Self-Organised Criticality in Animal Collectives

Principal Investigators:

Jens Krause
Pawel Romanczuk

Team Members:

Yunus Sevinchan (Postdoctoral researcher)
What do animal collectives and the brain have in common?
Research Unit 1, SCIoI Project 41

Previous research has suggested that collective biological systems, ranging from the brain to animal groups, should operate in a special parameter region, where the system behaviour undergoes a qualitative change, at a so-called critical point (Munoz 2018). In animal collectves, for example, groups make rapid transitions from disorder to order (Figure 1) at a critical point and this often happens in response to an increase in predation risk. Mechanistically, the assumption is that small changes in the interaction dynamics between group members can result in these rapid transitions. Functionally, this has been linked to the fact that critical systems exhibit unique properties like maximal responsiveness to external stimuli and optimal propagation of information within the group which could provide protection from predators.

However, direct empirical support for critical dynamics and its functionality in animal groups (in particular in the wild) is limited. In this project we will test whether large, naturallyoccurring schools of the freshwater fish Poecilia sulphuraria operate close to criticality across different ecological contexts thereby maximizing their capability to discriminate among different environmental stimuli. We will combine experimental observations with computational approaches to understand how the fish tune the responsiveness of the school in a selforganised way through individual-level changes in sensitivity to predator stimuli and social cues.

These insights on the analytic side will be used to develop on the synthetic side novel bioinspired algorithms for decision-making in unpredictable and noisy environments. Our project will help us to gain understanding of fundamental mechanisms underlying self-tuning of macroscopic collective states of animal groups via adjustment of behaviour of individuals with limited local information and provide intelligent decentralized synthetic algorithms for SCIoI’s third target behaviour.