Intelligent Goal-Oriented Feedback for Java Programming Assignments

Within computer science education, goal-oriented feedback motivates beginners to be engaged in learning programming. As the number of students increases, it is challenging for teaching assistants to cater to all the doubts of students and provide goals. This problem is addressed by intelligent visual feedback which guides beginners formulate effective goals to resolve all the errors they would incur while solving a programming assignment.

Most current automated feedback mechanisms provide feedback without categorization, prioritization, or goal formulation in mind. Students may overlook important issues, and high priority issues might be hidden among other issues. Also, beginners are not well equipped in formulating goals to resolve the issues provided in the feedback.

In this research, we address the problem of providing an effective, intelligent goal-oriented feedback to student's code to resolve all the issues in their code while ensuring that the code is well tested. The goal-oriented feedback would eventually implicitly navigate the students to write a logically correct solution. The code feedback is summarized into four categories in the descending order of priority: Coding, Student's Testing, Behavior, and Style. Each category is further classified into subcategories, and a simple visual summary of the student's code is also provided.

Each of the above-mentioned categories has detailed feedback on each error in that category to provide a better understanding of the errors. We also offer enhanced error messages and diagnosis of errors to make the feedback very useful.

This intelligent feedback has been integrated into Web-CAT, an open-source automated grading tool developed at Virginia Tech that is widely used by many universities. A user survey was collected after the students have utilized this feedback for a couple of programming assignments and we obtained promising results to claim that our intelligent feedback is effective.