Bolt Measurement Note 01 — A Supported Answer Is Not the Same Measurement as an Independent Answer

Bolt Measurement Notes · Supplementary to the Bolt 01–40 Core · Note 01

Wait, What? A student can get the answer right and still leave us unsure what they can do alone

A teacher asks a difficult question.

The student hesitates.

The teacher points to the relevant line, asks one carefully chosen question, and the student completes the solution correctly.

Was that a successful teaching moment?

Very possibly.

Was it proof that the student can now perform the same operation independently?

Not yet.

The supported answer and the independent answer are both useful evidence. They are simply evidence about different performance conditions.

Quick Answer

The amount and kind of help present during a performance changes what that performance can justify us believing. A prompt, worked step, hint, calculator, AI explanation, peer suggestion, teacher question or visible example may make successful performance possible. That can be excellent teaching. But if our question is whether the learner can perform independently, the support conditions must be recorded and then reduced or removed in a later check.

Bolt’s job here is not to decide whether scaffolding is good or bad. It is to calibrate the inference:

Who carried which part of this performance, under what conditions, and what later evidence would show that responsibility transferred to the learner?

Owned Calibration Job

This article owns one narrow Bolt problem:

How should schools, teachers, parents and students interpret performance when assistance was present?

It does not own the design of the learning scaffold itself. That belongs downstream to the learner-operation layer. It does not own the learner-facing task handoff. That belongs to the Student/Studying Interface. It does not claim that all assisted work is invalid. Quite the opposite: assisted performance can reveal what kind of support unlocks progress and what the learner may be ready to do next.

The calibration mistake is treating assisted and independent performances as interchangeable measurements.

The same correct answer can represent different performances

Imagine four students all produce the same correct algebraic solution.

Student A solves it independently from a blank page.
Student B needs the teacher to say, “What should you isolate first?”
Student C needs the first two steps demonstrated and then finishes.
Student D copies an AI-generated solution and can explain only parts of it afterwards.

All four pages may end with the same answer.

But the performance evidence is not the same.

The visible product tells us what appeared on the page. The support history tells us how that product came into existence.

This is why a polished output should not be confused with independent capability.

Why support is not a problem — but invisible support is

Education depends on support.

Teachers model, prompt, question, show examples, correct misconceptions and structure practice.

The Education Endowment Foundation’s recent work on metacognition and scaffolding describes scaffolds as temporary supports that should help pupils move toward increasing independence. It recommends providing the least amount of support first where appropriate, then prompting, clueing, modelling or correcting when more assistance is needed.

A support can therefore do two things at once: help the learner succeed now, and change what the current success tells us about independent performance.

The teaching question is, “Did the support help learning?” The Bolt question is, “What can this supported performance legitimately tell us?”

A practical observation structure: the support ladder

For ordinary classroom use, teachers do not need a complicated score. They can record the level of assistance required.

Level 0 — Independent: learner initiates and completes the target operation without task-specific help.
Level 1 — General prompt: “Check your first step.” “What is the question asking?”
Level 2 — Targeted cue: attention is directed to a specific concept, representation or location.
Level 3 — Partial model: one part of the method is demonstrated or supplied.
Level 4 — Heavy support: most of the method is carried by the teacher, peer, worked example or technology.

This is not a universal psychometric scale and should not be treated as one. It is a practical observation structure. The important information is not the number; it is the changing relationship between support and performance.

If the learner succeeds with Level 3 support today, Level 1 next week, and Level 0 later under changed conditions, the pattern tells us something meaningful about transferred responsibility.

Dynamic assessment shows why assistance itself can be evidence

Educational research has long examined approaches in which the assessor does not merely observe a static score but also studies how performance changes when teaching, feedback or graduated prompts are introduced.

A systematic review of dynamic assessment in reading described common approaches including test–teach–retest designs and graduated prompting. In the latter, the amount of assistance required can itself become part of the evidence about learning and responsiveness.

That research does not mean every classroom hint should be converted into a formal diagnostic instrument. Dynamic assessment has specific methods, populations and evidence boundaries. But it supports a durable principle:

Performance under assistance and performance without assistance can answer different questions.

A K–12 review of teacher-implemented scaffolding similarly emphasised diagnosis, responsiveness and fading while noting that these contingent processes are often underreported. That limitation matters: support is not one uniform treatment, and evidence about transfer depends on how it was given and removed.

Competing explanations: why did the student succeed after help?

The student knew the concept but failed to retrieve the first step.
The student understood the method but misread the task.
The student needed a representation change.
The prompt reminded the student of a strategy they could then execute independently.
The teacher’s cue performed the crucial discrimination the student could not yet perform.
The worked step reduced task difficulty enough for the remaining operations to become manageable.
The learner followed the support without understanding the underlying relation.

These explanations demand different next moves. The supported performance narrows the investigation. It does not finish it.

The discrimination test: reduce help and change the conditions

After a supported success, ask a second question that requires the same underlying operation but changes the surface features. Then reduce the help.

Can the learner start without the original cue?
Can they explain why that first step is appropriate?
Can they recognise when the method does not apply?
Can they perform after a delay rather than immediately?
Can they succeed when the example is removed?
Can they perform when the teacher is silent?

If performance collapses as soon as the support disappears, the original success was still useful—but independent capability has not yet been demonstrated. If performance survives, responsibility has moved toward the learner.

Technology makes this distinction more important

AI, search, calculators, adaptive platforms and worked-solution systems can all increase the quality or speed of an output. That does not automatically mean the learner’s unaided capability increased by the same amount.

Which cognitive operation did the learner still have to perform?

If AI identifies the method, generates the reasoning and checks the answer, the learner may have performed only selection or transcription. If the learner first attempts, explains, compares, revises and later reproduces the reasoning without the tool, the evidence is different.

The MindOS AI Assistance Gradient asks whether technology is scaffolding learning or substituting for it. Bolt does not own that learning mechanism; Bolt records what the assisted performance can and cannot establish.

For schools: record conditions, not just outcomes

whether the task was independent, guided or collaborative;
whether notes, examples, calculators, AI or other tools were allowed;
whether prompts were generic or task-specific;
whether the learner had seen an almost identical worked example;
whether the result came from practice, formative assessment or an independent check;
whether the same capability later survived reduced support.

This does not require turning every lesson into an audit trail. It requires enough context to stop unlike performances from being interpreted as though they were equivalent.

For teachers and coaches: support, then deliberately test ownership

Predict: ask the learner what they think they can do unaided.
Observe: let the learner attempt before intervening where safe and appropriate.
Support minimally: add the smallest prompt likely to unlock the next useful action.
Record the critical help: what did the support actually supply?
Fade: remove or weaken that support on a new example.
Change conditions: vary the surface, delay, representation or context.
Return: check what survives independently.

The goal is not to make the teacher disappear as quickly as possible. The goal is to know when the teacher’s contribution has become the learner’s capability.

For students: ask a harder question than “Did I finish?”

What part could I do before the help arrived?
What exactly did the hint give me?
Could I produce that step myself next time?
Could I solve a different version without looking back?
If the support disappeared tomorrow, what would still remain?

Common Misconceptions

“If help was used, the work does not count.”
Wrong. Supported performance can be highly informative and educationally valuable. It simply answers a different question from independent performance.

“A correct answer after one hint proves the student knew it.”
Not necessarily. The hint may have supplied the exact operation that was missing.

“Independence means never asking for help.”
No. Independent learners often seek help intelligently. The issue is whether the target capability can later be performed under the conditions that matter.

“If scaffolding works, we should keep it.”
Only while it remains useful. Educational scaffolding is normally intended to transfer responsibility, not permanently carry the same operation.

How Do We Know?

The evidence base is broad but not uniform. Scaffolding research varies by age, subject, support type and setting. A review of K–12 studies found substantial variation in how teacher scaffolding is implemented and noted that ongoing diagnosis, responsiveness and fading are not always reported in detail. A recent meta-analysis of digital learning prompts found positive average effects on learning achievement, while effects varied by prompt type, target group, domain and study context. EEF guidance treats scaffolds as temporary supports that should gradually enable greater pupil independence.

Dynamic assessment provides a complementary line of evidence by explicitly examining response to teaching or graduated prompts. A systematic review in reading found that dynamic measures could add predictive information beyond static measures in the studies reviewed, although effects varied by construct and outcome.

These sources support the principle that support conditions matter. They do not justify a universal formula converting “number of hints” into ability, and this article does not propose one.

Evidence Boundary

Do not infer a clinical or psychological condition from support needs.
Do not treat every request for help as dependency.
Do not assume performance without support is always the only educationally meaningful outcome.
Do not claim transfer until performance survives an appropriate reduction of assistance and a meaningful change in conditions.

The Bolt → Interface → MindOS → Bolt Handoff

Bolt calibrates: “The learner succeeded, but only after a targeted cue supplied method selection. Independent method selection is not yet demonstrated.”

The Student/Studying Interface makes that conclusion operable: the next task states the goal, removes the method label, allows one general help request, and makes the learner choose the first step before assistance is released. See the Performance Handoff and Human–Technology Handoff.

MindOS runs the learning operation: strategy selection, retrieval, representation, scaffold fading or another operation is chosen only after the evidence justifies it.

Return to Bolt: later independent performance under changed conditions tests whether the target capability transferred to the learner.

Parent and Tutor Guide

When a child finishes difficult work with help, avoid both extremes. Do not say, “That does not count because I helped.” And do not say, “Great, you can do it now,” before checking.

“That support helped you get through it. Let’s see which part is yours now.”

Then use one fresh example with less support. That small change turns help from a vague feeling into useful performance evidence.

Bolt Direction Graph

SUPPORTED PERFORMANCE
→ identify the target operation
→ record what assistance supplied
→ bound the inference
→ reduce assistance
→ change task / representation / delay where appropriate
→ observe independent return
→ recalibrate
→ Student Interface translates next action
→ MindOS runs the justified learner operation
→ new performance returns to Bolt

Authoritative Sources and Further Reading

Durable Bolt rule: Support can improve a performance and still change what that performance proves. Record the help, fade it deliberately, and let later independent performance carry the stronger claim.