Advancing Automated Exam Generation: Toward Scalable and Adaptive Solutions [before doctoral defense]

Abstract

This dissertation investigates the design, optimization, and practical application of automated assessment generation systems, with a particular focus on the Exercise Generation Algorithm+ (EGAL+). Manual exam construction is a complex and time-consuming task requiring educators to balance curriculum coverage, difficulty, cognitive complexity, and question diversity while maintaining consistency across multiple test versions. Automated assessment systems offer a promising solution by improving efficiency, objectivity, and scalability in examination design. Through a comprehensive review of existing literature, this research identifies key limitations in current automated assessment approaches and positions EGAL+ within the category of optimization-based test composition systems. To address these limitations, the dissertation presents a systematic redesign of the EGAL+ architecture. The redesigned system was evaluated through benchmarking and deployment in authentic university teaching environments. Quantitative and qualitative findings demonstrate improvements in computational efficiency, assessment quality, and usability. The research contributes novel insights into balancing pedagogical parameterization with operational scalability and outlines future directions for the development of intelligent, adaptable assessment generation systems in educational technology.