Towards an Adaptive Approach to a Personalised Design of Intelligent Learning Assistants (ILAs)

This paper presents an innovative approach to designing Intelligent Learning Assistants (ILAs) named GITTE (Great Individual TuTor Embodiment), focusing on their visual representation and personalisation. The significance of visual representations for ILAs is explored, underlining their impact on learning outcomes and student perceptions. The presented approach introduces a novel, exploratory, and participatory method for identifying influential design features in ILAs using the Pedagogical Agents-Levels of Design (PALD) model and generative AI models like DALL-E and ChatGPT. This method allows for the exploration of design features not previously considered, capturing unconscious student preferences. The paper documents the initial test involving a manual approach to evaluate the applicability of generative AI models in designing GITTE, emphasizing the interaction between students and the generative model. The results indicate the potential of this approach in enhancing the personalisation and effectiveness of ILAs, suggesting new directions for research in this emerging field. The concept presented is poised to make significant contributions to the development of more engaging and personalized digital learning aids.