Observed, Inferred, Unknown | Marie Curie Series | How to Read Learning Evidence Carefully

Quick Read

A student pauses before answering.

That is an observation.

“She does not understand” is an inference.

Whether the pause came from confusion, careful checking, uncertainty about wording, tiredness, anxiety, distraction or a deliberate attempt to compare two methods may still be unknown.

The Marie Curie Series treats this distinction as a safety rail: observe first, infer cautiously, and keep the unknown visible until better evidence arrives.

The One-Sentence Answer

A trustworthy learning dashboard separates what happened, what the evidence reasonably suggests, and what we still do not know before deciding what to do next.

Why This Matters So Much in Education

Education invites fast stories.

A child scores 58. We say the child is weak.

A learner works slowly. We say the learner does not understand.

A student asks for help twice. We say the student is dependent.

A careless-looking error appears. We say the child did not check.

Sometimes those explanations are right. Sometimes they are only convenient.

The problem is not that adults make inferences. We have to. Teaching requires judgement. Parenting requires judgement. Students themselves need to interpret what is happening.

The problem begins when an inference quietly changes category and starts being treated as if it were an observation.

“The student paused” is evidence. “The student is weak” is a claim that needs more evidence.

That difference changes the quality of the next educational decision.

The Three Boxes: Observed, Inferred, Unknown

Box	Question	Example
Observed	What actually happened?	The student used the wrong operation in four of six ratio questions.
Inferred	What might this pattern mean?	The learner may be misrepresenting the relationship before calculation.
Unknown	What have we not yet established?	Whether the break is language, representation, ratio concept, or a learned keyword habit.

This is intentionally simple. The value comes from refusing to skip the third box.

Unknown does not mean failure. Unknown tells us what evidence would be useful next.

Observation Should Be Narrower Than We Think

Good observation describes the event without smuggling the explanation into the sentence.

Too interpretive	Cleaner observation
The student was careless.	The student copied 36 as 63 in two separate steps.
The student does not understand inference.	The student selected a plausible inference but did not support it with textual evidence.
The student is dependent.	The student began two unfamiliar questions only after a teacher prompt.
The student has poor memory.	The student could not retrieve the method three days after successful immediate practice.
The student is weak in Science.	The student recalled the facts but did not connect them into a causal explanation.

The cleaner version is more useful because it is more repairable.

“Careless” is a broad label. A transposition error is a specific event. Once the event is specific, we can ask whether it repeats, under what conditions, and what reduces it.

Inference Is Necessary — But It Must Stay Correctable

A tutor cannot teach by collecting observations forever.

At some point we need a working explanation.

The strongest educational inference usually has three properties:

1. It explains more than one observation. A repeated pattern is stronger than one isolated mistake.
2. It predicts what should happen next. If the proposed weak link is real, a carefully chosen new task should expose it.
3. It can be contradicted. If new evidence does not behave as predicted, the explanation must be allowed to change.

This is the spirit of the Evidence Loop. An educational explanation is useful because it helps choose the next test or repair, not because it sounds certain.

Unknown Is Not an Empty Box

Adults can feel pressure to provide an answer immediately.

Parents want to know why the mark fell. Students want to know what is wrong. Tutors want to help quickly.

But a premature explanation can create unnecessary work.

Suppose a Secondary 1 student is making algebra errors. If we assume “weak algebra” and assign more algebra, we may miss the fact that the errors are almost entirely produced by unstable negative-number operations. The visible problem is upstairs. The weak link is earlier.

Writing “unknown: signed-number control not yet isolated” is not indecision. It is a precise statement about what should be checked next.

One Event Can Support Several Stories

Consider a student who leaves three homework questions blank.

Possible stories include:

• the learner did not understand the topic;
• the learner understood the concept but could not start independently;
• the learner ran out of time;
• the questions required a method not yet taught;
• the learner avoided difficult work;
• the learner was unsure whether a particular method applied;
• or the learner made a sensible decision to stop rather than fabricate an answer.

The blank page is real. The story requires investigation.

A Mark Is Observed; Its Meaning Is Inferred

A 62 is real.

But “the student knows 62% of the subject” is not what the mark literally means.

The mark came from one paper with a particular topic mix, question design, time limit and marking scheme. The result may contain conceptual weakness, execution weakness, language difficulty, time management, an unusual concentration of errors, or a combination.

Our reflective article The Mark Arrives After the Problem already explores the delay between cause and visible result. The Curie Series adds a stricter reading discipline: the mark is a sensor return, not a complete diagnosis.

Slow Is an Observation Only After We Define the Task

Even “slow” needs context.

Slow compared with what? On a familiar task or a new one? While reading, choosing, calculating or checking? With a teacher beside the learner or independently?

A learner may be slow because basic operations are unstable. Or because the learner is carefully verifying. Or because the question is genuinely difficult. Or because they have not yet compressed several beginner operations into one mental chunk.

This is why the existing reflection Slow Is Not a Diagnosis fits naturally beneath the Curie architecture.

Observed, Inferred and Unknown Across Subjects

English

Observed: the learner gives a plausible inference but cannot identify supporting evidence.
Inferred: the learner may be reasoning from general impression rather than textual grounding.
Unknown: whether the difficulty is evidence location, question interpretation, vocabulary or explanation structure.
Next useful check: ask the learner to highlight the exact phrase that changed the interpretation.

Mathematics

Observed: method choice is correct but sign errors recur.
Inferred: signed-number control may be fragile.
Unknown: whether the errors come from negative-number operations, copying, algebraic notation or rushing.
Next useful check: isolate the sign operations outside the larger algebra problem.

Science

Observed: the learner states the correct concept but ignores the experimental evidence.
Inferred: the learner may be retrieving a memorised answer instead of reasoning from the setup.
Unknown: whether the student failed to notice the evidence, did not know how to use it, or misunderstood the question demand.
Next useful check: ask the learner to explain only what can be concluded from the given observations.

Additional Mathematics

Observed: the learner can execute a familiar function routine but cannot begin when the representation changes.
Inferred: procedural familiarity may be stronger than conceptual transfer.
Unknown: whether the break is function meaning, algebraic manipulation, representation or method selection.
Next useful check: compare equivalent representations before adding more procedural practice.

Not Everything Important Is Easy to Observe

The Curie Series is not a promise that every valuable feature of a learner can be measured.

Curiosity, courage, intellectual honesty, resilience, imagination and judgement can show themselves through behaviour, but none should be collapsed into one crude score.

Sometimes the correct dashboard entry is qualitative:

The learner now stays with unfamiliar problems longer before asking for help.

That may be educationally important even when it does not produce an immediate mark increase.

A Better Parent Language

Parents can use the three boxes without becoming teachers.

Instead of:

“You are careless.”

try:

“I can see the same copying error happened three times. Do you know what was happening at that step?”

Instead of:

“You forgot everything.”

try:

“You could do this last week but not today. Let’s find out what part did not survive.”

The language becomes calmer because it describes a correctable event rather than attacking identity.

The Learner Should Eventually Use the Same Discipline

One purpose of Tutor is to help the learner internalise better self-observation.

Instead of:

“I am bad at algebra.”

a learner can learn to say:

“I understand the method when the form is familiar, but I am not yet choosing it reliably when the question changes.”

That sentence is longer. It is also far more useful.

How This Connects to Voyage and Darwin

Voyage supplies developmental context. A behaviour that is expected during early formation may become a meaningful signal later when greater independence should have emerged.

Darwin uses the evidence to ask whether the method or learner strategy now needs to adapt.

Curie keeps both honest by separating what is actually visible from what we think the visible evidence means.

Boundaries

• Do not medicalise ordinary learning variation. Classroom observations do not establish medical or psychological diagnoses.
• Do not overread one event. A single mistake may be noise.
• Do not ignore qualitative evidence. Not everything worth noticing is numerical.
• Do not convert uncertainty into a label. “We do not know yet” can be the most accurate answer.
• Do not let a dashboard replace conversation. Learners often provide evidence by explaining what they were trying to do.

The RFE of Observed, Inferred, Unknown

Keep every important claim about a learner proportionate to the evidence, clearly separated from what remains uncertain, and open to correction when new evidence returns.

Frequently Asked Questions

Does this mean tutors should never make judgements?

No. Teaching requires judgement. The point is to make the judgement traceable to evidence and revisable when the evidence changes.

How much evidence is enough?

Enough to justify the educational decision being made. A small low-risk practice change may need less evidence than a major claim about the learner’s capability.

Is uncertainty a weakness in the system?

No. Visible uncertainty prevents false confidence and tells us what to observe next.

---

Marie Curie Series · Tutor · Tutorial · Evidence Loop · Voyage · Darwin

eduKate Sengkang | Small groups of up to 3 | 1.5-hour tutorials | Properly Taught Kids Shine a Bright Light Into the Future.