Wednesday, November 16, 2016

Consensus-based grading: more accuracy and less stress

We just introduced a new method for grading, called Consensus-based grading. In consensus-based grading, the grades students assign are compared on each rubric item separately, to obtain a consensus for each rubric item independently. Subsequently, the consensus grades for each entry are combined in an overall grade for the submission.

Rubric-based consensus increases precision, as errors are eliminated on each rubric entry, before the totals for the entries are combined. 

A more streamlined, less stressful student experience

Corresponding to this change, we made a few other changes to CrowdGrader, based on our experience of running large classes:

  • Students see only consensus grades. If you choose rubric-based consensus, students will see only the conseusus grades for their submission, rather than the grades individually submitted by students. In our experience, this increases student satisfaction, as students no longer see anxiety-inducing, but harmless, outlier grades in their evaluation.
  • The review feedback phase is no longer present. We found that the review feedback phase was only mildly useful, and often worked as a dis-incentive for students to submit honest evaluations, in spite of the fact that CrowdGrader does eliminate tit-for-tat behavior in feedback.