Evolution of bistability and collective fission-fusion dynamics in a population of agents searching a heterogeneous evironment
Animal groups such as fish schools and bird flocks often exhibit dynamic fission-fusion behavior; large groups split into smaller groups (fission) and small groups or individuals band together to form larger groups (fusion). While fission and fusion occur at the population level, these phenomena result from real-time movement decisions made by individual animals. In general, it is unclear what decision rules lead to fission-fusion dynamics, and precisely how such rules could evolve through natural selection on individual behavior. We investigate these phenomena with a simple model of social decision-making, based on experimental studies of schooling fish, in which agents search their environment for a resource using a combination of social and environmental information. We find the parameters of the model that yield optimal group performance in a homogeneous population, and we use evolutionary simulations to study the rules that emerge when each agent is rewarded based on its individual performance.
This numerical study reveals that individuals readily locate and track dynamic resource patches by splitting and fusing to form groups that match the sizes of resource patches. This occurs even though individuals have no direct way of evaluating the sizes of resource patches or determining the size of the group to which they belong. The agents are able to achieve this when the parameters of the model permit two distinct stable states, a dense 'liquid' state and a dilute 'gas' state, and when the population can undergo a transition between states near the boundary of a resource path. Density differences between the two states allow agents to determine the location of resource patches from afar, making no use of gradient information. This study presents an example of how collective computation and fission-fusion dynamics can arise from an evolutionary process.