Bolt 16 — When Feedback Helps

Bolt Series · School–Teacher–Student Performance Calibration · Article 16

Wait, What? A Corrected Answer Is Not Yet Evidence That Feedback Worked

A teacher points out an error. The student fixes it. The page now looks better.

That proves the feedback changed this response. It does not yet prove that the feedback improved the learner’s later performance, helped the teacher identify the right mechanism, or produced a change that survives when the correction is no longer present.

Bolt therefore treats feedback as a performance-calibration problem: what did the feedback actually change, under what conditions, and what later receipt would justify believing that the change matters?

Quick Answer

Owned Bolt calibration job: distinguish feedback that merely changes an immediate output from feedback that produces a measurable improvement in later performance, while keeping teacher judgement and learner interpretation open to correction.

Feedback is not one intervention. Its effect depends on what information is given, when it is given, what the learner already understands, how the learner uses it, and what later performance is measured. The right question is not “Was feedback given?” but “What changed after it, and how do we know?”

Four Different Things Can Look Like “Feedback Worked”

• Output correction: the current answer becomes correct.
• Task improvement: the learner performs better on another closely matched item.
• Learning improvement: the change survives after time has passed.
• Transfer: the improvement survives when the task surface or selection demand changes.

These are not equivalent receipts. A school or teacher that measures only the corrected page can overstate the effect of feedback.

Observable Signatures

• The student corrects the marked answer but repeats the same error on the next independent question.
• The student improves immediately but loses the improvement after a delay.
• The student can use the feedback when the teacher names the weak step but cannot locate that step alone.
• The teacher changes feedback wording but performance does not move.
• The learner’s prediction becomes more accurate even before total scores rise.
• The same feedback helps one learner and confuses another because their underlying errors differ.

Competing Explanations for Improvement After Feedback

• The feedback correctly identified and changed the weak mechanism.
• The learner copied the correction without internalising it.
• The later task was easier or unusually similar.
• Extra teacher attention, time or motivation changed performance.
• The original poor result was anomalous and performance was returning toward baseline.
• The learner already knew the material and the feedback merely cued access.

One improved answer cannot discriminate among these explanations. The next performance must be designed to do that job.

School–Teacher–Student Calibration

School

A school should judge feedback policy by what it improves, not by how much marking is produced. Policies that mandate frequency or volume can create visible activity without proving better learning. The school should preserve teacher judgement about timing and format while requiring evidence that feedback is usable and connected to subsequent performance.

Teacher or Coach

The teacher should first locate the performance gap accurately. Feedback about the wrong mechanism can be specific, beautifully worded and still unhelpful. After giving feedback, the teacher needs a return task that can show whether the targeted performance changed.

Student

The student should know what the feedback is claiming: what was wrong, what standard matters, and what later performance will show whether the issue has changed. Bolt does not require the student to run the learning operation themselves; it requires the resulting performance to remain interpretable.

Worked Example: The Sign Error

A student repeatedly changes a sign incorrectly when expanding algebraic expressions. The teacher writes, “Be careful with signs.” The learner corrects the marked line.

The feedback is too broad to tell us what changed. A stronger calibration sequence is:

1. Name the exact transition where the sign changes.
2. Give one fresh matched problem.
3. Remove the explicit cue.
4. Return later with a changed surface form.
5. Compare whether the same error pattern remains.

If the error disappears only while the cue is present, the feedback improved supported performance. If it remains absent later and under changed conditions, the evidence for learning is stronger.

How Do We Know?

The Education Endowment Foundation’s Teacher Feedback to Improve Pupil Learning guidance is based on an international evidence review and stresses that feedback should rest on high-quality instruction and formative assessment, be appropriately timed, focus on moving learning forward, and be planned so pupils can use it. The guidance also warns that not all feedback has positive effects and that feedback carries real teacher workload.

EEF’s Feedback evidence summary synthesises 155 studies and rates the evidence base as high, while emphasising variation by context and the need for professional judgement rather than a universal delivery rule.

A 2024 meta-analysis of digitally delivered instructional feedback, Brummer and colleagues, further shows why context, feedback content and task factors matter when interpreting effects. This supports Bolt’s insistence that “feedback” is not a single uniform treatment.

Evidence Boundary

Feedback effects are heterogeneous. No single timing rule, medium, amount or wording is optimal for every learner and task. An average effect does not prove that a particular comment caused a particular learner’s improvement.

Bolt therefore does not prescribe a universal feedback technique. It calibrates the claim from the observed return performance. If a specific learner operation is needed to convert feedback into learning, that operation belongs to MindOS.

Common Misconceptions

• “More feedback is better.” More volume can add workload without adding useful information.
• “Immediate correction proves learning.” It proves immediate correction.
• “Written feedback is better than verbal feedback.” The evidence does not support a universal medium rule.
• “If the student ignores feedback, the student is the problem.” The feedback may be badly timed, unclear, unactionable or aimed at the wrong gap.
• “A better next score proves the feedback caused it.” Other explanations may remain plausible.

What Should Change Next?

After feedback, define the smallest fair return performance that tests the claimed change. Match conditions to the inference: remove the cue if independence matters, delay the task if retention matters, and alter the surface if transfer matters.

RFE: Did the feedback change later performance in the way school and teacher predicted, and is the evidence strong enough to recalibrate what we believe about the student?

Bolt Direction Graph

Observed performance gap → teacher interpretation → feedback target → feedback delivered under known conditions → fresh matched performance → delayed/changed-condition return if relevant → compare predicted with observed change → recalibrate feedback, teaching and learner-performance model.