Study of Primates Social Dynamics from Temporal Co-presence data

Non-human primates have complex and flexible social relationships that can rapidly change through time. However, the study of their social network is mostly based on data coming from direct observational measures and limited to static networks, which provide an aggregated picture and do not describe the dynamics of the network (emergence of leaders, growth or stability of communities, etc.). The use of data coming from automated systems for social network analysis is quite recent but opens the possibility of a temporal network approach.
The present study exploits 3 years of data originating from the Automated Learning Device for Monkeys (ALDM) system in which a group of captive guinea baboons (Papio papio) have free access to computerized cognitive testing. A previous study [1] showed that the social network obtained from the co-presence of the baboons in ALDM testing (simultaneous presence in adjacent booths) represents a good proxy for the social organization of the group and we here extend this approach to develop a weighted temporal social network. To control for different amounts activity and presence of different individuals in ALDM, we compare the co-presence data to a null model based on random permutations of the individual names to build monthly weighted, signed co-presence networks: the positive links reflect preferential choices of the individuals to locate in adjacent booths, while the negative links can be interpreted as individuals purposefully avoiding each other. The positive links, defining “positive” social networks, are validated by direct comparison with observed grooming behavior over one specific month. As we lack data on avoidance behaviors, we provide an indirect validation of the negative links by observing that social balance theory is obeyed in the signed networks (Fig. 1A).
We then consider the temporal evolution of the network. We study the stability of the strongest links and build similarity matrices between months, highlighting periods of stability and moments of discontinuity (Fig. 1B). In the latter case, we study the network on a finer temporal scale, studying community dynamics and group composition on a weekly basis.
Our study shows that the use of automatic systems to study the social network of non-human primates can provide a better understanding of the group dynamics and of the emergence of strongest bonds. Moreover, observing the network at different timescales can suggest methods to identify changes in the group's hierarchy or the emergence of leaders.