Misconception-Based Peer Feedback: A Pedagogical Technique for Reducing Misconceptions

Abstract

Developing high quality pedagogical materials and techniques is a challenging but important task. We leverage prior work identifying student misconceptions and difficulties in introductory computing courses to design misconception-based feedback (MBF) to address these difficulties. In MBF, peers working in pairs use prompts to guide their discussion of a recently completed coding assignment. A human autograder group (HAG) simulates the behavior of a typical autograder program, supplying only test cases and their correct outputs, allowing us to factor out the effect of the medium (computer vs. human). Participants completed conceptual pre-tests and post-tests that asked them to explain their reasoning, and we captured screen and audio recordings of the sessions. Our mixed-methods analysis looked for statistical differences in frequency counts of misconceptions and qualitatively analyzed audio/video data and language used in pre/post-test written responses to look for explanations of these differences. We use a mixed-methods approach: looking for statistical differences in frequency counts of misconceptions and qualitatively analyzing audio/video data and language used in pre/post-test written responses for explanations of these differences. Significant benefits of MBF were seen on questions that required students to comprehend differences between pass-by-value and pass-by-reference. Not only did these questions show a greater reduction in misconceptions for the MBF group than the HAG group, but the qualitative analysis provided evidence that the students' improvement in language and understanding of the concepts from pre-test to post-test was directly tied to the MBF intervention. This study presents a promising, low resource technique to address misconceptions about several important computer science concepts.