Group Cohesion and Performance in Computer-Supported Collaborative Learning (CSCL): Using Assessment Analytics to Understand the Effects of Multi-attributional Diversity

Many online learning environments offer opportunities for computer-supported collaborative learning (CSCL). Although the effectiveness of CSCL has been studied extensively, researchers have only recently begun to systematically examine student diversity in CSCL. Building on a theoretical integration of social psychology research with the CSCL literature, this chapter reports some key findings from an ongoing series of coordinated studies designed to assess and manage diversity effects in CSCL through social network analyses and planned interventions. We base our analyses on the observation that a defining characteristic of CSCL groups in many online learning contexts is the multi-attributional diversity of learners, that is, a combination of diversity in terms of learners’ sociodemographic characteristics and task-relevant attributes and competencies. We then present the results of a coordinated series of three empirical studies using social network analysis with digital behavioral data from 4628 distance learners in 930 groups. We use an unobtrusive assessment analytics approach to collect and analyze learner data from both formative and summative assessments, thereby mitigating the effect of potential self-report biases. Study 1 shows that the combination of high sociodemographic diversity and high task-related diversity has a significant and negative effect on group cohesion. Experimental interventions designed to attenuate this interaction effect through student grouping on single diversity attributes (Study 2) or through communication instructions (Study 3) were unsuccessful, testifying for a relatively robust and pervasive phenomenon. In our conclusion, we highlight the theoretical potentials of unobtrusive assessment analytics for understanding multi-attributional diversity effects and suggest directions for developing interventions using visualizations of group network structures for monitoring and supporting learning in CSCL groups featuring multi-attributional diversity.