Student self-governance network positions and civic participation outcomes
Jing Liu, Putu Kerti Nitiasih, Made Hery Santosa, Putu Nanci Riastini
Scientific Reports
Network design
Nodes
237 students participating in formal self-governance bodies at one regional university in eastern China
Ties
1,738 reported governance links covering information exchange, decision coordination, and cross-unit collaboration
Methods
The authors characterized density, communities, centralities, brokerage and E-I mixing, fitted demographic-adjusted multivariate regressions, and compared logistic regression, support-vector, random-forest and XGBoost classifiers with network-derived features.
Source: Scientific Reports
Reviewed summary
A study of 237 student-governance participants at an eastern Chinese regional university built a 1,738-edge network with six functional communities. Eigenvector centrality and the community E-I index predicted civic outcomes, while XGBoost classified engagement with 0.781 accuracy and 0.842 AUC. The study makes unequal structural access within student governance visible and suggests that civic-development opportunities depend on cross-community connectivity, not just holding a formal title.
The network had density 0.062 and six communities in a hierarchical hub-and-spoke form; the central student government bridged peripheral groups, and XGBoost reached 0.781 accuracy with AUC 0.842. The study is cross-sectional, single-site and based on governance participants rather than all students. Network position, civic attitudes and participation can influence one another, while classifier performance does not establish causal effects or justify ranking individuals for opportunities.
SNA
How SNA was used
The study defined 237 students participating in formal self-governance bodies at one regional university in eastern China as nodes and 1,738 reported governance links covering information exchange, decision coordination, and cross-unit collaboration as ties. The authors characterized density, communities, centralities, brokerage and E-I mixing, fitted demographic-adjusted multivariate regressions, and compared logistic regression, support-vector, random-forest and XGBoost classifiers with network-derived features.
Source
Scientific Reports, 16, Article 4141
DOI: 10.1038/s41598-025-34205-x
