Beyond the dyad: uncovering higher-order structure within cohesive animal groups

Abstract

Revealing the consequences of social structure in animal societies is largely determined by our ability to accurately estimate functionally relevant patterns of social contact among individuals. To date, studies have predominantly built up social structure from dyadic connections. However, many associations or interactions can involve more than two individuals participating together, which current approaches cannot distinguish from independent sets of dyadic connections. Here we demonstrate the application of higher-order networks to detect the central roles of dominance and sex in structuring social groups of vulturine guineafowl (Acryllium vulturinum). Specifically, we find that while females and low-ranking group members engage in more dyadic interactions, males and more dominant group members are substantially more likely to be observed forming hyperlinks--edges that contain more than two individuals (e.g. a triad). These results demonstrate how higher-order networks can provide a deeper understanding of the multidimensionality in the difference of centrality among group members.

Not your private tête-à-tête: leveraging the power of higher-order networks to study animal communication

“…We can model these behavioural states as contagions on social networks to quantify transitions from individuals (a) not calling to calling and (b) being present in the group to departing. Because social contagions are often best considered complex contagions and animal groups frequently contain higher-order social structures, such as subgroups [11] or family units [36], these systems are suited to modelling as a directed, dyadic signal (vocal communication) across a higher-order network structure. Incorporating this higher-order social structure could make meaningful differences to the predictions made about group coordination, helping to elucidate how partial- and full-consensus decisions are reached.…”

“…Accounting for the multi-body nature of such interactions prevents losing relevant information. Powerful new higher-order network approaches [11][12][13][14] encode these non-pairwise interactions between agents, helping us quantify the importance of multi-body interactions in driving group dynamics and wider social coordination. By explicitly representing multi-body interactions, higher-order approaches capture the rich set of dynamics introduced by including the non-dyadic components of communication networks.…”

Comparative approaches in social network ecology

“…This loses information on interaction size that can be captured using higher‐order network approaches (Silk et al., 2022). While these have only rarely been used in behavioural ecology (Musciotto et al., 2022), they are gaining popularity as a tool in network science (Battiston et al., 2021). It would be valuable to move towards also storing higher‐order network data in repositories (e.g., as group‐by‐individual or incidence matrices) to facilitate approaches that explicitly incorporate this higher‐order structure.…”

Generation and applications of simulated datasets to integrate social network and demographic analyses

“…Finally, we focus here on dyadic social networks, however many of the social interactions studied are non-dyadic and may include higher-order interactions (Battiston et al, 2021;Greening Jr et al, 2015). While there has been limited focus on higher-order interactions in animal societies (Musciotto et al, 2022), theory suggests they will impact infectious disease transmission and social contagions (Battiston et al, 2021;Iacopini et al, 2022;Noonan & Lambiotte, 2021) among other ecological and evolutionary processes.…”