A Pedagogical Framework for Enhancing Inclusion in STEM Higher Education Through Large Multimodal Models

The rapid advancements in Large Multimodal Models (LMMs) and Generative AI (GenAI) have opened new possibilities for enhancing accessibility in higher education, particularly in STEM disciplines. However, current AI applications in education often prioritize direct problem-solving over fostering deep learning experiences, potentially bypassing essential cognitive struggles that are critical to mastering complex concepts. This study proposes a pedagogical framework for leveraging LMMs to support inclusive learning without replacing the cognitive engagement necessary for students' conceptual development. The research is structured around a two-fold approach, where three interdisciplinary working groups—special pedagogy, mathematics, and engineering- collaborate to design and test an AI-driven methodology that supports both professors in creating accessible lectures and students in developing customized learning pathways. The framework will be evaluated through two pilot studies, focusing on mathematics and electronics university courses, to assess its impact on both teaching practices and student learning experiences. By integrating AI-driven adaptations with evidence-based pedagogical strategies, the project aims to strike a balance between leveraging AI for accessibility and preserving the cognitive challenges necessary for deep learning in STEM disciplines.