Visualization of student activity patterns within intelligent tutoring systems

Abstract

Novel and simplified methods for determining low-level states of student behavior and predicting affective states enable tutors to better respond to students. The Many Eyes Word Tree graphics is used to understand and analyze sequential patterns of student states, categorizing raw quantitative indicators into a limited number of discrete sates. Used in combination with sensor predictors, we demonstrate that a combination of features, automatic pattern discovery and feature selection algorithms can predict and trace higher-level states (emotion) and inform more effective real-time tutor interventions. © 2012 Springer-Verlag.