Two 45-Minute Sessions, 142 Students, and a Citation Audit
At a 1,150-student high school on a 4x4 block, the 10th-grade ELA team is launching an argumentative research paper. Five sections (142 students) will visit the library twice, in 45-minute blocks. The team lead wants students to "just search Google" and turn in a Works Cited page by Friday. The librarian, Mr. Reyes, has a district subscription to a Gale database suite, a state-funded news database, and a print reference set last weeded in 2019. A citation audit of 120 papers last spring found 78% of URLs were content farms or AI-generated summaries, 9% of students used a database, and 31% of sources could not be located again at check-in. The school-improvement plan targets a 10-point gain in research-based writing, and the district reports database authentication logins to the board monthly. Four students in the group have IEPs requiring read-aloud and extended time.
1. Design the two 45-minute sessions so they produce a measurable outcome tied to the writing goal. Explain the sequencing choices.
Read the worked answer
Start from the artifact, not the tool. The outcome should be an annotated source set - three sources per student, each with a working permalink, a one-sentence relevance claim, and a one-sentence authority justification - submitted at the end of session two. That artifact is observable, gradable, and it is the input the teacher needs before drafting.
Session one (45 min): 8 minutes whole-group modeling on one live research question, showing a database search, a subject-tagged result, and the permalink/e-mail export; 25 minutes of student work on their own question with Mr. Reyes and the teacher conferencing; 7 minutes to save sources and complete an exit ticket naming one search strategy that worked; 5 minutes for the IEP group's read-aloud setup and extended-time plan. Session two: 10 minutes on verification - lateral reading, tracing a claim to its primary source, checking ownership and dates - using one credible and one content-farm article on the same topic; 25 minutes to complete the annotated set; 10 minutes of peer check against the rubric.
Sequencing rationale: searching is a small part of 45 minutes, so orientation must be asynchronous (a six-minute recorded walkthrough) and class time must be spent producing. Because the audit shows the failure is source quality and source loss, the second session targets corroboration and stable links - the two things a Works Cited page by Friday cannot detect. Save the reading aloud/extension modifications by building the listen feature and permalinks into the task itself.
2. The team lead says database instruction "takes time away from writing." How do you respond, and what compromise do you offer?
Read the worked answer
Take the concern seriously rather than defending the database. The lead is right that a 45-minute tour of search boxes does not improve writing. Agree on the shared goal - fewer dead links and fewer fabricated sources in final drafts - and then argue from the local audit rather than from principle: 31% of sources could not be relocated at check-in, which means the teacher was grading claims that could not be verified, and 78% of cited URLs were content farms or AI summaries, which means the evidence base under the argument was weak. Those are grading burdens, not librarian complaints.
Then concede the real point: not all open-web searching is bad, and lateral reading is a skill students need on the open web, not just inside a paywalled database. The compromise is a division of labor. The library handles mechanics asynchronously - a short recorded walkthrough, a guide with direct links to the two subscribed databases, and a permalink cheat sheet - so that no class minute is spent on buttons. The two visits are then spent on producing the annotated source set, and the teacher's class time is spent on claim and evidence. Mr. Reyes co-scores the source-quality portion of the rubric so feedback is not duplicated.
Make the offer measurable and time-bound: if the audit of the fall papers shows source quality no better than the spring baseline, revise the design. That gives the lead a genuine voice and makes the collaboration an evidence question rather than a turf question.
3. How would you assess source evaluation so that the score is defensible and resistant to AI shortcutting?
Read the worked answer
Assess the process artifacts and require a live explanation, because a polished annotation can be generated in seconds. Three components work together.
First, a submitted source set with a permalink, relevance claim, authority justification, and one corroborating source per claim. Permalinks make verification possible and, per the audit, that was the missing link. Second, one short in-class check using a common pair of sources (one credible, one content farm or AI-summarized article) where students write two or three sentences on which source they would cite and why. Same stimulus for everyone, completed in class, no devices: it measures transfer rather than production of text. Third, a 60-second oral defense with two or three randomly selected students per section: "Show me where this claim comes from and how you checked it." Sampling makes it feasible in a 45-minute window.
A single rubric should carry the scoring across all three: relevance to the claim, authority of the source and how it was established, corroboration, and stability of the link. Keep the rubric to four criteria and three levels so it is usable by both the librarian and the teacher.
On AI-generated summaries specifically: teach students to trace a claim to the primary source, check the "about" page and ownership, inspect the publication date, and look for the absence of named authors and citations. Then require that every citation resolve to a page a reader can open - if it does not resolve, it is not a source.