Automated Code Assessment for Education: Review, Classification and Perspectives on Techniques and Tools

Abstract

Automatically assessing code for learning purposes is a challenging goal to achieve. On-site courses and online ones developed for distance learning both require automated ways to grade learners’ programs to be able to scale and manage a large public with a limited teaching staff. This paper reviews recent automated code assessment systems. It proposes a systematic review of the possible analyses they can perform with the associated techniques, the kinds of produced feedback and the ways they are integrated in the learning process. It then discusses the key challenges for the development of new automated code assessment systems and the interaction with human grading. In conclusion, the paper draws several recommendations for new research directions and for possible improvements for automatic code assessment.