Refining Competency-Based Grading in Undergraduate Programming Courses

This paper investigates different methodologies for implementing competency-based grading (CBG) in programming courses. In competency-based grading (CBG) students focus on mastering the individual topics, therefore, students must show their level of competence in the various topics or sub-topics by the end of the semester. For example, a topic User-defined Functions is graded on a scale 0 No Submission, 1 Beginner, 2 Competent, and 3 Proficient in one of the courses used in this study. A characteristic of CBG is flexibility; to accomplish flexibility these courses combined a flipped-classroom and unlimited resubmissions so that students had the freedom to move through the entire course at their own pace, and so that students can learn from their feedback. One disadvantage of this approach was the burden of grading on the instructor; with a large number of submissions to grade, it can be difficult to give students feedback quickly. In this study we started with the initial implementation of competency-based grading in two Junior-level programming courses in the Electrical Engineering and Computer Engineering programs in 2020, this initial version included unlimited resubmissions and the flipped-classroom. This study also tested the following refinements to the initial CBG approach (1) only one resubmission per assessment and (2) unlimited resubmissions with a live-coding lab demonstration. We compared the performance - overall grades and numbers of students receiving grade D, F, or Withdraw - for the initial implementation and our two refinements. In the case of allowing only one resubmission per assessment, we observed no overall change in performance. This showed that moving from unlimited to one resubmission does not affect performance while still allowing some flexibility and the opportunity for students to master the topic after receiving feedback. Additionally, the instructors' burden is reduced with fewer submissions to grade. In the case of adding in the live-coding lab demonstration, instructors were able to quickly grade the lab exercise and give feedback directly to the students in real-time, decreasing the time taken to grade the lab exercises. Students also perform better when they have a live-coding demo or exam since they know that they will have to answer questions on their code in real-time. In future work we plan to combine live-coding demonstrations with limited resubmissions.