Exploring Predictive Indicators of Reading-Based Online Group Work for Group Formation Teaching Assistance

Abstract

Using digital systems to group students according to their indicators provides opportunities for better group work implementation. However, how these indicators can affect group work performance remains unclear. Teachers tend to feel confused about which indicators should be considered when creating groups using learning log data. Capitalized on the datadriven environment under GLOBE, we conducted a preliminary study to explore predictive indicators for algorithmic group formation in a reading-based group learning context. This study presented our effort to explore the key factors that correlated to a desirable group work via factor analysis and correlation analysis. We found that reading engagement and previous peer rating scores suggest a higher potential to predict desirable group work performance in the readingbased online group work, which aims to help teachers set appropriately in future student model data-based group formation.