Beyond boundaries: capturing social segregation on hypernetworks

In recent years, the study of complex social systems has been fueled by the renewed interest in higher-order topologies, thus leading to the emergence of hypernetwork science. A critical and interesting phenomenon often characterizing social complex systems is segregation, i.e., the extent to which network entities are separated or clustered based on certain semantic attributes or features. This paper introduces a novel approach to studying segregation in hypernetworks. Firstly, we propose a general framework to extend classical segregation measures from dyadic to polyadic network structures. Then, we introduce a novel segregation measure called ``Random Walk HyperSegregation'' (RWHS), which exploits random walkers to estimate segregation at multiple scales. Through an extensive experimental study involving synthetic and real-world case studies, we illustrate the applicability and effectiveness of our measure. Moreover, we highlight the limits of classical segregation measures when extended to high-order topologies---conversely from RWHS, which effectively captured highly-segregated scenarios.