Using Social Network Analyses to Model Relation-Intensive Data on Student Engagement

• In Event: Affordances and Modeling of Intensive Data

Abstract

Theoretical/Conceptual Frame
Engagement may be operationalized as energized action (Skinner, 2016), including individual-level phenomena such as emotional investment and learning behavior. Burgeoning research operationalizes engagement as a complex phenomenon that emerges between people, or as a dynamical form of energized interaction (Jarvela et al., 2016), including group-level phenomenon such as information exchange and shared goals. The individual and group level approaches complement each other, representing ontologically distinct levels of analysis in engagement research.
Multi-level, emergent relationship between micro (individual) and macro (group) levels of human behavior is the hallmark of complex systems research in education (Hilpert & Marchand, 2018). Social network analysis (SNA) is an appropriate methodological technique for measuring these relation intensive phenomena because its focus on simultaneously assessing endogenous and exogenous effects (Sweet, 2016). SNA examines how individual-level engagement shapes group-level engagement (bottom up), and how group-level engagement shapes individual engagement (top down).

Purpose
The purpose of the study is to investigate collaborative, self-organizing group processes in the classroom as they unfold over time in a naturalistic setting as a form of collaborative engagement.

Primary Research Question(s)
RQ1: What are the characteristics of collaborative classroom networks in computer science undergraduate courses and how do these characteristics change over time? RQ2: To what extent does individual perceptions of engagement predict the formation of different types of collaborative engagement ties cross-sectionally and over time?

Measurement Approach
A brief weekly, end-of-class survey was developed based on existing instruments. The 5-items measuring behavioral and emotional engagement were modified from a brief measure developed for use with undergraduate students in STEM fields (Skinner et al., 2017). Students were also presented with a list of classmates and asked whom they interacted with and four additional network questions related to exchange of ideas, collaborative quality, shared understanding and the nature of the collaboration. Data were collected weekly over 6-7 weeks.

Participants
The study was conducted in two courses. Twenty-two undergraduate students in a computer science course completed 6 weeks of survey data. Twenty-nine students in an undergraduate computer engineering course completed 7 weeks of survey data.

Analyses
Data from each course was analyzed using exponential random graph modeling (ERGM; Lusher, et al., 2013) using the statnet package in R and Separable Temporal ERGMs for modeling discrete relational dynamics over time with statnet.

Findings
Network characteristics at each time point for the two courses revealed that network density was low with increased tie formation over time. Results from the ERGM analyses (see Figure 1 for example) demonstrated that over time, affective/emotional engagement was the strongest predictor of tie formation. The number of transitive network configurations varied depending on the nature of the instructional task(s) for that week, with active instruction contributing to more collaborative engagement patterns. The findings include discussion of network characteristics, longitudinal network analyses, and comparisons across classes.

Discussion
The use of SNA to model relation-intensive data offers the affordance of including interdependence of individuals in classes to simultaneously investigate the relations among micro level engagement and macro level collaborative engagement patterns.

Authors

• Jonathan C. Hilpert, University of Nevada Las Vegas

• Gwen C. Marchand, University of Nevada - Las Vegas