Source-linked AI summary
Gender homophily from spatial behavior in a primary school: a sociometric study
J. Stehlé, F. Charbonnier, T. Picard, C. Cattuto, A. Barrat
TL;DR
It remains open whether friendship-based patterns are reflected in children’s observed proximity behavior and how gender homophily varies with age and tie strength. The paper analyzes wearable-sensor proximity data and finds contrasting grade trends for strong and weak ties.
Problem
It remains open whether friendship-network patterns are reflected in children’s observed proximity behavior and how gender homophily varies with age and tie strength.
Source: Introduction · pp. 2–3Method
The study uses wearable sensors to measure children’s face-to-face proximity over time and defines tie strength by cumulated contact duration.
Source: Introduction · p. 2, §4 · p. 19Results
For strong ties, individual homophily tends to increase with age faster for boys than girls; for weak ties, it correlates positively with grade for boys and negatively for girls.
Source: §4 · pp. 18–19Takeaways & Limitations
Same-gender ties are more similar across days than mixed-gender ties, which show higher interday variability.
Source: §4 · p. 18Proximity is a behavioral proxy, not direct information about social relationships, and does not distinguish relationship types.
Source: §4 · p. 19
Abstract
from arXiv · showhide
We investigate gender homophily in the spatial proximity of children (6 to 12 years old) in a French primary school, using time-resolved data on face-to-face proximity recorded by means of wearable sensors. For strong ties, i.e., for pairs of children who interact more than a defined threshold, we find statistical evidence of gender preference that increases with grade. For weak ties, conversely, gender homophily is negatively correlated with grade for girls, and positively correlated with grade for boys. This different evolution with grade of weak and strong ties exposes a contrasted picture of gender homophily.
Introduction
The paper examines whether gender homophily in children’s observed proximity behavior varies with age and tie strength, using wearable-sensor data to compare behavioral patterns with friendship-based findings.
Introduction
- Gender homophily can be difficult to disentangle from social influence because traits such as smoking may both shape and be shaped by friendships.Gender is treated as an almost immutable trait, allowing the authors to avoid this particular entanglement when studying homophily.Source: Introduction · p. 2
- The study asks whether friendship-based patterns are reflected in children’s observed proximity behavior and how that behavior depends on gender and age.The authors also compare proximity patterns among same-gender and opposite-gender peers.Source: Introduction · pp. 2–3
- Wearable sensors record face-to-face proximity over time, enabling a weighted behavioral network in which links represent children’s contact.Link weights reflect time spent in contact; this spatial measure complements surveys, which capture broader subjective relationships.Source: Introduction · p. 2
- The paper reports gender homophily findings and examines how its association with age differs between behaviorally defined strong and weak ties.Its stated aim is a fine-grained quantitative assessment of gender- and age-dependent proximity behavior.Source: Introduction · pp. 2–3
1. Background
Prior work describes gender homophily across childhood and adulthood, while the paper situates sensor-based proximity as a behavioral complement to friendship networks. It uses time spent in proximity to distinguish strong from weak behavioral ties.
1. Background
- Prior research reports same-gender preferences in young children, an increase toward ages 8–11, and a later decline associated mainly with romantic relationships.The review also notes that adult homophily appears relatively stable, though studies of that life period are scarce.Source: §1 · p. 3
- Studies report gender differences in network size, friendship nominations, and age-related interaction patterns, with findings varying by age and gender.La Freniere et al. (1984) report earlier increases for girls followed by stronger preferences among boys; Martin and Fabes (2001) find a significant age correlation for boys only.Source: §1 · pp. 3–4
- Relationship stability is described as increasing during primary school, while same-sex relationships are reported as more stable than mixed relationships.The review says evidence remains too shallow to conclude about gender differences in stability.Source: §1 · p. 4
- Wearable sensors measure face-to-face proximity as a behavioral network, complementing friendship networks that capture broader, partly qualitative social relations.These measures focus on time spent together and can provide an objective, reproducible tie definition without continuous observers.Source: §1 · p. 4
- Tie strength is defined by cumulated face-to-face time, with a chosen threshold separating long-engagement strong ties from short-engagement weak ties.The classification depends on an arbitrary threshold; the authors choose total time because it is simple and measures time spent in proximity.Source: §1 · pp. 5–6
2. Data and methodology
The study analyzes anonymized face-to-face proximity data from children in a French primary school. It restricts the network analysis to playground and canteen contacts, aggregates contacts across two days, and weights ties by their total duration.
2. Data and methodology
- Badges recorded face-to-face proximity during two consecutive school days, detecting contact when devices exchanged packets within a 20-second window.The sensors assessed proximity at a 1–1.50 m range; measured contact durations were multiples of 20 seconds.Source: §2 · p. 6
- The school had 10 classes across five grades, with children aged 6–12; 232 children participated and metadata was available for 227.The analyzed sample comprised 112 girls and 115 boys; teachers were excluded.Source: §2 · p. 6
- Table 1 reports the numbers of girls and boys in each class for which contact data and gender were collected.The authors restrict the sample to the 227 children with usable metadata.Source: §2 · pp. 6–7, Table 1 · p. 7
- To reduce possible seating-arrangement bias, the analysis uses playground and canteen contacts, forming a two-day weighted network of 227 children and 7,070 edges.These settings account for 32,027 contact events, or 41% of recorded contacts; each edge weight is the pair’s cumulative contact duration.Source: §2 · p. 7
3. Results
The results examine gender homophily in children’s within-class ties through class-level, individual, and day-to-day analyses of same-gender mixing and neighborhood stability. They also describe variation by age and tie strength, and isolation among children who skipped or repeated a grade at a five-minute threshold.
3. Results
- Most recorded playground and canteen contacts occurred within the same class, with the class-contact matrix showing a grade-level block structure.The passages attribute much of the structure to class-break and canteen schedules.Source: §3 · pp. 7–8, Figure 1 · p. 8
- Mean edge weights were 118 s for mixed-gender versus 242 s for same-gender edges; girl–girl edges averaged 265 s versus 225 s for boy–boy edges.Average degree was 67.1 for boys and 57.3 for girls in the aggregated network, and 9.0 versus 7.0 for strong ties (at least 5 minutes).Source: §3 · p. 8, Figure 2 · p. 9
3.1. Statistical evidence of gender homophily
- For within-class strong ties (at least 5 minutes), mixed-gender edge fractions fell below H0’s 95% acceptance intervals in every class, and a fixed-topology null likewise supported same-gender preference throughout.The fixed-topology test preserves the observed network structure and controls for the empirical average degrees of boys and girls.Source: Figure 3 · p. 11, §3.1 · p. 11
3.2.1. Within class ties
- Within classes, individual homophily-index baselines usually fell outside the empirical interquartile ranges, indicating same-gender homophily even in lower grades.The index compares the share of a child’s same-gender neighbors with a baseline that accounts for the class gender ratio.Source: §3.2.1 · p. 12, Figure 4 · p. 13
3.2.2. Inter- and intra-class ties
- Across whole-school strong ties, same-gender homophily was high in every grade for both genders, while boys’ higher individual index was only marginally significant in grades 4 and 5.The descriptive pattern depends on interactions observed at school; gender-correlated opportunities outside school may partly account for it.Source: §3.2.2 · pp. 12–14, Figure 5 · p. 14
- Boys averaged a higher two-step homophily index than girls (one-sided Wilcoxon p-value < 5·10^-3); among egos with a majority of opposite-sex neighbors, neighbors averaged more ties with children of the ego’s sex than egos did.Source: §3.2.2 · pp. 14–15
3.3. Similarity of neighborhoods in different days
- Same-gender neighborhoods were more stable from day 1 to day 2 in six classes at the 5% threshold; neighborhood stability did not significantly differ by gender in almost all classes.The other second-grade class had p = 0.125 for same- versus opposite-gender stability; gender comparisons had p-values above 0.3 except in one third-grade class (p = 0.06).Source: Figure 7 · p. 16, §3.3 · p. 16
3.4. Evolution of individual homophily with age
- Figure 8 reports Pearson correlations between age and the individual same-gender homophily index across edge-weight thresholds, separately for boys and girls.The supplied passages describe the plotted measure and 90% bootstrap confidence intervals but do not provide correlation values or directions.Source: Figure 8 · p. 17, §3.4 · p. 17
- With ties lasting at least 100 seconds, individual homophily correlated positively with age for both genders, more strongly for boys; for weaker ties, the correlation was negative for girls.For contacts under three minutes, same-gender neighbors decreased with age for both genders, while opposite-gender neighbors decreased faster for boys and increased for girls. Above five minutes, same-gender neighbors decreased for boys and stayed almost constant for girls, while opposite-gender neighbors decreased for both.Source: §3.4 · pp. 17–18
- At the five-minute threshold, 5 of 23 children who skipped or repeated a grade became isolated, versus 7 of 204 classmates; their relative risk was 6.33 (odds ratio 7.8).The authors note that data collection took place early in the school year, when some children were still new to others, so social exclusion might be weaker later.Source: §3.4 · p. 18
4. Discussion
The discussion highlights gender- and age-dependent patterns that differ between strong and weak ties, alongside greater day-to-day stability of same-gender ties. Wearable-sensor proximity data provide behavioral measures, but these remain a proxy for social relationships.
4. Discussion
- Wearable sensors provide quantitative face-to-face proximity measures and daily interaction networks without recall biases.The method also supports access to weak ties, which surveys and diaries may measure differently because of cognitive and recall biases.Source: §4 · pp. 18–19
- Gender homophily appears in all primary grades and reaches higher levels for boys than girls in grades 4 and 5.Source: §4 · p. 18
- Same-gender ties are more similar across days than mixed-gender ties, which show greater interday variability.Source: §4 · p. 18
- Strong-tie homophily increases with age faster for boys, while weak-tie homophily rises by grade for boys and falls for girls.Source: §4 · p. 19
- Proximity patterns are a behavioral proxy rather than direct information about social interactions, and they disregard relationship types.Source: §4 · p. 19
- The authors call for testing robustness across schools, school-year periods, and longer follow-up durations.Source: §4 · p. 19