Item-based Bayesian Student Models

Many intelligent educational systems require a component that represents and assesses the knowledge state and the skills of the student. We review how student models can be induced from data and how the skills assessment can be conducted. We show that by relying on graph models with observable nodes, learned student models can be built from small data sets with standard Bayesian Network techniques and Na ve Bayesian models. We also show how to feed a concept assessment model from a learned observable nodes model. Different experiments are reported to evaluate the ability of the models to predict item outcome and concept mastery