Training adaptation, adaptive teaching, implementation adaptation, contextual tailoring, evidence-based practice, conceptual fidelity, active ingredients, responsive instruction, differentiated instruction, learner needs, school implementation and educational change all describe versions of the same difficult problem: how do we change a training plan when reality changes without accidentally changing the mechanism that made the plan worth using?

No serious educational system can remain perfectly fixed. Learners differ. School timetables change. A child improves. A child falls ill. A parent becomes unavailable. An examination approaches. A representation becomes a bottleneck. A task turns out to be too easy. Another is too difficult. A supposedly minor support cue removes the very decision the learner was supposed to practise. Good teaching must adapt.

But adaptation has a danger. Every change can quietly weaken the active ingredient. Mixed practice becomes blocked. Retrieval becomes rereading. Self-explanation becomes teacher explanation. Independent initiation becomes adult prompting. A fresh retest becomes repeating the same corrected item. The system still looks educational. The mechanism has changed.

Current implementation research therefore treats adaptation as something to manage rather than something to celebrate blindly. AERO’s 15 September 2026 Context first: Tailoring implementation to each school setting states that implementation is not one-size-fits-all and should be adapted to local context and need. EEF’s current implementation guidance likewise frames change as iterative, contextual and responsive rather than a single launch-and-follow process.

Training adaptation is the deliberate change of delivery form, timing, support, representation, dosage, sequence or context in response to learner and environmental evidence while preserving the active ingredient and target capability that define the training mechanism.

Adaptation is not drift.

It is not improvisation without a record.

It is not “whatever works today.”

It is controlled change around a protected learning function.

---

Quick Read: Change the Form, Protect the Function

Identify the Active Ingredient → Identify the Context Problem → Decide What Can Move → Adapt One Meaningful Dimension → Observe Learner Response → Check Fidelity → Keep, Reverse or Refine the Adaptation

A useful adaptation process asks:

• What is the training job?
• Which active ingredient must remain?
• What exactly is not fitting the learner or context?
• Is the problem difficulty, timing, representation, support, dosage, environment or acceptability?
• Which part can change safely?
• What evidence will tell us whether the adaptation helped?
• What evidence would show that the adaptation damaged the mechanism?

The Current Apex Implementation Frame

AERO’s Context first guide, published 15 September 2026, is directly relevant to this article. It argues that implementation should respond to local context rather than assume a fixed delivery model. AERO uses examples from different school settings to show how implementation can be tailored while remaining connected to evidence-based practice.

AERO’s companion guide Staying on track provides the second half of the logic: once adaptation occurs, teams should monitor feasibility, acceptability, fidelity, reach and sustainability. Adaptation without monitoring is merely change. Adaptation plus implementation evidence becomes a testable design decision.

EEF’s Evaluate and sustain your strategy, published 3 September 2026, explicitly states that approaches may need to be sustained, adapted, scaled or stopped. The guidance describes implementation as an ongoing cyclical process rather than a one-way march toward a fixed endpoint.

A 2026 school-based implementation study in BMC Public Health is also useful conceptually. The study examines teacher-led adaptation while maintaining conceptual fidelity. Its domain is physical activity rather than academic tuition, so its specific results should not be transplanted mechanically. The general implementation lesson is relevant: adaptation can coexist with fidelity when the core conceptual function is protected.

Adaptation Owner Boundary

This article owns the question:

How should the delivery of a training mechanism change when learner or context conditions change, while preserving the function the mechanism was designed to perform?

It does not re-own:

• Training Fidelity, which asks whether the active ingredient was preserved.
• Training Feasibility, which asks whether the plan can operate under real constraints.
• Training Acceptability, which asks whether people will use the plan.
• Training Sustainability, which asks whether the system keeps functioning over time.
• The Replanning Trigger, which asks when the current strategy should be abandoned and replaced.
• Training Branching, which chooses the next task based on learner response.

Adaptation sits between branching and replanning.

Branching changes the next instructional route inside an existing training architecture.

Adaptation changes how the architecture is implemented in a context.

Replanning changes the strategy when the current strategy no longer deserves continuation.

The Tight–Loose Principle

Every useful adaptation begins by separating what must remain tight from what can remain loose.

Tight elements are active ingredients.

Loose elements are delivery choices.

The distinction is not universal.

It depends on the training job.

Mathematics Tight–Loose Example

Target: mixed quadratic method selection.

Tight:

• methods must be mixed;
• Mira must choose before calculating;
• the discriminating feature must be attended to;
• fresh verification must occur.

Loose:

• whiteboard or paper;
• specific coefficients;
• number of practice items within a useful range;
• whether explanation is written or spoken;
• whether the fresh check occurs now or next session.

If the learner dislikes writing explanations, switching to verbal explanation can be a fidelity-consistent adaptation.

If the learner finds mixed questions difficult and the tutor responds by labelling every method, the adaptation destroys the target decision.

English Tight–Loose Example

Target: evidence-bounded inference.

Tight:

• learner identifies evidence;
• learner proposes an inference;
• claim strength is compared against evidence;
• fresh context follows.

Loose:

• narrative or article;
• oral or written explanation;
• sentence length;
• whether confidence is recorded numerically or verbally.

Science Tight–Loose Example

Target: evidence-to-mechanism explanation.

Tight:

• evidence identified;
• pattern described;
• mechanism connected;
• conclusion bounded by evidence.

Loose:

• apparatus context;
• graph or table;
• oral causal map or written paragraph;
• number of examples.

Adaptation Type 1: Timing

The same training operation can occur at different times.

A low-energy weekday may not support a full writing task.

The family can move the full composition to Saturday and keep one paragraph-development block on Wednesday.

The active ingredient remains.

Timing changes.

Timing Adaptation Should Not Become Delay Forever

If difficult work is always postponed, the plan is not adapting.

It is avoiding.

Adaptation should specify when the full task returns.

Adaptation Type 2: Dosage

High-capacity week:

six fresh method-selection cases plus transfer.

Low-capacity week:

three classifications plus one solve.

The mechanism remains.

Dosage falls.

Dosage Adaptation Needs a Minimum Viable Mechanism

If reduction removes the active ingredient, the plan has crossed from adaptation into substitution.

Five minutes of closed-book retrieval can preserve retrieval.

Five minutes of rereading does not preserve the same operation.

Adaptation Type 3: Support

The learner struggles to begin.

Support can increase temporarily.

But support should be designed around the target decision.

Training Cue Hierarchy gives the local mechanism: add the smallest cue that restores the route.

Support Adaptation Example

Mira cannot choose a method.

Unsafe adaptation:

Use factorisation.

Safer adaptation:

Which structural feature would make factorisation attractive?

The second cue supports attention without making the method decision for her.

Adaptation Type 4: Representation

A learner may understand the relationship in one form and fail in another.

Adaptation can temporarily return to a clearer representation.

Then the learner should cross back.

Representation Adaptation Example

Nadia struggles with an unfamiliar experiment diagram.

The tutor rewrites the apparatus relationships as a simple table.

Nadia identifies variables correctly.

Then the tutor returns to the diagram.

The table adaptation reduced representational load without replacing the target experimental reasoning.

Adaptation Type 5: Language

When language is incidental to the target, simplify or clarify it.

When language is the target, preserve it.

This distinction prevents adaptations from accidentally lowering the construct.

Language Adaptation Example

Mira is solving a Mathematics problem with unusually dense wording.

The target is algebraic structure, not reading complexity.

The tutor paraphrases the context while preserving the mathematical relationships.

This can improve construct access.

Language Adaptation Counterexample

Jonas is training inference from nuanced wording.

The tutor replaces every difficult phrase with simple synonyms before Jonas reads.

The adaptation removes part of the reading demand.

If nuanced vocabulary is part of the intended construct, this adaptation threatens fidelity.

Adaptation Type 6: Response Format

Learners can sometimes demonstrate the same reasoning through different response formats.

Oral explanation.

Written causal map.

Annotated diagram.

Short paragraph.

Format adaptation can improve access when the format is not itself the target.

Response Format Boundary

If the examination requires formal writing, oral reasoning cannot permanently replace written performance.

It can be an intermediate adaptation.

Authentic format must return before exit.

Adaptation Type 7: Task Size

A full paper may be too large for a narrow repair.

Use one representative cluster.

After repair, recombine into whole performance.

Task-Size Adaptation Example

Jonas repeatedly overclaims inference.

Instead of another full comprehension paper, use four carefully chosen inference items.

Once calibration improves, return to a full passage.

Adaptation Type 8: Sequence

Some learners need pre-training before the full task.

Others are ready to attempt first.

Sequence can adapt to readiness.

Sequence Adaptation Example

Nadia encounters unfamiliar apparatus terminology.

Pre-train the apparatus names.

Then ask for experimental reasoning.

The target relationship remains untouched.

Adaptation Type 9: Example Selection

The next example can be changed to reveal structure more clearly.

Training Example Selection owns choosing the example that reveals the next structure.

Adaptation decides when the current example family is not fitting the learner.

Adaptation Type 10: Context

A routine can be moved across home, tuition or school preparation depending on support needs.

High-support repair may belong in tuition.

Low-support maintenance may belong at home.

Context Adaptation Should Match Support Demand

Do not assign a fragile new skill to independent home practice merely because home time exists.

Do not keep a stable routine inside expensive tutor time when the learner can run it alone.

Adaptation Type 11: Social Format

Individual work.

Pair comparison.

Three-student discussion.

Tutor modelling.

The social form can change.

The target cognitive operation must remain visible.

Social Adaptation Example

Mira, Jonas and Nadia compare three solutions.

Each student first commits independently.

Then they discuss why methods differ.

Peer interaction extends contrast without letting one student decide for everyone.

Adaptation Type 12: Technology

Paper can become digital.

Digital can become paper.

AI can generate examples.

A tutor can replace an app.

Tool changes are adaptations when the active learning operation remains.

Technology Adaptation Example

A spaced-repetition app creates too much login friction.

Move the same retrieval schedule to index cards.

Mechanism preserved.

Delivery adapted.

Technology Adaptation Counterexample

A learner is supposed to generate a paragraph.

AI produces the paragraph first.

The learner edits it.

The activity may support editing.

It no longer preserves independent generation.

Adaptation Type 13: Difficulty

Difficulty can move up or down.

The key is to change the dimension that matters.

Difficulty Adaptation Example

Mira’s method selection is stable on clean cases.

Add competing methods.

Do not merely use uglier numbers.

Difficulty Adaptation Counterexample

Nadia struggles with causal reasoning.

The tutor adds more obscure vocabulary.

The task becomes harder without becoming more diagnostic.

Adaptation Type 14: Pacing

Some learners need longer before support.

Others need a cue sooner to avoid unproductive search.

Pacing can adapt while the active ingredient remains.

Pacing Should Be Evidence-Based

Do not equate slow with weak.

Do not equate fast with ready.

Observe accuracy, reasoning and support dependence.

Adaptation Type 15: Feedback Form

Feedback can be:

• verbal;
• written;
• coded;
• comparative;
• question-based;
• worked-example based.

Choose the form that enables repair.

Feedback Adaptation Must Preserve Action

If feedback becomes so detailed that the learner simply copies the corrected answer, the repair operation disappears.

Adaptation Type 16: Monitoring Intensity

Early training may require close observation.

Later training should require less.

Monitoring intensity should adapt to competence.

Monitoring Adaptation Example

At first, tutor records Mira’s cue level on every mixed case.

Later, tutor samples one or two per session.

The learner now self-monitors.

Adaptation Type 17: Assessment Format

Multiple choice can be replaced with open response when deeper explanation needs to become visible.

Open response can be replaced with short classification when rapid sampling is needed.

The assessment format should match the measurement job.

Adaptation Type 18: Frequency

New skill: frequent contact.

Stable skill: wider spacing.

Frequency should adapt to retention and maintenance state.

Adaptation Type 19: Intensity

Some periods justify intensive repair.

Then intensity should fall.

Permanent high intensity is rarely sustainable.

Adaptation Type 20: Ownership

Perhaps the most important adaptation is who controls the system.

Tutor-controlled.

Shared.

Learner-controlled.

Ownership should adapt upward with competence.

Adaptation and Training Fidelity

Training Fidelity provides the boundary.

Every adaptation should ask:

What active ingredient did we protect?

If no one can answer, the change may be drift.

Conceptual Fidelity vs Surface Fidelity

Surface fidelity asks whether the lesson looks the same.

Conceptual fidelity asks whether the mechanism still functions.

Two lessons can look different and be conceptually faithful.

Two lessons can look identical and be conceptually different because tutor cues changed the mental work.

The Adaptation Log

A simple adaptation log records:

• what changed;
• why;
• which active ingredient was protected;
• what outcome was expected;
• what happened.

One sentence can be enough.

Reduced Wednesday dose from six mixed questions to three classifications because CCA reduced capacity; preserved method-selection decision and delayed retest.

Why Record Adaptations?

Without a record, adults later remember the intended plan rather than the enacted plan.

Outcome interpretation becomes confused.

Adaptation and Drift

Drift is adaptation without deliberate reasoning or fidelity control.

It often begins with convenience.

One skipped retest.

One extra hint.

One easier worksheet.

Repeated often enough, the intervention changes identity.

The Drift Test

Did this change happen because of learner/context evidence, or because it was easier in the moment?

Convenience is not always wrong.

But convenience should not silently decide the learning mechanism.

Adaptation and Training Reach

Training Reach often creates the reason to adapt.

If the mechanism does not reach home because parent support is unavailable, adapt the delivery.

If the mechanism does not reach timed performance, adapt practice conditions.

If one representation blocks access, adapt the representation temporarily.

Adaptation and Feasibility

Low feasibility is one of the most common triggers for adaptation.

Shorten dose.

Move timing.

Reduce administrative burden.

Change materials.

Protect the active ingredient.

Adaptation and Acceptability

Low acceptability can often be solved by changing form rather than mechanism.

Written explanation becomes verbal.

Full essay becomes paragraph drill.

Public explanation becomes private note.

Same learning job.

Adaptation and Sustainability

Sustainable systems adapt as support needs fall.

Keeping novice scaffolds forever is not high fidelity.

It is failure to adapt.

Adaptation and Branching

Branching is micro-adaptation inside a session.

The learner response selects the next task.

Adaptation can operate at larger scales:

• session design;
• weekly dosage;
• home delivery;
• representation mix;
• monitoring intensity.

Adaptation and Replanning

Not every problem can be solved by adaptation.

If the active mechanism itself is wrong, use the Replanning Trigger.

Do not adapt a failed strategy forever.

The Adapt-or-Replan Test

Adapt when:

• the mechanism remains plausible;
• the problem is context or delivery;
• fidelity can be preserved;
• outcomes show some useful signal.

Replan when:

• high-fidelity implementation repeatedly fails;
• new evidence changes the bottleneck;
• the target itself was misdiagnosed;
• a superior mechanism becomes available.

Adaptation and Training Review

Training Review is where larger adaptations should be reconsidered systematically.

Adaptation and Measurement Validity

An adaptation can change what a task measures.

If vocabulary is removed from an English task, the construct may change.

If time limits are removed, performance conditions change.

Record these changes before comparing scores.

Adaptation and Comparability

Training Comparability protects interpretation across adapted tasks.

Do not claim improvement from an easier adapted form without acknowledging the change.

Adaptation and Parallel Forms

Parallel forms are adaptation with controlled comparability.

Surface changes.

Blueprint remains.

Adaptation and Measurement Noise

If delivery changes every session, outcomes become harder to interpret.

Adapt one meaningful dimension when possible.

Then observe.

Adaptation and Triangulation

When an adaptation appears helpful, confirm across more than one evidence window.

Fresh task.

Learner report.

School work.

Process trace.

Adaptation and Observability

Good adaptation requires enough visibility to know what problem is being solved.

Otherwise adults adapt to symptoms.

Adaptation and Measurement Resolution

Too-coarse diagnosis creates blunt adaptations.

“Weak English” produces generic extra work.

“Evidence selection stable; claim calibration weak” produces targeted adaptation.

Adaptation and Error Propagation

One source error may justify adapting a single upstream step rather than changing the whole programme.

Adaptation and Ceiling Effects

Raise challenge when the current task can no longer show growth.

Change the relevant dimension, not random difficulty.

Adaptation and Floor Effects

Lower incidental difficulty until the target operation becomes visible.

Then rebuild upward.

Adaptation and Contamination

If repeated items inflate performance, adapt to fresh forms.

If cues contaminate independence, fade them.

Adaptation and Baselines

When delivery conditions change substantially, note the new conditions before comparing with baseline.

Adaptation and Retests

Retests themselves may need adaptation for context while preserving the repaired capability.

Adaptation and Progression

Progression is planned adaptation of challenge.

It should respond to evidence rather than calendar alone.

Adaptation and Readiness

Readiness determines whether to simplify, cue, pre-train or progress.

Adaptation and Dependencies

If a dependency blocks the target, temporarily adapt the task to isolate it.

Then recombine.

Adaptation and Pre-training

Pre-training is a proactive adaptation that reduces incidental novelty before complex performance.

Adaptation and Decomposition

Decomposition is appropriate when the whole task hides the weak link.

Do not stay decomposed forever.

Adaptation and Recombination

After micro-repair, return to the whole.

Authentic performance is the final fidelity test.

Adaptation and Training Environment

Environment can be adapted:

• quiet space;
• device mode;
• materials prepared;
• shorter travel chain;
• different study time.

Environment changes should support the learning operation rather than become an end in themselves.

Adaptation and Family Coordination

Home schedules change.

A family plan may need to move the same training job between days.

Protect the core.

Adaptation and Parent Role

A parent can move from marker to time protector.

That is an ownership adaptation.

Adaptation and Tutor Role

Tutor can move from explainer to diagnostician to verifier as competence rises.

Adaptation and Learner Role

Learner can move from recipient to co-designer to independent operator.

Mira Case Study: CCA Week

Mira’s Wednesday capacity falls because CCA ends late.

Original plan:

• six mixed classifications;
• three full solves;
• one reflection.

Adapted:

• three mixed classifications;
• one full solve;
• fresh retest on Saturday.

Active ingredient preserved.

Dosage reduced.

This is adaptation.

Mira Case Study: Too Much Cueing

Mira repeatedly freezes at method selection.

Tutor adapts by naming the method.

Performance improves immediately.

Fidelity falls.

Better adaptation:

provide a structural cue and ask Mira to choose.

Mira Case Study: Representation Bottleneck

Mira can solve equations but fails graph-linked quadratics.

Adapt:

• extract graph features first;
• map them to algebra;
• recombine into original task.

Do not permanently remove graphs.

Jonas Case Study: Full Essay Burden

Jonas needs idea development.

Full essays twice weekly are creating avoidance.

Adapt:

• one idea chain;
• one developed paragraph;
• one weekly full composition.

Mechanism remains.

Burden falls.

Jonas Case Study: Vocabulary Interference

An unfamiliar word blocks a comprehension inference.

If vocabulary is incidental to the inference target, clarify the word.

Then retest inference.

If vocabulary itself is the target, do not remove it.

Jonas Case Study: Oral to Written Bridge

Jonas can explain a paragraph idea orally but cannot write it.

Adaptation sequence:

• oral explanation;
• bullet structure;
• sentence construction;
• full paragraph.

The destination remains writing.

Nadia Case Study: Apparatus Novelty

Nadia’s reasoning disappears in unfamiliar apparatus diagrams.

Adapt:

• pre-train apparatus labels;
• use clean diagram;
• return to unfamiliar diagram;
• fresh transfer.

Nadia Case Study: Time Pressure

Nadia explains accurately untimed but writes too slowly.

Adapt:

• compress explanation structure;
• practice one-minute causal chains;
• preserve evidence and mechanism;
• then return to whole paper timing.

Evan Case Study: Peer Delivery

Evan and Mira study together.

The peer format improves motivation but risks answer leakage.

Adapt social sequence:

independent attempt → compare → explain difference → fresh independent item.

Adaptation in a Three-Student Classroom

Small groups require constant adaptation.

Same topic.

Different current states.

Mira needs contrast.

Jonas needs cueing.

Nadia needs transfer.

The tutor should not create three unrelated lessons.

Use shared core and adaptive branches.

Shared Core, Adaptive Branch

Example:

• shared explanation of quadratic structure;
• Mira: mixed method classification;
• Jonas: supported execution;
• Nadia: unfamiliar graph transfer;
• shared review of why routes differed.

This is personalisation without curricular fragmentation.

Adaptation and Pacing in Small Groups

One learner finishes early.

Do not automatically give more of the same.

Adapt into variation or explanation.

Adaptation and Peer Learning

Peer explanation can help one learner and over-support another.

Require prediction before listening.

Adaptation and Dignity

Adaptation should not visibly mark a learner as deficient.

Different next steps can be normalised.

Adaptation and Age

Older learners need greater ownership and less childish scaffolding.

Younger learners may need stronger external routines.

The active ingredient can remain while delivery matures.

Adaptation and Expertise

Supports can reverse in value as expertise grows.

What helps a novice may burden an advanced learner.

Adaptation should track competence, not age alone.

Adaptation and Expertise Reversal

The broader HPL owner on Expertise Reversal explains the principle.

Punggol’s implementation question is practical: when should a previously useful scaffold be removed?

The Scaffold Removal Test

Remove or reduce support when:

• the learner starts correctly without it;
• accuracy remains stable;
• fresh transfer remains stable;
• support begins slowing or annoying the learner.

Adaptation and Accessibility

Change response format or presentation where needed to provide access.

Do not remove the target construct unless the learning goal changes.

Adaptation and Known Learner Needs

Known accommodations can be integrated into training.

Tutors should not invent medical diagnoses.

Work within documented needs and observable learning requirements.

Adaptation and Anxiety

High-pressure simulation may need gradual introduction.

Begin with partial timing or shorter exposure.

Build toward authentic performance.

Adaptation and Confidence

Do not lower challenge merely to create confidence.

Create reachable progress with valid task demands.

Adaptation and Boredom

Boredom can mean:

• ceiling;
• repetition without progression;
• purpose unclear;
• low interest.

Adapt after diagnosis, not from boredom alone.

Adaptation and Low Motivation

Change burden or relevance where appropriate.

Do not remove the learning operation merely to increase engagement.

Adaptation and High Motivation

Motivated learners can still be overtrained.

Do not use willingness as justification for unnecessary volume.

Adaptation and Exam Proximity

As examination nears:

• increase authenticity;
• reduce novelty;
• increase time realism;
• preserve feedback loops;
• avoid radical strategy changes without evidence.

Adaptation and Post-Exam Period

Reduce exam-specific routines.

Return to underlying capabilities and recovery.

Adaptation and Holidays

Use reduced core.

Repair dependencies.

Allow genuine rest.

Adaptation and Illness

Rest may be the correct adaptation.

After recovery, use re-entry rather than debt repayment.

Adaptation and School Transition

New stage changes:

• independence;
• volume;
• representation;
• timing;
• subject complexity.

Re-specify delivery while preserving durable routines.

Adaptation and Curriculum Change

Underlying mechanisms can transfer across syllabus updates.

Examples and terminology may change.

Adaptation and Exam Format Change

When exam format changes, update practice distribution.

Do not assume old templates remain authentic.

Adaptation and AI

AI makes adaptation easier because examples, explanations and feedback can be generated quickly.

This increases opportunity and risk.

AI Adaptation Risk 1: Over-support

AI solves the target operation.

Learning looks smoother.

Fidelity falls.

AI Adaptation Risk 2: Infinite Variation

AI produces endless examples.

Quantity expands beyond useful dosage.

Adaptation becomes noise.

AI Adaptation Risk 3: Difficulty Drift

Generated variants change difficulty unpredictably.

Comparability falls.

AI Adaptation Risk 4: Answer Style Drift

AI feedback changes expected language or method beyond school norms.

Verify against syllabus and assessment demands.

AI Adaptation Opportunity 1: Parallel Forms

Generate fresh items from a defined blueprint.

Human verifies construct and difficulty.

AI Adaptation Opportunity 2: Reading-Level Adjustment

When language is incidental, simplify instructions while preserving reasoning demand.

AI Adaptation Opportunity 3: Case Families

Generate prototypes, near-misses and transfer cases around one invariant.

AI Adaptation Opportunity 4: Feedback Format

Convert a long explanation into a cue, question or comparison depending on learner state.

AI Adaptation Rule: Specify What Must Not Change

When asking AI to adapt a task, define the protected construct.

Example:

Change surface context and numbers, but preserve the need to choose between factorisation and quadratic formula without naming either method in the question.

Adaptation and Measurement

Whenever task conditions change, ask whether the metric remains comparable.

Do not use adapted easier tasks to prove broad improvement.

Adaptation and Outcome Interpretation

If performance improves after adaptation, several explanations remain:

• the adaptation improved access;
• the task became easier;
• support increased;
• the learner genuinely improved;
• measurement changed.

Use fresh comparable verification.

The Adaptation Hypothesis

Before adapting, write one sentence:

If we change X while preserving Y, then Z should improve because…

Example:

If we move Jonas’s idea-generation practice from full essays to paragraph-scale work while preserving independent idea generation, then completion should increase without reducing transfer because current resistance is driven by task length rather than inability to generate ideas.

This makes adaptation testable.

The Adaptation Reversal Rule

Some adaptations should be reversible.

If they do not help, restore the previous form.

Reversibility lowers experimentation risk.

Low-Risk Adaptations

• change time of day;
• change paper to whiteboard;
• reduce example count;
• change surface context;
• switch verbal/written explanation when writing is not the target.

Higher-Risk Adaptations

• remove time pressure;
• remove entire representations;
• increase prompting;
• change scoring criteria;
• replace independent generation with model study;
• change the target capability.

Higher-risk changes deserve stronger monitoring.

The One-Dimension Rule

When possible, adapt one important dimension at a time.

If timing, support, representation and difficulty all change, it becomes hard to know what helped.

When Multiple Adaptations Are Necessary

Real life sometimes requires several changes.

Illness week may reduce dosage, timing and difficulty simultaneously.

Record the bundle honestly.

Do not overinterpret outcomes.

Adaptation and Research Translation

Evidence from one study rarely transfers unchanged into every learner context.

Implementation requires translation.

But translation should preserve the mechanism supported by evidence.

Adaptation and Evidence-Based Practice

Evidence-based practice is not copy-and-paste practice.

It combines:

• research evidence;
• professional judgement;
• learner evidence;
• context.

Adaptation is where those sources meet.

Adaptation and Professional Judgement

Teacher expertise matters because not every adaptation can be pre-scripted.

Professional judgement should be made visible enough to review:

• what changed;
• why;
• what evidence supported it.

Adaptation and Standardisation

Standardisation can protect active ingredients.

Over-standardisation can block necessary adaptation.

Use standard principles and flexible delivery.

The Standard–Adaptive Balance

Standardise:

• diagnostic logic;
• active ingredients;
• support boundaries;
• verification requirements.

Adapt:

• examples;
• timing;
• dosage;
• format;
• context;
• pacing.

Adaptation and Scaling

As an intervention moves across learners and settings, adaptation becomes unavoidable.

The companion Training Scale article owns how systems should change when a successful intervention expands.

Adaptation and De-implementation

Sometimes the correct adaptation is removal.

If a routine no longer earns its cost, de-implement it.

The companion Training De-implementation article owns that process.

Adaptation Failure Mode: Change Everything at Once

Outcome improves.

No one knows why.

Repair:

change fewer dimensions where possible.

Adaptation Failure Mode: Protect the Surface, Lose the Function

Worksheet looks identical.

Tutor now gives leading cues.

Mechanism changed.

Repair:

track active ingredient, not appearance.

Adaptation Failure Mode: Protect the Function Too Rigidly

Adult refuses harmless delivery changes because “that is not the programme.”

Repair:

identify loose elements.

Adaptation Failure Mode: Learner Preference Becomes Curriculum

Learner chooses only comfortable formats.

Repair:

allow choice where equivalent; preserve authentic demand where necessary.

Adaptation Failure Mode: Difficulty Always Goes Down

Every struggle triggers easier tasks.

Repair:

distinguish useful difficulty from irrelevant overload.

Adaptation Failure Mode: Difficulty Always Goes Up

Every success triggers immediate escalation.

Repair:

stabilise before increasing multiple dimensions.

Adaptation Failure Mode: Support Never Fades

Temporary cue becomes permanent.

Repair:

include a fade rule.

Adaptation Failure Mode: Support Fades by Calendar

Scaffold removed because Week 3 arrived.

Learner not ready.

Repair:

fade from evidence.

Adaptation Failure Mode: Adaptation Is Never Recorded

After six weeks no one knows what plan actually ran.

Repair:

keep one-line adaptation notes.

Adaptation Failure Mode: Every Session Is Custom

No stable routine forms.

Learner cannot internalise process.

Repair:

preserve stable core and adaptive edges.

Adaptation Failure Mode: No Adaptation at All

Same task despite repeated evidence of mismatch.

Repair:

review learner response.

Adaptation Failure Mode: Adapt to Marks Only

One bad score triggers easier work.

Repair:

inspect process, task difficulty and measurement noise.

Adaptation Failure Mode: Adapt to Emotion Only

Every complaint removes difficulty.

Repair:

listen, diagnose, preserve useful challenge.

Adaptation Failure Mode: Ignore Emotion Entirely

Plan produces repeated dread and conflict.

Repair:

inspect acceptability, burden and dignity.

Adaptation Failure Mode: Technology Drives the Curriculum

Tool only supports multiple choice, so all training becomes multiple choice.

Repair:

choose tool after defining the construct.

Adaptation Failure Mode: AI Makes Everything Easier

AI supplies method, structure and wording.

Repair:

sequence assistance after learner attempt.

Adaptation Failure Mode: Parent Adaptation Removes Independence

Parent explains every difficult question to help homework finish.

Repair:

use one agreed cue and flag unresolved work for tuition.

Adaptation Failure Mode: Tutor Adaptation Becomes Exam Coaching Only

Near exams, every lesson becomes paper drilling.

Repair:

preserve diagnostic repair and targeted weak-link work.

Adaptation Failure Mode: School Alignment Becomes Duplication

Tuition copies school homework because school topic changed.

Repair:

complement school by targeting the bottleneck or transfer gap.

The Parent Adaptation Audit

• What changed in the child’s week?
• What part of the training became difficult to use?
• Can timing change?
• Can volume change?
• Can format change?
• What learning operation must remain?
• Is parent help removing that operation?
• Is the adaptation temporary?
• When will the normal task return?
• How will we know the adaptation helped?

The Tutor Adaptation Audit

• What is the active ingredient?
• What learner/context evidence triggered adaptation?
• What exactly will change?
• What must not change?
• What outcome do I expect?
• Can I adapt one dimension?
• What validity or comparability risk appears?
• What fidelity check is needed?
• What is the reversal rule?
• When will authentic performance return?

The Learner Adaptation Audit

• What part of the task is blocking me?
• Do I need more explanation, a smaller step or a different format?
• What part do I still need to do myself?
• Am I asking for help or asking for the answer?
• Will this adaptation help me return to the real task?

The Adaptation Traffic Light

• Green: active ingredient preserved; adaptation improves access or implementation.
• Amber: mechanism probably preserved but measurement, support or difficulty changed materially; monitor closely.
• Red: target decision, generation, retrieval or transfer operation removed; the intervention has changed identity.

The Adaptation Ladder

Before changing the target mechanism, try lower-risk adaptations:

• timing;
• materials;
• format;
• dosage;
• pacing;
• support level;
• representation;
• task size;
• sequence;
• context.

If high-fidelity implementation still fails, reconsider the strategy.

The Adaptation Field Manual: Eighty Questions

• What is the training job?
• What is the active ingredient?
• What part of delivery is failing?
• Is the learner ready?
• Is the task too easy?
• Is the task too hard?
• Is difficulty relevant?
• Is representation blocking access?
• Is language blocking access?
• Is timing poor?
• Is energy low?
• Is dosage too high?
• Is dosage too low?
• Is support too little?
• Is support too much?
• Is support removing the decision?
• Is the task too large?
• Is the task too fragmented?
• Is the learner bored?
• Is boredom caused by ceiling?
• Is the learner anxious?
• Is anxiety coming from authentic demand or unnecessary threat?
• Is the plan acceptable?
• Is the plan feasible?
• Is the setting appropriate?
• Should the task move from home to tuition?
• Should it move from tuition to home?
• Should the format become oral?
• Should it become written?
• Should a diagram be added?
• Should the diagram be removed?
• Should terminology be clarified?
• Would clarification remove the target?
• Should examples become cleaner?
• Should examples become more varied?
• Should distractors become stronger?
• Should support fade?
• Should support temporarily increase?
• Should the learner attempt first?
• Should modelling come first?
• Should pre-training occur?
• Should the whole task be decomposed?
• When will it recombine?
• Should the session shorten?
• Should frequency change?
• Should spacing widen?
• Should timing become more authentic?
• Should authentic pressure be reduced temporarily?
• Should a fresh form be used?
• Should an anchor remain?
• Should monitoring intensity fall?
• Should confidence ratings be used?
• Should think-aloud be used briefly?
• Should AI generate variants?
• Should AI feedback be delayed until after attempt?
• Is AI removing generation?
• Is technology increasing friction?
• Would paper work better?
• Does the learner have a meaningful choice?
• What choice can be added safely?
• What element must remain non-negotiable?
• Did school demands change?
• Did family logistics change?
• Did tutor availability change?
• Did learner competence change?
• Did the weak link change?
• Is the evidence still current?
• What adaptation is lowest risk?
• What adaptation is easiest to reverse?
• What evidence would show benefit?
• What evidence would show harm?
• Will the outcome remain comparable?
• What fidelity risk appears?
• What should be recorded?
• When will the adaptation be reviewed?
• When will the original condition return?
• When should adaptation become replanning?
• What can be de-implemented instead?
• What will become more sustainable because of this change?
• Did we change the form or the function?

The Adaptation Design Template

Training job: what capability changes?

Active ingredient: what must remain?

Problem: what learner/context evidence shows mismatch?

Adaptation dimension: timing, dosage, support, representation, language, format, difficulty, sequence, context?

Protected function: what will still require the learner to perform the target operation?

Expected effect: what should improve?

Fidelity risk: what could be lost?

Measurement risk: what becomes less comparable?

Review point: when will we inspect the adaptation?

Reversal rule: when do we undo it?

Progression rule: when does authentic difficulty return?

Worked Adaptation Design: Mathematics

Job: method selection.

Ingredient: choose from structure before calculating.

Problem: CCA day leaves only fifteen minutes.

Adapt: reduce six full solves to three classifications + one solve.

Protect: decision remains independent.

Expected: mechanism preserved despite low capacity.

Risk: execution dose lower.

Review: Saturday fresh full item.

Worked Adaptation Design: English

Job: idea development.

Ingredient: learner-generated claim-explanation-example-consequence chain.

Problem: full essays create avoidance and consume too much time.

Adapt: weekday paragraph-scale work, weekly whole composition.

Protect: independent idea generation.

Expected: more frequent high-quality attempts.

Risk: integration practice could fall.

Review: weekly whole composition.

Worked Adaptation Design: Science

Job: evidence-mechanism explanation.

Ingredient: explicit evidence-to-mechanism link.

Problem: unfamiliar apparatus terminology blocks reasoning.

Adapt: pre-teach apparatus terms.

Protect: learner still identifies evidence and generates mechanism.

Expected: reduce irrelevant vocabulary load.

Risk: if terminology itself is assessment target, construct changes.

Review: fresh apparatus with reduced support.

The Family Adaptation Protocol

When a week changes:

• name the constraint;
• protect one or two active ingredients;
• reduce volume before mechanism;
• move tasks to higher-capacity windows;
• flag what was postponed;
• return to normal mode when capacity returns.

The Tutor Adaptation Protocol

When learner response changes:

• preserve first attempt;
• identify the mismatch;
• choose the lowest-risk adaptation;
• keep target decision visible;
• observe response;
• verify on fresh work;
• record material change.

The Learner Adaptation Protocol

When stuck:

• name what is blocking you;
• ask for the smallest change that restores work;
• keep doing the important part yourself;
• return to the full task later.

Research Foundations and Evidence Boundaries

AERO’s 15 September 2026 Context first: Tailoring implementation to each school setting is the most direct current practice anchor for this article. It explicitly frames implementation as flexible and context-sensitive rather than one-size-fits-all. The companion Staying on track guide provides the monitoring frame needed after adaptation: feasibility, acceptability, fidelity, reach and sustainability.

EEF’s Evaluate and sustain your strategy reinforces an iterative view of educational change and explicitly asks whether approaches should be sustained, adapted, scaled or stopped. EEF’s broader implementation resources emphasise active ingredients, context and knowing where to be tight and loose.

The 2026 BMC Public Health study on teacher-led adaptation and conceptual fidelity provides a contemporary empirical example of adaptation being studied alongside fidelity in a school-based programme. Its specific intervention is not academic tuition and should not be treated as direct evidence for the exact adaptations recommended here. The relevant conceptual contribution is the distinction between preserving core conceptual function and allowing teacher-led adaptation.

The evidence boundary matters. This article does not claim that every adaptation improves learning. Adaptations can help, harm or merely change the construct. The practical standard is therefore conservative: state the active ingredient, state the reason for change, adapt the smallest necessary dimension, preserve valid measurement where possible, and verify the result on fresh performance.

Final Principle

Education that never adapts stops fitting real learners.

Education that adapts without boundaries loses its mechanism.

The skill is not choosing between rigidity and flexibility.

The skill is knowing what must remain stable while everything else is allowed to move.

Change the time.

Change the number of questions.

Change the representation.

Change the response format.

Change the setting.

Change the support.

Change the sequence.

But keep asking:

Is the learner still doing the mental operation this training was built to strengthen?

If yes, adaptation may have improved the implementation.

If no, the plan has not merely adapted.

It has become something else.