Resources

• Slides—Download PPTX
• LC blog post introducing the AI Planning Framework: https://blogs.loc.gov/thesignal/2023/11/introducing-the-lc-labs-artificial-intelligence-planning-framework/.
• LC AI use case risk assessment worksheet: https://github.com/LibraryOfCongress/labs-ai-framework/blob/main/Understand/Use_Case_Assessment_Worksheet_2023-11-15-draft.docx.
• Dikow, Rebecca B., Corey DiPietro, Michael G. Trizna, et al. 2023. "Developing responsible AI practices at the Smithsonian Institution." Research Ideas and Outcomes 9 (October), Article e113334. https://riojournal.com/article/113334/. 

Presenters

• Abigail Potter, Senior Innovation Specialist, Library of Congress
• Caroline Saccucci, Chief, U.S. Programs, Law, and Literature Division, Library of Congress

Description

Exploring Computational Description (ECD) is an experiment at the Library of Congress that is testing the effectiveness of machine learning models in creating key MARC record fields from ebooks. It builds on several years of investigating how machine learning could support the Library’s programs, collections, and services. ECD is expressly designed to support informed decision making about future system design and requirements. The goal of this work is to develop clear performance expectations, quality guidelines and technical requirements for using artificial intelligence (AI) to assist the creation of MARC records. Catalogers are co-leading this initiative, which utilizes new vendor documentation requirements while also refining quality review methods and techniques.

ECD is the first AI experiment moving through the LC Labs AI Planning Framework. The Framework offers a repeatable and iterative process to evaluate the impacts of AI tools. The steps help the Library develop policy, guidance, and staff fluency while responsibly adopting AI for specific use cases. This presentation shares the Framework, provides updates on experiments-in-progress, and invites feedback on the findings so far and the possibilities for the future.

This webinar will be of interest to anyone seeking information and guidance on how to responsibly implement AI tools in libraries, archives, and museums, and those interested in the results of experiments testing AI tools that could potentially assist catalogers. 

All affiliates of OCLC Research Library Partnership organizations are invited to participate.