A growing interest has been directed toward including a fourth dimension in the engagement construct: the social dimension. The aim of this study is twofold: first, to explore how teachers talk about the social dimension of student engagement in online learning, and second, to explore the possibilities of using computational methods for interview data analysis. A longitudinal intervention (interview-diary-interview) was conducted with teachers who actively teach online classes in hybrid, remote, and distance modes of delivery. Natural language processing methods, more specifically topic modelling, were used to extract and analyse topics discussed in the data. Analysis of topic overlap and distinctions were made. Key results reveal that co-creation and shared cognition are core concepts when teachers talk about social engagement. However, results also show that individual engagement is critical for social engagement, and that teacherstudent interaction can, potentially, be viewed as a separate component. Interestingly, the teachers’ talk also reflected phases in learning such as process and product-based focuses. Following the results, we suggest that computational methods can be combined with traditional (human) analysis to contribute to a richer and more nuanced understanding of abstract topics. Understanding social engagement may inform theory development and is vital for researchers, practitioners, and decision-makers.