Diagnostic teaching begins with a simple but demanding principle: do not choose the next teaching move until you have a reasonable model of the learner’s present state.
A wrong answer does not tell you enough. A low score does not tell you enough. A learner saying “I don’t understand” does not tell you enough. Even a correct answer may not tell you enough, because the answer may have been produced through guessing, prompting, copying, pattern recognition or temporary familiarity.
Diagnostic teaching asks a more precise question: what mechanism produced this performance? Only after that question is explored can the teacher decide what to explain, what to practise, what to retrieve, what to simplify, what to vary, what to challenge and what support to remove.
This article is part of the eduKate Sengkang How Teaching Works series and connects directly with How Learning Works, How Understanding Works in Learning, How Cognitive Load Works in Learning, How Metacognition Works in Learning, and the measurement principle that a supported answer is not the same measurement as an independent answer.
What Diagnostic Teaching Is
Diagnostic teaching is the continuous process of gathering evidence about a learner, forming a working explanation of the learner’s current state, selecting an instructional response, and then checking whether that response produced the expected change.
It is not merely a diagnostic test administered before teaching begins. It is not a label attached to a student. It is not a fixed profile such as “weak in mathematics,” “visual learner,” “careless,” or “not motivated.” Diagnostic teaching is dynamic because learner states change from task to task and over time.
A learner may understand fractions but fail to retrieve multiplication facts quickly enough to complete a multi-step problem. The same learner may appear strong in algebra when examples are blocked by topic but fail when the problem type is mixed. A student may explain a science concept accurately in conversation yet fail to write it under examination conditions. Another may produce excellent homework only because every step was guided at home.
Diagnostic teaching therefore treats performance as evidence that requires interpretation.
The Core Diagnostic Cycle
- Observe: collect a piece of learner performance.
- Localise: identify where success begins to break down.
- Differentiate: distinguish between plausible causes.
- Hypothesise: form a working explanation of the learner’s state.
- Intervene: choose the smallest useful teaching move.
- Re-test: check whether the predicted change occurred.
- Update: keep, refine or reject the diagnosis.
- Fade: remove support and verify independent performance.
This is why diagnosis is not separate from instruction. Every teaching move can also be a test of the teacher’s model. If a short retrieval prompt immediately restores performance, the problem may have been access rather than understanding. If another example does not help but a prerequisite explanation does, the problem may sit deeper in the knowledge structure. If performance collapses only after support is removed, the issue may be dependence on scaffolding rather than lack of comprehension.
1. Start With the Learner’s Actual Performance, Not the Teacher’s Assumption
Teachers naturally form expectations. A student has a history, a grade, a reputation and a pattern of strengths and weaknesses. Those expectations can be useful, but they can also become a filter that distorts diagnosis.
The learner who “always rushes” may actually be confused by the wording of one question type. The learner who “does not revise” may have revised repeatedly using methods that produce familiarity rather than retrieval. The learner who is “weak in English” may understand the passage but lack the vocabulary needed to express the inference. The learner who “cannot do algebra” may be failing because negative-number fluency is unstable.
Diagnostic teaching therefore begins with observable evidence before broad character explanations. The first job is not to decide who the learner is. It is to locate what happened in this performance.
2. A Wrong Answer Is the Beginning of Diagnosis, Not the End
Two learners can produce the same wrong answer for completely different reasons.
One learner may never have learned the relevant concept. Another may know the concept but misread a condition. Another may choose the wrong formula because several similar formulas are not well discriminated. Another may understand the method but make an arithmetic slip. Another may know everything required but lose track because the task overloads working memory.
If the teacher responds to all five learners with the same explanation, four may receive the wrong intervention.
The practical implication is important: after an error, the teacher should often ask one or two discriminating questions before teaching again. “Tell me what this question is asking.” “Which information matters?” “What rule did you think applied?” “What would you do first?” “Can you solve the same idea with simpler numbers?” “Can you explain why this answer cannot be correct?”
Each answer narrows the possible state.
3. The Major Learner States Diagnostic Teaching Must Distinguish
A useful diagnostic system does not need hundreds of labels, but it does need enough categories to prevent different problems from collapsing into one.
- Missing knowledge: the learner has not acquired a required fact, concept, rule or procedure.
- Weak retrieval: the knowledge may exist but cannot be accessed reliably when needed.
- Misconception: the learner holds an internally coherent but incorrect model.
- Discrimination failure: the learner knows several ideas but cannot tell when each one applies.
- Execution weakness: the learner understands the method but performs steps inaccurately.
- Working-memory overload: the task requires more simultaneous coordination than the learner can currently manage.
- Language or representation barrier: the concept may be understood in one form but not another.
- Transfer failure: the learner succeeds in familiar contexts but does not recognise the same structure elsewhere.
- Prompt dependence: performance succeeds only when external cues organise the thinking.
- Metacognitive miscalibration: the learner inaccurately judges what is known, unknown or worth practising.
- Motivational withdrawal: the learner has the capacity to act but is not investing sufficient effort under the present conditions.
- Knowledge-structure gap: a visible problem is being caused by an earlier missing prerequisite.
These categories can overlap. A learner with weak retrieval may become overloaded because too much basic knowledge must be reconstructed. A misconception may be strengthened by repeated practice. A language problem can appear to be a conceptual problem. Diagnosis is therefore a process of progressively reducing uncertainty rather than forcing each learner into a single permanent category.
4. Missing Knowledge: When the Learner Does Not Yet Have the Required Idea
Missing knowledge is the most straightforward state, but even here the teacher should be precise. What exactly is missing?
In mathematics, the missing element may be a fact, definition, relationship or procedure. In science, it may be a causal mechanism. In English, it may be vocabulary, sentence structure, genre knowledge or a model for inference. In humanities, it may be background knowledge necessary to interpret a source.
The strongest response is usually explicit teaching of the missing element followed quickly by an opportunity to use it. The teacher should avoid reteaching the entire topic if the gap is narrow. Diagnostic precision makes instruction efficient.
A useful test is to simplify the task until only the suspected knowledge remains. If the learner still cannot explain or apply that element, the diagnosis becomes stronger.
5. Weak Retrieval: When Knowledge Exists but Cannot Be Reached Reliably
Weak retrieval can be mistaken for lack of knowledge because both produce hesitation and errors. The difference matters because the intervention is different.
A learner with weak retrieval may recognise the answer once shown, complete the problem after a small cue, or say, “I knew that.” The teacher should not stop at recognition. The question is whether the learner can generate the knowledge independently and repeatedly.
Retrieval-focused teaching uses low-stakes recall, spaced reactivation, cumulative review and repeated production from memory. The aim is not to re-explain content that is already understood but to strengthen access.
This is also why immediate reteaching can sometimes hide retrieval problems. If the teacher explains first, the learner never has to reveal whether the knowledge could have been accessed independently.
6. Misconceptions: When the Learner Has a Model, but the Model Is Wrong
Misconceptions are more difficult than missing knowledge because the learner is not empty. The learner already has an explanation, rule or pattern that feels plausible.
A science learner may believe heavier objects fall faster. A mathematics learner may believe multiplication always makes numbers larger. A language learner may assume every paragraph needs the same rigid structure regardless of purpose. A student may believe rereading is effective because familiarity feels like learning.
Simply stating the correct answer may not dislodge the existing model. Stronger teaching often creates a contrast between the learner’s prediction and the observed result, then explains the mechanism that resolves the contradiction.
A diagnostic teacher therefore asks the learner to predict before correcting. Prediction exposes the active model. Without that evidence, the teacher may correct the surface answer while leaving the underlying misconception intact.
7. Discrimination Failure: When the Learner Knows the Rules but Chooses the Wrong One
Many advanced errors are not failures of procedure. They are failures of selection.
A learner may know how to calculate percentage increase and percentage points but confuse them. The learner may know several algebraic methods but not know when substitution is preferable to elimination. The learner may understand cause and correlation separately but not discriminate them in a data question. The learner may know several literary devices but identify whichever one was taught most recently.
Repeated blocked practice will not necessarily solve this because the method has already been named by the worksheet context. The learner needs contrast, mixed examples and questions that require method selection.
The teacher can diagnose discrimination by asking, “How did you decide which method applies?” A learner who can execute but cannot justify selection needs a different intervention from a learner who cannot execute the method at all.
8. Execution Weakness: When the Plan Is Right but the Performance Is Unstable
Some learners understand what to do but execute inconsistently. They lose signs, omit units, skip reasoning steps, copy numbers incorrectly or fail to check a final answer.
Calling all of this “careless” is rarely sufficient. The teacher should identify the stage at which execution breaks down. Does the learner rush because the procedure is not fluent? Is notation disorganised? Is checking absent because the learner has no checking routine? Does the learner lose information when shifting between representations?
Execution improves through targeted routines, deliberate slowing at critical points, fluency practice, structured checking and repeated correction of the same error type. The goal is not simply more practice. It is practice focused on the unstable component.
9. Working-Memory Overload: When Too Much Must Be Coordinated at Once
A learner can possess the necessary knowledge and still fail because too much has to be held and manipulated simultaneously. This is a central diagnostic problem because overload often looks like weak understanding.
The teacher can test for overload by reducing task complexity without changing the underlying concept. Break the problem into stages. Provide one intermediate result. Remove irrelevant information. Offer a diagram. Let the learner externalise a step instead of holding it mentally.
If performance improves sharply, the diagnosis may shift from “does not understand” to “cannot yet coordinate all components at full complexity.” That leads toward sequencing, worked examples, partial scaffolds and fluency-building rather than complete reteaching.
For the broader mechanism, see How Cognitive Load Works in Learning.
10. Language and Representation Barriers
Knowledge can be available in one representation but inaccessible in another. A learner may understand a science process orally but not interpret the formal diagram. A mathematics learner may solve an equation when written symbolically but fail when the same relationship appears in a word problem. A student may understand a historical event but struggle to express causation in academic prose.
Diagnostic teaching tests the same concept across representations. Ask the learner to draw it, explain it, symbolise it, give an example, compare it with a non-example or translate it into a different form.
If performance varies sharply by representation, the teacher has learned something important. The problem may not be the concept itself. It may be the translation layer.
11. Transfer Failure: When Learning Does Not Travel
Transfer failure appears when the learner succeeds in a familiar format but fails when surface details change.
This is common because learners often encode examples together with their superficial features. A formula may become attached to the page layout on which it was practised. A grammar rule may work in isolated exercises but disappear in free writing. A scientific principle may be remembered only in the apparatus used during instruction.
To diagnose transfer, preserve the underlying structure while changing the surface. Use different wording, context, numbers, diagrams, order or irrelevant details. Ask the learner what remains the same beneath those changes.
If the learner fails only when the surface changes, the next move is not necessarily more identical practice. The teacher must make the invariant structure more visible and then vary examples deliberately.
12. Prompt Dependence: When Support Is Doing More Work Than the Learner
Prompt dependence is one of the most important states in intensive teaching and tuition because helpful teachers can accidentally hide it.
A learner may succeed because the teacher points to the relevant line, reminds the learner of the formula, asks a leading question, supplies the first step, confirms every move or narrows the options until the answer becomes obvious.
The resulting performance looks strong, but the evidence is contaminated by support.
Diagnosis requires deliberate support removal. Ask the learner to repeat the same type independently after a short delay. Change the surface slightly. Remove the cue. Ask for the first step before giving any hint. If performance collapses, the learner may understand parts of the process but not yet control it independently.
This measurement problem is developed in A Supported Answer Is Not the Same Measurement as an Independent Answer.
13. Metacognitive Miscalibration: When the Learner Misjudges Learning
Learners do not merely know or not know. They also make judgments about what they know. Those judgments guide revision, help-seeking, checking and effort allocation.
A learner who feels fluent after rereading may stop too early. Another who underestimates knowledge may over-study easy material and avoid challenge. A learner may repeatedly say “I know this” because the page looks familiar, then fail when asked to retrieve the idea without notes.
Diagnostic teaching therefore compares predicted performance with actual performance. Ask the learner to rate confidence before answering. Then examine whether confidence tracks accuracy. Over time, the learner can be taught to use better internal evidence: successful retrieval, explanation, transfer and independent problem solving rather than familiarity.
See How Metacognition Works in Learning for the wider planning, monitoring and evaluation system.
14. Motivation and Withdrawal: Diagnose Before Moralising
Low effort can be real, but the explanation for low effort matters. A learner may not see value in the task. The learner may expect failure after repeated unsuccessful attempts. The task may be too difficult to enter. The learner may not know what action to take. Anxiety may consume attention. Or the learner may simply be choosing not to invest sufficient effort.
Calling all of these states “lazy” produces poor instructional information.
A diagnostic response asks whether the learner can begin when the first step is made clear, whether success changes engagement, whether challenge level is appropriate, whether the task goal is understood and whether effort changes when the learner sees progress.
Motivation should not be used to explain away an instructional problem. It should be diagnosed as part of the learning system.
15. Hidden Prerequisite Gaps
Visible failure often occurs above the true point of weakness.
A student failing simultaneous equations may have unstable manipulation of negatives. A student struggling with chemistry calculations may have weak ratio reasoning. A learner failing reading comprehension may lack vocabulary or background knowledge. A student unable to write a strong science explanation may understand the science but lack causal language.
Diagnostic teaching moves backward through the dependency chain until the first unstable component is found.
This is more efficient than repeatedly practising the final task. If the foundation is unstable, practice at the top of the chain may simply rehearse failure.
16. The Smallest Discriminating Question
A powerful diagnostic habit is to ask: what is the smallest question I can ask that separates the two most likely explanations?
If you are unsure whether a learner cannot remember a formula or does not know when to use it, provide the formula and ask for method selection. If performance improves, retrieval was part of the problem. If not, the issue may be discrimination or understanding.
If you are unsure whether a learner misunderstands a concept or is overloaded by the full problem, reduce the number of steps while preserving the concept. If the learner succeeds, complexity rather than conceptual absence may be the limiting factor.
If you are unsure whether the learner understands or is following prompts, remove the prompt and slightly vary the case.
Good diagnosis does not always need more data. It often needs better-designed data.
17. Think-Alouds: Useful but Imperfect Windows Into Reasoning
Asking learners to explain what they are thinking can reveal strategy choice, interpretation, uncertainty and misconceptions. It can show whether the learner is attending to the relevant feature or using an inappropriate rule.
But think-aloud evidence should be interpreted carefully. Learners do not always have access to the causes of their own performance. Some reasoning is too automatic to verbalise. Verbal explanation can itself change the task. A learner may also produce a plausible explanation after the fact.
For that reason, think-aloud evidence is strongest when combined with performance evidence. What the learner says should be compared with what the learner actually does.
18. Error Patterns Matter More Than Isolated Errors
An isolated mistake may be noise. A repeated pattern is much more informative.
If a learner repeatedly changes signs incorrectly during transposition, omits units only in multi-step science questions, confuses inference with summary across several texts or misreads graphs whenever the scale is non-standard, the pattern begins to reveal a stable mechanism.
Teachers should therefore maintain a small mental or written registry of recurring error classes rather than treating every wrong answer as new. This allows teaching to move from correction toward prevention.
The goal is not to create a permanent defect list for the learner. The registry should be updated as errors disappear, change or re-emerge under harder conditions.
19. Use Contrast to Diagnose Concept Boundaries
Single examples often reveal whether a learner can imitate. Contrasting examples reveal whether the learner understands boundaries.
Ask why one case belongs to a category and another does not. Compare two graphs that look similar but imply different relationships. Compare two essay paragraphs where only one actually explains the evidence. Compare two algebraic moves where one preserves equivalence and one does not.
Contrast is diagnostic because it forces attention onto the feature that controls the classification. When the learner can explain that feature, the teacher obtains stronger evidence of conceptual structure.
20. Use Prediction Before Explanation
Prediction exposes the learner’s current model before new information overwrites the evidence.
In science, ask what will happen before showing the result. In mathematics, ask whether an answer should be larger or smaller before calculation. In reading, ask which interpretation is more plausible before revealing commentary. In learning strategy, ask which study method the learner believes will produce better recall before testing it.
Prediction is especially useful for misconceptions because the learner commits to a model that can later be compared with evidence.
21. Use Confidence as Data, Not as Proof
Confidence can help diagnosis, but confidence is not knowledge.
A correct answer given with low confidence may indicate fragile retrieval or guessing. A wrong answer given with high confidence may indicate a misconception. A correct answer given with high confidence repeatedly across varied conditions is stronger evidence.
The teacher can therefore collect both accuracy and confidence. The gap between them is informative. Overconfidence calls for stronger independent testing and contrast. Underconfidence may call for evidence of successful performance and carefully calibrated challenge.
22. Diagnostic Teaching in Mathematics
Mathematics is especially suited to diagnostic teaching because errors often preserve traces of the learner’s reasoning.
A teacher can inspect not only the final answer but the representation chosen, the first operation, the transformation between lines, the use of notation, the handling of signs, and the reasonableness check.
If the learner chooses the right method but makes arithmetic errors, the intervention differs from a learner who cannot identify the method. If the learner solves correctly with a labelled chapter heading but fails in mixed practice, the problem may be selection. If a complicated algebraic expression fails but a structurally identical simpler one succeeds, the bottleneck may be fluency or working-memory demand.
A strong mathematics diagnosis therefore moves between concept, method selection, execution, fluency and transfer rather than collapsing every error into “needs more practice.”
23. Diagnostic Teaching in Science
Science diagnosis often requires distinguishing factual recall from causal understanding, data interpretation and experimental reasoning.
A learner may know that increasing light intensity can affect photosynthesis but fail to explain why. Another may explain the mechanism but misread a graph. Another may identify a trend but overclaim a causal conclusion. Another may understand variables but confuse a control set-up with “the set-up where nothing happens.”
Diagnostic questions should therefore separate knowledge of terms, mechanism, evidence and inference. Ask for prediction, explanation, data reading and transfer into an unfamiliar experimental set-up.
24. Diagnostic Teaching in English
English performance is often multidimensional. A weak answer may reflect comprehension, vocabulary, inference, evidence selection, sentence control, organisation or task interpretation.
If a learner gives a weak written response, ask for the answer orally. If the oral explanation is strong, the problem may lie in expression rather than comprehension. If the learner can explain the passage but chooses poor evidence, the bottleneck is selection. If the learner identifies evidence but cannot explain how it supports the claim, the weakness is analytical linkage.
This is why generic advice such as “write more,” “add detail” or “improve vocabulary” is often diagnostically weak. The teacher needs to know which component is limiting performance.
25. Diagnostic Teaching in Study Skills
Study problems should also be diagnosed mechanistically.
A learner who studies for hours but remembers little may be using passive rereading. A learner who knows what to do but never starts may have a planning or motivation problem. A learner who keeps revising familiar material may be miscalibrated. A learner who completes many questions without reviewing errors may have a feedback-loop problem.
The question is not simply “Does the learner study?” but “What actions occur during study, what evidence guides those actions, and what happens to performance afterward?”
This links diagnostic teaching directly to The Study Runtime and How Metacognition Works in Learning.
26. Diagnostic Teaching in a Small Group
Small-group teaching creates a major diagnostic opportunity: the teacher can compare several learners responding to the same task in real time.
One learner may answer immediately but with shallow reasoning. Another may be slower but conceptually stronger. A third may require prompting at the first step but then continue independently. These differences provide information that a single class average would hide.
The teacher can keep the same learning goal while varying the route. One learner may need a worked example. Another may need a contrast case. Another may need a transfer problem. The group remains shared, but the diagnosis controls who receives which support.
This is why three students can share a class without receiving the same teaching, and why a small class still needs a strong teaching system.
27. Do Not Diagnose From One Measure Alone
Tests, classroom observation, oral explanation, student confidence, homework and teacher judgment each reveal different aspects of performance. None is complete.
A test may show what the learner produced under one set of conditions but not why. Homework may be contaminated by support. Oral explanation may overestimate what the learner can write independently. Classroom participation may reflect confidence more than mastery. A student survey may reveal experience but not instructional effectiveness.
Strong diagnosis triangulates evidence when the decision matters. The larger the instructional consequence, the stronger the evidence should be.
Related measurement cautions appear in A Classroom Observation Rubric Can Miss the Teaching It Was Meant to Measure and Student Feedback About Teaching Is Evidence, Not a Verdict.
28. The Intervention Itself Is a Diagnostic Test
One of the most powerful ideas in diagnostic teaching is that intervention generates evidence.
If a vocabulary clarification immediately unlocks a reading question, language was likely part of the barrier. If a worked example enables the next problem but performance disappears when the example is removed, the learner may still depend on scaffolding. If one contrast pair corrects method selection across several problems, discrimination was likely the limiting factor.
The teacher therefore predicts what should happen if the diagnosis is correct. The learner’s response updates the model.
This turns teaching into disciplined inquiry rather than repeated delivery.
29. Choose the Smallest Useful Intervention
When diagnosis is precise, the teaching response can be smaller.
If the learner needs one prerequisite idea, teach that idea. If retrieval is weak, retrieve. If selection is weak, compare and interleave. If overload is the problem, reduce simultaneous complexity. If a misconception is active, surface and contrast it. If prompt dependence is the problem, fade support.
The smallest useful intervention is often better than complete reteaching because it preserves learner agency, reduces wasted time and gives clearer evidence about what changed.
30. Re-test After Teaching
A teaching move is not complete when the explanation ends. It is complete when new evidence shows whether the learner changed.
The re-test should target the diagnosed problem. If the issue was retrieval, ask for independent recall. If it was discrimination, mix nearby problem types. If it was transfer, change the surface. If it was prompt dependence, remove the support. If it was a misconception, use a new example that would have triggered the old model.
The teacher should also test after delay. Immediate success can reflect freshness rather than durable change.
31. Keep Diagnoses Provisional
A diagnosis in teaching is a working model, not a permanent identity statement.
“The learner currently appears to confuse these two concepts” is better than “the learner is weak at this subject.” “Retrieval is unstable under mixed conditions” is better than “the learner has bad memory.”
Provisional language protects against overconfidence and keeps the teacher open to new evidence.
The purpose of diagnosis is not to classify the learner. It is to improve the next decision.
32. Common Diagnostic Errors by Teachers
- Diagnosing from the final answer alone.
- Using personality labels instead of task evidence.
- Assuming a low score means missing knowledge.
- Reteaching before testing retrieval.
- Giving hints too early and destroying the evidence.
- Confusing slow performance with weak understanding.
- Confusing fluent performance with deep understanding.
- Ignoring prerequisite chains.
- Using one test as a complete learner model.
- Failing to verify whether the intervention worked.
- Keeping a diagnosis after the learner has changed.
- Measuring only supported performance.
33. A Practical Diagnostic Sequence for a Wrong Answer
- Ask the learner to explain what the question is asking.
- Ask what knowledge or method seems relevant.
- Ask for the first step without giving a hint.
- Inspect where the reasoning first diverges.
- Simplify the task if overload is plausible.
- Provide a cue only if retrieval is plausible.
- Use a contrast case if selection or misconception is plausible.
- Teach the smallest missing component.
- Ask the learner to complete a fresh example independently.
- Return later with a varied example.
This sequence is not rigid. Its value is the logic: preserve evidence as long as possible, discriminate between states, intervene minimally and verify independently.
34. Diagnostic Teaching Before a Lesson
Before a lesson, diagnosis can identify prerequisites and reduce wasted teaching.
A two-minute retrieval check may reveal that yesterday’s knowledge is not secure. A short example may show that the new topic’s prerequisite is unstable. A prediction question may expose an existing misconception before the new explanation begins.
This allows the teacher to adjust sequence early rather than discovering the problem twenty minutes later.
35. Diagnostic Teaching During a Lesson
During a lesson, diagnosis is continuous.
The teacher watches response latency, explanation quality, recurring errors, confidence, method choice and the amount of prompting needed. Questions are selected not merely to check participation but to test specific hypotheses about understanding.
The teacher may change examples, slow the sequence, remove a scaffold, return to a prerequisite or move faster because the evidence shows the learner is ready.
36. Diagnostic Teaching After a Lesson
After a lesson, the teacher asks whether the apparent improvement survived.
Homework, retrieval at the next lesson, an independent question or a transfer task can reveal whether learning was durable and whether support had been doing hidden work.
This is where diagnosis connects with How Homework Works in Learning. Homework can serve as a delayed evidence point when it is designed to reveal independent performance rather than simply increase volume.
37. The Diagnostic Teacher Builds Better Questions Over Time
Expertise includes a growing library of questions that quickly expose learner states.
Teachers learn which wrong answers are associated with particular misconceptions, which examples separate two similar concepts, which prompts reveal dependence, which transfer cases expose shallow learning and which prerequisite questions locate hidden gaps.
This library becomes one of the teacher’s most valuable professional assets because it reduces diagnostic time while increasing precision.
Learners can also be taught to generate diagnostic questions themselves. See How Question Generation Works in Learning.
38. Diagnosis Should Eventually Transfer to the Learner
The long-term aim is not for the teacher to remain the learner’s permanent diagnostic system.
Learners should gradually learn to ask their own questions: Do I know this or only recognise it? Is this an understanding problem or a retrieval problem? What was the first step where I went wrong? Which prerequisite am I missing? Can I do this without a prompt? Can I solve the idea in a new context? What evidence would prove that I have actually learned it?
This is the point at which diagnostic teaching begins to become metacognition and self-regulated learning.
The transition is central to From External Marks to Self-Regulated Learning and How Metacognition Works in Learning.
39. Diagnostic Teaching and AI
AI can support diagnosis by generating discriminating questions, varying examples, comparing solution paths and helping classify error patterns. But AI also creates a major measurement problem: it can improve the answer without improving the learner.
The diagnostic principle remains unchanged. Any AI-supported performance should eventually be followed by an independent check. If the learner cannot reproduce, explain or transfer the knowledge after AI assistance is removed, the output should not be treated as evidence of independent mastery.
See How AI-Assisted Study Works | Help That Must Leave the Learner Stronger.
40. Diagnostic Teaching Is a Form of Intellectual Humility
The teacher never sees the learner’s mind directly. Teaching operates through evidence, inference and revision.
A diagnostic teacher therefore becomes comfortable saying, “This is my current explanation, and I need one more piece of evidence.” That stance is not uncertainty in the weak sense. It is disciplined uncertainty.
The teacher avoids overclaiming, preserves alternative explanations and changes the model when evidence changes.
The Diagnostic Teaching Standard
Before choosing what to teach next, ask:
- What exactly did the learner do?
- Where did performance first break down?
- What are the two or three most plausible causes?
- What question would distinguish between them?
- What is the smallest intervention that follows from the evidence?
- What should improve if the diagnosis is correct?
- How will I test that improvement independently?
- How will I know the learning survives time and transfer?
If those questions are answered well, teaching becomes more precise, more efficient and more humane. The learner receives less unnecessary repetition, fewer generic labels and more instruction aimed at the actual barrier.
The Core Principle: Find the State Before Choosing the Move
Diagnostic teaching can be summarised in one sentence:
Do not respond to the visible error until you have asked what learner state could have produced it.
The teacher’s advantage is not simply knowing more than the learner. It is being able to interpret evidence, distinguish between causes, select an appropriate response and then verify whether the learner became stronger.
That is why diagnostic teaching sits near the beginning of the full How Teaching Works cycle. Before explanation, before practice, before feedback and before assessment, the teacher must first know what problem is actually being solved.
