Personalised Programming Education with Knowledge Tracing

In traditional programming education, addressing diverse student needs and providing effective and scalable learning experiences is challenging. Conventional methods struggle to adapt to varying learning styles and offer personalised feedback. AI-based Programming Tools (AIPTs) have shown promise in automating feedback, simplifying programming concepts, and guiding students. Their widespread adoption is hindered by limitations related to accuracy, explanation, and personalisation. Conversely, AIPTs tailored for expert programmers, such as ChatGPT and Copilot, have gained popularity for their productivity-enhancing capabilities, but they still fall short in terms of personalisation, neglecting individual students' unique knowledge and skills. Our research aims to leverage AI to create AIPTs that offer personalised feedback through adaptive learning, accommodating diverse student backgrounds and proficiency levels. In particular, we explore using Knowledge Tracing (KT) to anticipate specific syntax errors in programming assignments, addressing the challenges novices face in acquiring syntactical knowledge. The findings suggest the KT's potential to transform programming education by enabling timely interventions for students dealing with specific errors or misconceptions, automating personalised feedback, and informing tailored instructional strategies.