DEVELOPMENT OF AN APPLICATION FOR DIAGNOSTIC ASSESSMENT IN EARLY CHILDHOOD EDUCATION USING ARTIFICIAL INTELLIGENCE RESOURCES

Abstract

This article presents the development and application of a digital diagnostic assessment tool for Early Childhood Education, aimed at identifying and monitoring literacy predictor skills based on the analysis of children’s drawings and early writing, mediated by Generative Artificial Intelligence. The proposal is grounded in the Science of Reading, Psychopedagogy, and classical and contemporary studies on graphic development and children’s writing. The method originated from a pedagogical practice initially developed manually in the context Early Childhood Education and the early grades of Elementary School through systematic diagnostic assessments conducted throughout the school year. Although these practices yielded positive pedagogical results, they revealed limitations related to interpretative subjectivity in teachers’ analyses, difficulties in standardizing evaluative criteria, and the resulting overload of the validation process. In response to these challenges, the method was transposed into an educational application based on computer vision techniques with pattern recognition and large language models (LLMs). The tool automates the classification of drawing and early writing phases, enables longitudinal monitoring of children’s development, and generates individualized reports with pedagogical recommendations, contributing to greater diagnostic accuracy and supporting decision-making at the levels of education systems, schools, classrooms, and individual students.