Individual decision-making and collective animal behavior

A swarm of ants, a flock of birds, and a school of fish are all examples of collective behavior in animals. These complex behaviors emerge without any clear leader, as animals respond to the environment and movements of their neighbors.

I’m Matthew Hardcastle, and in this episode of Science Sessions we’ll explore advances in the understanding of collective animal behaviors. Animals in a group have traditionally been modeled as particles influenced by physical forces. While these simple models can successfully reproduce many natural behaviors, they provide little insight into how the decision-making processes of individual animals influences the actions of the collective. In a recent PNAS article Conor Heins, a machine learning researcher at Verses AI, a cognitive computing company in Canada, and the Max Planck Institute of Animal Behavior in Germany and his colleagues incorporated cognitive processes into a model of collective behavior.

Heins: In our model, we actually consider each little particle in, say, a school of moving fish or flocking birds, and we imagine that they're an agent that's actually undergoing a process of ingesting sensations, doing some kind of cognitive processing on their sensations, on their perceptions, and then taking decisions to act.

What you can quickly reveal is that these agents are capable of the exact same kind of collective patterns of motion that we see in the earlier physics-based models, things like directed motion where all the agents are moving together. You also see milling, where they're kind of moving with high angular momentum around some center of mass. Now, we can also control and understand the actual causes of these behaviors more from a cognitive level.

Let's say one individual at the edge of the school is suddenly informed about the presence of a rapidly approaching predator. Then, often with the older self-propelled particle models, the rest of the school will not be able to respond in time. Only agents that are endowed with this ability to update their beliefs about the statistics of the environment are able to sensitively respond to fluctuations.

PNAS: Classical models often place particle-like animals on undefined planes, but physical boundaries are a facet of both laboratory settings and the natural world. In another PNAS article, Eva Kanso, a mechanical engineer at the University of Southern California in Los Angeles, and her colleagues analyzed how confinement influences collective behavior.

Kanso: The immediate behavior that occurs when you add confinement is they go around the tank wall. Not super surprising, but it's good that the model reproduces experimental observations. The double milling was very surprising, because now the school kind of splits randomly into two groups. Maybe 60% of them will go in one direction, and the other 40% will go in the opposite direction.

It has been observed experimentally that with a group of fish put in a tank with geometric confinement, they had seen this ability to switch back and forth between schooling and milling. What we see from the model, it's the emergent behavior at the collective level that is switching back and forth, without requiring changes in how the individual responds to its neighbor. We could speculate that maybe this could be useful for a group of fish that could transition its behavior from exploring the environment, going around, seeing what's out there, and then coming back and going into those milling states.

PNAS: Apart from the physical environment, animals in a group are also influenced by the movements of their immediate neighbors. In a PNAS article, Andreu Puy, a physicist at the Polytechnic University of Catalonia in Spain, and his colleagues considered the role of speed in the leader–follower dynamics of schooling fish.

Puy: We use fish of the species black neon tetra, which is a small freshwater species. We recorded their movements with an overhead camera. Then, we digitized their trajectories. We found a pattern where fish appear to align only with faster neighbors and ignored slower neighbors. Additionally, we analyzed leader–follower relationships in the experimental data and demonstrated that faster neighbors transmitted information about their direction and speed to slower neighbors.

There are several biological reasons why fish might pay more attention to faster-moving neighbors. First, focusing on quicker neighbors can reduce the cognitive load of fish, allowing them to simplify their decision-making. Secondly, faster fish often carry more relevant information. Fish that speed up might be responding to an urgent situation, such as the presence of a predator or discovery of food.

PNAS: In addition to visual stimuli, animals perceive a variety of sensations from their environment that can influence their individual and collective decision-making. In another PNAS article, Daniel Kronauer, an evolutionary biologist at Rockefeller University in New York, and his colleagues explored how a colony of clonal raider ants collectively responds to rising temperatures.

Kronauer: The ants are usually settled by default, right? They form a pretty dense nest cluster. There's always a few ants that kind of explore the arena. Then, when you increase the temperature, you'll see that the ants start to move around. They become restless inside the nest cluster. And then at some point, when the temperature becomes too high, you'll see that there's some kind of collective decision emerging, in which, then, all the ants decide to pack up and move the nest somewhere else.

It's a collective response, in the sense that the ants are highly correlated in their response. It's an all-or-nothing response. Either the colony evacuates the nest or they don't evacuate the nest. When they evacuate the nest, they all move in the same direction. The larger the colony size is, the higher the temperature perturbation has to be in order to make the ant colony evacuate in a collective way. So, there seems [to be] some kind of balance between the temperature that the ants perceive and then a kind of settling or inhibitory force that scales with colony size.

PNAS: Moving in a group provides several benefits to individual animals, such as defense from predators. Birds flying in a V-formation also save on energy expenditures, due to aerodynamic forces. In a recent PNAS article, Sonja Friman, an environmental physicist at Lund University in Sweden, and her colleagues quantified the energy savings of starlings, which fly in more complex formations.

Friman: So, starlings obviously don't fly in V-formation. They fly in much larger flocks, like up to a few thousand birds, and to the untrained eye, they actually look very chaotic. We used the wind tunnel, where we had two or three birds flying at the same time. We had a camera set up around the wind tunnel, in order to get their positions and their location relative to each other. We had a little backpack on the birds that was measuring their acceleration. We also measured their metabolic cost.

They had a very dynamic way of flying together. They didn't fly at the same spots as if they would be flying alone, which we also tested. Relative to each other, they were actually, on average, finding this V-formation. So, they found a spot relative to the bird that was flying in front of them that they like to fly at. We also found that the birds that flew in this position actually were using less energy than if they were flying alone.

PNAS: In another example of collective behavior, some ant species form structures out of the living bodies of individual ants. In a PNAS article, Daniele Carlesso, a behavioral biologist at the University of Konstanz in Germany, and his colleagues modeled how weaver ants decide to form chains to explore their environment.

Carlesso: We essentially model the chain as being a simple structure composed of the number of ants joining the chain minus the number of ants leaving it. They seem to join at a quite fixed rate, but instead, they are modulating the probability of leaving. So, when they are farther away from the ground, they seem to leave more often.

Ants seem to stop building chains over gaps above 9 centimeters in length. This seems to be modulated by the visual stimuli that the ants perceive at the end of the chain. By adding these individual decisions all together, the ants can actually reach a trade-off between building the chain or not, depending on the cost and benefit that the structure provides them.

We built a very simple apparatus in which the ants try to reach a little platform that was put on a microscope slider. We could lower the platform as the chain grew. So, the ants arriving at the end of the chain were always thinking that they were very close to the ground. Through this mechanism, we could actually trick the ants into building very long chains, even 12.5 centimeters long.

PNAS: By observing their neighbors, animals in a group can extend their own sensory perceptions. However, if one animal perceives a threat where there is none, a false alarm can quickly spread. In another PNAS article, Ashkaan Fahimipour, a community ecologist at Florida Atlantic University in Boca Raton, and his colleagues explored how reef fish minimize the spread of misinformation.

Fahimipour: A really, really common form of misinformation that these animals experience are false alarms. How do animals usually respond to true threats but avoid making these false alarms? We had to know what these animals were taking in, in the first place, from their sensory systems. We actually employed this algorithm to reconstruct the first-person view from our overhead cameras.

When we look at our cameras from across these coral reefs, we see that these false-alarm events are basically happening all of the time. Almost always, these mistakes that propagate spread to a really small number of fish. These animals are sensitive to information that they get from their social ties, but they're tuning this sensitivity up and down, depending on what's happening in the environment. They're tuning their sensitivity to social information way down when their groups just get too large and things are moving too fast.

PNAS: Sometimes, animals that appear to be moving as a collective are not actually responding to their neighbors. In an article published in Scientific Reports, Clare Doherty, an evolutionary ecologist at Ulster University in Ireland, and her colleagues explored the individualism of terrestrial hermit crabs moving in groups.

Doherty: The species that we conducted this study in was Coenobita compressus. They carry out daily migrations from the forest down to the beachfront in the early morning to eat and forage. We were able to basically form sham aggregations of these groups. We used 60 shells, which have been used as proxies, and we were able to attach 60 stand-ins for our crabs to clear fishing lines that were then pulled synchronously in one direction down the beach.

We videoed the experiments overhead by a drone. We would measure the starting point of our focal individual and the end point to get the direction that they traveled, with respect to the direction of the synchronously moving group. What we found from these experiments is that, surprisingly, the group didn't bias the direction of the individual. The interesting thing about this species is that they do live in these portable shells, and that therefore grants them a certain level of autonomy in their movement.

PNAS: The individual decisions made by animals can produce complex, emergent behaviors at the group level. Considering the cognitive processes of individual animals can enhance our understanding of these fascinating behaviors. Thanks for tuning into Science Sessions. If you liked this episode, please consider leaving a review and helping us spread the word.