Aplicação de IA Generativa para Verificação de Conformidade de Planos de Ensino

Abstract

This work aims to evaluate the effectiveness of using artificial intelligence-based language models, specifically LLaMA 3.1 8B, in verifying the compliance of the teaching plan with institutional guidelines. To achieve this, a Python application was developed to extract textual content from PDF documents and submit it to the model for analysis, using a structured prompt based on the guidelines for developing the teaching plan and the course syllabus. The results were compared with manual evaluations previously conducted by faculty members, revealing a high degree of agreement between human and automated assessments, particularly in standardized sections such as identification, syllabus, objectives, course content, and bibliography. Some discrepancies were observed in areas requiring greater textual interpretation, such as justification and methodology. It is concluded that the LLaMA model demonstrated satisfactory performance in the task of automated verification, showing potential to support pedagogical review processes, provided it is used alongside human supervision to address subjectivities and contextual interpretations.