AI-Generated Citations in a Senior Capstone
You are the library information specialist at a high school where all 92 seniors complete a yearlong capstone paper. The capstone coordinator asks you to deliver one 50-minute lesson on 'credible sources' after several students cited articles that do not exist; the citations had plausible authors, journals, and dates, and at least two students used an AI chatbot to generate annotated bibliographies. The district's acceptable-use policy predates generative AI. Teachers disagree: some want AI banned, others want students taught to use it responsibly, and one asks you to run every senior draft through an AI detector. Your library has 30 laptops, a projector, and no dedicated technology coach. The capstone rubric currently awards points for 'using academic sources' but does not define source quality or disclose AI use. You have one class period per section and no additional staffing.
1. What should the single 50-minute lesson prioritize?
Read the worked answer
The lesson should prioritize one transferable routine: verifying whether a cited source exists and whether it says what the student claims. Hallucinated citations are a symptom of students treating AI output as a finished product. I would use a 10-minute demonstration: take a plausible-looking citation generated by a chatbot, then check the journal site, the DOI, the author's institutional page, and a library database. Then students would practice with two citations-one real, one fabricated-and record where they looked and what they found. I would explicitly address AI use: it can help brainstorm or locate keywords, but the student is responsible for verifying every citation and for disclosing AI assistance per a class norm. I would not spend the period debating whether AI is good or bad, and I would not use an AI detector, which is unreliable and can damage trust. The capstone coordinator should get a one-page verification checklist to embed in the assignment.
2. How would you assess whether students can evaluate sources after this lesson?
Read the worked answer
Assess performance on a short, authentic task rather than a quiz about source types. I would give students two source sets drawn from capstone topics: one set with a real peer-reviewed article and a predatory or irrelevant article; another with a fabricated citation and a real one. Students would have 15 minutes to decide which source they would use and write a justification citing at least two verification moves. I would score with a 3-point rubric: verification action, evidence about the source, and fit to the research question. I would also collect a random sample of annotated bibliographies four weeks later and check citation existence and claim-source alignment. Pre/post comparison would use the same rubric, not the same sources. This measures transfer and gives the capstone committee data for revising the rubric, while avoiding the false precision of an AI detector.
3. What should you recommend to the capstone committee on AI and source quality?
Read the worked answer
Recommend a two-part policy: disclosure and verification. Students may use generative AI for brainstorming, keyword development, or explaining a concept, but they must disclose the tool and the purpose, and they may not submit AI-generated citations or annotations without verifying each one. The capstone rubric should add a source-verification criterion and a short reflection on how sources were found and checked. The library should provide a citation-verification guide and a short asynchronous module, since one 50-minute lesson cannot carry the whole skill. I would also recommend that the committee not ban AI outright: bans are hard to enforce, drive use underground, and do not teach judgment. Instead, require a process that makes verification visible. Finally, I would propose a small pilot with two teachers, collect examples, and revise the policy before schoolwide adoption.