Measuring Collaboration Quality Through Audio Data and Learning Analytics

Abstract

Collaboration is an important twenty-first-century skill. Collaboration quality detection can help to support collaboration. This chapter addresses the collaboration quality detection and measurement: (1) to define collaboration quality using audio data and unobtrusive learning analytics measures; (2) to explain the design of a sensor-based set up for automatic collaboration analytics; (3) to move toward quantifying the quality of collaboration by using this set up and show the analysis using meaningful visualizations. Furthermore, we address the challenges and issues at hand and how solutions can be built upon the work already done. To elaborate the different chapter’s objectives, we use the terminology of indicators (i.e., the events) and indexes (i.e., the process) to define the components to detect collaboration quality. In one study, during collaborative brainstorming, higher was the equality (i.e., the index) of total speaking time (i.e., the indicator), lower was the dominance of each group member (in terms of total speaking time), and better was the quality of collaboration. However, quality of collaboration is dependent on the context of collaboration and the actual content of the discussion. During collaboration content analysis has been mostly on the surface level by using certain representative keywords to model different topic clusters. Therefore, we develop a sensor-based setup for automatic collaboration analytics to understand collaboration quality holistically in a learning context. Here, our aim is to understand “how” group members speak (i.e., speaking time indicator) and “what’” (i.e., the content of the conversations) group members speak to move toward collaboration quality measurement.