Abstract:

This paper describes a teaching experience that brings generative AI into a Control Theory course. The goal was to help teachers design assessment materials and make learning more engaging. Using generative AI tools, the team created conceptual and problem-based questions based on real laboratory activities. A total of approximately 160 questions were generated across several core topics of the course, including Laplace analysis, first-order systems, state-space models, and PID control, using a large language model provided by Mistral AI.

Each question was labeled by type and cognitive level. The main categories were scenario-based, cause–effect, and numerical problems. The questions were exported in GIFT format, which made it easy to upload them directly into the MOODLE platform used for the course. Before deployment, all questions were reviewed by the teaching staff following four criteria: clarity of the language, accuracy of the explanations, correct use of the GIFT format and removal of potentially confusing items. Students then answered the questions without knowing that AI had written them. This avoided any bias or preconceptions about AI content. Student responses and perceptions were later analysed using a Likert-scale survey focused on clarity, comprehension, and perceived didactic value.

Results show that generative AI can produce questions of high educational quality. Survey results indicate high levels of student agreement regarding the clarity of the questions and their usefulness for consolidating knowledge. In particular, the inclusion of explicit explanations for incorrect answers was highly valued by students. It also saves instructors a great deal of time and effort. With less time spent on repetitive tasks, teachers can focus on improving feedback, discussion, and experimentation in class. The proposed workflow is novel in that it combines automated question generation, structured pedagogical review, and direct integration into official university platforms, making it easy to replicate and scale in other technical courses. The study suggests that AI-assisted assessment design is a practical way to make control engineering courses more efficient, consistent, and creative. It lowers one of the main barriers to innovation in higher education: lack of preparation time.

Keywords:

Generative AI, control theory education, automated question generation, GIFT format, MOODLE integration, engineering education.