Division of labor enables efficient collective decision-making under uncertainty

TL;DR

This work investigates how leaderless animal groups can make efficient collective decisions under environmental uncertainty. It shows that a simple division of labor, with a small, heterogeneous minority of scouts and a synchronized majority of deliberators, allows decentralized collectives to match centralized performance using threshold-based rules. The key insights are that the number of explorers scales sublinearly with group size (roughly $\mathcal{O}(\log N)$) and that structured heterogeneity emerges most strongly under intermediate ecological pressures, collapsing to homogeneity in extreme conditions. The findings illuminate how ecological constraints shape strategy distributions and offer a mechanistic account of emergent informational leadership without central coordination.

Abstract

How do social animals make effective decisions in the absence of a leader? While coordination can improve accuracy, it also delays responses as information propagates through the group. In changing environments, these delays can outweigh the benefits of centralized control, making decentralized strategies advantageous in large groups. This raises a key question: how can groups implement efficient collective decisions without central coordination? We address this question using a model of collective foraging in which individuals choose whether to invest in costly exploration or remain idle, while sharing information and rewards across the group. We show that decentralized collectives can match the performance of centrally controlled groups through a division of labor: a small, heterogeneous subset explores even when expected rewards are negative, while a synchronized majority forages only when expected rewards are positive. Information redundancy causes the optimal scout number to grow sublinearly with group size, so that larger groups need proportionally fewer explorers. The heterogeneity of the group is maximized at intermediate ecological pressures, but optimal groups are homogeneous when costs or environmental contrasts or fluctuations are extreme. Crucially, these group-level policies do not require central coordination, emerging instead from agents following simple threshold-based decision rules. We thus demonstrate a mechanism through which leaderless collectives can make effective decisions under uncertainty and show how ecological pressures can drive changes in the distribution of strategies employed by the group.

Figures (9)

• Figure 1: Overview of model and setup. Individuals in a collective observe returning foragers to update their belief about the environmental state and select actions. (A) The belief, $g^t$, is updated based on the observed outcomes of foraging attempts. (B) Individuals forage when their belief that the environment is in a high-reward state exceeds a threshold, $\theta$; these thresholds can differ between individuals. (C) Foraging is successful with probability $\gamma^t$, which switches between high and low values with the environmental state. (D) Under centralized control, a planner decides on the optimal number of foragers to maximize returns. (E) Under decentralized control, coordination emerges from individual foraging decisions.
• Figure 2: Division of labor enables centralized and decentralized coordination.(A) In decentralized collectives, individuals have different foraging thresholds. Risk-averse "deliberators" share a common threshold, $\theta_c,$ and forage only when the expected reward is positive. Risk-tolerant "scouts" have lower, heterogeneous thresholds and forage even when expected rewards are negative. (B) Under centralized control, a planner implementing the optimal policy, $\pi(g),$ deploys an increasing number of foragers as the belief increases. An appropriate distribution of individual thresholds allows a decentralized collective to match this allocation (compare panels A and B). (C) The optimal number of risk-tolerant scouts, $N_s,$ grows sublinearly with population size, $N$, so an increasingly small fraction of explorers suffices for efficient foraging. (D) The collective belief, $g^t,$ tracks environmental fluctuations; scouts continue exploring during lean periods, enabling rapid re-engagement when conditions improve.
• Figure 3: Effects of perturbing the distribution of decision thresholds.(A) Optimal threshold distribution for a collective of $N=100$ individuals. (B-D) Perturbed strategy distributions: (B) desynchronizing the risk-averse majority by spreading their thresholds; (C) synchronizing risk-tolerant individuals at a high threshold; (D) synchronizing scouts at a low threshold. Colors indicate strategy types. (E) A trade-off plot of return and accuracy for the optimal strategy and perturbations shows that returns are maximized at intermediate accuracy (gray dot; panel A), reflecting weakened coordinated exploitation under desynchronization (brown–gray curve; panel B), improved accuracy at increased cost under moderate heterogeneity (pink–gray curve; panel C), and degraded returns under excessive risk tolerance (blue–gray curve; panel D). Returns are normalized by the maximum possible return $\rho_{\max} = (\gamma_+ - \lambda)/2$, achieved by committing the entire collective only in the high reward state.
• Figure 4: Ecological modulation of collective heterogeneity.(A) Heterogeneity of optimal decision thresholds varies non-monotonically with commitment cost $\lambda$, peaking at intermediate values where foraging decisions depend sensitively on belief; extreme costs collapse heterogeneity through uniform inactivity (high cost) or uniform foraging (low cost). (B) Heterogeneity also peaks at intermediate reward contrasts $\Delta\gamma$ (shown for two commitment costs): small contrasts provide little information, while large contrasts saturate inference, both reducing heterogeneity. (C) Rapid environmental switching suppresses heterogeneity by limiting achievable inference accuracy. (D) Joint variation of cost and reward contrast reveals a wedge of intermediate ecological conditions in which division of labor and structured heterogeneity are selectively favored.
• Figure SI.1: Sublinear scaling of risk-tolerant individuals across ecological conditions.(A) The number of risk-tolerant individuals as a function of group size $N$ for different commitment costs, $\lambda$. (B) Scaling of risk-tolerant individuals with group size under varying reward conditions, $\Delta \gamma = \gamma_+ - \gamma_-$ (difference in high- and low-reward probabilities). (C) Scaling under different environmental switching rates, $\epsilon$. Across all conditions, most individuals are risk-tolerant in small groups, whereas in larger groups the number of risk-takers increases sublinearly and logarithmically with group size.