Education · feedback · judgement · correction

Evaluation

Producing useful information about a capacity without confusing evaluation, measurement, ranking, certification and sanction.

Introduction

Evaluation serves several functions that must not be confused.

Grading, ranking, certifying, diagnosing a pedagogical difficulty and helping someone learn do not require exactly the same devices.

In the interdisciplinary education report, evaluation functions primarily as a way to make visible what is understood, what remains fragile, what type of support is needed and what should be revisited next.

Central thesis: Integrative evaluation produces a situated judgement about a real capacity, distinguishes the dimensions it observes without flattening them into a single score, keeps its criteria discussable and turns the result as far as possible into information for correction and transfer.

Central question

What does this evaluation actually allow us to know — and what decision does it legitimately support?

The same grade can correspond to very different understandings, strategies and degrees of autonomy.

In short

Evaluation is not only assigning a value. Understanding, strategy, execution, verification, transfer, level of support and stability over time must remain distinct. Some dimensions are qualitative and should not be aggregated artificially. Formative evaluation remains useful when it leads to feedback, another attempt and a precise point to revisit.

01 · Grading

Evaluating is not only grading.

A grade can condense a performance, but it rarely explains what was understood, what remains fragile or what type of help is still needed.

A useful evaluation therefore first seeks to produce information about a real capacity rather than a global verdict.

A summarised performance is not yet an understood trajectory.

02 · Measurement

Measurement and evaluation are not identical.

Measurement assumes a magnitude and a unit. Evaluation may also require qualitative, situated and reasoned judgement about a production, strategy or capacity.

The education report refuses to convert heterogeneous dimensions automatically into one score. Understanding, strategy, execution, verification and transfer do not become more comparable simply because numbers are assigned to them.

03 · Same result, different capacity

The same success can hide different capacities.

Two learners can obtain the same final result with different levels of support, strategies and degrees of autonomy. Conversely, a wrong final answer can contain robust partial understanding.

Evaluation therefore benefits from preserving a trace of the type of help needed and the exact point where action became disorganised.

04 · Error

Error should produce feedback, not identity.

The report distinguishes conceptual, execution and attention errors while requiring that these categories not become identities of the learner.

Evaluation becomes formative when error indicates what should be revisited and permits another attempt rather than immediately closing judgement.

05 · Alignment

Evaluate what was actually taught.

Evaluation is not only about the learner. The teacher must also ask whether the instruction was clear, time sufficient, decisive cues actually taught and the task genuinely testing the target capacity.

A mismatch between teaching and assessment can produce a weak result without correctly informing us about learning.

06 · Second attempt

A second attempt changes the function of judgement.

The report recommends a new attempt after feedback and asks the learner what they would change. Evaluation then stops being only terminal and becomes part of the learning cycle.

This return helps distinguish a temporary difficulty from what remains genuinely unstable.

07 · Transfer

Transfer must be verified.

Success on a familiar exercise is not enough to establish that a capacity is available elsewhere. The report proposes evaluation across several contexts and checking stability over time.

The task may therefore need to change form in order to test whether the learner recognises the relevant structure without depending entirely on the initial model.

08 · Different institutional functions

Ranking, certification and learning are different functions.

An institution may need to rank, certify or decide. Those functions exist, but they should not be confused with the formative function of evaluation.

A device designed for selection does not automatically provide the same information as one designed to support correction of learning.

09 · AI

AI should not automate educational judgement.

The report allows AI to help compare traces, vary exercises or prepare questions, but rejects AI deciding orientation or sanction on its own.

Observed data, hypothesis and inference must remain distinct. Fluent output or automatic classification is not sufficient evidence about a learner.

10 · Contestability

Evaluation must remain contestable and revisable.

Because it produces judgement, evaluation creates power. Its criteria must therefore be explicit enough to be understood, discussed and corrected.

An isolated result should not become a durable identity when later work shows something different.

11 · Proof-act

Make the result usable for correction.

A good evaluation allows a precise answer: which capacity is available, under which conditions, with what level of support and what should be revisited next?

Its proof-act is not the score itself, but the ability of feedback to orient correction, a second attempt or genuinely observable transfer.

Condensation question: Does this evaluation actually show what is available and what must be revisited — or does it mainly produce a ranking?