Intelligent Feedback for Individualized Introductory Programming Exercises

Abstract

This paper proposes a method to improve programming instruction in an interdisciplinary Bachelor of Science and Technology program by integrating Artificial Intelligence (AI) into the code assessment process. Programming skills are a fundamental part of Engineering and Computer Science degrees. Using the Virtual Programming Lab plugin for Moodle, students complete parameterized exercises that generate unique problem instances and test cases for each individual. An embedded AI system, powered by a selection of seven widely used Large Language Models, analyzes students' code and provides automated feedback upon each correction request. This AI-based approach promotes scalable assessment while supporting student autonomy. Preliminary results indicate significantly positive student perceptions across key pedagogical dimensions, including idea generation, clarity of explanations, learning autonomy, and feedback timeliness.