Assessment alignment: a proposal until a named human approves it

Assessment alignment is a judgement, not a fact: how to align items with learning outcomes, what CEFR alignment means, and who approves each tag.

Assessment alignment is the match between what a programme says learners should be able to do and what its assessment actually asks of them — and for a single item, every alignment is a judgement that someone has to own. Tagging an item "B1" or "Bloom's: analyse" looks like stating a fact. For programme directors and language centre managers, this post covers what alignment claims, how to align assessment with learning outcomes item by item, what CEFR alignment means, and the record NUADU keeps: a proposal, any change, and a named approver.

What is assessment alignment?

Assessment alignment means each task measures the outcome it is meant to evidence: the same content, at the same cognitive demand, at the level the framework names. Course designers call the wider version constructive alignment, in which intended outcomes, teaching activities and assessment all point at the same thing.

Why should assessment tasks be aligned to the learning outcomes? Because a result is only as meaningful as the match between the task and the outcome it is reported against. A learner can look weak on a task that asks for something the outcome never mentioned, and strong on one that asks for less than it promised.

For an item, alignment takes the form of a tag, and every tag is a claim:

  • A Bloom's level claims a cognitive demand: whether the question asks for recall, application or analysis.
  • A CEFR level claims the item is designed to elicit performance at that level, as the framework's can-do descriptors describe it.
  • An EQF level locates a learning outcome on a qualifications ladder.

All three are judgements: reasonable, evidenced and open to disagreement. That is why "proposal" is the right word for a tag nobody has confirmed yet, and why the record has to say who confirmed it.

The proposal does not become the alignment until a human says so.

How do you align assessment with learning outcomes?

Start from the outcome, decide what evidence would show it, and read each item against that outcome and its level before the item is used. A rubric written first does much of this work, because its criteria are the outcomes in observable form — the method is in our guide on how to write a rubric.

On NUADU, alignment is proposed when an item is created or imported, as part of exam authoring on the Create pillar. There are two routes to a proposal, and both end with the same person.

  • By hand. At review, the author or QA reviewer places the item against the framework and approves the alignment manually. No AI is required.
  • With the Standards mapper. When AI assistance is switched on, this workflow assistant proposes an alignment — Bloom's, CEFR, EQF or your custom framework — and sends it to the author or QA reviewer.

The mapper's human/never pair on NUADU's AI in assessment page is short: the alignment is approved by a named human, and the assistant never writes a mapping without approval. Major frameworks are supported, custom institutional frameworks can be added, and like every AI capability on the platform the mapper is opt-in: everything works with it switched off.

Alignment says nothing about whether an item works. Whether an aligned item separates stronger from weaker candidates is a statistical question, answered after the sitting by item analysis in assessment.

What is CEFR alignment?

In its strict sense, CEFR alignment describes a whole test: the claim that its results can be read against the CEFR levels, established through a documented linking procedure that the Council of Europe sets out on its page on relating examinations to the CEFR. It is a property of an examination, earned through a process, not a label added to it.

A CEFR reference on a single item is a smaller claim. It says the item is designed to elicit performance at a level, and that an educator agrees it does. It does not say that a learner who answers correctly holds a B2. With AI assistance on, NUADU's Content Scanner drafts CEFR-tagged candidate items from your materials for you to review, edit and approve; the tag arrives as part of the draft and is a proposal like any other.

NUADU's English assessments are CEFR-referenced, reported per skill with CEFR levels and can-do statements; what that looks like for an institution choosing a CEFR-referenced English test is covered separately, and the product lines are on the online English tests for institutions page. The boundary, in that page's own words: "NUADU delivers and scores English assessments for placement, progress, preparation and institutional requirements. It is not a national certification body and does not issue state-recognised language certificates — it complements them."

The record an external examiner can read

The alignment audit record shows three things, and it sits on the item rather than in a separate file.

  • The proposal: what was suggested, and by which route.
  • The change: whether the approver accepted the proposal, moved it or replaced it.
  • The approver: a named educator or QA reviewer.

An illustrative case shows how small the record is. The mapper proposes B2 for a reading item. The reviewer reads the item against the B2 can-do descriptors, judges that answering it needs nothing beyond B1, and moves the tag. The record now holds the proposal (B2, from the mapper), the change (to B1) and the reviewer's name. Nobody has to reconstruct that decision from memory when the examiner asks.

Because every item carries a named approver and an audit record, a reviewer reads alignment item by item, not as one sweeping sign-off. An accreditation reviewer who asks who decided that an item evidences B2 gets a name. One who asks whether the human simply accepted the proposal can see whether it changed. One who asks what the mapping looked like last year finds it in the item's version history, because the bank versions every item, with its approvers and its change history.

Three things are deliberately absent. No alignment is written without approval. No tag is presented as a verdict on a learner. And no accuracy figure is claimed for the mapper's proposals: the record, item by item, is the evidence that matters. The alignment then travels with the item into delivery, marking, reporting and certification, so a result is reported against the alignment a named educator confirmed, not against a second mapping made later.

→ Talk to us about your assessment workflow (30 min)