Automating Detection of Good Reflective Responses in Discourse

In this paper we demonstrate the successful implementation of a learning theory- based computational technique to capture students' abilities to engage in reflection-on-action. Grounded in Epistemic Network Analysis, we operationalize reflection-on-action in discourse in terms of connections between important concepts within a given domain. We also demonstrate that we can use the detection of connections between concepts to determine the degree to which students rely on group discourse as a scaffold for reflection-on-action. We argue that these results will enable CSCL environments to accurately provide on-time and continuous monitoring of student learning.