- KiltHub Repository
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1184/r1/35029162
- Primary Topic
- Library Collection Development and Digital Resources
- Type
- article
- Field-Weighted Citation Impact
- 0.00