As AI changes archivists’ work, U-M is helping guide the future

For many people, archives can feel intimidating: boxes of records, unfamiliar search systems and materials that do not always appear with a simple keyword search.

But as more researchers turn to generative AI to brainstorm questions, generate search terms and imagine what records might exist, archivists are facing a new challenge: How do they guide researchers through archival collections when AI is shaping what people ask, how they search and what they expect to find?

At the University of Michigan School of Information and U-M Bentley Historical Library, this question sits at the center of a broader conversation about how AI is changing the way people seek, interpret and use information. 

“Even if you don’t want to use it yourself, people are still going to be coming in, asking you about it,” says Jesse Johnston, clinical assistant professor at UMSI. “It’s going to change the way that you do your work.” 

For Johnston, the question is not only how AI might be used in archives, how archival values might shape the future of AI. Archives often hold personal or official records, and archivists have developed systems and practices to provide trustworthy and ethical access to materials with sensitive information. 

“Archivists also can bring their professional and ethical perspectives into the conversation about what AI is and how it’s developed and implemented,” said Johnston. “So it’s a really important time to discuss how archival values can inform AI, too.” 

Johnston is one of the organizers of Archivists in the Loop, a June 4–5 symposium at the University of Michigan focused on the implications of generative AI for archival reference services. Sponsored by the Bentley Historical Library and UMSI, the event brings together archivists, technologists and researchers for workshops and discussions on description, archival reference, access and technology.

The event comes at a moment when AI tools are influencing the role and operations  of libraries and archival systems, but the role of archivists in shaping those tools remains unsettled.

For Johnston, the question began with a problem that predates generative AI: searching archival collections is not like searching Google or a library catalog.

Archival collections are often described at the box or folder level, not item by item, and most material is not digitized. A researcher may find a box of correspondence spanning 10 years, but not know whether it contains letters, photographs, telegrams or other materials until they open it. Because collections can be enormous, archivists often do not have the time or resources to describe every document individually.

“So often you can just find, like, here’s a box with records of interest,” Johnston said. “The box or folder title has the approximate dates or names that I’m interested in, and it probably has the stuff I’m looking for, but no keyword searching because that’s not super effective.” 

That is where AI becomes both promising and complicated. Tools that allow people to ask natural language questions of archival finding aids could make collections easier to navigate, especially for people who cannot visit an archive in person or do not know where to begin.

Johnston began experimenting with that possibility with Meghan Courtney, associate director for public engagement at the Bentley Historical Library. Together, they asked students in Johnston’s information organization course to compare traditional keyword searching with a natural language interface that searched archival finding aids.

The experiment helped them see AI not simply as a back-end technical tool, but as something that could reshape archival reference itself.

“AI is not just going to affect those back-end aspects of what archivists do,” Johnston said. “It’s actually also going to affect how they work with users.” 

Courtney said she and Johnston have been thinking about how technology might help archivists better communicate what they have, facilitate access and help people who are new to archives feel more comfortable asking questions.

At the Bentley Library, that work aligns with a broader effort to reach beyond academic audiences and support community members, students and first-time archive users.

“Whether we’re afraid or not, the truth is that our audience and our researchers are using these tools,” Courtney said. “We can’t ignore this moment when a lot of researchers are experimenting with AI.”

Courtney sees potential in AI as an entry point, particularly for people who may not see archives as places meant for them.

“If their way in is through generative AI, then great,” Courtney said. “If it makes them feel like, ‘I have discovered that there’s something in archives that’s exciting enough for me to want to go see it,’ then we have succeeded.” 

For her, AI is not replacing the role of archivists. Instead, it makes the archivist’s interpretive role more visible.

AI tools may help users generate keywords or find a starting point, but they do not always understand the structure of archival data. Archives rely on context, hierarchy and structured description in ways that differ from the open web.

“That’s one way that I think archivists will always have to translate,” Courtney said. “We’re doing something that’s a little different from what people who use the internet are used to in terms of searching.”

For Courtney, the goal is to make archival expertise more accessible, not less necessary.

“We want to be a guide, not a barrier,” she said.

That guiding role may become even more important as researchers begin to expect AI-supported access to archival materials.

Julia Corrin, a UMSI alumna and associate dean for distinctive collections and university archivist at Carnegie Mellon University, said archivists need to help define that future before others define it without them.

“AI is coming whether we want it to or not,” Corrin said. “If we as a field don’t start thinking further ahead and thinking further outside the box with how AI can assist in archival access, our users are going to define it instead of us.”

Corrin, who graduated from UMSI in 2012 and worked at the Bentley as a student, said archival principles must be part of the conversation as AI tools are built for historical research.

“Context is the most important thing about archives,” Corrin said. “The relationship and placement of two documents near each other is in and of itself a carrier of information.”

Without archivists at the table, she worries that AI systems could flatten or misunderstand what makes archives meaningful: original order, provenance, relationships among records and the interpretive work required to understand historical materials.

At the same time, Corrin sees opportunity. Archives contain enormous bodies of information that are not already built into large language models. That makes them valuable not only as historical repositories, but as places where researchers can encounter materials that have not already been summarized, scraped or absorbed into existing AI systems.

“This is actually one of the most exciting opportunities for archives in a long time,” Corrin said. “We are repositories full of untouched data sets that no one has been able to crunch before.”

The challenge, she said, is to help researchers use emerging tools without losing the context that makes archival research distinct, and without favoring content that is easiest for a large language model to parse.

“It’s a scary time,” Corrin said. “But it’s really a chance for archives to push themselves forward in the conversation if we can take advantage of the moment.”

That is where UMSI’s role and its collaboration with the Bentley Historical Library becomes especially important.

The School of Information brings together expertise in archives, libraries, human-computer interaction, data science, user experience, digital collections and the ethics of information systems. Johnston said that combination makes the school a natural place to lead conversations about AI and archival practice.

“I think UMSI is kind of unique in the sense that we’re a connective piece between the libraries, some of the people doing technical development, people understanding archival theory and archivists,” Johnston said.

Corrin said that interdisciplinary training shaped her own career. At UMSI and the Bentley, she learned to connect archival theory with hands-on practice and to approach new technologies with curiosity rather than fear.

“Even in 2010, UMSI had a really forward-thinking approach to digital records,” Corrin said. “I was really prepared to experiment, to take a hands-on approach and really just jump in.”

That willingness to experiment, she said, is what the field needs now.

“These conversations are so complex and are at such a nascent level,” Corrin said. “We need people who are thinking about archives in different ways to get in a room together and just talk and work.”

For Courtney, that is one of the most important parts of Archivists in the Loop: It brings together people who do not always get to speak directly to one another. Some participants work closely with researchers. Others build the systems that shape how search and access happen.

“We rarely get to be in the same space,” Courtney said. “We almost never get to talk to each other directly.”

The symposium is not designed to promote AI as an easy solution. Instead, organizers hope it creates space for archivists and technologists to think carefully about what should change, what should be preserved and what researchers need from archives in an AI-mediated world.

Johnston said he does not see the moment as a crisis, but as a call to participate.

“There are a lot of things to be concerned about, from energy use to intellectual property to the way we design and teach classes, as well as the ways we manage and provide access to archives,” Johnston said. “So we need to figure out how to be in the conversation. Now is the time to do that.”

RELATED

Learn more about UMSI’s Library and Information Science expertise by visiting UMSI’s website, and see  more research about archives and digital curation at UMSI. 

Read about Jesse Johnston’s research and interests by visiting his UMSI faculty profile. 

Visit the Bentley Historical Library at U-M to browse their collection. 

Check out Archivists in the Loop at U-M on June 4-5.