In this paper, we present an empirical study with simulated students that allows to compare the accuracy of two models based on Bayesian networks in the context of student modelling: a dynamic model versus an static model. The results show that the performance of both models is very similar, being the dynamic much faster and easier to implement. A second study evaluates the use of adaptive item selection criteria, that can provide an increase on accuracy and a big reduction in test length.
展开▼