NARA Building Capacity Through Responsible Use of AI

NISO Member News

Washington DC | October 17, 2024

Last week Archivist of the United States Dr. Colleen Shogan announced a new Strategic Framework for the National Archives. The framework, a template that will guide the development of a full Strategic Plan, charts a course for the agency that emphasizes building digital capacity, scalability, and responsibly embracing technological innovation.

“Our mission is both straightforward and complex: We preserve, protect, and share the historical records of the United States to promote public inquiry and strengthen democratic participation,” said Shogan. “The goals outlined in this framework will guide our efforts as we successfully navigate the complexities of the rapidly evolving digital landscape and strive to engage all Americans in meaningful ways.” 

One of the agency’s key objectives outlined in the framework is to incorporate artificial intelligence (AI) into archival recordkeeping and information-sharing practices to make it easier for everyone to use the records held by the National Archives. As a nonpartisan institution dedicated to making the nation's history accessible, the National Archives does not change the records within its holdings or interpret them. Making technology tools including AI and machine learning available to researchers and the public can enable more Americans to have greater success navigating the agency’s vast holdings.

NARA's early AI projects have showcased the technology's strengths by improving response times for records requests and making information from holdings more easily understood. One of the first uses of AI at NARA, in 2022, helped identify names in the 1950 Census before the records were released. Census records are a rich resource for genealogists—but it can be difficult and time-consuming to find names. These names were handwritten by census takers and can be difficult to read. NARA was able to use AI to identify names and make the records searchable within the National Archives Catalog, making it easier for the public to search and find family members in the census on the day that it was released.

While working to eliminate a backlog of National Personnel Records Center records requests from veterans and their families that had built up early in the pandemic, NARA also conducted a promising proof of concept for the use of AI-driven Robotic Process Automation (RPA), which may help the agency more efficiently manage routine operations in the future. In both projects, AI directly supported the work of archivists and made smaller pieces from large quantities of information accessible to the public more quickly.

Additional projects are exploring how AI can improve the efficiency of conducting Freedom of Information Act and other document reviews, capture metadata in microfilm digitization, safeguard personally identifiable information (PII), and perform natural-language search queries in digitized records. A pilot project is in development to test the capacity of AI to perform user-directed search queries. Known as ArchieAI, the pilot is slated to be opened to the public for testing and feedback in December 2024.

Every AI project at the National Archives depends on the expertise of multidisciplinary teams of  employees to establish use cases, document testing, set parameters, and validate results. The agency is conducting these projects within the context of a larger U.S federal framework for trustworthy use of AI, including guidance outlined in the Executive Order on the Safe, Secure, and Trustworthy Development and Use of AI and an AI risk management framework from the U.S. National Institute of Standards and Technology (NIST). Project results will be used to further develop AI governance at the National Archives, which will weigh innovation, risk management, and resources. 

AI is also being tested as an administrative business tool to help NARA employees work more efficiently in day-to-day tasks. Employees in the pilot project can access Google Gemini AI capabilities within the Google applications used at NARA for help summarizing documents, writing emails, and creating presentations and data visualization. The Gemini pilot greets employees with a message that reminds pilot users that no data will be shared outside of the National Archives environment and will not be used to train Google’s AI model.

National Archives Chief Information Officer Sheena Burrell stated, "AI technology has the potential to revolutionize the way we work at NARA. By automating routine tasks and providing us with new tools to analyze and understand our data, AI can help us to be more efficient, effective, and responsive to the needs of our customers."

NARA’s inventory of AI use cases to date are listed on its website.

“Like any new tool, we have to assess how best to use artificial intelligence and ensure we are prepared to use it smartly. I have said from my first day here that access is one of my top priorities for the National Archives. We have literally billions of records, and we will never be able to provide access broad enough to meet the needs of all Americans without thoughtfully embracing new tools like AI,” Shogan said.