SOFTWARE AND PEDAGOGICAL FOUNDATIONS OF THE USE OF MULTIMODAL ARTIFICIAL INTELLIGENCE TECHNOLOGIES IN THE EDUCATIONAL PROCESS

The article analyses the software and pedagogical foundations of using multimodal artificial intelligence (AI) technologies in the educational process. The study aims to substantiate the correspondence between the technical architecture of multimodal models and the principles of learning theory, and to identify the pedagogical conditions for their integration into teaching. The theoretical analysis describes the software basis of multimodal AI systems through a three-stage scheme: encoding of modalities, fusion in a shared semantic space, and response generation. R. Mayer’s cognitive theory of multimedia learning and J. Sweller’s cognitive load theory are applied as the pedagogical basis, with an interpretation of the modality, redundancy, coherence and segmenting principles under multimodal AI conditions. UNESCO’s AI competency frameworks for teachers and for students are compared. A four-stage implementation model is proposed, together with five key risks and pedagogical solutions for mitigating them.