Avaliação de Juízes: Um Modelo Estatístico para Perfilação de Avaliadores

Resumo

In this article we present a model to audict judges behaviour in the evaluation of essays according to ENEM's principles. Our model quantifies the reliability and the concordance among judges, based on the Pearson correlation coefficient applied both the general notes and, also, on the skills of the correction grid. In the obtained results we highlight the tendency of the judges to be more judicious in Competence 2, with a high degree of concordance, and with a low agreement degree, -0.58, among the judges for Competence 3. Competencies that often denote a sharp discrepancy are a sign of the need for training to align the evaluators.