EXPLAINABLE NEURAL NETWORKS AND EDUCATIONAL ROBOTICS: REVIEW AND ANALYSIS OF THE OPEN ROBERTA LAB FOR TEACHING ARTIFICIAL INTELLIGENCE

Abstract

This research presents partial results on the teaching of Artificial Intelligence in high school, with an emphasis on Explainable Neural Networks and the use of the Open Roberta Lab simulator as a pedagogical resource. The study included a literature review and a functional and pedagogical analysis of the Open Roberta Lab simulator, considering its compatibility with the BNCC Computing and constructionist approaches. The findings indicate that the platform offers visual block programming, simulation of 32 systems including microcontrollers, physical robots, and AI modules, and resources for creating and training simple Explainable Neural Networks with real-time visualization. These features favor the understanding of concepts such as supervised learning and data-driven decision-making, overcoming the black-box logic of Artificial Intelligence. The Open Roberta Lab simulator proved to be accessible, free, and multiplatform, enabling practical activities even in school settings with limited infrastructure. The proposal is articulated with BNCC competencies, such as critical understanding of AI systems, computational modeling, and the development of integrative projects. The Open Roberta Lab simulator represents a viable and inclusive alternative for the popularization of Artificial Intelligence in basic education, being promising for the development of teaching sequences that combine theory and practice.