Redesign canvas — the 26-minute sprint
Used in the workshop’s redesign sprint (minutes 42–70: the five steps below are time-boxed to 26 minutes, leaving ~2 minutes of buffer in the 28-minute slot), and afterwards on your own. Pairs work on one real assignment. The blank version for printing is templates/redesign-canvas-blank — this page adds the prompts and examples for facilitated use.
Box 1 — What learning must this assessment show? (4 min)
One sentence: a passing grade certifies that the student can ___.
- Not “writes good essays” — what does the essay certify? (argue from evidence? find and verify sources?)
- Example (Case A): “…can select and verify scholarly evidence and defend an argumentative choice under questioning.”
Box 2 — Where is it vulnerable? (4 min)
For each rubric criterion, ask: could a competent AI produce this artefact at this standard without the learning? Mark each criterion:
| Mark | Meaning |
|---|---|
| Pass | AI could satisfy it easily — the criterion no longer discriminates |
| Fragile | AI could approximate it; only careful marking tells the difference |
| Holds | The criterion needs the student’s own process, context or live defence |
Example (Case A, original): thesis/argument Pass; use of evidence Fragile→Pass (formatting rewards, fabrication risk); analysis Fragile; writing Pass.
Box 3 — AI-use level for each stage (6 min)
List the stages of your task and assign an AIAS 2.1 level to each — icon + text:
| Stage | Level (1–5) | Level name | Why |
|---|---|---|---|
| e.g. topic selection | 2 | AI Planning | ideas cheap; commitment is the student’s |
| e.g. sources | 1 | No AI | verification is a core capability |
| e.g. draft | 3 | AI Collaboration | feedback allowed, logged and critiqued |
| e.g. defence | 1 | No AI | secure lane |
Rule of thumb from the cases: open the practice, secure the proof.
Box 4 — Design moves (8 min)
Pick 1–3 moves from the eight approaches:
- Staged submissions → checkpoints you can verify cheaply
- Process portfolio (log/notebook) → the journey becomes evidence
- Personal/local data → generic AI output can’t match it
- Critique-the-AI-output → evaluation as the assessable skill
- Open AI + reflection → AI-collaboration as a real outcome
- Oral/written defence (5 min, sampled, with written alternative) → assurance
- Weighted secure checks → the certificate lives here
- In-class writing → strong link, watch the equity notes
Finish step 4 early? Run the decision tool as a cross-check (3–5 minutes — an early-finisher option, not part of the timed step).
Box 5 — Sketch the rubric change (4 min)
- Which criterion dies (it certified fluency)?
- Which criterion replaces it (verification? defence? AI-use judgment)?
- Does the total grading time go up or down? Estimate minutes × students. If up: cut something now.
Box 6 — What this does NOT solve
Name it honestly — every case in this kit does. What can still be gamed? Who is disadvantaged if you don’t add the alternative format? What will you check after the first run?
Box 7 — Commitment
One next step, within two weeks, with a date:
By [date] I will ___.
Facilitators: photograph/collect canvases only with the pair’s explicit permission. Anonymised canvases are welcome evidence for the kit’s evaluation pages.