The Missing Citation: An AI Summary and the Moving County Seat
Mrs. Nkemelu's eighth graders are investigating why their county's seat moved from a river town to a railroad town in 1832. Working in pairs, they gathered five sources. First, an AI chatbot summary that states confidently that a courthouse fire in 1874 destroyed all early records and that the move was driven by flooding. Second, an unsigned local history blog post that repeats the flooding story. Third, an 1885 county history, which says the legislature acted after a petition from landowners and mentions land speculation by three named men. Fourth, a digitized 1832 newspaper notice announcing the election of commissioners to build a new courthouse. Fifth, the county treasurer's ledger, showing a payment in 1833 for a new courthouse and no entries mentioning fire damage. One pair has cited the chatbot's "1874 fire" claim in their draft, complete with a citation to a book title that no one can find in the state archive catalog. The drafts are due in two days.
1. What should the pair do about the citation that cannot be found in the catalog, and what does the episode teach about apparently authoritative summaries?
Read the worked answer
The pair should not quietly delete the citation to save the paragraph, and they should not keep it because the summary sounded confident. They should treat it as an unverified claim and go looking for independent confirmation in the two places that would settle it: the county's own records, meaning the treasurer's ledger, and the state archive catalog for the session laws that created the new seat. The ledger they already have shows a courthouse payment in 1833 and no fire damage entries, which is evidence against the fire story rather than neutral silence. The broader lesson is about how generated summaries fail: they produce fluent prose with plausible titles and dates, and the plausibility is the problem, because a fabricated title looks exactly like a real one until someone checks. The routine the class should adopt is boring and effective, which is verifying every citation in an independent catalog before it goes in the draft.
2. The class must decide whether the evidence supports the claim "flooding caused the county seat to move." Design the corroboration check and state the verdict with reasoning.
Read the worked answer
The corroboration check starts from independence, not quantity. The chatbot summary and the blog post are not two sources if the blog's content is the likely origin of the summary, so the class should look for a chain of borrowing rather than counting hits. That leaves three independent items, and each has an interest to identify. The 1885 county history is the closest thing to a full account, but many such volumes were subscription publications in which families paid for favorable biographies, so its land-speculation explanation may reflect the politics of its own moment as much as 1832. The 1832 newspaper notice is closest in time and shows only that commissioners were elected to build, not why. The treasurer's ledger is the strongest independent check on money and shows activity consistent with a new building. The verdict should be that the flooding claim is unsupported, since its only apparent support traces back to one recycled sentence. The class should also note honestly that speculative motives are the hardest to document, and that the 1885 account needs corroboration before it is treated as established.