AI Insight
Researchers collected a detailed dataset of face-to-face interactions during a wedding cocktail hour, where 95 participants wore proximity sensors that detected when people were within 1.5 meters of each other every 5 seconds. The dataset captures 7,213 contact events over approximately 2,760 unique pairs of individuals, with participants self-reporting their relationship to the wedding couple (e.g., bride's family, groom's friends). This is the first publicly available temporal network dataset from an unstructured social event with relationship-based group labels, in contrast to previous datasets from institutional settings like conferences or workplaces.
Why it matters
This dataset enables researchers to study whether social interaction patterns observed in structured environments (hospitals, schools, conferences) also occur in informal social gatherings. The data has practical applications for contact tracing during disease outbreaks, optimizing social event spaces, and understanding how relationship groups naturally mix in social settings.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Objectives: We captured a fine-grained dataset of unstructured social interaction with socially meaningful group labels to fill a gap in the study of face-to-face interaction. Prior interaction data from conferences, classrooms, hospitals, and workplaces exhibit network signatures such as heterogeneous contact rates, clustering, and bursty dynamics. However, schedules, room assignments, and authority roles in these settings may obscure unstructured social group dynamics. Studies on group mixing often rely on demographic proxies like gender, or assigned categories like school classes, rather than relationship-based groups. We aim to understand if temporal network signatures of institutionally structured settings generalize to unstructured social interaction. Data description: We present the first public temporal proximity network dataset of a privately hosted social event with contextual relationship-based group membership. At the outdoor cocktail hour of a wedding, 95 participants wore proximity sensor badges that detected other badge-wearers within approximately 1.5 m in 5 s intervals. This dataset, coarsened to 10 s temporal bins, contains 7,213 contact events over 2,760 observed dyads. Participants self-reported their relationship category with respect to the wedding couple, enabling group mixing analysis. Beyond implications for the generalizability of interaction patterns, this dataset supports social event modeling for applications from contact tracing to social-space design.
Source: A temporal proximity network dataset from a wedding cocktail hour with relationship category labels