Unfolding Teachers’ Orchestration Load in Computer-Supported Collaborative Learning: Factors and Multimodal Data Analytics

This dissertation advanced the understanding of teacher orchestration load in scripted CSCL by introducing a novel multimodal data-based approach. We aimed to address the following main research question: What can multimodal data reveal about teachers' orchestration load in scripted CSCL settings? To address this research question, three objectives were defined: to study the potential of multimodal data analytics to estimate teacher orchestration load, to compare this load across learning settings and dashboard designs, and to understand contextual factors influencing teacher orchestration load in CSCL. Seven studies have been conducted to achieve these objectives. Across all the studies, multimodal data were collected and analyzed, including physiological data (e.g., EDA and HRV), objective (e.g., log data, dashboard screen recordings with audio, and observation notes), and subjective data (e.g., self-reported questionnaires) to study teacher orchestration load in scripted CSCL. The investigation shows that the multimodal data-based analytics offers a robust and comprehensive approach to estimating teacher orchestration load by combining subjective and objective data. The findings show that teachers experience more orchestration load in co-located settings than in online settings, as evidenced by both the coding scheme used to code orchestration actions and the triangulation of physiological data (EDA) with the self-reported questionnaire, highlighting the real-time demands of managing co-located CSCL activities. Moreover, the dashboard design can influence teacher orchestration actions in CSCL. While mirroring dashboards primarily support situation evaluation (evaluating and monitoring student responses and collaboration), alerting dashboards prompt more frequent and diverse orchestration actions. The findings suggest that real-time alerts can guide teachers more effectively in orchestrating CSCL activities. Finally, the thesis identifies contextual factors such as the number of students, activity duration, task type, and teacher experience as critical factors influencing the load. This dissertation contributes to the field of CSCL by analyzing teachers’ orchestration load, leading to the following contributions: (1) a novel multimodal data-based approach to facilitate the analysis of teacher orchestration load, (2) providing empirical evidence of variations in teacher orchestration load across different learning settings and dashboard designs, and (3) insights into contextual factors that influence teacher orchestration load.