Using Case-Based Reasoning to Automatically Generate High-Quality Feedback for Programming Exercises

My research explores methods for automatic generation of high-quality feedback for computer programming exercises. This work is motivated by problems with current automated assessment systems, which usually provide binary (“Correct”/“Incorrect”) feedback on programming exercises. Binary feedback is not conducive to student learning, and has also been linked to undesirable consequences, such as plagiarism and disengagement. We propose a Case-Based Reasoning approach to utilize knowledge created by human instructors in order to automatically generate comparable responses for students that submit incorrect solutions to programming exercises. Such a system would offer significant labor savings for instructors, without sacrificing the quality of student learning. Preliminary experiments have demonstrated the strength of our Case-Based Reasoning approach and its potential impact, especially in MOOCs. Further research is being conducted in order to refine the procedure and to evaluate it effect on student learning.