A Qualitative Study of Insults in Collaborative Learning

Abstract

Abstract-In computer-supported collaborative learning, automatic coding procedure strategies are necessary for teaching because of the large amount of dialogue acts that must be evaluated. In addition, the characterization of a student's social identification for collaborative and learning behaviors might affect a student's learning outcomes in a variety of ways. An effective learning analysis of the interactive processes cannot dissociate cognitive from social factors. We present a qualitative study of social behavior for insults (flaming) in an anonymous, text-based, collaborative learning dialogue protocol. The application of a nuanced framework of miscommunication for 'flaming' conveys new outcomes for social behavior, as the effect of insults, in collaborative learning processes. This study reinforces the importance of conflict as a variable to understand what, when, and how agents can intervene in collaborative learning dialogues in order to monitor and mediate when necessary, thus keeping the conversation progressing in a productive direction.