The global structure of classroom social networks predicts academic and psychosocial adjustment.

Humans are deeply social, and in early life, classroom friendships shape academic and socio-emotional development. Classroom peer ecologies describe the social structure in which individual relationships are situated and which form the context for student learning and development within these classrooms. Yet, research has only begun to explore how these global social network structures are related to classroom-level outcomes. Using a large-scale dataset of 35,038 students in 1,433 classrooms, we extracted 314 social network metrics from four different network types (i.e. friendship, rejection, help, and break-time contacts) and applied machine learning to predict various classroom-level outcomes, including academic achievement, motivation, and social integration. Classroom networks predicted both academic achievement (up to 2% of variance) and well-being (up to 18% of variance), beyond school tracks, socioeconomic status, and demographic factors. However, they did not significantly improve predictions for other characteristics of the teaching and learning environment-such as instructional methods and motivational characteristics. The predictive power for different outcomes was highly dependent upon the type of network and categories of network metrics. These insights position social networks as a powerful tool for understanding aspects of educational environments and for informing network-based interventions.