Grade-then-justify

Assign your own grade first, then defend it against an independent grading, so calibration happens in argument, not in passive receipt of a number.

When to use this: When checking yourself
Learning impact: Reliable gain
Learning evidence: Credible with caveats
Where you are with the topic: Checking readiness
How often to use it: Worth repeating often

When to use it

You chronically under- or over-shoot marks and only find out when results land.
You self-mark dimensions but never commit to an overall grade and defence.
You want disagreement with an independent read to force evidence quotes.

What to ask AI

State your grade and a short rubric-tied justification first. Forbid the AI from grading until you do.

Then ask for an independent grade that does not match you on purpose.

Debate mismatches with quotes from the work. End with your revised grade, do not force agreement.

How it helps

Students under- and over-shoot marks chronically. Stating a grade and defending it forces evidence.

Meeting an independent grade creates productive conflict: either you update, or you show the other grading missed rubric evidence.

Blind self-marking scores dimensions; this adds an overall grade plus defence.

How to run it

Finish the work; self-assign a grade plus a five-line defence tied to the rubric.
Have AI grade independently with instructions not to peek or match.
Compare; argue mismatches with quotes from the work.
Settle on a revised self-grade you can still defend.
Log calibration error over weeks.

Example exchange

You

Here's the rubric and my draft. I'm giving myself 62, claim is clear, but evidence is thin on the counter-argument. Grade it yourself without matching me.

Independent read: 55. I score counter-argument lower than you did, and structure higher.

You

I'll take the counter-argument hit. On structure, though, here's a quote that should put me mid-band…

Mutual read lands around 58. What's your revised self-grade? Don't force agreement, just log the gap.

Copyable prompt

Rubric: [ ]. My work: [ ].
I will state my grade and justification FIRST. Only then grade independently:
do not match me. Then debate mismatches with evidence quotes. End with my
revised grade; don't force agreement.

The Tell

You always “agree to disagree” without changing either side’s evidence.

An AI that politely matches your grade also kills the method, demand an independent read.

Grades without rubric defence are just numbers with feelings attached.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Assigning a grade then defending it is calibration plus self-assessment: commit before seeing an external judgment. The learning is in the disagreement negotiation.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

An AI “independent” grade is not authoritative. Use the debate to improve your evidence habits; verify against real marks when you can.