Bugs as Features: Describing Patterns in Student Code through a Classification of Bugs

Code puzzles can be an engaging way to learn programming concepts, but getting stuck in a puzzle can be discouraging when no help or feedback is available. Intelligent tutoring systems can provide automatic individualized help, but they rely on having a robust and useful representation of student state. One common challenge for Intelligent tutoring systems in the programming domain is a large state space of possible students states. We propose a constrained set of features of student code based on detecting and classifying the bugs present in the code.