Bolt 32 — The Teacher’s Final Job Is to Become Less Necessary

Bolt Series · Human Performance Calibration · Article 32

A strange way to judge a successful teacher

Imagine two classrooms.

In the first, the students perform beautifully whenever the teacher is present.

The teacher catches every mistake.

Chooses every task.

Decides when the student is ready.

Explains every misunderstanding.

Tells the learner what to revise next.

The work looks excellent.

In the second classroom, the teacher is also highly involved.

But over time, something changes.

Students begin catching some of their own errors.

They can explain which parts of a topic are secure and which are not.

They can decide when a hint is enough and when a deeper explanation is needed.

They can compare a result with criteria before the teacher marks it.

They can choose a reasonable next practice target.

The teacher is still important.

But the students carry more of the learning system themselves.

Which classroom has produced the stronger learners?

The teacher’s job is not to disappear

The title of this article is deliberately provocative.

“Become less necessary” does not mean:

stop teaching;
withdraw support too early;
leave novices to discover everything alone;
pretend expertise is unnecessary;
replace teachers with self-study.

It means something more precise.

A learner should gradually need less external help for operations they have genuinely learned to carry themselves.

The teacher remains available for harder problems, new domains, recalibration, deeper expertise and situations where the learner’s internal model is no longer enough.

But the teacher should not remain permanently responsible for every judgement the learner could eventually make.

Support and dependence are not the same thing

Educational research on scaffolding helps clarify this distinction.

A 2023 meta-analysis of 46 studies found a moderate positive overall effect of regulated-learning scaffolding on regulation strategies and academic performance.

That is an important reminder that learners often need external structures before they can regulate effectively alone.

But support should have a developmental purpose.

A scaffold is useful because it helps someone do something they could not yet carry alone.

If the scaffold becomes permanently inseparable from the performance, we have to ask whether the underlying capability was transferred.

The key question is: who is carrying the regulation?

Recent research in digital learning has made this question explicit.

In system-regulated learning, the system can decide the sequence, timing and regulation of learning for the student.

In self-regulated learning, more of those decisions belong to the learner.

A 2025 study comparing self- and system-regulated learning with and without scaffolding found benefits from scaffolding and highlighted the importance of understanding who carries the regulation in a learning environment.

This is one of the most important educational questions in an age of AI.

If the teacher, tutor or AI always decides:

what the learner should do next;
which answer is weak;
whether the learner is ready;
when help is needed;
what error occurred;
which strategy should be chosen;

then the learner may produce efficient work while remaining dependent on an external regulator.

Performance can improve without ownership of the performance system improving equally.

A teacher should transfer the work of seeing

At first, the teacher may see the problem before the learner does.

“You are losing marks because you are not reading the qualifier in the question.”

Later, the teacher can ask:

“What pattern do you notice across these three errors?”

Later still:

“Before I mark this, what are you checking for?”

Eventually the learner says:

“I caught the same qualifier mistake before submitting.”

The teacher has not merely fixed a repeated error.

The teacher has transferred part of the error-detection process.

A teacher should transfer the criteria too

Students cannot evaluate their own work if the standard lives entirely inside the teacher’s head.

The teacher says:

“This is a strong explanation.”

Why?

What made it strong?

What distinguishes description from explanation?

What makes an argument complete?

What makes a mathematical solution trustworthy?

When criteria become visible, the learner can begin using them before the teacher speaks.

This is one of the most important transitions from external judgement to self-regulation.

A teacher should transfer prediction too

Early learners often wait for the score to discover what happened.

A stronger learner begins predicting before the score arrives.

“I think this is around 75. The concepts are secure, but I probably lost marks in the last section because my pace dropped.”

Now the teacher does not need to supply the first model of the performance.

The learner brings one.

The teacher can compare it with the evidence.

At this stage, teaching has moved up one level.

The teacher is no longer only calibrating the performance.

The teacher is calibrating the learner’s calibration.

The teacher becomes more valuable at the boundaries

As learners become more independent, the teacher’s role does not shrink to nothing.

It shifts toward the places where external expertise adds the most value.

Novel tasks the learner has never calibrated against.
Persistent blind spots the learner cannot diagnose alone.
High-stakes decisions where another observer reduces risk.
Advanced standards the learner does not yet know exist.
Contradictory evidence that needs a discriminating test.
Plateaus where the learner’s current strategies no longer produce progress.

This is similar to elite coaching.

The athlete may be highly self-aware and still benefit enormously from a coach because the coach sees different information and can challenge a model that has become locally blind.

Independence does not eliminate collaboration.

It changes its quality.

The learner should know when the teacher is still necessary

One of the strongest signs of independence is not refusing help.

It is seeking help accurately.

The learner can distinguish:

“I can solve this with another attempt.”
“I need one hint.”
“I understand the method but cannot see why my answer keeps failing.”
“I have reached the limit of my current model and need another perspective.”
“This decision matters enough that I should not rely on my own judgement alone.”

That is not dependence.

It is calibrated use of expertise.

How do we know whether teaching has transferred?

Do not judge only by supported performance.

Remove some support.

Change the task.

Delay the feedback.

Ask the learner to predict.

Ask the learner to diagnose the error before seeing the teacher’s diagnosis.

Ask the learner to choose the next practice target and justify it.

Research on self-regulated learning in simulation has made a similar point: learning should not be reduced to simply removing instructors, but should include designed supports that prepare people for future learning and progressively greater self-regulation.

The test of good support is not whether the learner looks good while supported.

It is what remains when the support changes.

AI makes this principle urgent

AI can become an extraordinary external regulator.

It can identify errors instantly.

Choose the next task.

Explain the concept.

Rewrite the answer.

Suggest a strategy.

That can reduce friction dramatically.

But if the learner never has to detect uncertainty, generate a first attempt, judge whether an answer is secure or decide when help is necessary, the system may become more capable while the learner remains less independent than the output suggests.

The right question is therefore not:

“Can AI help the student do this?”

It is also:

“Which part of this performance should still belong to the learner after the AI is removed?”

The teacher’s final job is a transfer problem

At the beginning, the teacher may carry most of the system.

The teacher sets the goal.

Defines the standard.

Chooses the task.

Finds the error.

Interprets the result.

Decides the next move.

Over time, more of those functions should become available inside the learner where developmentally appropriate.

Not all at once.

Not perfectly.

Not without continued access to expertise.

But enough that the learner is no longer helpless when the external regulator leaves.

The teacher’s deepest success is not that the learner always needs the teacher. It is that the learner carries forward more of the seeing, judging, checking and choosing that good teaching once supplied from outside.

Why this matters for education

Every child eventually enters situations where no teacher is standing beside them.

The examination hall.

University.

Work.

Parenthood.

Professional judgement.

Learning something nobody assigned.

They have to decide whether they know enough, whether the work is trustworthy, whether more evidence is needed and when another human perspective should be invited in.

Education should prepare them for that world.

A great teacher does not make expertise unnecessary. A great teacher makes permanent dependence unnecessary.

That is a very different ambition.

And it may be one of the clearest tests of whether learning has truly transferred.