Creating high-quality programming exercises remains a critical bottleneck in computer science education. Educators face significant challenges in designing pedagogically sound content with clear problem statements and robust test cases at the scale required for large student populations. Current approaches suffer from synchronization issues when multiple instructors collaborate, leading to conflicts and lost work. Additionally, the lack of structured feedback mechanisms and version control creates inefficiencies in the review process and risk when modifying exercises. While artificial intelligence offers potential solutions, the creative and nuanced nature of exercise design requires a “human-in-the-loop” approach rather than full automation.
This thesis proposes a collaborative system within the Artemis learning platform that combines human expertise with AI assistance. The solution features real-time synchronization for co-creation, AI-assisted review for quality assurance, and comprehensive versioning support. By implementing this collaborative environment, the thesis aims to significantly improve both the quality and scalability of programming exercise creation, ultimately enabling more personalized learning experiences for students.