Scored Dimensions

A single response can be strong in some skills and weak in others. Scored dimensions let one answer be evaluated across several independent competencies — each scored dimension receives its own score from the evaluator (or AI), instead of one overall correctness score. Example: an open question measures Aptitude/Deductive Reasoning, Aptitude/Inductive Reasoning, and Aptitude/Attention to Details — the one response scores 80% / 50% / 30% respectively, and the question score is derived from those three.

Associated vs Scored Dimensions

• Associated dimensions (classic) — mirror the question's overall score: whatever the answer earns flows into each associated dimension, scaled by its multiplier. See Scoring.
• Scored dimensions — evaluated individually: each gets its own score for the same answer.

Both kinds can live on one question — a question with scored dimensions can still feed a shared taxonomy used by classic questions in the same test. The one rule is no path overlap, because dimension results roll up path segments (a score on Aptitude/Deductive Reasoning also increments Aptitude) and overlap would double-count:

• ✅ Valid: scored Aptitude/Deductive Reasoning + associated Overall/Aptitude — separate trees.
• ❌ Invalid: scored Aptitude/Deductive Reasoning + associated Aptitude — the editor blocks this and explains why.

Enabling Scored Dimensions

In the question editor, open the Dimensions chip and switch to the Scored dimensions tab. Enable the switch and add your dimensions — pick paths from your definition bank or type them freely. Each dimension gets its own points (its weight in dimension reports) and, in derived mode, a weight.

Question Score Strategies

Derived (default)

The question score is the weighted average of the dimension scores — the evaluator enters only the dimension scores, and the overall score is computed. Weight presets: Equal (simple average), Same as points (points-weighted — with this preset, nominal dimension points typed by the evaluator sum exactly to the question points), or edit weights freely. A dimension with weight 0 is analysis-only: scored and reported, but excluded from the question score. At least one dimension must have a weight above zero.

Independent

The question score is judged separately — the evaluator enters the dimension scores plus an overall score, and the overall is not computed from the rows. Use it for holistic judgment: an essay rated as a whole while Communication / Structure / Vocabulary feed the profile, or code where the question score is “does it solve the task” and Style / Efficiency / Readability are diagnostic — a non-working but beautiful solution must not earn question points. Independent mode is also the way to make all dimensions analysis-only.

Scored dimensions have no negative marking — every dimension score is 0–100% (or 0 to the dimension's points when entering nominally). The question's own positive/negative multipliers still apply to the final question score as usual.

How Evaluation Works

• Manual — the evaluation dialog shows one input per scored dimension (plus the separate overall input in independent mode). A per-dialog toggle switches all rows between percentage and points entry; an existing evaluation reopens in the mode it was entered in.
• AI — AI evaluation returns one score and written feedback per scored dimension alongside the overall result.
• Precedence — a manual score overrides the AI score per dimension; approval applies to the evaluation as a whole.
• Simulation — the question preview's scoring simulator accepts a score per dimension and shows exactly the question %, test points, and per-dimension points a real result would produce.

Things to Know

• Pending until complete — the question score stays undefined until every scored dimension is evaluated; partially evaluated answers count as “waiting evaluation”.
• Renames orphan evaluations — renaming or removing a scored dimension after sessions were evaluated orphans those dimension scores: recalculated results drop them and the answers revert to waiting. Ongoing sessions keep the configuration they started with; edits affect future sessions only.
• Rubrics are unavailable while scored dimensions are enabled — the rubric region is locked with an explanation, and such questions always evaluate through the score dialog. An existing rubric is kept but not used.
• Tests fully override — a test's copy of the question overrides its scoring including the scored dimension list; page builders can force one configuration onto every drawn question.
• A single scored dimension is valid — under derived it simply equals the question score; under independent it is one diagnostic measurement.

Reach for scored dimensions when one answer genuinely measures several things — an interview response, an essay, a case analysis. When a question measures one thing that several reports should see, associated dimensions are all you need.

When to use it