Learn Better With AI

Methods · Assessment & diagnosis · Short guide

H7

Feedback-on-feedback

Ask what the feedback itself missed, so you do not treat every comment as complete or authoritative.

When to use this
At the end of a session
Learning impact
Clear help, modest gain
Learning evidence
Thin or mixed evidence
Where you are with the topic
Checking readiness
How often to use it
Useful once for a topic
What AI is doingEvaluate your work

When to use it

  • You just got feedback (human or AI) and feel pressure to obey every line.
  • A comment is vague (“deepen analysis”) and you need to know what to verify.
  • You suspect the feedback focused on polish and missed a rubric dimension.

What to ask AI

Paste your work and the feedback. Ask for a critique of the feedback, not a rewrite of your work.

Ask which parts are vague, possibly mistaken, or missing versus the rubric.

Ask what to verify before you revise. Then you decide what to change.

How it helps

Feedback can be partial, wrong, or oddly focused. Learners either obey all of it or ignore all of it.

Feedback-on-feedback is a second-order check: given this feedback, what is underspecified, what might be mistaken, what should I verify?

Essential with AI feedback, which can sound sure while being wrong. Useful with human markers too.

How to run it

  1. Receive feedback on a piece of work.
  2. Before revising, critique the feedback: gaps, possible errors, vague directives.
  3. Verify contested points against sources, rubric, or calculation.
  4. Revise only what survives scrutiny plus your own judgment.
  5. Ask clarifying questions where the feedback is foggy.

Example exchange

You

I got this feedback on my paragraph. Before I rewrite anything, can you check whether the feedback actually matches the rubric?

AI

Two issues: “add more theory” never names which criterion, and it praises a claim your rubric does not weight. Which criterion do you think “more theory” is aiming at?

You

Criterion 3, but that one wants application, not more named theorists. So I'll skip the theory binge.

AI

Right. Revise the application gap it correctly flagged; park the rest until a marker clarifies.

Copyable prompt

My work: [ ].
Feedback I received: [ ].
Do not rewrite my work. Critique the feedback: vague parts, possible mistakes,
missing dimensions vs rubric [ ]. Tell me what to verify before I obey. Then I
will decide revisions.

The Tell

You implement every AI suggestion, including the wrong ones.

Obedience without a second-order check is not the method.

Using this only to dismiss all criticism also fails, verify, then decide.

Principle evidence

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

The underlying learning idea is rated weak. Questioning the feedback itself is sensible metacognition, but there is thin named literature treating it as a standalone method.

AI delivery evidence

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

AI self-critique of its own feedback is especially speculative. Prefer verifying contested points against rubric and sources rather than trusting a second AI opinion alone.

Related methods