Phase Diagram and Optimal Information Use in a Collective Sensing System.

We consider a population of individuals living in an uncertain environment. Individuals are able to make noisy observations of the environment and communicate using signals. We show that the model shows an order-disorder transition from an ordered phase in low communication noise in which a consensus about the environmental state is formed to a disordered phase in high communication noise in which no consensus is formed. There are different consensus states: informed consensus in which consensus on the correct belief about the environmental state is formed, and misinformed consensus in which consensus on a wrong belief is formed. Based on the consensus state reached, the ordered phase is decomposed into multistable states separated by first order transitions. We show that the inference capability of the population in a changing environment is maximized on the edge of bistability: on the border between an informed consensus phase and a bistable phase in which both informed and misinformed consensuses are stable. In addition, we show that an optimal level of noise in communication increases the responsiveness of the population to environmental changes in a resonancelike phenomenon. Furthermore, the beneficial effect of noise is the most crucial in a fast changing environment.