ABS 2023
Learning within groups: capturing higher-order dynamics using hypernetworks
Matthew Hasenjager, Nina Fefferman. University of Tennessee, Knoxville, Knoxville, TN, United States

Social interactions are often polyadic, involving three or more simultaneous participants. For example, signaler-receiver interactions can be modified by a third-party audience or intercepted by unintended receivers. Typical network analysis techniques assume pairwise interactions and are therefore ill-suited for modeling such scenarios. Here, we introduce hypernetworks as a means to efficiently describe and investigate multi-way social interactions involving any number of participants. Hypernetworks encode interactions as hyperedges that can contain any number of individuals, allowing for both dyadic and polyadic effects. Using computational simulations, we illustrate how hypernetworks can offer insight into the impact of polyadic interactions on social processes, focusing on social information flow. We evaluate how centrality measures specifically designed to leverage hypernetwork data relate to individuals’ access to novel information and their effectiveness in transmitting information to others and compare them to metrics derived from pairwise approaches. Our models thus offer guidance regarding the conditions under which accounting for polyadic structure can be worthwhile.