“This Grade 6 English assignment asks students to argue for a library change using one supplied source. Keep claim, evidence and reasoning. Replace vague labels with observable descriptors. Use Developing, Meeting and Extending. Do not add requirements we have not taught. Flag any assumptions.”
Advanced work within a learning partnership
A rubric becomes a tool.
A team learns how.
This is where sustained learning can lead: staff move from directing AI to building, testing and maintaining something useful together.
A fictional worked example, not a completed district project or a workshop outcome. The pathway is shaped around readiness and district needs, not a fixed timeline.
The engagement behind the artifact
- LMS learning
- Context, privacy, output evaluation and iterative prompting.
- Applied work
- A teacher improves a rubric and documents the decisions behind each revision.
- Consultation
- I help the team scope the tool, challenge weak assumptions and work through testing.
- The aim
- A small, teacher-reviewed tool the team understands, not software it blindly depends on.
Watch the 20-second overview
A staged AI exchange showing the idea in miniature. The fuller process below includes the review, revision and testing omitted from the short clip.
Prepared examples. Not a live AI generator. No student records, uploads or saved responses. This is not a one-prompt production build.
The video could not load. The written walkthrough below covers the process.
Read the video transcript
A reviewed rubric for claim, evidence and reasoning is attached to a staged AI conversation. The request asks for grade, subject and assignment fields with editable criteria. A form appears, followed by a Grade 6 English rubric draft. A follow-up changes the example to Grade 8 Science and the evidence descriptor changes from a source example to measured plant-growth results. The teacher still reviews the result.
Start with real work
A document that leaves too much unsaid.
The team uses claim, evidence and reasoning. Its descriptors are too vague to guide consistent feedback.
Argument rubric.docx
| Criterion | 3 | 2 | 1 |
|---|---|---|---|
| Claim | Strong | Mostly clear | Unclear |
| Evidence | Good | Some | Little |
| Reasoning | Thorough | Adequate | Limited |
The learning task
Explain what each vague label fails to tell a learner. Gather the assignment and what was actually taught before asking AI to revise anything. No student work or personal information is needed.
Guided LMS practice
Improve it with AI.
Do not accept it untouched.
The important evidence is the teacher’s judgment, not the polished table.
“Synthesizes several independent sources and evaluates their credibility.”
That exceeds the task. Students have one supplied source. More sources would change the assessment, not improve the rubric.
“Keep the single-source constraint. Make Extending about weighing the evidence and acknowledging a limitation, not collecting additional sources.”
“Selects relevant evidence from the supplied source and explains a limitation in how it supports the claim.”
Ongoing consultation
Turn the reviewed structure into a tool.
I help the team decide what stays consistent, what must change and what a first version should leave out.
“Build a browser-based rubric editor. Start with our reviewed criteria. Let teachers choose a teaching context, inspect all three performance levels, edit descriptors and copy the rubric. Label every output as a draft. No student records, accounts, automatic scoring or automatic publishing.”
Changing “English” to “Science” changes the heading, but the evidence row still asks for a quotation from a text.
A relabeled rubric is not an adapted rubric. What counts as evidence in this investigation? What have these learners practiced?
“For this plant-growth investigation, use measured results, units and the link to the scientific claim. Keep descriptors editable. Test both contexts and verify that editing one does not overwrite the other.”
What stays consistent, and what changes?
The layout, three performance levels and review workflow can be reused. The criteria, learning expectations and acceptable evidence need subject-specific review. A grade selector alone does not establish appropriate rigor or standards alignment.
Try a small working prototype
A form, a full draft and room for judgment.
Two prepared contexts, not live AI generation. Switch examples, edit the descriptors and copy a draft. Changes remain in this open preview only. Do not enter private information.
Grade 6 English draft
Prepared draft. Review against your learning goals before use.
Evidence
- Developing
- Includes an example but does not show how it supports the claim.
- Meeting
- Uses relevant evidence from the supplied source and explains its connection to the claim.
- Extending
- Selects relevant evidence and explains a limitation in how it supports the claim.
Review before sharing
The test is whether the team can explain and improve it.
- Curricular fit: check every descriptor against the task and what was taught.
- Transfer: test another subject. Look for expectations carried over incorrectly.
- Usability: have a colleague use it without coaching, including on a phone.
- Reliability: test context switching, edits and copied output. Keep notes on what failed.
- Ownership: agree who maintains the criteria and reviews future changes before team use.
What the LMS can make visible
The original rubric, the prompt, the rejected AI suggestion, the teacher’s explanation, a prototype and its test notes. Facilitator feedback identifies the next revision.
That is evidence of applied learning, not just a completed module. It does not by itself prove classroom impact or production readiness.
A pathway shaped around your team.
Not everyone starts here. We build the foundations first, then choose applied projects that fit your staff and district priorities.
Talk about a learning partnership