The current study explored how to improve learning outcomes for over 25000 learners on MOOC. Results showed that performance is predicted by the frequency of viewing lecture materials, and this link is mediated by interaction with others in class and moderated by their first approach to tackling the course materials.

Past studies have documented numerous benefits of online learning (Dolence & Norris, 1995). During COVID-19, online learning (including that of MOOCs) has gained momentum and has become a necessity rather than an option. On Coursera, the enrolment increased from 26million to 189million within a five-year time-frame (2016-2021). There is a huge urgency to understand the learning process in MOOCs. The current study seeks to uncover the psychological determinants that enhance the learning process and improve learning outcomes in a MOOC course hosted on Coursera.

Past studies showed that video viewing and final grades are highly correlated in MOOCs (Kim et al., 2021). However, in the MOOC learning environment, primary facilitators in face-to-face learning – interaction with peers and teachers- is largely ignored. MOOCs often boost autonomous learning, and collaboration does not occur spontaneously; hence, the benefit of interaction within MOOC courses is side-lined. Nonetheless, past studies have shown children's potential can be realized in learning from peers (Vygotsky, 1978), and learning communities can instill a sense of 'group inquiry' where learners share a common identity (Lipman, 1991). Because MOOCs are plagued by high drop-out rates, strategies that enhance student sustainability throughout the course are highly valuable. One such strategy is to engage student to make an early approach to course materials, and past study has shown such initiative can predict higher sustainability of students' effort (Chen et al., 2019). Therefore, we hypothesized that total viewing frequency on course lectures predicts course performance. More importantly, this performance is mediated by interaction with peers or course instructors and further moderated by an early attempt to tackle course materials.

Big data from a computer science course on Coursera were analyzed. Data on their lecture viewing frequency, all graded materials, and interaction exchanges on forum posts were recorded. In total, 90 lectures were offered in this 15 weeks course, with 25,745 total enrolments. The final performance is an average of total graded materials in the 15 weeks course, total viewing frequency is the times they viewed the lectures altogether, interaction is measured by each learner's forum post frequency, and the first approach is measured by their assignment grade taken in week one after viewing the first week of materials. In total, the course generated over 2,857,695 data points.

The results are consistent with our hypotheses in that the total viewing frequency on course lectures significantly predicts course outcome (β=0.0336,SE=0.0024,p<0.001), and more importantly, this effect is mediated by the frequency of interaction (95%CI[-0.015524599, -0.008326667]). We regressed the first approach in the model to test for the proposed moderation effect. Consistent with our hypothesis, first approach moderates the relationship between total viewing frequency and interaction (β=0.0209,SE=0.0010,p<0.001). 

Our results suggested that interaction is crucial for MOOCs, and this relationship can be enhanced when learners are motivated enough to take the first approach to their learning. The study uses empirical data to test the importance of interaction and first approach in a large sample, and the findings can readily be applied to enhancing pedagogical development.