“Careless.” “Weak in Mathematics.” “Not motivated.” “Not a language child.”
These phrases feel useful because they are compact. The problem is that they are often too compact to guide good teaching.
A working hypothesis is deliberately less final:
“The learner may understand the concept but lose control when the question surface changes. Let us test that.”
That sentence can be wrong. That is one of its strengths.
A label closes. A hypothesis opens.
“Careless” does not tell us what to teach next.
“The learner is accurate on routine questions but often drops units and signs during the final two minutes of timed work” does.
The first phrase sounds like a property of the child. The second describes a performance pattern under conditions.
Once the pattern is specific, we can design a next observation. Does the same error occur untimed? Does a checking routine reduce it? Does the problem appear after cognitive load rises? If not, the explanation changes.
The aim is not to avoid judgement
Good teaching requires judgement. Tutors have to decide what to teach, what to ignore for now, when to simplify, when to increase difficulty and when a student is ready for less help.
The discipline is to keep the judgement at the right size.
“This student currently struggles to retrieve multiplication facts fluently” is a useful educational claim.
“This student is weak at Mathematics” is much larger.
The more specific the model becomes, the less reason there is to turn it into identity.
Working hypotheses protect us from the first plausible story
A low result can support several explanations at once.
A child who cannot answer a Science question may lack the concept, misunderstand the changed variable, know the mechanism but fail to express it, or simply misread the command word.
If we decide too early that the problem is “weak content”, we may spend weeks reteaching knowledge the learner already has while the actual break remains untouched.
A working hypothesis keeps alternatives alive long enough for evidence to distinguish them.
Good hypotheses are testable
Compare these two statements:
“She lacks confidence.”
and:
“She predicts low performance even on topics where recent independent work is strong.”
The second statement creates something we can compare: prediction against actual performance.
Or compare:
“He is dependent.”
with:
“He can continue once the tutor chooses the first representation, but rarely chooses it independently.”
Now we can remove the prompt and observe what happens.
The hypothesis should update when the child changes
Children are moving systems.
A learner may need close support in January and much less in June. A transition may temporarily destabilise performance. A foundation repair can change the whole profile of a subject. Examination pressure can expose a weakness that ordinary homework did not reveal.
If the description never changes, even when the evidence changes, it has stopped functioning as an educational model and started functioning as a label.
This is why Which Student Are You? treats learner states as temporary rather than permanent identities.
Parents can use the same discipline at home
When a pattern appears, try replacing identity language with observation language.
| Instead of | Try |
|---|---|
| “Lazy” | “Starts easily when the task is clear, but avoids tasks where the first step is uncertain.” |
| “Careless” | “Most lost marks are late execution errors after the main method is correct.” |
| “Weak in English” | “Understands the passage but gives vague answers that do not use enough evidence.” |
| “No confidence” | “Predicts failure more often than recent performance warrants.” |
| “Dependent” | “Usually waits for an adult to identify the first step.” |
This is not softer language for its own sake. It is more operational language.
A tutor should be willing to say “we do not know yet”
Uncertainty is not a failure of expertise.
Sometimes the evidence genuinely does not distinguish two explanations. A learner may perform poorly in a timed task because retrieval is weak, because time pressure changes strategy, or because the material is not as secure as it seemed.
The responsible answer can be:
“We have two plausible explanations. One small comparison would tell us more.”
That is often more useful than a confident diagnosis built from weak evidence.
When labels are legitimate, they belong to the right professional context
Some formal labels have defined professional purposes in education, health or developmental support. Ordinary academic tutors should not casually manufacture or replace those assessments from tuition performance.
Our boundary is straightforward: public academic observations can describe learning behaviour and performance. They should not be inflated into medical, psychological or developmental diagnoses.
What this page does not mean
This page does not mean every statement about a learner must be tentative forever. Repeated, comparable evidence can support strong conclusions about a current capability. If a student repeatedly cannot retrieve a fact or execute a skill independently, teaching should respond rather than hiding behind uncertainty.
The rule is proportionality: make the claim no larger than the evidence, and leave a route for revision when new evidence appears.
Where to go next
If you are collecting the evidence that should come before a hypothesis, use What Should Parents Bring to a Tuition Consultation? and What Can One School Paper Tell Us—and What Can’t It Tell Us?.
For the deeper public evidence architecture, the Marie Curie Series owns observation, uncertainty and evidence-testing mechanisms. For the full set of parent decision routes, return to the Parents’ Guide.
eduKate Sengkang · A useful educational model should be specific enough to act on and provisional enough to be corrected.
eduKate Sengkang 