From Discourse Analytics to Design
Carolyn Penstein Rosé
Computational models of social interaction in textual form reveal layer upon layer of insight about student orientation towards one another as well as towards their experiences in the environment. These insights allow us to make sense of patterns of attrition and learning that occur in online courses and inform design of interventions to support improved outcomes. In this talk I will discuss a methodology for text mining applied to discourse in a MOOC context that allows us to estimate measures of student attitudes, motivation, cognitive engagement, and confusion. Using survival modeling techniques I will illustrate how these measures make predictions about student dropout over time, and thus how we might use these measures to identify students who are particularly vulnerable to dropout at a time point so that available human resources can be dispatched judiciously. Finally, I will describe two interventions deployed in a recent edX MOOC that were inspired by these analyses, and offer an analysis of the positive impact of these interventions in the MOOC environment.