Related Experiment Video
Updated: Jun 27, 2026
Visualization of Intensity Levels to Reduce the Gap Between Self-Reported and Directly Measured Physical Activity
Published on: March 7, 2019
Survey-calibrated agent-based modeling of peer targeting for promoting physical activity among adolescents
Weixuan Long1, Binbing Zheng1, Zhihan Wu2
1Department of Sport Science, College of Education, Zhejiang University, Hangzhou, Zhejiang, China.
Objectives:
This study aims to compare the effectiveness of various centrality-based peer-targeting strategies in promoting physical activity (PA) among adolescents and identify practical principles for school-based social network interventions.
Study Design:
A modeling study combining cross-sectional survey data with an agent-based model (ABM) analysis.
Methods:
Questionnaire data from 1,692 middle-school students were used to parameterize PA distributions, friendship ties, and co-activity probabilities. Maximum-entropy class networks were generated, and a peer-influence diffusion process inspired by the Susceptible-Infected-Recovered (SIR) framework was implemented. Four targeting strategies including random, degree-, betweenness-, and closeness-centrality were evaluated based on diffusion speed and changes in the proportions of low-, moderate-, and high-PA students.
Results:
Centrality-based strategies generally accelerated PA diffusion relative to random targeting, although differences among centrality rules were small. Closeness-centrality showed a more concentrated diffusion pattern, but no single rule consistently dominated across all outcomes. Specific advantages varied by outcome: degree-centrality tended to strengthen already active clusters, while betweenness-centrality showed limited advantage in improving PA composition. Across all strategies, over 60% of adolescents who remained low active had below-average friendship degrees, indicating persistent structural disadvantages.
Conclusion:
Centrality-based peer-targeting strategies offer advantages over random allocation in school-based PA interventions. However, their effects on PA composition were limited, and socially peripheral adolescents remained difficult to engage through peer influence alone. No single centrality rule should be interpreted as universally optimal. These findings suggest that network-informed peer targeting should be combined with low-threshold, teacher- or coach-supported opportunities to improve inclusiveness and overall effectiveness.
