Bolt 24 — Building an Internal Coach

Bolt Series · School–Teacher–Student Performance Calibration · Article 24

Wait, What? The Best Evidence of Coaching May Be That the Coach Becomes Less Necessary

A teacher can stand beside a learner and make performance look excellent.

The harder question is what remains when the teacher steps away.

Bolt calls this a calibration problem. If coaching is meant to improve future independent performance, then support should eventually leave observable evidence inside the learner’s later performance: better prediction, earlier error detection, more accurate use of criteria, better judgement about readiness, or more appropriate help-seeking.

This page does not own the learner’s self-regulation machinery. It owns the measurement question: how do school and teacher know whether coaching has actually transferred rather than merely improving supported performance?

Quick Answer

Owned Bolt calibration job: measure the transfer of monitoring and judgement from external teacher/coach support into later learner performance.

A strong coaching interaction is an early receipt, not the final one. The stronger evidence arrives later, when the learner encounters another task and can carry more of the relevant observation, checking or decision work without the coach supplying it again.

Supported Performance and Transferred Performance Are Different Objects

• Supported performance: what the learner can do while coaching, prompts, cues or feedback are present.
• Immediate independent performance: what remains after support is removed right away.
• Delayed return: what remains after time has passed.
• Transfer performance: what survives when the task surface, selection demand or context changes.
• Calibration performance: whether the learner can increasingly predict, inspect and judge the performance accurately.

A coaching programme should not claim transfer merely because the supported session looked good.

Observable Signatures That Coaching Has Transferred

• The learner identifies a recurring error before the teacher points it out.
• The learner uses criteria accurately without waiting for the teacher to restate them.
• The learner’s prediction becomes more closely aligned with later performance.
• The learner can explain which support is needed rather than asking for the whole solution.
• The same improvement appears on a fresh task, not only the coached example.
• The teacher can reduce cue specificity without performance collapsing.
• The learner recognises when a previous diagnosis no longer fits new evidence.

These are performance receipts. They should be observed, not assumed.

Competing Explanations for Better Performance After Coaching

• The learner genuinely acquired a more independent monitoring or judgement capability.
• The next task was easier or unusually similar to the coached example.
• The learner remembered the teacher’s exact cue without understanding when it applies.
• The teacher was still providing subtle support.
• The learner improved in the subject skill but not in self-monitoring.
• Practice effects or familiarity contributed.
• The apparent improvement is one anomalous result.

Bolt keeps these possibilities alive until the evidence discriminates among them.

School–Teacher–Student Calibration

School

The school should evaluate coaching with more than attendance, teacher satisfaction or the quality of the coaching conversation. It should ask whether the intended change becomes visible in later teaching or learner performance and whether that change is stable enough for the decision being made.

Teacher or Coach

The teacher should distinguish helping from measuring. During learning, direct support can be appropriate. During a return check, the teacher must know which support remains present or the resulting performance cannot be interpreted cleanly.

Student

The student should not be judged by whether they become completely independent immediately. The useful question is whether the amount and type of external support needed for the same performance job changes over time.

Worked Example: The Teacher Who Always Says “Check Your Sign”

A Mathematics student repeatedly makes sign errors. The teacher catches each one by saying, “Check your sign.” Accuracy rises during lessons.

If the examination still contains the same errors, the supported classroom performance was real but the monitoring function did not transfer far enough.

A stronger sequence of evidence would be:

1. The teacher names the error directly.
2. Later, the teacher indicates only that an error exists.
3. Later, the learner checks against a criterion without location help.
4. Later, the learner detects the error before submission.
5. On a changed problem days later, the monitoring still appears.

Bolt is interested in the evidence of transfer across those stages. MindOS owns the learner mechanism and practice operation used to build it.

How Do We Know?

The 2023 meta-analysis Effects of Regulated Learning Scaffolding on Regulation Strategies and Academic Performance synthesised 46 studies and found a moderate overall effect of regulated-learning scaffolding, with important variation by context and scaffold type. This supports measuring outcomes rather than assuming that any scaffold automatically transfers.

A 2025 experimental study, Who Regulates? Effects of Scaffolding in System- and Self-Regulated Learning, found benefits of scaffolding for learning performance and self-efficacy in its sample, while finding no simple main effect for who carried regulation. That is a useful caution against turning “independence” into a slogan: the design and learner conditions matter.

The Standards for Educational and Psychological Testing provide the wider measurement discipline: the interpretation and use of evidence must be justified. If the claim is “coaching transferred,” the evidence should come from a task that can actually show transfer.

Evidence Boundary

This page does not claim that support should always be reduced, that independent performance is always the only educational goal, or that a learner who still needs help has failed. Some tasks are legitimately collaborative or tool-supported.

The claim is conditional: when the educational objective includes future independent monitoring or judgement, the measurement must eventually test what remains when the relevant support is reduced or removed.

Common Misconceptions

• “Good coaching means the learner needs no help.” No. It means support is appropriate to the current job and does not masquerade as independent capability.
• “If performance improves during coaching, transfer is proven.” Not yet.
• “Fading support is always best.” Not for every task or learner state.
• “The student should become their own teacher.” Too broad. Bolt measures specific transferred functions.
• “Independence means never asking for help.” Skilled help-seeking can itself be part of calibrated performance.

What Should Change Next?

When coaching appears successful, the next Bolt move is not automatically more coaching. It is a return performance designed to test the claimed transfer at the right difficulty and with the support conditions recorded.

RFE: Did coaching change what the learner can later observe, judge or perform without the coach carrying the same function—and is that conclusion supported by repeated, relevant performance?

Bolt Direction Graph

Coaching target → supported performance → specify what the coach is carrying → reduce or remove that support in a fair return task → observe what survives → delay/change conditions → compare with prediction → update the school–teacher–student performance model.

Useful neighbours: Bolt — When Feedback Makes You Dependent; Bolt Measurement Note 26 — A Coaching Conversation Is Not Yet Improved Teaching; and MindOS — Scaffold Fading State.