Using Machine Learning Techniques to Analyze and Support Mediation of Student E-Discussions

Students are starting to use networked visual argumentation tools to discuss, debate, and argue with one another about topics presented by a teacher. However, this development gives rise to an emergent issue for teachers: how do they support students during these e-discussions? The ARGUNAUT system aims to provide the teacher (or moderator) with tools that will facilitate effective moderation of several simultaneous e-discussions. Awareness Indicators, provided as part of a moderator’s user interface, help monitor the progress of discussions on several dimensions (e.g., critical reasoning). In this paper we discuss preliminary steps taken in using machine learning techniques to support the Awareness Indicators. Focusing on individual contributions (single objects containing textual content, contributed in the visual workspace by students) and sequences of two linked contributions (two objects, the connection between them, and the students’ textual contributions), we have run a series of machine learning experiments in an attempt to train classifiers to recognize important student actions, such as using critical reasoning and raising and answering questions. The initial results presented in this paper are encouraging, but we are only at the beginning of our analysis.