Stochastic dynamics of social patch foraging decisions

TL;DR

This paper introduces a mechanistic model of social patch foraging that incorporates stochastic evidence accumulation and different communication modes, revealing how social information sharing influences foraging efficiency.

Contribution

It develops a novel model of group foraging behavior with pulsatile and diffusive communication, enabling analysis of their effects on efficiency and robustness.

Findings

Pulsatile coupling's efficiency depends strongly on coupling strength.

Diffusive coupling is more robust to parameter variations.

Model can distinguish communication modes and fit real animal data.

Abstract

Animals typically forage in groups. Social foraging can help animals avoid predation and decrease their uncertainty about the richness of food resources. Despite this, theoretical mechanistic models of patch foraging have overwhelmingly focused on the behavior of single foragers. In this study, we develop a mechanistic model describing the behavior of individuals foraging together and departing food patches following an evidence accumulation process. Each individual's belief about patch quality is represented by a stochastically accumulating variable coupled to others' belief, representing the transfer of information. We consider a cohesive group, and model information sharing as either intermittent pulsatile coupling (communicate decision to leave) or continuous diffusive coupling (communicate online belief). Foraging efficiency under pulsatile coupling has a stronger dependence on the coupling strength parameter compared to diffusive. Despite employing minimal information transfer, pulsatile coupling can still provide similar or higher foraging efficiency compared to diffusive coupling. Conversely, diffusive coupling is more robust to parameter detuning and performs better when individuals have heterogeneous departure criteria and social information weighting. Efficiency is measured by a reward rate function that balances the amount of energy accumulated against the time spent in a patch, computed by solving an ordered first passage time problem for the patch departures of each individual. Using synthetic data we show that we can distinguish between the two modes of communication and identify the model parameters. Our model establishes a social patch foraging framework to parse and identify deliberative decision strategies, to distinguish different forms of social communication, and to allow model fitting to real world animal behavior data.