Statistics Seminar: Susu Zhang
Speaker: Susu Zhang, Columbia University
Title: Modeling Learner Heterogeneity: A Mixture Learning Model with Responses and Response Times
Abstract: Studies in education and psychology have shown that short- and long-term outcomes for learning and for psychological interventions can rely on a number of noncognitive factors, including personality, learning style, interests, and affects. In online learning settings where educators cannot interact face-to-face with the learners, measuring these noncognitive constructs can be especially useful for understanding how students actually learn and, further, for providing personalized stimuli or interventions. In this presentation, a mixture learning model, which jointly models observed responses and reaction times to assessment items over time, is introduced. Based on the mixture hidden Markov modeling framework, such a model accounts for the heterogeneities in learning styles among learners, or, similarly, the heterogeneities in reactions to psychological interventions among patients. The proposed model is evaluated through a simulation study and is applied to a data set collected from a computer-based learning system for spatial rotation skills, where it is used to identify disengaged learners.