A narrative review toward a Multimodal Intervention for continuous and always-accessible education for children with Dyslexia using Multimodal Large Language Models

This review article clearly presents a MLLM-based application designed for dyslexic children with a pre-existing diagnosis, acting effectively as an intelligent assistive tutor that personalizes adaptively the learning path using multimodal behavioral data, without replacing the specialist's diagnosis. While the use of MLLMs during the diagnostic phase has been explored previously in the existing literature, topics that will be addressed, the focus here is on the training phase for children with dyslexia who have already received an initial diagnosis from a specialist. The proposed system uses systematically multimodal large language models to provide continuous multimodal education. The system is designed to provide personalized education for each child. It incorporates weekly and monthly monitoring to track progress regularly based on each child’s individual profile and, when necessary, adapt instruction to enhance improvement. Based on literatures dyslexia is a multifaceted disorder affecting areas such as speech, handwriting, and working memory. Therefore, a system aimed not at clinical diagnosis but at classifying educational levels and personalizing a child’s profile should rely on multimodal inputs. The proposed framework gathers such data from relevant tools and channels, analyzes it using a MLLM, and automatically generates a personalized educational program based on the integrated multimodal profile.

کلیدواژه ها:

Dyslexia ، Multimodal Large Language Models ، Learning Disorder ، Cognitive Science ، User Experience