Stress Propagation in Group Project Social Networks

Technical Writing (ENG307) is a required course for many science students at Western Kentucky University (WKU). A group project is required in this course. To gain insight into how the social networks between these groups affect individual grades through stress propagation, we obtained anonymous logs from these group projects. We constructed “Friend Approximation Networks” (FANs) by connecting students if they (1) are in the same project group or (2) have the same major and college year. We adapted sand-pile models to stress test the FANs: we randomly dropped “grains of sand” (representing stress) onto the network; when a node exceeds its threshold, it divides its sand evenly among its neighbors; an avalanche or cascade typically follows if any of those neighbors subsequently exceeds its threshold. We measured the size and frequency of cascades at different locations to test correlations between stress propagation and individual project grades. With this limited dataset, we have not produced a correlation stronger than ρ=0.35; however, this data has provided ample support for adapting these preliminary methods for larger social networks. Additionally, we note that although these stress tests are closely related to eigenvector centrality measures, they provide further parameter control and insight into behavior. The current research is supported by a WKU FUSE grant and the dataset was provided by Dr. Angela Jones.