Can an Intelligent Tutoring System Predict Math Proficiency as Well as a Standarized Test?

It has been reported in previous work that students' online tutoring data collected from intelligent tutoring systems can be used to build models to predict actual state test scores. In this paper, we replicated a previous study to model students' math proficiency by taking into consideration students' response data during the tutoring session and their help-seeking behavior. To extend our previous work, we propose a new method of using students test scores from multiple years (referred to as cross-year data) for determining whether a student model is as good as the standardized test to which it is compared at estimating student math proficiency. We show that our model can do as well as a standardized test. We show that what we assess has prediction ability two years later. We stress that the contribution of the paper is the methodology of using student cross-year state test score to evaluate a student model against a standardized test.