A Technique for Conflict Detection in Collaborative Learning Environment by Using Text Sentiment

摘要

Computer-Supported Collaborative Learning (CSCL) can give many benefits to students such as promoting creativity and sense of community, sharing abilities, etc. However, when groups of people work together, conflict is inevitable. Generally, conflict in any CSCL situation is uncomfortable, time consuming and counterproductive. It is hard to characterize a conflict because it can involve many factors -e.g., environmental factors, member's differences, etc. This paper proposes a technique to recognize conflicts in a group and the members involved in them by focusing in the socio-emotional interactions. As disagreements between group members generally cause negative emotions, and members can induce negative emotions to other members; then, a conflict between two or more members can be recognized when there are bidirectional negative messages in the same conversation thread. The proposed technique represents chat interactions as a digraph in which the nodes represent users and the edges indicate the transference of negative sentiments during the interactions. Then, a matrix of scaled commute times is used to detect clusters (subgroups having conflict). The validation of the technique shows promising results. The proposed technique is able to detect conflicts automatically, reducing the human effort required to detect these conflicts by other means.