Get access

Using Graph Theory Metrics to Infer Information Flow Through Animal Social Groups: A Computer Simulation Analysis

Authors


C. Vital, Department of Biology, Indiana University, Bloomington, IN 47405, USA. E-mail: cvital@indiana.edu

Abstract

Single individuals, termed ‘gatekeepers’, can have a profound impact on information flow in a group, whether that gatekeeper be a large male despot which strictly controls access to resources and mates, or an infant that is freely passed around among members of a social group. Graph theory offers powerful tools for considering larger aspects of social dynamics such as information flow, and their impact on phenomena such as social learning, social roles, foraging skills transfer and eavesdropping. Here, we use computer simulation to test the abilities of several social network metrics to estimate the proportion of gatekeepers in a social group with sample sizes and study designs typical of animal behavior studies. We find that most network metrics are sensitive to the amount of sampling (number of recorded interactions), and did not give good estimates when fewer than 10 interactions were recorded for each animal in the group. Metrics were also quite sensitive to variation in group size, yielding the full range of possible values for groups varying from 20 to 50 animals. We thus recommend against their use with animals that move in and out of groups seasonally. Individual values estimated by each of the metrics were often quite different from each other such that a combination of metrics chosen from each of the following groups provides the most comprehensive description of information flow: (1) Closeness or Degree Centrality, (2) Betweenness Centrality, (3) Density or Clustering Coefficient, and/or (4) Diameter, Average Degree or Information Centrality. Finally, we introduce a software package, SocANet, that can be used to conduct similar simulations to determine the best metrics for a particular group of animals and set of conditions.

Ancillary