Investigating Practice Schedules of Multiple Fraction Representations Using Knowledge Tracing Based Learning Analysis Techniques

The goal of this paper is to use Knowledge Tracing to augment the results obtained from an experiment that investigated the effects of practice schedules using an intelligent tutoring system for fractions. Specifically, this experiment compared different practice schedules of multiple representations of fractions: representations were presented to students either in an interleaved or in a blocked fashion. The results obtained from posttests demonstrate an advantage of interleaving representations. Using methods derived from Knowledge Tracing, we investigate whether we can replicate the contextual interference effect, an effect commonly found when investigating practice schedules of different task types. Different Knowledge Tracing models were adapted and compared. A model that included practice schedules as a predictor of students’ learning was most successful. A comparison of learning rate estimates between conditions shows that even during the acquisition phase, students working with interleaved representations demonstrate higher learning rates. This finding stands in contrast to the commonly found contextual interference effect when interleaving task types. We reflect on the practical and theoretical implications of these findings.