Bayesian Student Modeling Improved by Diagnostic Items
Yang Chen, Pierre-Henri Wuillemin, Jean-Marc Labat
Bayesian network (BN) has been successfully applied in hierarchical student models. Some researchers used diagnostic strategies to improve the evidence level of student models. But test items are typically related to a dichotomous response model, namely students’ answers are scored as right or wrong. As we know, wrong answers result from lacking one or more relevant concepts in students’ knowledge states. This diagnostic information of wrong answers is ignored. To maximize the precision of student model, this paper presents an approach using diagnostic items, which are designed to provide the information about which concepts are probably lacked in students’ knowledge states when they give wrong answers. A modified NIDA (Noisy Input, Deterministic AND) model is built to represent the relations between students’ answers and their knowledge states. We use simulated students to evaluate our model and the results show that the efficiency and accuracy of student modeling are improved.
The final publication is available at Springer via https://doi.org/10.1007/978-3-319-07221-0_17.