The Tutor Handbook Vol No.0065 | The Concurrency Problem

The Tutor Handbook · Volume 0065 · Series ID THB-0065

The tutor changes one thing.

The school changes three.

The parent changes the evening routine.

The learner starts using a new revision app.

A new teacher introduces a different answer format.

Then the marks move.

What caused what?

The Concurrency Problem appears when several relevant parts of the learner’s world change during the same observation window, making it difficult to tell which change improved performance, which change created cost, which change was irrelevant, and which apparent effect is really the interaction between two or more changes.

The previous volume, The Tutor Handbook Vol No.0064 | The Change Queue, asked how the tutor orders several valid changes instead of activating all of them at once.

This volume handles the harder reality:

The tutor can control the tuition route, but cannot freeze the rest of the learner’s life. How should evidence be read when several systems move anyway?

Quick Read

Students live inside several learning systems at once.
School, tuition, home routines, peers, tools, assessments and the learner’s own behaviour can all change concurrently.
Concurrency makes simple before-and-after stories dangerous.
A better result after tuition change does not prove tuition alone caused it.
A worse result after tuition change does not prove the change failed.
Record major concurrent changes that could plausibly affect the same outcome.
Do not record everything. Track only changes with meaningful causal relevance.
Preserve timestamps. Order matters.
Separate exposure changes from support changes, measurement changes and learner-state changes.
Look for repeated receipts across different conditions rather than one dramatic result.
Use school-generated work as an independent observation when appropriate.
When several changes collide, reduce certainty before increasing intervention.
Freeze non-essential tuition changes if the outside environment is already highly unstable.
Do not fight school changes merely to protect a clean experiment.
Do not let a parent add a parallel intervention without understanding what the tutor is trying to observe.
AI and digital tools can silently change the amount of support available between lessons.
Three-student classes create shared instructional changes but individual learner responses.
Concurrency is an interpretation problem, not a governance owner. Who gets to decide remains a different question.
The learner should increasingly report external changes that affect the route.
The ethical goal is not perfect causal certainty. It is appropriately cautious attribution and better next decisions.

1. What This Volume Owns

This volume owns interpretation under simultaneous change.

It asks:

When several relevant conditions changed during the same period, what can the tutor responsibly infer from the learner’s next performance?

It does not own authority between parent, school, tutor and learner. The wider eduKateSG article How Governance Works | Who Gets to Decide? owns decision rights.

It does not own handoff design. How Interfaces Work owns the broader problem of information and responsibility crossing boundaries.

This page owns the evidence problem created when those systems all change near the same time.

2. The Learner Is a Shared Environment

A student may receive:

school teaching;
tuition;
parent support;
peer explanations;
online videos;
AI assistance;
revision guides;
new assessment demands;
new personal study routines.

Each source can alter the learner’s behaviour.

The tutor observes the result after all of them have had some opportunity to act.

3. Before and After Is Not the Same as Cause and Effect

The tutor introduces mixed algebra practice on Monday.

On Tuesday, school revises the same algebra topic.

On Wednesday, the parent supervises homework more closely.

On Friday, the student scores better.

The score is good news.

It is weak evidence for the claim “the tuition change caused the improvement.”

4. Concurrency Can Produce False Credit

A tutor changes the worksheet sequence and later sees improvement.

What the tutor does not know is that the learner’s school teacher also reteached the prerequisite and the parent stopped interrupting study every fifteen minutes.

Claiming full credit would exceed the evidence.

5. Concurrency Can Produce False Blame

The tutor fades a prompt.

The next paper is worse.

During the same week, the school paper became substantially harder and the learner slept badly before the test.

The prompt fade may still have been too aggressive.

But the result does not isolate that explanation.

6. Concurrency Can Produce Real Interaction

Sometimes two changes only work because they occur together.

A tutor reduces prompting at the same time school increases independent classwork.

The learner improves because both environments now demand the same independent operation.

The combined effect is not a nuisance to eliminate. It is part of the real educational system.

7. The Goal Is Not Laboratory Purity

Tuition is not a controlled experiment.

School cannot pause while the tutor measures.

Parents cannot suspend family life so one variable stays clean.

The tutor’s job is not to manufacture perfect causal identification.

The job is to notice enough concurrent change to avoid confident stories that the evidence cannot support.

8. Record Only Plausibly Relevant Changes

Do not create a diary of the learner’s entire life.

Track changes that could plausibly affect the outcome currently being interpreted.

If the question is why algebra timing improved, relevant concurrent changes may include:

new timed school practice;
reduced calculator dependence;
extra tuition practice;
a change in prompt policy;
a new test format;
substantial changes in sleep or workload.

A new art project may be irrelevant unless it materially changed available study capacity.

9. Time Order Matters

A change that happened after the observed improvement cannot explain the earlier improvement.

This sounds obvious.

Without timestamps, retrospective narratives regularly violate it.

Record enough sequence to distinguish:

school reteach → tuition change → first independent attempt → home routine change → delayed retest.

10. Distinguish Measurement Change From Learning Change

The paper changes from untimed topical questions to a timed mixed section.

Scores fall.

The learner may not have become weaker.

The measurement condition became more demanding.

Concurrency analysis begins by separating changes in what the learner can do from changes in how that capability was sampled.

11. Distinguish Support Change From Knowledge Change

A learner suddenly answers more accurately.

During the same period, a parent began checking every homework step.

The output improved.

The learner’s independent knowledge may or may not have improved to the same degree.

Support level belongs in the interpretation.

12. Distinguish Exposure Change From Strategy Change

The learner starts using retrieval practice and scores better.

But the learner also doubled the total amount of contact with the material.

Was the improvement caused by retrieval, more exposure, or both?

The responsible answer may be “the current evidence cannot separate them yet.”

13. Distinguish Learner-State Change From Intervention Change

A learner can mature, recover from illness, regain sleep, become less anxious, or simply become more familiar with school routines.

Not every performance change is intervention-driven.

The tutor does not need to explain every state change. The tutor needs enough humility to keep non-tuition causes open when the evidence supports them.

14. Distinguish Task Difficulty From Capability

A learner gets 78% on one paper and 62% on the next.

If the second paper contains more unfamiliar problem forms, the score movement is not a clean measure of deterioration.

Compare like with like where possible, and use multiple observations when the tasks are not directly comparable.

15. The Concurrency Log

A lightweight concurrency log can contain five fields:

date or period;
change;
source: school, tuition, home, learner, tool or assessment;
likely affected outcome;
confidence that it matters: low, moderate or high.

This is enough to stop a surprising amount of retrospective simplification.

16. Keep the Log Event-Based, Not Minute-Based

Do not record every homework session.

Record meaningful changes:

new teacher;
new assessment format;
new tuition route;
new parent support rule;
new AI/tool use;
new major timetable constraint;
major illness or recovery affecting study;
new independent habit that changes exposure or support.

17. Use Independent Receipts

If tuition changes method selection, look for evidence beyond the tuition worksheet.

Does the learner select methods better on school work?

On a delayed fresh set?

Under a different representation?

Independent receipts do not prove a single cause, but they make it harder for one local support condition to masquerade as broad learning.

18. Repetition Across Contexts Raises Confidence

One improvement after five concurrent changes is ambiguous.

The same improvement returning across several later tasks, after some of those temporary conditions disappear, is stronger evidence that the underlying capability really changed.

Concurrency does not make learning unknowable.

It makes repeated evidence more valuable.

19. Freeze Non-Essential Tuition Change During External Instability

The learner changes school teacher, enters exam week and begins a new household schedule.

This may be the wrong week for tuition to redesign the study system unless the redesign is itself necessary for stability.

When the environment is already moving, holding a known-good tuition route can improve interpretability and reduce burden.

20. Do Not Freeze Necessary Support for the Sake of Measurement

The learner is struggling under a new school demand.

The tutor notices a clear need for support.

Do not say, “We cannot help because it would contaminate our data.”

Education serves the learner before it serves causal neatness.

Make the necessary change and lower the certainty of later attribution.

21. The Change Queue and Concurrency Work Together

The tutor can reduce self-created concurrency by using the Change Queue.

External concurrency will still happen.

That distinction matters.

The tutor should control what can be controlled without pretending the rest of the world is static.

22. School Changes Often Dominate the Environment

A new chapter can change the prerequisite load.

A new teacher can change explanation style and marking expectations.

A prelim period can change timing pressure and total study load.

School is not noise around tuition. It is one of the learner’s primary environments.

23. Home Changes Can Alter the Measurement

A parent starts sitting beside the learner during every homework session.

Completion improves.

Independent initiation may not.

Home support can be valuable and still change what later homework evidence means.

24. Peer Changes Can Matter

A learner begins studying with a capable friend.

Explanations become better.

Is the learner producing the reasoning or borrowing it?

Peer learning is not a contaminant to remove. It is a support condition to understand.

25. Digital Tools Can Create Invisible Concurrency

A student begins using AI to check explanations between lessons.

The tutor sees cleaner homework.

If the tool use is invisible, the tutor may infer a capability gain that has not yet been demonstrated independently.

The answer is not automatic prohibition.

Name the tool’s job, then preserve independent checks.

26. Assessment Changes Can Look Like Learner Changes

A school moves from topical quizzes to integrated papers.

Scores fall.

This may reveal a real method-selection weakness that the old format hid.

The learner did not necessarily lose knowledge. The assessment began sampling a different capability.

27. Tutor Style Changes Are Concurrent Conditions Too

A new tutor uses different wording, pacing and examples even when the official route is unchanged.

Volume 0063 separated teaching style from route change.

Concurrency analysis remembers that style differences can still affect performance during a handoff period.

28. The Concurrency Card

Outcome being interpreted: What changed?
Observation window: Over what period?
Tuition change: What did tuition alter?
School changes: What relevant teaching, assessment or workload conditions changed?
Home changes: What relevant support or routine changed?
Learner changes: What self-directed behaviour changed?
Tool changes: What digital or AI support changed?
Task changes: Did difficulty, format or timing change?
Likely contributors: Which changes plausibly affect the same outcome?
Confidence: What can be claimed strongly, moderately or only tentatively?
Next discriminating check: What future task would reduce uncertainty?

29. Use Confidence Language

Strong language:

Across three delayed independent tasks, method selection remained better after the route change even when school topic cues were absent.

Moderate language:

The change is consistent with the improvement, but school reteaching occurred in the same period.

Weak language:

The learner improved, but several conditions changed and the contribution of each is currently unclear.

Good tutoring can say “we do not know yet” without losing authority.

30. The Learning Claim Still Applies

The Learning Claim already limits what can be said about progress, mastery and cause.

Concurrency explains one major reason those claims need limits: multiple changing conditions create alternative explanations.

31. The Decision Record Should Capture Concurrent Conditions

A decision record does not need a full causal analysis.

It should note major concurrent changes when they matter.

Timing routine began this week. School also moved into timed mixed revision. Improvement cannot yet be attributed to tuition alone.

32. Class 0 · Homework Helper Concurrency

Homework completion can improve because the learner became more organised, because school assigned less, because a parent helped more, or because the tutor installed a better routine.

Class 0 should distinguish completion from independence before claiming success.

33. Class 1 · Explainer Concurrency

A concept becomes clearer after tuition explanation.

The school teacher taught the same concept differently the next day.

The correct educational response is not to compete for explanatory ownership.

Use the convergence. Then test whether the learner can reconstruct the concept independently.

34. Class 2 · Drill Builder Concurrency

Practice volume rises in tuition and school at the same time.

Accuracy improves but fatigue also rises.

The Drill Builder should read total practice load rather than treating tuition dose as the only dose.

35. Class 3 · Diagnostic Tutor Concurrency

The Diagnostic Tutor should become more cautious when the observation environment changes rapidly.

A new error pattern after a new assessment format may reflect the format, not a new knowledge defect.

Change the evidence condition before changing the diagnosis when possible.

36. Class 4 · Route Designer Concurrency

The Route Designer must account for other routes already acting on the learner.

School is running one sequence.

Home may be enforcing another.

The tutor should avoid adding a third sequence that fights both unless the evidence demands deliberate repair.

37. Class 5 · Performance Coach Concurrency

Performance outcomes are especially exposed to concurrent variables: sleep, paper difficulty, time limits, stress, topic mix and recent practice.

Do not diagnose a timing collapse from one unusually hard paper.

Do not celebrate one fast paper if accuracy conditions were easier.

38. Class 6 · Learning Architect Concurrency

The Learning Architect sees interacting systems.

The danger is over-explanation.

Seeing many plausible causes does not mean the tutor knows which one dominated.

Class 6 maturity means preserving multiple hypotheses without paralysis.

39. Repair Mode Under Concurrency

If school simultaneously reteaches the same prerequisite the tutor is repairing, that is often beneficial.

Do not reject the help.

Instead, lower causal attribution and focus on whether the repaired capability now survives independently.

40. Alignment Mode Under Concurrency

Alignment Mode naturally contains more school-tuition interaction.

The goal is not separation.

The goal is coherence: school demand and tuition support should reinforce one another without making it impossible to see what the learner can carry alone.

41. Frontier Mode Under Concurrency

Frontier work can be especially vulnerable to hidden concurrent cost.

The learner may handle extension well in tuition while school workload simultaneously increases.

Do not interpret local success as proof that the total system has spare capacity.

42. Three-Student Tutorials Add Shared Concurrency

One group change can affect all three learners differently.

The tutor introduces more mixed questions.

Alicia benefits.

Beatrice becomes overloaded.

Ciara is unaffected because the mixed task does not target her active job.

Shared change does not imply shared effect.

43. Alicia · Mathematics: School Reteach and Tuition Mixing

Alicia begins mixed algebra practice in tuition.

School independently reteaches algebraic representation during the same fortnight.

Her method selection improves.

The tutor records both changes and avoids claiming that mixed practice alone produced the gain.

The next useful check is a delayed mixed set using unfamiliar surface forms. If the gain persists, the tutor can raise confidence in the capability change even if causal credit remains shared.

44. Beatrice · English: New Teacher, New Marking Language

Beatrice’s school teacher begins using a more explicit evidence-explanation marking code.

Tuition simultaneously tightens evidence discrimination.

Her answers improve.

The tutor should welcome the alignment, not compete over whose language caused the change.

The practical question is whether Beatrice can use the underlying reasoning without needing either adult’s code visible.

45. Ciara · Science: AI Enters the Homework Loop

Ciara begins using an AI tool to compare her Science explanations with model causal chains.

Homework quality improves sharply.

The tutor does not ban the tool merely to protect measurement.

Instead, the tutor records the support change and adds fresh in-class explanation tasks without the tool. If quality survives, confidence in independent learning rises.

46. Denise · Additional Mathematics: Harder Paper, Worse Score, Better Learning

Denise moves into a harder school paper while tuition increases timing pressure.

Her score falls slightly.

But item-level analysis shows she attempts more questions, preserves accuracy on familiar methods and recovers better after difficult items.

The total mark alone hides that the measurement changed and one performance capability improved.

47. Emily · Studying: Home Schedule Changes Mid-Plan

Emily’s study planner appears to fail for two weeks.

During the same period, family transport shifts and two evenings disappear.

Do not conclude that the planning method is ineffective before accounting for the changed capacity constraint.

The correct response may be replanning under the new world, not abandoning the planning mechanism.

48. Faith · Independence: Parent Rescue Masks the Fade

Tuition fades Faith’s task-restatement prompt.

At home, a parent begins restating every difficult question.

Homework remains smooth.

The learner has not actually experienced the intended reduction in total support.

Concurrency analysis reveals why the tutor’s fade appears to have no effect.

49. Parent Communication: Ask About Changes, Not Just Results

When marks move, useful parent questions include:

Did the paper format change?
Did school reteach the topic?
Did homework support change?
Did study time change materially?
Did a new tool enter the process?
Did sleep, illness or workload change?

These questions protect the family from overreacting to one output.

50. Tutor Communication: Do Not Claim Exclusive Credit

A trustworthy tutor can say:

Her method selection is clearly better. Tuition changed the practice condition, but school also revisited the same topic, so I would not assign the full gain to one source. What matters now is that the capability is showing up independently across different tasks.

This language is not modesty for appearance.

It is evidence discipline.

51. Learner Communication: Help Them Notice Their Own Concurrent Changes

Ask the learner:

What changed in how you studied, what changed in school, and what changed in the task?

A learner who can answer begins to separate personal strategy, external support and task conditions.

That is a metacognitive skill.

52. Research Foundation: Formative Assessment Is Ongoing Adaptation

The OECD’s Unlocking High-Quality Teaching describes formative assessment and feedback as an ongoing process of setting learning goals, diagnosing learning, providing feedback and adapting to student thinking.

Concurrency is one reason adaptation needs judgement. Student responses arrive from a moving system, not a frozen classroom.

53. Research Foundation: Monitoring Progress Should Inform the Next Move

AERO’s Monitor Progress guide emphasises checking what students know and can apply, identifying gaps and adjusting teaching where necessary.

The concurrency addition is simple: when several conditions moved, the tutor should interpret the evidence with that context before choosing the adjustment.

54. Research Foundation: Feedback Must Be Acted On

The EEF’s feedback evidence summary emphasises specific information linked to learning goals and opportunities for pupils to act on feedback.

When feedback arrives from several sources, the tutor should look for coherence. Two adults can both give high-quality feedback and still create a poor combined system if the learner cannot reconcile the instructions.

55. Research Foundation: Metacognition Supports Independence

EEF’s 2026 guidance on metacognition and self-regulation highlights planning, monitoring and evaluating learning, with modelling and guided practice before independent use.

A learner who can report concurrent changes is beginning to monitor not just performance, but the conditions under which performance was produced.

56. Common Failure: One Result, One Story

The mark rises.

The tutor says the new strategy worked.

The mark falls.

The parent says tuition failed.

Both stories may be too simple.

57. Common Failure: Record Everything and Understand Nothing

A massive log can create false precision.

Track only changes that plausibly touch the outcome being interpreted.

The purpose is better judgement, not surveillance.

58. Common Failure: Treat School as a Confound to Eliminate

School is not noise.

The learner is supposed to function there.

If school and tuition changes reinforce one another, that may be an educational success even if causal credit cannot be separated cleanly.

59. Common Failure: Compete for Ownership of Improvement

Teacher says school caused it.

Tutor says tuition caused it.

Parent says home discipline caused it.

The learner becomes evidence in an adult credit contest.

Use the result to improve the next decision instead.

60. Common Failure: Ignore Tool Assistance

Homework quality rises after AI enters the process.

If the tutor does not know the tool is present, the learner model becomes inaccurate.

Tool use should be visible enough to interpret the work honestly.

61. Common Failure: Overcorrect for Concurrency

The tutor sees many possible contributors and refuses to make any judgement.

That is also a failure.

Education rarely offers perfect certainty.

Use graded confidence, repeated evidence and reversible next steps.

62. The Tutor’s Concurrency Audit

What outcome am I trying to interpret?
What exactly changed in tuition?
What changed in school during the same period?
What changed at home?
What changed in the learner’s own study behaviour?
Did any new tool enter the process?
Did the assessment format or difficulty change?
Did support level change?
Did total exposure change?
Did sleep, illness or total workload change materially?
Which changes plausibly affect this outcome?
What is the time order?
What can I claim with high confidence?
What is only consistent with the evidence?
What remains genuinely uncertain?
What next task would reduce uncertainty?

63. The Parent’s Concurrency Audit

Did anything else change when the marks changed?
Was the new paper comparable with the old one?
Did school reteach the topic?
Did we change how much help we give at home?
Did the learner begin using a new resource or AI tool?
Did sleep or total workload change?
Are we asking one result to prove too much?
What evidence would make the interpretation stronger?

64. The Learner’s Concurrency Audit

What did I do differently?
What did my teacher do differently?
What did my tutor do differently?
What help did I use outside lessons?
Was this task harder, easier or simply different?
Did I have more or less time?
Did I feel more prepared because I had seen similar work?
Can I still do it when the extra support is removed?

65. What the Concurrency Problem Is Not

It is not a reason to refuse adaptation.
It is not a reason to freeze school or home support.
It is not a demand for laboratory control.
It is not a licence to say “we can never know.”
It is not the owner of governance or decision rights.
It is not the owner of interfaces or handoffs.
It is not a reason to deny genuine learner progress.
It is a discipline for lowering attribution confidence when multiple plausible changes acted on the same learner during the same period.

66. The Ethical Standard

Adults like clean stories.

“This worked.”

“That failed.”

“The child improved because of us.”

The learner’s real world is messier.

Professional tutoring does not need exclusive causal ownership. It needs enough evidence to protect the learner from false conclusions, enough humility to keep alternative explanations alive, and enough courage to act when the next responsible move is clear even though the past can never be reconstructed perfectly.

Evidence and Connected Reading

Final Compression

The learner lives in more than one system.

School changes.

Tuition changes.

Home changes.

Tools change.

The learner changes.

Then performance changes.

Do not force one cause onto a many-cause world.

Record the important concurrent changes.

Keep time order.

Separate measurement from capability.

Separate support from independence.

Use repeated independent receipts.

Lower certainty when the world is moving.

Do not lower care.

The Concurrency Problem does not ask the tutor to control the whole world. It asks the tutor to remember that the whole world is allowed to affect the learner—and to make claims, changes and next decisions that remain honest under that reality.

That is the Concurrency Problem.

That is Tutor Handbook Volume 0065.

Small Group Tutorials

The Tutor Handbook Vol No.0065 | The Concurrency Problem — How a Tutor Reads Learning When School, Tuition, Home and the Learner Are All Changing the System at Once