Adaptive Learning Systems in Programming Education: A Prototype for Enhanced Formative Feedback

Abstract

Formative feedback is crucial in programming education, yet many learning systems fall short, concentrating mostly on pinpointing errors rather than guiding learners on how to resolve them. This is particularly unhelpful for novices who often lack advanced skills like debugging. Feedback is considered more valuable when it addresses error causes rather than just symptoms. However, this is challenging using only conventional methods like unit testing. Identifying error causes requires detailed information about both the error and the learner. Our proposed prototype introduces a new approach to integrating programming exercises into adaptive learning systems. It directly categorizes student code into so-called answer classes using a combination of static and dynamic code analysis. When integrated with data derived from a learner model, this approach enables tailored feedback that lowers the barrier to learning programming while keeping motivation high.