Automatically detectable indicators of programming assignment difficulty

Abstract

The difficulty of learning tasks is a major factor in learning, as is the feedback given to students. Even automatic feedback should ideally be influenced by student-dependent factors such as task difficulty. We report on a preliminary exploration of such indicators of programming assignment difficulty that can be automatically detected for each student from source code snapshots of the student's evolving code. Using a combination of different metrics emerged as a promising approach. In the future, our results may help provide students with personalized automatic feedback.