Checking out AI: UC Berkeley Library puts emerging technology to the test
UC Berkeley Library is testing AI tools to enhance cataloging, metadata generation, and sensitive data detection while emphasizing human oversight and legal compliance in academic research.
The UC Berkeley Library is evaluating AI tools to assist with cataloging, such as a custom interface built by Haiqing Lin that uses Google’s Gemini AI models to generate metadata for Chinese film posters. The tool, still in development, aims to speed up the slow process of cataloging by translating text and creating descriptive records, though Lin emphasizes that AI will support, not replace, librarians’ expertise. The project reflects the Library’s cautious approach to integrating AI amid broader discussions about its role in academic settings.
Carolyn Caizzi, associate university librarian for digital initiatives, is leading the Library’s efforts to assess AI’s potential and limitations in academic libraries. She highlights the need for a deliberate, human-centered approach to adopting these tools, noting that librarians’ expertise is essential in guiding AI applications. The Library is testing where AI can assist, where it falls short, and how to use it responsibly, positioning itself to adapt to technological changes while maintaining its role as an information professional.
Becky Miller, the university’s natural resources librarian, used an AI-powered metadata tool, JSTOR Seeklight, to process decades-old agricultural extension publications that had been inaccessible to researchers. The tool generated searchable metadata for 79 records, which were added to the Library’s Digital Collections portal. Miller plans to digitize more materials, emphasizing the value of making these resources discoverable for scientists studying climate change or historians examining California’s agricultural history.
At The Bancroft Library, Christina Velazquez Fidler employs AI-generated Python scripts to detect sensitive information, such as Social Security numbers, in born-digital archives. The scripts help isolate and protect personal data before materials are made public, accelerating the processing of collections. Meanwhile, the Scholarly Communication and Information Policy team, led by Rachael Samberg, works to remove legal barriers that prevent scholars from using AI for research, such as negotiating publisher contracts to ensure researchers retain necessary rights.