Language Technologies Institute
Student Research Symposium 2006

Using Dynamic Bayes Nets to Model Students in Intelligent Tutoring Systems

Kai-Min Chang

This paper describes an effort to model a student's changing knowledge state during skill acquisition. Dynamic Bayes Nets (DBNs) provide a powerful way to represent and reason about uncertainty in time series data, and are therefore well-suited to model student knowledge. Many general-purpose Bayes net packages have been implemented and distributed; however, constructing DBNs often involves complicated coding effort. To address this issue, we introduce a tool called BNT-SM. BNT-SM inputs a data set and a compact XML specification of a Bayes net model hypothesized by a researcher to describe causal relationships among student knowledge and observed behavior. BNT-SM generates and executes the code to train and test the model using the Bayes Net Toolbox (Murphy, 1998). BNT-SM can be used to simulate Knowledge Tracing (KT; Corbett & Anderson, 1995), an established technique for student modeling. Compared to the original KT code, the DBN trained by BNT-SM does a better job of modeling and predicting student performance (Area Under Curve = 0.610 > 0.568), due to differences in how it estimates parameters. We also used the BNT-SM to construct a novel model framework that simultaneously assess the student's knowledge and evaluate the impact of tutorial interventions. This model distinguishes between two types of tutor help: scaffolding immediate performance vs. teaching persistent knowledge that improves long term performance. We infer from the parameters of the trained model to suggest that students do benefit from both the scaffolding and teaching effects of help. For instance, students are roughly one sixth as likely to make a careless mistake after receiving help (P(slip|helped) = 0.009) than after not receiving help (P(slip|not helped) = 0.058). In a more recent effort, we combine evidence from both the student model and automatic speech recognizer (ASR) to better assess if a student is reading correctly. Whereas ASR assesses student's reading via recent acoustic evidence, student model assesses such via student's history with the particular word.