Inferring daily college peer networks from campus behavioral data

Behavioral records for 4,738 undergraduates entering in 2018 were used to infer daily peer networks from co-occurring attendance, dining, residence, and bathing activity. Communities shifted from residence-centered to classmate-centered patterns, while peer achievement and scholarship status were associated. The large longitudinal trace dataset shows how campus routines can reveal changing opportunity structures, while also making explicit the need to validate what a digital co-occurrence tie actually represents.

Early communities were more residence-centered and later became more classmate-centered; geography and living areas shaped ties, and high-scholarship students were more likely to connect with similarly awarded peers. Co-location and similar routines are proxies, not confirmed friendship, advice, or collaboration. One university cohort, institutional systems, privacy filtering, unmeasured prior selection, and many daily observations limit generalization and causal claims about peers or scholarships.