Towards understanding the effective design of automated formative feedback for programming assignments

Background and Context: automated feedback for programming assignments has great potential in promoting just-in-time learning, but there has been little work investigating the design of feedback in this context.Objective: to investigate the impacts of different designs of automated feedback on student learning at a fine-grained level, and how students interacted with and perceived the feedback.Method: a controlled quasi-experiment of 76 CS students, where students of each group received a different combination of three types of automated feedback for their programming assignments.Findings: feedback addressing the gap between expected and actual outputs is critical to effective learning; feedback lacking enough details may lead to system gaming behaviors.Implications: the design of feedback has substantial impacts on the efficacy of automated feedback for programming assignments; more research is needed to extend what is known about effective feedback design in this context.