Ai-Assisted Multiple-Choice Questions Generation with Multimodal Large Language Models in Engineering Higher Education

This paper presents an AI-assisted approach that leverages Multimodal Large Language Models (MLLMs) to automate the generation of Multiple-Choice Questions (MCQs) for modules in engineering education. The system introduces a LOs extraction to MCQs generation pipeline, which extracts Learning Outcomes (LOs) from provided lecture notes and generates relevant MCQs with solutions and explanations based on the extracted LOs. By harnessing MLLMs' capabilities in vision and text comprehension, coupled with carefully crafted prompts from human educators, the tool efficiently produces context-relevant MCQs that can streamline teaching material development. The effectiveness of this AI-powered MCQ generation pipeline is investigated through experiments across a number of engineering modules with evaluations on the quality of the generated MCQs by human educators. The analysis of the evaluation results shows the AI tool's ability to generate MCQs that are well-aligned with LOs and exhibit strong contextual relevance, demonstrating the potential of AI-assisted approaches to enhance the efficiency of creating high-quality MCQs in engineering education. However, the variability in quality ratings across different aspects underscores the continued need for human expertise and oversight in the assessment design process. The findings provide useful insights into the capabilities and limitations of state-of-the-art multimodal language models in supporting assessment development in engineering education.