Nearly every rubric includes a row about evidence, yet few define what good evidence actually means. Descriptors like "uses strong evidence" leave students guessing and graders improvising. The thinking in unSpun, which treats evidence as something to be tested rather than simply collected, suggests a more precise way to write these criteria.

Strong evidence is not just evidence that is relevant to the topic. It is evidence that comes from a credible source, is accurately represented, is specific enough to be checked, and genuinely supports the claim it is attached to. Breaking the concept into these components lets a rubric measure each part and gives students a clearer path to improvement.

Consider separating the evidence row into two or three smaller rows. One can address the quality of sources, another the accuracy and specificity of the details, and another the way evidence is connected to the claim through explanation. This avoids the common situation where a student with excellent sources but weak explanation receives a middling score with no indication of why.

Describe Levels With Observable Features

Performance levels should distinguish between actual features of the writing, not between adjectives. For source quality, a high level might describe sources that are credible, current, and independent of each other, while a lower level might describe reliance on a single source or on sources of unclear origin. Students can compare their own work to these descriptions and see exactly where they stand.

• Source quality: credibility, independence, and relevance of the sources used
• Accuracy: facts, numbers, and quotations are represented faithfully
• Specificity: details are precise enough for a reader to verify
• Connection: the writer explains how each piece of evidence supports the claim
• Balance: significant contrary evidence is acknowledged and addressed

Avoid Counting Sources as a Proxy for Quality

Many rubrics reward the number of sources, which encourages padding rather than depth. A student with three excellent, carefully analyzed sources may deserve a higher score than one with ten superficial ones. Describe the quality of use and variety of perspective instead of setting a minimum count as the main measure.

If a minimum number is helpful for logistical reasons, keep it separate from the quality criteria. Treat it as a requirement to meet, not as evidence of excellence. This distinction helps students understand that the point is the thinking behind the sources, not the size of the bibliography.

Test the Rubric on Real Essays

A rubric looks good in theory until it meets actual student writing. Score a handful of real essays and notice where you hesitate or where two levels seem indistinguishable. Those moments point to descriptors that need revision.

Tools such as GraideMind let you load a custom rubric and see how it performs across a batch of essays, which can reveal unclear language or criteria that overlap. If the feedback it generates seems off, that usually signals a rubric problem rather than a tool problem. Refining the language improves both automated and human scoring.

Share the Rubric Early

A rubric is a teaching tool as well as a grading tool, and it works best when students see it before they start writing. Walk through the evidence criteria in class using a sample paragraph and ask students to score it. The discussion that follows often does more to build understanding than any lecture.

Revisit the rubric after grading and invite student feedback about which parts were unclear. Their perspective can reveal confusing language you had not noticed. Over several rounds, you will develop criteria that are precise, fair, and genuinely useful for improving student writing.