Pedagogical Companions to Support Teachers’ Interpretation of Students’ Engagement from Multimodal Learning Analytics Dashboards

Abstract

Teaching demands educators adapt and improvise their instruction for each student’s unique needs and capabilities across contexts. This requires teachers to observe their students, evaluate their ongoing learning, and offer individualized scaffolding and feedback and foster sustained growth and development. This challenging practice has been made even more difficult by the recent emphasis on data-driven instructional decision making from dashboards, further requiring teachers to become both pedagogical and data experts. Despite the development of dashboards to alleviate some of the load of collecting and aggregating complex multimodal student data, there is a need to provide support for teachers in analyzing, interpreting, and applying their students’ real-time multimodal learning analytical data (e.g., metacognitive accuracy, negative emotions) in the form of pedagogical companions. Before we can begin the design and development of these agents, we must first understand how educators are currently approaching multimodal learning analytics (MMLA) that report on more than just performance- based outcomes. In this on-going work, we begin by briefly reviewing MMLA in teacher dashboards, teacher data literacy, and the role of pedagogical companions in teacher augmentation technologies. We then describe the development of an in-progress study exploring how three teachers currently use fictious MMLA on self-regulated learning (SRL) processes and the emerging trends we see from their data. Finally, we postulate what these results suggest about the needs that embedded intelligent pedagogical companions may fill in future dashboard and agent design.