coding-exam: Enhancing Academic Integrity and Code Understanding through Personalized Assessments with Generative AI

Abstract

While Generative AI has made code generation effortless, it has also compromised academic integrity through "AI-assisted ghostwriting." This paper introduces "coding-exam," an LTI-compliant tool designed to analyze studentsubmitted source code and automatically generate personalized multiple-choice quizzes. By extracting and questioning critical logic within the students' own code, the system validates their genuine understanding. Evaluation results indicate significant improvements in student motivation and self-perceived understanding.