A general mechanism for collective foraging: percolating recruitment cascades coupled to a maximum-entropy

Abstract

Foraging recruitment is a contagion: an agent finding a resource triggers a linear-threshold cascade -- the model for rumour and adoption spread on social networks -- embedded in space. Because activation is degree-normalized, the same rule that spreads the cascade also stifles it as neighbourhoods crowd (density-dependent blocking), so the swarm aggregates on a resource, then releases to explore again. The cascade runs on the communication network, but acting on it moves agents in physical space, and that motion rewires the network: network and physical space are one coupled system, and this feedback -- not tuning -- makes the coordination self-limiting. Because recruitment chains along the graph rather than near neighbours alone, whether a discovery stays local or spans the swarm is a matter of percolation, and past a critical connectivity the swarm holds itself in a cascade window -- locally near-critical though globally subcritical, a self-organized, sandpile-like drive-avalanche cycle. Nothing is tuned, yet behaviours foraging biology documents separately emerge together -- recruitment, marginal-value patch-leaving, and Levy search with a switch to local exploitation. These are not coincidences but one least-committal principle: the resource field and the search both follow from maximum entropy -- Levy search matched to the field's sparsity, recruitment to its clustering, so active-matter order and optimal foraging are one system in two domains. The gains are real but conditional: up to 7.6x the harvest of solitary search when resources are sparse and clustered, shrinking to nothing when search is already easy. Percolating cascades yield a large return on minimal communication. And that sparse, clustered field is the generic one -- resources almost inevitably aggregate, and maximum entropy makes it the least-committal one once they do -- so one untuned rule should recur widely.