Design Rubrics Around Evidence and the Task
Connect criteria to the assignment goal and inspect real responses for unclear wording. A rubric does not automatically ensure objectivity or rater agreement.
Publication & corrections: infostoa · Editorial standards
Individual author and specialist reviewer information is not provided for this article.
Decide what evidence matters first
Before writing a rubric, state what knowledge or performance the task is meant to reveal. If a science explanation should use evidence, check that the rubric values the link between a claim and data rather than letting spelling dominate. Moskal (2003) recommends aligning rubric criteria with task requirements and stated goals, then using anonymized example papers to clarify levels and ambiguous wording.
Write observable criteria and performance descriptions
A criterion such as “use evidence” is hard to apply. A more observable description might say, “Cites a relevant data point and explains how it supports the claim.” Describe differences between levels through evidence visible in the response, not judgments about a student’s character or effort. This is a design illustration, not a validated rubric.
Try the wording on actual responses
If possible, compare a few anonymized responses that show different levels and flag ambiguous language. When raters disagree, discuss what evidence each criterion is meant to capture and revise unclear descriptions instead of forcing agreement. Analytic rubrics are not always preferable, and there is no universal number of criteria. Rubrics can support interpretation but do not remove bias or guarantee reliability.
Use a sample rubric to make the criteria concrete
Carnegie Mellon’s Eberly Center describes a rubric through criteria, performance descriptors, and levels, and notes that rubrics can support feedback and help instructors identify class-wide strengths and weaknesses when adjusting instruction. Choose an example for a similar assignment and see how its three parts map to visible work, then keep only criteria your own task needs. A sample is a starting point, not a ready-made answer key.
Review questions
Does each criterion assess evidence required by the objective? Can a learner tell what to improve from the level descriptions? Can raters explain a difference by pointing to the stated evidence?
Moskal, Developing Classroom Performance Assessments and Scoring Rubrics—Part II (2003), ERIC ED481715, pp. 2–4 · Carnegie Mellon Eberly Center: Creating and Using Rubrics