01
What triggers feedback
Traditional
A number gap. The student sees “you over-rated this” and a canned explanation.
With Calibrate
The student's own written reasoning. The coach reads the justification and names the specific move that went wrong.
How it works & impact
Feedback triggered by reasoning, not by a score gap.
Calibrate treats a rubric score as the tip of an argument. The student says where the work lands and why; the coach reads the why and pushes back on the specific reasoning move that went off — not the number.
The loop, at a glance
Six beats. Steps 1–5 are what one student does with the coach. Step 6 is what the class produces together — and it feeds the next cohort's exemplars.
01Student
Watch
An exemplar presentation from an earlier cohort.
02Student
Rate & justify
1–5 on four rubric criteria, with a short written justification each.
03Coach
Read the “why”
Each justification compared against the instructor's benchmark rationale; the misconception is named.
06Instructor
Class-wide alerts
Justifications aggregate into ranked misconceptions with student quotes and a suggested mini-lesson.
05Student
Re-rate & reveal
Original, revised, and benchmark scores side by side, with the instructor's own comments.
04Coach ↔ Student
A few questions, your replies
A bounded Socratic exchange — up to three questions per criterion, then a chance to re-rate.
Alerts inform teaching — and next term's exemplars come from this data.
01Student
Watch
An exemplar presentation from an earlier cohort.
02Student
Rate & justify
1–5 on four rubric criteria, with a short written justification each.
03Coach
Read the “why”
Each justification compared against the instructor's benchmark rationale; the misconception is named.
04Coach ↔ Student
A few questions, your replies
A bounded Socratic exchange — up to three questions per criterion, then a chance to re-rate.
05Student
Re-rate & reveal
Original, revised, and benchmark scores side by side, with the instructor's own comments.
06Instructor
Class-wide alerts
Justifications aggregate into ranked misconceptions with student quotes and a suggested mini-lesson.
↺ Alerts inform teaching — and next term's exemplars come from this data.
Before & after
What changes when the coach reads the reasoning, not the score.
02
Shape of the exchange
Traditional
One-way. The student reads; the conversation ends.
With Calibrate
Socratic. Up to three questions per criterion with the student's replies, then an invitation to re-rate — bounded so it stays honest.
03
Instructor workload
Traditional
High. Someone has to pre-write feedback paths for every plausible score.
With Calibrate
Low. The instructor authors the rubric and one benchmark; the coach handles the variations.
04
What the analytics show
Traditional
Averages and variance. You know the class is off, not why.
With Calibrate
Ranked misconceptions drawn from justification text, each with anonymised student quotes.
Who gets what
Two sides of the same conversation.
For the student
From guessing to defending.
- 01
Evaluative judgement
Students practise deciding what quality looks like — on someone else's work first, then on their own.
- 02
Feedback literacy
Rubric language stops being decorative. Students have to use it to defend a score.
- 03
Meta-cognitive awareness
Writing a justification blocks the guess-and-move-on reflex. The reasoning becomes visible — first to the student, then to the coach.
For the instructor
From averages to alerts.
- 01
Actionable interventions
Instead of a general read of the class, you get named misconceptions — “delivery confidence read as evidence quality” — with the student quotes behind them. You know exactly what to teach next.
- 02
A diagnostic mirror for the rubric
If a criterion mis-calibrates across the whole class, that's a signal about the rubric, not just the students. The dashboard makes ambiguity in your own language visible.
- 03
Dialogic feedback at scale
Carless and Boud describe feedback as a dialogue. Doing that with 100+ students is impossible by hand. The coach carries the first pass so instructors can step in where nuance actually needs them.
The five-minute test
Rate one presentation. Let the coach hear you.
The fastest way to understand the loop is to run it. Pick an exemplar, rate it honestly, and see what the coach names in your reasoning.