Using AI-powered multiple-choice question generation for self-regulated learning

This study examines the integration of generative AI in education, specifically evaluating AI-generated multiple-choice questions (MCQs) and their role in supporting self-regulated learning (SRL). Using AIQUIZ, an open-source AI-driven platform, 325 of the 593 enrolled students (54.8%) across four computing courses (Web Technologies and Databases) used the platform and generated 38,752 MCQs over two years. An explanatory sequential mixed-methods design analysed student performance, error reports, survey insights, and expert evaluations. Results showed a 70.79% overall student performance (79.45% in Databases, 66.84% in Web Technologies). Only 0.85% of questions were flagged by students as potentially incorrect, a figure that reflects user perception rather than a verified error rate. Surveys indicated strong student acceptance, engagement, and motivation, which are vital for the forethought phase of SRL. However, error analysis of flagged items revealed recurring issues like incorrectly marked answers and flawed distractors. These findings suggest that AI-generated MCQs may support the SRL cycle by facilitating forethought, performance control, and self-reflection. While Large Language Model (LLM) tools provide scalable opportunities for practice and self-assessment, our results confirm that human validation remains essential to ensure content quality and maximize learning benefits.