Enhancing Automated Peer Code Reviews in Software Engineering Education with Context-Aware Generative AI

Abstract

This study investigates the enhancement of peer code reviews in software engineering education through the integration of Generative Artificial Intelligence (GenAI) with contextual awareness. Previous implementations of GenAI, such as ChatGPT, lacked detailed context about the educational content and assessment goals, limiting their effectiveness. Our work-in-progress research addresses this gap by providing GenAI with additional static and dynamic contextual information, including project overviews, pull-request descriptions, and comments. In a controlled study involving 26 students from a 12-week software engineering course, we compared the efficacy of the original GenAI system with a context-enhanced version. Results demonstrated that the context-aware GenAI provided more accurate and useful feedback, as perceived by the students. These findings suggest that incorporating contextual information improves the quality of automated peer reviews, offering a promising tool for educators to enhance student learning and engagement in code review activities.