Rubrics: How the Output Will Be Judged

Unit ID: PAW-M06-U03 Estimated active time: 20-25 minutes

A rubric is success criteria with levels

Module 2 introduced success criteria: checkable statements about what a good answer contains.

A rubric goes one step further and describes what weak, acceptable and strong look like on each criterion. That extra step is what makes it useful for judging rather than only for asking.

Two uses, and both matter

Before — include it in the prompt, so the tool knows the standard.

After — use it to review the output, yours or someone else's, consistently.

The same document does both, which is why it repays the ten minutes it takes to write.

A rubric you can reuse

OUTPUT RUBRIC

Criterion        Weak                  Acceptable            Strong
--------------   -------------------   -------------------   --------------------
Accuracy         claims not in source  all traceable         traceable + gaps named
Completeness     misses a key point    covers the points     covers + says what is
                                                             out of scope
Usefulness       generic advice        specific to the case  specific + names the
                                                             next action
Honesty          hidden assumptions    assumptions listed    assumptions listed and
                                                             ranked by impact
Tone             wrong register        appropriate           appropriate and
                                                             consistent throughout

Five criteria is a good working size. Ten becomes a form nobody fills in.

Ask the tool to score against it

> Score your own draft against the rubric above. For any criterion below "strong", say what you would change.

The self-score is not evidence — it is generated like everything else. But it reliably surfaces the weakest criterion, which is a useful starting point for your own review.

Keep the rubric shorter than the output

If the rubric is longer than the thing it judges, it is not being used; it is being written.

For a short email, three criteria. For a report that will be circulated, five or six.

Mini practice

Write a five-criterion rubric for a kind of output you produce often.

Score two past outputs against it.

If both score identically on a criterion, that criterion is probably not discriminating — replace it.

Mark complete