Discovering Processes and Patterns of Learning in Collaborative Learning Environments Using Multi-modal Discourse Analysis

Multimodal learning analytics, with a focus on discourse analysis, can be used to discover, and subsequently understand, the processes and patterns of learning in complex learning environments. Our work builds upon and integrates two types of research: (a) process analytic approaches of dynamically captured video and computer-screen activity and (b) learning analytics. By combining previous analyses of a dataset with new analyses of the processes of learning, patterns of successful and unsuccessful collaboration were identified. In this paper, the results of the application of a heuristics miner to utterances coded with the Decision-Function Coding Scheme, are combined with the results of First Order Markov transitions and in-depth linguistic analysis of the discourse to analyse the processes of collaborative problem solving within a scenario-based virtual world. The analysis of dependency graphs extracted from students’ event logs revealed problem solving actions enacted by students, as well as the dependency relationships between these actions. The addition of in-depth linguistic analysis explained the micro-level discourse of students, producing the observable patterns. Integration of these findings with those previously reported added to the depth of our understanding about this complex learning environment. We conclude with a discussion about the design of the tasks, the processes of collaboration, and the analytic approach that is presented in this paper.