Five Suspicious Bibliographies and No AI Policy
Northgate High School, grade 11, co-taught AP-style seminar with a social studies teacher. The summative task is a 2,000-word argument paper with an annotated bibliography of eight sources; it is 60% of the quarter grade, and there are 28 students. You and the teacher co-designed the unit, and you taught one 50-minute session on source evaluation and citation in September. After the drafts come in, five papers show problems: three contain citations to sources that cannot be found, one student's prose is markedly different from her fall benchmark writing sample, and one student tells the teacher privately that he used a chatbot to "help with the wording" and then cites its output. The teacher wants to assign zeros to all five and file academic-integrity reports. The district has no AI policy; the student handbook prohibits plagiarism and fabrication.
1. How do you distinguish permissible assistance from misconduct here, and what do you recommend for the five students?
Read the worked answer
With no AI policy, the students had no notice of a rule, and a zero on 60% of a quarter grade for an unannounced standard is not defensible even where the conduct was dishonest. But "no AI policy" does not license everything: fabricating citations already violated the assignment's stated requirements and the handbook, and that is what I would anchor any consequence to.
Evidence tiers matter. A style shift against a fall benchmark is weak - writing changes, and AI detectors are unreliable. Fabricated sources are verifiable, so I would check every entry in those bibliographies against the actual source. Then I would run short conferences with the five students asking each to walk me through how they found and evaluated source 4. A student who cannot locate their own source tells me something; a student who explains it well usually did the work.
For the student who admitted chatbot use, I would recommend the conversation center on process: what the tool did, what he learned, a rewrite with a written reflection, and an integrity referral only if the code covers misrepresentation of authorship. For the fabricators, refer under the existing code with a consequence proportional to what the school already does for invented sources. What I would not do is let five zeros rest on detector scores or on prose that "sounds too good."
2. How would you redesign the assessment so that this situation is visible earlier next time, and what formative data would you collect?
Read the worked answer
A single 2,000-word product at the end tells you nothing until it is too late to teach. I would redesign so the process is visible and assessable: a research-question check-in in week one; a five-source annotated bibliography in week two with one paragraph per source on authority, currency, and how it changes the argument; a 10-minute source conference in week three where the student defends two sources; a draft with a revision memo in week four.
The annotated bibliography is the load-bearing assessment. It makes evaluation visible, it is hard to fake convincingly, and it is teachable. The conference is where I collect formative data: can the student say why this author is credible for this claim? I would log it on a three-point rubric - cannot explain, explains with prompting, explains independently - and hand the roster to the teacher so she can group students for the next lesson.
I would also change the prompt so sources matter: require at least one source the student had to obtain through interlibrary loan or a database the open web does not surface, so "find a quote" stops being the task. And I would write the AI expectation into the assignment itself - what is allowed (brainstorming, grammar checks) and what is not (generated prose, generated citations) - because that is the only notice students will actually read.
3. The teacher says the library should have taught citation better. How do you respond?
Read the worked answer
Part of the criticism is fair, and I would say so first: our citation instruction was one 50-minute session in September, and I have no evidence about what students retained. That is a data problem, not only a public-relations one.
Then the facts: which sessions each class attended, what the exit tickets showed, and the co-designed rubric we agreed the bibliography would be scored against. If that rubric never reached the students, that is a shared fix, not a library failure.
The productive move is to convert blame into a shared assessment. I would propose that we each score ten randomly selected bibliographies against the same rubric and compare; the disagreements between our scores are the real finding, and they tell us what to re-teach. I would offer a 30-minute department session on source evaluation with a five-minute task teachers can reuse, and volunteer to co-score the next round.
What I would avoid is accepting the frame that the library "should have taught citation better," as though citation were a hygiene habit. It is a reasoning skill that needs more than one September session, and the paper is where it gets practiced.