Enhancing students’ learning outcomes through AI integration: an exploratory study in teaching interface design accessibility

Universal digital accessibility is relevant to equity and sustainability, yet its pedagogical integration in computing curricula and the AI's educational value of AI lacks empirical evidence. This study investigates the pedagogical effectiveness of multi-modal LLMs in enhancing software development students’ understanding and implementation of accessible User Interface (UI) design. A 12-week teaching experiment involving 42 senior computer science students in an HCI course examined the impact of structured engagement with an LLM on their accessibility knowledge and design skills using a mixed-methods approach. Findings suggest observed improvements (51.7%) in the heuristic evaluation scores of students’ post-intervention UI designs and their conceptual understanding of accessibility principles. Qualitative analysis highlights the LLM’s role as a collaborative design partner and information resource. Quantitatively, significant positive effects (p < .001) were observed in students’ implementation of accessible UI and self-efficacy, though gains in foundational knowledge, competence, and motivation were limited. Epistemic network analysis (ENA) indicated that high-achieving students strategically used the LLM for augmenting design expertise, particularly in accessibility standards, while lower-achieving students used it for reactive correction. The research provides preliminary evidence for the judicious integration of intelligent language models into design curricula, suggesting a reconceptualization of pedagogical frameworks in HCI and software engineering education.