Measuring the higher-order homophily in a general social hypergraph

Area of Science:

• Social Network Analysis
• Network Science
• Computational Social Science

Background:

• Homophily, the tendency to interact with similar others, is crucial for social dynamics.
• Traditional homophily measures use k-uniform hypergraphs, limiting analysis to fixed group sizes (k≥2).
• Real-world social interactions often involve varying numbers of participants, necessitating new analytical tools.

Purpose of the Study:

• To propose a novel descriptive homophily measure for general social hypergraphs with non-uniform group sizes.
• To establish theoretical constraints for monotonic and majority homophily in two-class labeled networks.
• To provide a framework for analyzing homophily patterns in complex social structures.

Main Methods:

• Developed a new homophily measure applicable to general social hypergraphs.
• Defined and established constraints for monotonic and majority homophily.
• Conducted experiments on diverse datasets to validate the measure's effectiveness.

Main Results:

• Identified systematic deviations in hyperedge composition based on node class labels.
• Demonstrated the measure's ability to capture homophily patterns in non-uniform hypergraphs.
• Showcased differences compared to label-independent baselines.

Conclusions:

• The proposed measure effectively quantifies homophily in social hypergraphs with arbitrary group sizes.
• This work bridges the gap between theoretical homophily concepts and practical applications.
• Advances the understanding of social and structural dynamics in complex networks.