Every semester, we deliver a capstone course on software engineering where students undertake a real-world project in three iterations. By the end of each iteration, students are graded in several dimensions: software quality, project management, and peer assessment. The latter is the only grade assigned individually; therefore, students who are penalized by their teams (e.g., for being perceived as low contributors to the team effort) are not severely affected. This results in little incentive for improvement, which potentially jeopardizes the overall quality of the project outcome. Envisaging to promote team cohesion, we devised a new grading rule: if the peer assessment of a student is lower than a threshold, that would be their final grade in the iteration. This paper reports the results of studying the effectiveness of the proposed rule. We recorded peer assessments over six consecutive semesters: (1) the first three as the baseline measure; (2) the semester where we introduced the new grading rule; and (3) the following two semesters, as a contrast. When the rule was first introduced, peer assessments resulted low and heavily spread at the beginning, but they consistently improved toward the end of the semester. When the instructional team already trusted the rule and explicitly emphasized its application, peer assessments consistently grew along the semester but resulted in fewer outliers. The study results show that exposing peer assessments earlier on helps promote team reflection. They also made evident the positive impact of teamwork for producing quality products in a software engineering capstone course.
Read full abstract