Abstract:

We know that employing collaborative learning strategies does not lead to productive collaborative learning per se. In fact, some groups are dysfunctional and might have a detrimental influence on group members. This issue has methodologically been studied through self-reported surveys, transcripts coding, and observational methods. Although these methods are informative, they are also time intensive, exhausting and not practical to be applied in real practice. Social network analysis (SNA) and learning analytics, on the other hand, open the door for using automatic and non-intrusive methods that can help us monitor group interactions. Here we study if SNA combined with machine learning techniques can be employed in order to better understand and predict how different type and quantities of collaborative interactions in online environments affect individual and group performance. More specifically, we study the correlation between group interaction parameters as measured by SNA and groups and individual’s performance. Using interaction parameters and machine learning methods, we identify the indicators that best predict groups that will perform and gain and groups that will not, as well as individuals who gain in performance and those who do not. The article provides support for the idea that learning analytics can help automatically monitor group performance and offer an opportunity for educators and learners to support productive collaborative learning.

Keywords:

Learning analytics, Collaborative learning, small groups, interaction analysis, social network analysis.