University Libraries receives grant to develop AI-assisted metadata workflows
The Digital Public Library of America awarded University Libraries a grant to pilot an AI project that keeps human expertise at the center of cultural heritage work. The Metadata Remediation and AI-Enhanced Workflow Pilot will develop practical workflows for metadata remediation while preparing collections for the Digital Virginias service hub, which connects regional cultural institutions to DPLA's national platform.
The project investigates how tools like GPT and OpenRefine can assist with metadata remediation across diverse digital collections. Unlike many AI initiatives focused on full automation, this work positions people, not systems, as the primary decision-makers.
Human judgment remains essential
Wen Nie Ng, digital collections and emerging formats librarian at University Libraries, said the project intentionally centers human expertise. "Instead of replacing catalogers or metadata professionals, it looks at how AI can help reduce repetitive tasks and elevate the intellectual labor that humans do best - making contextual, ethical, and community-informed decisions about collections," Ng said.
The pilot asks a fundamental question: How can cultural heritage professionals prepare for a world where AI is ubiquitous, yet still requires thoughtful human oversight?
Student training emphasizes critical thinking
Library student employees work directly with real metadata sets, performing tasks such as subject enrichment, vocabulary mapping, and evaluating description fields. The training goes beyond technical exposure to build professional judgment.
"Students learn not only how to use AI tools, but also when and why not to use them," Ng said. "The training is intentional. It builds professional judgement by asking students to analyze AI-generated outputs, identify inconsistencies, question assumptions, and understand the limitations of automated systems."
Ng used an analogy to explain the approach. "A spoon helps you eat, but it doesn't teach you how to use it. You still need to learn the angle, grip, and movement yourself. Across cultures, tools are used differently. AI functions the same way. It can assist with part of the process, but it cannot replace the ability to interpret context, apply standards, make ethical choices, or understand cultural nuance."
Students who recognize what AI cannot do naturally begin to question, reflect, and reason more deeply, Ng said.
Workflows designed for smaller institutions
A major outcome will be a reusable, lightweight AI-assisted metadata workflow designed for small museums, community archives, and under-resourced organizations that may lack specialized staffing. The workflow includes transparent documentation and clearly defined human oversight points.
"Like the spoon analogy, the workflow provides a shared structure, but each community can adapt it based on their unique collections, staffing, and values," Ng said.
This adaptability helps reduce barriers to metadata remediation while enabling more collections to be responsibly prepared for aggregation and discovery.
Positioning for regional leadership
The project will reduce processing bottlenecks and strengthen readiness for the Digital Virginias service hub. It also positions Virginia Tech as a leader in responsible, community-oriented AI work in libraries.
"For communities across Virginia and Appalachia, the project lowers barriers to sharing local histories, ensuring that more voices and stories can be represented in national digital ecosystems," Ng said.
Ng's motivation stems from firsthand experience with metadata remediation demands. "I have seen how much time and labor metadata remediation requires. It is foundational to creating access to materials, yet often overwhelming for institutions with limited staffing and resources. AI shouldn't replace expertise, but it can support it, creating more room for meaningful human work."
For professionals working with AI tools, understanding when and how to apply them responsibly is critical. Generative AI and LLM Courses and Prompt Engineering Courses can help develop these decision-making skills.
Your membership also unlocks:
