Leveraging student confusion in online forum posts to enhance student engagement using text-based learning analytics

Abstract

Online discussion forums serve as a natural repository of students’ emotions and feelings in online learning environments, as students often post to seek assistance with academic challenges and difficulties. Understanding the complexity of student emotions or affective states is crucial for fostering meaningful engagement and promoting academic success. This study focuses on confusion, a cognitive-affective state that emerges from complex learning processes and is recognised as potentially beneficial for deep learning when reaching an optimal state in terms of its duration and intensity. A text-based learning analytics (TLA) workflow grounded on the confusion and affect dynamics model was proposed. Six machine learning models were evaluated for their effectiveness in classifying confusion from online discussion forum posts collected from three courses. Results indicated that model performance was inconsistent. To demonstrate the viability of the proposed TLA workflow, it was applied to a comprehensive set of course data gathered from two study periods. Through systematic evaluation of students’ confusion states and transitions and their relationship with students’ academic performance, the study establishes the feasibility of the proposed workflow in facilitating educators’ early identification of students experiencing prolonged cognitive disequilibrium that may lead to unproductive struggle rather than meaningful learning gains.