Rubric Template and Grade Weight Check

Read the assignment the way students will, and list each thing it asks them to do. Each distinct demand that shows up in the finished work becomes a criterion. If the assignment asks for a recommendation, a feasible recommendation is a criterion. "Effort" is not, because no one can see it on the page.

Keep the criteria about learning, not about following instructions. Susan Brookhart, an assessment researcher at Duquesne University, warns that criteria like "has three sources" misdirect student effort and mis-measure learning¹. She offers "uses a variety of relevant, credible sources" instead. Her 2018 review of rubric studies in higher education concludes that appropriate criteria are the key to effective rubrics.

Next, check the list against your course outcomes. Fix any criterion no outcome calls for, and any outcome the assignment claims but no criterion measures, before you publish the rubric. If the outcomes themselves are vague, sharpen them with a learning outcomes template first.

Choose the number of levels first. Three, four and five levels were the most common in Brookhart's review. The University of Hawaiʻi at Mānoa's assessment office suggests an even number, such as four or six, because with an odd one the middle tends to become the catch-all².

Write every cell as something a reader can point to in the work. Brookhart recommends descriptive language over verdicts like "excellent" and "poor," because a description shows students where they are and where to go next. In ASCD's Educational Leadership in 2000, Heidi Andrade argued for describing the problems students run into³, such as not stating a claim clearly enough for a reader to understand.

Give the middle levels the most care. Clear excellence and clear failure are easy to place; the hard calls come in between. Cornell's Center for Teaching Innovation says the criteria must clearly differentiate one performance level from another⁴. Make each level differ from the next by one thing a reader can see.

Weight each criterion by how much it matters, and make the weights add up to 100. Carnegie Mellon's Eberly Center says grading against criteria weighted by the importance of the assignment's objectives⁵ helps keep your standards from drifting over time.

Turn levels into points in equal steps: on four levels, a criterion worth 20 earns 5, 10, 15 or 20. Put little weight on anything a submission system already enforces, such as file names or citation format. Almost everyone earns those points, so they barely separate anyone.

Hand the rubric out with the assignment, as the Hawaiʻi guide advises, so students know what counts before they start. In Brookhart's review, only 56% of the studies reported using rubrics with students; the rest used them only for grading. State the assignment's share of the course grade in your course syllabus too.

Before you grade a whole class, score a few papers with a colleague or teaching assistant. Hawaiʻi's assessment office calls this norming²: scoring the same work separately, comparing, and talking it through until everyone applies the rubric the same way. Grading alone? Score a few papers again a week later and compare.

1. 1

   Pick sample papers

   Choose about six that span the range. The Hawaiʻi guide suggests two low, two middle and two high.

2. 2

   Score them separately

   Each grader scores every criterion alone, before any discussion.

3. 3

   Compare row by row

   Find the criterion where your levels differ most. If you agree on the rest, its wording is the problem, not the grader.

4. 4

   Fix the wording and keep anchors

   Rewrite that row, then keep one paper per level as an anchor, asking the writer's permission before students see it.

Carnegie Mellon suggests tallying how many students fell below an acceptable level on each criterion⁵ to see which skills need more teaching. To test the weights, add up the points every paper lost on each criterion and divide by the total lost across all criteria. Like your published weights, these shares total 100.

Citation and formatting was weighted at 8% but drew only 1.1% of the points lost. The submission portal enforced it, so 21 of 24 papers reached the top level. Analysis drew 35.1% against its 25% weight, and its scores spread from top to bottom: that is where the class separated.

Close only part of a gap, and only for the next class, since one class is a small sample. Here, moving 6 points from formatting and 2 from structure lifts analysis to 33%, still below its 35.1%. Regrading work already returned is a separate decision, under your institution's policy.

If several people graded, compare them criterion by criterion. In the example, the two graders' averages differed by 6.27 points, 5.42 of them on one criterion whose top level read "well-supported and realistic." Everywhere else they agreed within a tenth of a level, so the wording was the problem.

The template on this page works through all of this for one assignment, and River can run both checks on your own scores. If the course is still taking shape, start with the course design template.