Exploring the Properties and Growth of Student Interaction Networks on Twitter: Insights on STEM Learning and Engagement

The objective of this research is to promote student engagement in STEM learning and demonstrating an opportunity for leveraging social media as a pedagogical platform. As such, the study utilizes social media interactions on Twitter from students in an online class of Construction Materials and Methods during Spring 2020. Traditional datasets have limited capacity to capture online information sharing and interactions of students and external peers in a STEM class with such details and coverage. The fully online class was consisted of 66 students enrolled in the fully online class and 49 of them provided Twitter handle through an end-of-semester survey. Twitter Application Programming Interface (API) was used to retrieve students' online interactions in five consecutive deliverables during the course of the semester. These deliverables included student social media interactions on several topics covered in class such as (i) construction; (ii) delivery methods, drawings and specifications, zoning regulations, building codes, construction standards; (iii) soil types, properties, gradation, testing and exploration; (iv) concrete ingredients, making, placing and reinforcing concrete, sitecast and precast concrete, wood products, masonry; and (v) steel construction. The requirement was to post any information of their choice (photo, video, text, website link or anything relevant) already covered in the class and comment to at least five other students. To maximize peer influence, students used several keywords (such as #course number, #university name, #construction) as hashtags and mentions. Interactions from external peers such as experts, alumni faculty, senior students among others were also observed. Analyses of such interaction networks showed significant growth of students' social media activity and connectivity over time i.e. the number of network agents and their connectivity increased significantly from the first deliverable to the final deliverable. These also exhibit small-world properties that are evident in many real networks. For example, within the first 25 days of assigning the initial deliverable, five times more participants and twenty-seven times more interactions were observed. We also observed that the degree (i.e. number of immediate neighbors) distribution follows power-law which means many nodes with less interaction and very few in the network are having most of the interactions. Within the first 25 days, the average degree of the participants increased from 2.29 to 12.92. The insights of this study can be useful in facilitating STEM learning and engagement with more efficient peer influence. The findings of this research will help to craft targeted information sharing strategies for various student groups based on their interaction, activities, and characteristics on how students react to what others post in social media on a related STEM topic.