AI can make large bodies of archival description accessible to new forms of questioning across collections, people, places, subjects, and relationships. It can also obscure the origins of an answer, flatten archival structure, or present unsupported claims with confidence. This workshop develops a practical approach to using AI in special collections research while keeping evidence, provenance, transparency, and verification central. 

The session places this approach within Yale Library’s existing discovery landscape. Participants will examine Quicksearch, Archives at Yale, and LUX as distinct systems for locating and exploring collections, then consider AI as an additional discovery pathway with different capabilities, assumptions, and limitations. Commercial and institutionally hosted tools will be considered alongside the possibility of running a model locally, with attention to privacy, control, and the reliability of results. 

Using a prepared Visual Studio Code environment, the instructors will demonstrate AI-assisted inquiry across more than 6,000 published finding aids in Yale Library’s public EAD3 dataset. The environment includes context specific to archival description and the dataset, helping the system interpret collection structures, retain connections to supporting records, and produce results that can be inspected and verified. Participants may follow the guided demonstration directly or work alongside selected portions on a laptop.

Instructors: Sandrine Guérin, Processing Archivist, Beinecke/Yale Special Collections

                   Will Nyarko, Metadata Specialist, Special Collections Metadata Services

Attendees will learn how to:

• Explain how Quicksearch, Archives at Yale, LUX, AI tools, and library staff can each support different parts of archival research. 

• Describe what would be needed to apply the same general approach to another library dataset or research interface. 

• Decide when to revise a question, consult the original finding aid, try another discovery tool, or ask library staff for guidance 

• Check AI-generated answers against finding aids and other source records, paying attention to provenance, archival context, and uncertainty. 

• Assess whether an AI tool is likely to answer a research question based on the information, context, and tools available to it. 

Audience: Students, researchers, and library staff

Registration: Yes

Workshop Incentive Program: Any student who attends at least 3 library workshops this semester will be eligible to receive a FREE Yale Library tote bag. Learn more.