Early Prediction of Cognitive Tool Use in Narrative-Centered Learning Environments

Abstract

Narrative-centered learning environments introduce novel opportunities for supporting student problem solving and learning. By incorporating cognitive tools into plots and character roles, narrative-centered learning environments can promote self-regulated learning in a manner that is transparent to students. In order to adapt narrative plots to explicitly support effective cognitive tool-use, narrative-centered learning environments need to be able to make early predictions about how effectively students will utilize learning resources. This paper presents results from an investigation into machine-learned models for making early predictions about students' use of a specific cognitive tool in the Crystal Island learning environment. Multiple classification models are compared and discussed. Findings suggest that support vector machine and naive Bayes models offer considerable promise for generating useful predictive models of cognitive tool use in narrative-centered learning environments.