DATA-DRIVEN MODELING OF ENGAGEMENT ANALYTICS FOR QUALITY BLENDED LEARNING

Engagement analytics is a branch of learning analytics (LA) that focuses on student engagement, with most studies conducted by computer scientists. Thus, rather than focusing on learning, research in this field usually treats education as a scenario for algorithms optimization and it rarely concludes with implications for practice. While LA as a research field is reaching ten years, its contribution to our understanding of teaching and learning and its impact on learning enhancement are still underdeveloped. This paper argues that data-driven modeling of engagement analytics is helpful to assess student engagement and to promote reflections on the quality of teaching and learning. In this article, the authors a) introduce four key constructs (student engagement, learning analytics, engagement analytics, modeling and data-driven modeling); b) explain why data-driven modeling is chosen for engagement analytics and the limitations of using a predefined framework; c) discuss how to use engagement analytics to promote pedagogical reflection using a pilot study as a demonstration. As a final remark, the authors see the need of interdisciplinary collaboration on engagement analytics between computer science and educational science. In fact, this collaboration should enhance the use of machine learning and data mining methods to explore big data in education to provide effective insights for quality educational practice.