Self-organized Regulation of Group Size and Number in Natural and Artificial Collectives
This paper proposes and validates a decentralized mechanism based on local perception and individual preferences that enables both artificial and natural collectives, such as white-nosed coatis, to self-organize and actively regulate their group sizes and numbers to match environmental demands.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
In the natural world, from the swirling schools of fish to the bustling colonies of ants, living things rarely act alone. They gather, split, and re-form into groups that shift and change like clouds. This ability to organize and reorganize is vital. A group can offer protection from predators or help find food, but if it grows too large, it becomes a burden, with members fighting over limited resources or spreading disease. Scientists have long understood that there is a sweet spot, an ideal size where a group functions best. However, a deeper mystery remains: how does a collection of individuals, each with only a limited view of their immediate surroundings, know when to stay together and when to break apart? They cannot see the whole population, nor can they count how many groups exist in total. Yet, in both nature and the machines we build, groups seem to find a way to regulate themselves, settling into a stable pattern of sizes and numbers without a central leader giving orders.
A team of researchers set out to solve this puzzle by asking a specific question: Can a group of individuals, using only local information and their own personal preferences, self-organize to control both the size of their groups and the total number of groups? To find the answer, they first looked at the simplest possible behavior: following. In many animal groups, individuals simply move toward a neighbor. The researchers built a mathematical model to see if this simple rule alone could create stable groups. They found that while following behavior could indeed cause individuals to clump together, it was not enough to actively regulate the group size. The groups would form, but their sizes would be a passive result of how crowded the environment was, not a conscious adjustment to reach an ideal number. The animals would not know if they were in a group that was too big or too small.
To fix this, the researchers proposed a new mechanism based on a simple internal feeling: a personal preference for a specific group size. Imagine each individual carrying a quiet, internal sense of what the perfect group size is for them. If the group they are in feels too small, they look for others to join. If it feels too large, they decide to leave and start a new group or find a different one. This idea was turned into a computer simulation where virtual agents, representing robots or animals, moved around and made decisions based on this internal preference. The researchers tested this system under three different conditions of communication. In the first, the agents could only see where others were, with no idea of how many were in a group. In the second, they could ask for the group size just once before deciding to join. In the third, they received a constant, updated stream of information about the group size.
The results showed that the system worked best when the agents had access to information about the group size. When the agents could only see positions, the groups oscillated wildly, growing too large and then shrinking too small, never finding a stable rhythm. When the agents could check the group size even just once, the system became much more stable, and the groups settled closer to the desired size. The most effective method was continuous communication, where agents constantly knew the size of the groups they were considering. In this scenario, the virtual robots successfully formed three perfect groups of fourteen individuals each, eliminating the chaotic swinging between too big and too small. The researchers measured how long the robots spent in groups that were larger than the target size, finding that continuous information reduced this "oversize" time by more than ninety-five percent compared to the position-only method.
To see if this artificial mechanism mirrored reality, the researchers turned to real-world data from wild white-nosed coatis, a mammal related to the raccoon. These animals live in social bands that frequently split and merge. The team used GPS tracking data from two different coati groups to calibrate their model. They did not program the coatis with a specific rule; instead, they inferred what the individual preferences might have been based on the actual group sizes the animals experienced in the wild. For one group that split evenly, the model assumed all individuals shared a similar preference. For another group that split unevenly, the model assigned different preferences to different individuals. When they ran the simulation with these real-world preferences, the virtual coatis reproduced the statistical patterns of the real animals. The model successfully matched the observed number of subgroups, the typical sizes of those subgroups, and the way the groups transitioned between different configurations.
The study suggests that the complex, shifting social structures seen in nature and the efficient organization of engineered systems might rely on the same fundamental principles. It is not necessary for every individual to know the total number of animals or the total number of groups. Instead, stability emerges when each individual simply compares their current situation to a personal standard and acts accordingly. The research demonstrates that a decentralized system, where everyone makes decisions based on local perception and a personal goal, can solve the difficult problem of regulating both group size and group number simultaneously. While the study confirms that such a mechanism is mathematically possible and consistent with real animal data, it does not prove that wild coatis use this exact method. Rather, it shows that this simple, local strategy is a plausible and powerful way for collectives to adapt and thrive without a central commander.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.