Download

Download Full Text (2.7 MB)

Description

Social Network Analysis (SNA) is a research method which quantitatively maps qualitative social interactions between individuals who comprise a ‘network’. Delineating these social networks yields highly valuable data, including SNA measures like centrality, which can be used to measure social influence, connectivity, and more. Further, these networks can also be visualized by graphing individuals as ‘nodes’, and then by drawing ‘edges’ (the lines that connect them) to produce sociograms. With these sociograms, researchers can concurrently conduct visual and statistical analysis of relations between node and network traits of interest. As a result of these capabilities and the growth of social learning, SNA has become increasingly popular in educational settings. However, the difficulties in consolidating students’ interaction data into quantitative networks has steered SNA researchers towards oversimplified social environments which do not exhibit ambiguous connections. For example, research is well established in observing students’ online interactions, where participant information is collected concurrent with interaction data. Similarly, SNA studies in face-to-face contexts are typically bounded to single classrooms which greatly reduces the number of participants’ possible ties. These examples observe environments that are easily monitored, and bar the observation of the true underlying social networks. Hence, a gap exists for those hoping to understand the true, non-course-bounded networks of undergraduate students. To this end, our research group is currently conducting a study comparing all participating freshmen and sophomore engineering students’ interactions to academic outcomes at USU using social network analysis. The current disambiguation process for this large (1000+ nodes), ambiguous (open response name identified ties), interaction data is manually intensive, intricate, and takes careful organization--increasing with network size. Therefore, charged with the task of interaction data disambiguation, I have organized the overarching disambiguation task into a hybrid blend of automation and manual stages, to take advantage of emerging network information throughout the process (i.e., previously ambiguous responses are connected to an entity). The procedural analysis I present in this poster then highlights the technicalities of these stages, which begin with simple spelling checks, and end with a sub-network comparison process (similar to agglomerative hierarchical clustering) to yield accurate and complete network data. My methodology proved effective in matching many ambiguous names with their counterparts found elsewhere in the network. However, several names were still unable to be consolidated and had to be de-identified in their present conditions. Therefore, this presentation also highlights procedures that could be refined through modern SNA clustering methods, including possibilities for defining supervised algorithm parameters and comparing our manual to automated results for such algorithms. As the need for understanding the relationship between interpersonal interactions and educational outcomes expands, so too are the needs for improved SNA methods. To meet these growing needs, researchers must develop new and effective procedures for the disambiguation of authentic interaction data. This presentation provides an example of such research, developing and disseminating more effective and efficient approaches for network development.

Publication Date

12-9-2021

City

Logan, UT

Keywords

social interaction, online interaction, Social Network Analysis

Disciplines

Engineering Education

Disambiguation of Large-Scale Educational Network Data for Social Network Analysis

Share

COinS