How Feedback Works in Learning | From Information to Better Independent Performance

Direct Answer: Feedback works when information about a learner’s performance helps them see a useful discrepancy, decide what to change and produce a better subsequent attempt. Feedback is not successful because it was delivered. It is successful when the learner can use it—and eventually needs less of it.

The simplest definition of feedback

Feedback is information that helps a learner compare current performance with a useful target and act on the difference.

In one line: Feedback should improve the next independent move, not merely explain the last mistake.

The core mechanism

GOAL → ATTEMPT → OBSERVE → COMPARE → INFORMATION → LEARNER INTERPRETS → REPAIR → REATTEMPT → FADE FEEDBACK → INDEPENDENT RETURN

The chain is longer than “mark the work and tell the student what is wrong.” A comment can be accurate and still fail educationally if the learner cannot interpret it, does not know what action follows, copies the correction without rebuilding the operation or becomes dependent on the adult to identify every problem.

The Education Endowment Foundation defines feedback as information about performance relative to learning goals or outcomes. Its evidence synthesis also stresses that feedback varies greatly in effect: the useful question is not simply whether feedback exists, but what it contains, when it arrives, whether the learner can act on it and whether it supports future learning.

The seven jobs of effective feedback

1. Feedback needs a clear target

“Improve this” is difficult to use if the learner does not know what good performance looks like. Feedback becomes more actionable when the goal or success criterion is visible enough to compare against.

This is why the Student/Studying Interface begins with a Goal & Criteria Interface. A learner cannot regulate toward a target they cannot see.

2. Feedback should locate the discrepancy

A score tells us how much credit was awarded. It does not automatically tell us where the performance departed from the target. Useful feedback narrows the difference: Was the concept missing? Was the evidence weak? Was the representation wrong? Was the method inappropriate? Was the answer correct but insufficiently explained?

The narrower the useful discrepancy, the easier it becomes to choose a repair.

3. Feedback should preserve the learner’s work

There is a major difference between helping a learner see what to do and doing the cognitive work for them. A hint, question or partial cue may preserve more learner production than a complete worked answer.

Our MindOS Feedback State therefore asks whether feedback leads to a learner-owned repair rather than an externally supplied correction.

4. Feedback has to survive the handoff

A teacher can intend one message while the student receives another. “Show more working” might be heard as “write more lines” instead of “make the reasoning visible.” “Be more precise” may mean nothing if the student cannot identify what precision looks like in this task.

The Feedback Handoff separates what was written or said from what the learner actually understood and converted into a next move.

5. Feedback needs a reattempt

Without another attempt, feedback remains information. The reattempt is where we discover whether the learner can use it.

This is one reason a corrected answer is not necessarily a repaired mistake. The page may now be right while the learner’s underlying operation remains unchanged. Correction State asks whether the learner can locate, explain and repair the error without copying the fix.

6. Feedback should become less necessary

External correction is valuable partly because learners cannot initially see everything about their own performance. But if every decision remains externally supplied, the learner may improve only while the feedback system is present.

Bolt 17 — When Feedback Makes You Dependent draws this boundary clearly. The long-term goal is better self-monitoring, not permanent dependence on a teacher, tutor, answer key or AI.

7. Feedback should improve the learner model

Good feedback teaches more than the answer. It can teach the learner what kinds of errors they make, which cues matter, when a strategy fails and when to seek help. Over time, external feedback can become part of an internal coach.

This is the bridge into Bolt 24 — Building an Internal Coach and the wider move from external marks toward self-regulated learning.

What feedback is not

Feedback is not merely a mark. A mark can signal an outcome without explaining the discrepancy.
Feedback is not the same as praise. Encouragement may matter, but “good job” does not necessarily tell the learner what was effective or transferable.
Feedback is not the same as correction. A correct answer supplied by somebody else may leave the learner unchanged.
Feedback is not automatically helpful because it is detailed. Too much information can exceed what a learner can use at that moment.
Feedback is not successful merely because the teacher worked hard on it. The educational receipt is what the learner subsequently does.

When feedback goes wrong

Visible situation	Possible failure underneath	Useful check
The student reads comments but repeats the same error	The feedback did not become a usable repair action	Ask the learner to explain what they will change before trying again
The student waits for marking before checking anything	Error detection remains external	Require a self-check before teacher feedback
The student gets the next question right only after a hint	The hint may be carrying method selection	Give a fresh problem with the hint removed
Feedback causes confusion	The comment may assume knowledge or vocabulary the learner does not possess	Ask the learner to paraphrase the feedback in their own words
The student becomes afraid of mistakes	Feedback may be experienced as judgement of the person rather than information about the performance	Return feedback to task, strategy, evidence and next action

Timing: when should feedback arrive?

There is no single timing rule for every task. Immediate feedback can prevent an error from being rehearsed when a learner is building a new route. Delayed feedback can sometimes preserve more independent thinking or provide a stronger test of what the learner can monitor alone. The EEF evidence summary notes that feedback can be effective during, immediately after and some time after learning.

The timing question should therefore be connected to the job. Ask: What information does the learner need now, and what useful thinking would disappear if we give it too soon?

The smallest useful feedback test

After giving feedback, do not judge its success from the student nodding. Give a new attempt that requires the same underlying operation but does not reproduce the original answer. If the learner can use the feedback there, we have stronger evidence that something changed.

How feedback works across subjects

English

“Improve your vocabulary” is weak feedback if the student cannot identify where precision is missing. Better feedback may locate one vague word, compare alternatives and then require the learner to revise another sentence independently. In comprehension, feedback should connect an answer to evidence and inference rather than merely provide the accepted wording.

Mathematics

Mathematics feedback should distinguish a conceptual error from an execution error. A wrong final answer can come from a correct representation followed by arithmetic failure, or from perfect arithmetic applied to the wrong model. The repair is different.

Science

Science feedback should help students connect claim, mechanism and evidence. A technically correct fact may still fail to answer the question if it does not explain the observed relationship. The learner needs to know which scientific operation was insufficient.

How feedback changes as learners develop

Early stage: feedback may be more explicit because the learner has not yet built reliable criteria or error-detection routines.

Developing stage: prompts can increasingly ask the learner to locate the problem before the adult identifies it.

Independent stage: the learner predicts, checks, explains, revises and decides when external expertise is genuinely needed.

The teacher has not become useless. The teacher’s job has changed from constant external controller toward calibrator, expert observer and source of higher-resolution feedback when the learner’s own model is insufficient.

How do we know feedback worked?

The learner can state what the discrepancy was.
The learner can name or demonstrate the intended repair.
The next attempt improves for the relevant reason.
The improvement survives a fresh problem rather than only the corrected original.
The learner becomes better at detecting similar issues without waiting for the adult.
Feedback can gradually become lighter without performance collapsing.

For parents: how should I read teacher comments?

Comments are most useful when they become action. Instead of asking only, “What did the teacher write?”, ask, “What does that mean you will do differently on the next piece of work?” If the child cannot answer, the handoff may not yet be complete.

A long page of marking is not automatically stronger than one precise comment that changes a recurring error.

For students: how should I use feedback?

Read the goal or criteria again before reading the comment.
Locate exactly where your performance differs.
Translate the feedback into one action.
Try the repair without copying the model.
Test the same idea on a new task.
Record recurring error patterns you want to detect earlier next time.
Ask for clarification when the feedback itself is unclear.

The complete feedback chain

MAKE THE TARGET VISIBLE → ATTEMPT → OBSERVE → LOCATE THE GAP → GIVE THE SMALLEST USEFUL INFORMATION → CHECK THE HANDOFF → LEARNER REPAIRS → FRESH REATTEMPT → REDUCE EXTERNAL PROMPTING → TEST SELF-DETECTION → RETURN LATER

Frequently asked questions

Is more feedback always better?

No. Feedback has to be usable. Too much information can obscure the highest-value discrepancy, while excessive prompting can reduce the learner’s need to monitor and decide independently.

Should teachers correct every mistake?

Not necessarily. Some errors require immediate correction; others can become valuable opportunities for the learner to detect, classify and repair the issue. The choice depends on the task, learner state, risk of rehearsing the error and what independence we are trying to build.

Can feedback make learning worse?

Yes. Research syntheses find a wide distribution of effects. Feedback can be unhelpful when it is confusing, person-focused, mistimed, excessive or when it replaces the learner’s own cognitive work. That is why feedback design matters.

What is the best kind of feedback?

There is no single best format for every situation. Useful feedback is specific enough to guide improvement, connected to a meaningful goal, understandable to the learner and followed by an opportunity to act. Written, verbal, peer and digital feedback can all be useful when those conditions are met.

Research bridges

Useful external evidence summaries include the Education Endowment Foundation Teaching and Learning Toolkit: Feedback and its Teacher Feedback to Improve Pupil Learning guidance report.