Modeling temporal cognitive topic to uncover learners' concerns under different cognitive engagement patterns.

Massive open online courses (MOOCs) provide learners with high-quality learning resources, but learners drop out frequently. Learners' concerns (e.g. the topics in course content or logistics) and cognitive engagement patterns (e.g. tentative or certain) are considered the essential factors affecting learners' course completions. However, it is still unclear what different learning achievement groups focus on in each cognitive engagement pattern. In this study, we adopted an unsupervised computational model, the temporal cognitive topic model (TCTM), to automatically investigate learners' cognitive engagement patterns in discussing different topics, as well as the changes under each cognitive engagement pattern over time. A data experiment of 4080 learners enrolled in a Modern Etiquette course revealed that the high-achievement group preferred to discuss on-task topics in an exclusive cognitive engagement pattern; the low-achievement group preferred to discuss off-task topics in a tentative pattern, including certificate acquisition and examination grades; the medium-achievement group showed less variation in different cognitive engagement patterns. Additionally, a moderation analysis showed that there was a significant moderating effect of discussion guidance, especially for instructor-led guidance, between the salient cognitive topics and learning achievements. The analytical results can help instructors to conduct (e.g. feedback and guidance) and timely intervention of cognitive knowledge construction in MOOCs. [ABSTRACT FROM AUTHOR]