Integrating Learning Analytics to Measure Message Quality in Large Online Conversations

Abstract

Research on computer-supported collaborative learning often employs content analysis as an approach to investigate message quality in asynchronous online discussions using systematic message-coding schemas. Although this approach helps researchers count the frequencies by which students engage in different socio-cognitive actions, it does not explain how students articulate their ideas in categorized messages. This study investigates the effects of a recommender system on the quality of students' messages from voluminous discussions. We employ learning analytics to produce a quasi-quality index score for each message. Moreover, we examine the relationship between this score and the phases of a popular message-coding schema. Empirical findings show that a custom CSCL environment extended by a recommender system supports students to explore different viewpoints and modify interpretations with higher quasi-quality index scores than students assigned to the control software. Theoretical and practical implications are also discussed.