Inferring signed social networks from contact patterns

Area of Science:

• Network Science
• Computational Social Science
• Statistical Inference

Background:

• Inferring social networks often relies on indirect data like proximity.
• Existing methods struggle to distinguish between no observed interactions due to lack of opportunity versus active avoidance (negative ties).

Purpose of the Study:

• To develop a method for inferring signed social networks from contact patterns.
• To differentiate between absent relationships and actual negative ties when no interactions are observed.

Main Methods:

• A Bayesian framework utilizing Markov Chain Monte Carlo (MCMC) inference.
• Modeling interaction groups to distinguish chance encounters from deliberate avoidance.
• Validation using synthetic data and application to real-world high school contact data.

Main Results:

• The proposed method significantly outperforms baseline approaches, especially in detecting negative edges.
• Analysis of French high school data revealed a network structure aligning with friendship survey findings.
• Posterior predictive checks confirmed the model's robustness and adequacy.

Conclusions:

• The developed Bayesian framework effectively infers signed social networks by distinguishing between lack of opportunity and negative relationships.
• This approach offers a more nuanced understanding of social network structures, crucial for fields like sociology and epidemiology.