Designing an Explainable Multimodal AI Framework for Identifying Creative Artistic Developmental Potential in Early Childhood: A Theoretical and Methodological Foundation

Abstract

Discovering artistic-creative potential in children in an early stage has remained critical even so largely underexplored research domain in educational research. Existing computational approaches in talent recognition primarily focus on academic advancement or developmental delays and disorders. This paper presents the theoretical and methodological foundation for the Explainable Multimodal Artificial Intelligence for Talent Identification (EMAI-T) framework.  EMAI – T is an interdisciplinary effort designed to enhance transparent and theory -driven recognition of artistic talent in children at their early stage. The proposed study initially outlines the design and preparation of a multimodal reference dataset from experts in the art field capturing textual, visual, behavioural of creativity across different age groups. Later, the study implements a conceptual and theoretical framework by integrating Gardener’s Multiple Intelligence Theory, Gagne’s Differentiated Model of Giftedness and talent (DMGT), and Explainable AI(XAI) tenets to translate psychological dimensions into suitable computational interpretations.  Ultimately, the paper introduces the architectural design of the EMAI-T framework, which adopts explainability techniques such as SHAP, LIME, and attention mechanism to ensure transparency and trust in AI interpretations.