Using multimodal learning analytics to study collaboration on discussion groups: A social network approach

Nowadays, companies and organizations require highly competitive professionals that have the necessary skills to confront new challenges. However, current evaluation techniques do not allow detection of skills that are valuable in the work environment, such as collaboration, teamwork, and effective communication. Multimodal learning analytics is a prominent discipline related to the analysis of several modalities of natural communication (e.g., speech, writing, gestures, sight) during educational processes. The main aim of this work is to develop a computational environment to both analyze and visualize student discussion groups working in a collaborative way to accomplish a task. ReSpeaker devices were used to collect speech data from students, and the collected data were modeled by using influence graphs. Three centrality measures were defined, namely permanence, persistence, and prompting, to measure the activity of each student and the influence exerted between them. As a proof of concept, we carried out a case study made up of 11 groups of undergraduate students that had to solve an engineering problem with everyday materials. Thus, we show that our system allows to find and visualize nontrivial information regarding interrelations between subjects in collaborative working groups; moreover, this information can help to support complex decision-making processes.