Predicting Task Persistence within a Learning-by-Teaching Environment

Abstract

We attempted to model task persistence, a student attribute reflecting one’s dispositional need to complete difficult tasks in the face of frustration, within a learning by teaching intelligent tutoring system (ITS) called SimStudent. We used the interaction logs of 32 students from the Philippines to develop a Naïve Bayes model to detect task persistence. Using forward feature selection, an optimized set of predictors was derived. Out of 11 candidate features, those that significantly predicted task persistence were time on task, time spent on resources after failure, number of re-attempts to unsolved problems, and proportion of difficult problems attempted.