Using process mining to identify models of group decision making in chat data

Abstract

This paper introduces process modeling and mining as an approach to process analysis for CSCL. This approach is particularly relevant for collaborative learning that takes a project-based form, and is applied in this study to online chat data from teams working on a complex task. The groups differed in terms of the number of members and the amount of scaffolding aimed at group processes and task requirements. The models, produced using the HeuristicsMiner algorithm, showed that the group with fewer members that received more instruction in the task requirements had a more linear decision-making process than the group that received instruction in group processes, however neither were an example of a linear, unitary phase model. This approach has relevance both for CSCL research methods and for providing feedback to students on their decision-making processes.

Discovering the Effects of Metacognitive Prompts on the Sequential Structure of SRL-Processes Using Process Mining Techniques

“…As an inductive approach, the validity of PM depends on the representativity and quality of the data stored in the event log (Reimann, Frerejean, & Thompson, 2009). It is possible that SRL processes, for example, those obtained by think-aloud protocols or log files, comprise a high variety of regulatory behaviour (Hadwin, Nesbit, Jamieson-Noel, Code, & Winne, 2007;Winne, 2014).…”

A survey on educational process mining

“…While t-pattern analysis can be used to explore longer, more temporally separated sequences than LAS and Markov models, all these techniques are best suited to relatively short recurring sequences and analysis of event transitions. 17 SPM techniques have been widely applied to analyze student learning behaviors. But, they are more indicated when trying to discover serial or simpler behavioral patterns than a process, for instance, learner activity paths in a course.…”

“…17 The goal was to use PM to identify models of decision-making groups that took place in a chat room. In a similar study, Bergenthum et al 9 proposed a modeling language for collaborative learnflows that specifically takes the actors, the roles, and the explicit representation of groups into account.…”

“…Reimann et al 17 To discover learning flows and provide feedback Bergenthum et al 9 To discover learning flows…”

Dialogue as Data in Learning Analytics for Productive Educational Dialogue

“…The first, Dyke, Kumar, Ai, and Rosé (2012), notes that, while earlier work (Chiu & Khoo, 2005;Jeong, 2005;Reimann, Frerejean, & Thompson, 2009) offers important insights in to sequences of events, because they are interested in relatively fine-grain regularities across time, they are poor at accounting for change over time, mid-range granularities and dependencies, and the multi-time-scale nature of many social processes involved in learning. They thus propose "interactive time-series visualizations, calculated from sliding window analyses [which] can make a methodological contribution to understanding the irregularities of the unfolding of collaborative processes over time" (Dyke et al, 2012, p. 2).…”