An observation from working with AI this the past week.
I was experimenting a bit, curious what Claude’s research mode would do when faced with a research task that might require digging into the internet archive or wayback machine, or might require access to organizational documents on an intranet. Or, in the case of the problem I was thinking of, many years of different intranets. I tasked it with researching the history of parking policies for students at my institution (for reasons) and summarizing any policies limiting parking.
Claude made a good stab at finding archives of what would have been old publicly accessible content on the topic but found very little. It did identify potentially relevant archives at my school that I might want to check. It did also come up with a variety of information about our recent growth in enrollment and the scholarship enabled by a generous gift behind that growth, and speculated that those pressures might lead to changes in parking policies over time (very astutely observed). It did a decent job noting what it could validly conclude and where it would only be making guesses based on a lack of evidence.
What I was pleasantly surprised by was that, from the start, Claude kept telling me that I really should talk to my college archivist about this. It looked up her name and email address, let me know what archival sources I might suggest she look at, and offered to help me write an email. I kept poking around at different angles I could use to task Claude to find some content it had overlooked, and by the end it said very explicitly that rather than asking it to spend more time doing research, the “single highest-value next step” I could take would be to talk to my archivist.
Going into this semester, I hope this is a response that my students encounter where relevant as well!
One thought on “AI likes archivists”