The wisdom of crowds of ants

Ant colonies operate without central control. If we understand how they do this, we may be able to understand other systems that operate without leaders, from the brain to the Internet / Deborah M. Gordon 

• Ant colonies operate without leaders. They organize their activity through simple interactions based on smell.
• The system of interactions used by the colony is related to its ecology.
• Insights into the collective behavior of ants can illuminate other systems that operate without central control.

The article is published with the approval of Scientific American Israel and the Ort Israel network

In the 2015 blockbuster film, Antman, a suit capable of shrinking a person to the size of an insect is described. The inventor of the suit, the scientist Hank Pym, claims in the film that ants can perform amazing feats, but they need a leader to tell them what to do. Pim wears a small device behind his ear that allows him to order the ants to form a warrior battalion that helps the hero of the plot, who has shrunk to the size of an ant, defeat an evil genius.

The idea that ants have female leaders who set their agenda and orchestrate their activities seems logical to us because of the hierarchy found in many human organizations, and it provides a convenient premise for Hollywood movies whose heroes are humans. There is only one difficulty: the idea is wrong. Ants never march in unison, in uniform obedience to a single command. In the real world, individual ants often act haphazardly and seemingly unsuccessfully, without any sense of common purpose. These actions together allow the colony to find and collect food, build nests, create paths and bridges, protect their surrogate plants from weed killers or cultivate gardens - all without supervision. Ants do not need a leader, and no ant tells its mate what to do.

Ant colonies are not the only systems in nature that operate without central control. Collective behavior, without instructions from on high, occurs everywhere, from the flight of a flock of starlings in the sky, to the networks of neurons in the brain that allow you to read this sentence, to molecules that work with genes to make proteins. All the many outcomes of collective behavior are obtained through simple interactions between individual participants, and one is whether they are ants, birds, neurons, or molecules.

When I was a PhD student and started researching systems that operate without central control, I was looking for a system where it was easy to observe the interactions taking place within it, and ants are everywhere. There are more than 14,000 biological species of ants scattered across every possible terrestrial habitat on Earth. They build nests in the ground, in hollow branches and acorns, under stones and leaves hanging high in the treetops. Their diet is very diverse, from fish to mushrooms to other insects. All ant species exhibit collective behavior, so they provide an excellent opportunity to study what evolutionary changes this behavior has undergone to solve the diverse ecological problems that ant colonies encounter.

The studies I conducted On different types of ants in a variety of ecological conditions, from the desert to the tropical forest, it is shown that each type uses interactions in a different way, for example to increase the level of activity, to slow it down or simply to continue as usual. These findings suggest a correlation between the ecological state and the way simple interactions affect collective behavior. It is possible that several different evolutionary pathways, in a variety of systems operating without central control, have produced similar algorithms to solve similar environmental challenges.

Simple interactions

All types of ants have several characteristics in common, including the way they perform their tasks. Ants live in colonies consisting of many sterile workers (the ants we see wandering around outside) and one or more fertile females that do not leave the nest. Although the fertile females are referred to as royalty, they have no political power - all they do is lay eggs. The queen, nor the other ants, have any ability to assess what should be done in a certain situation or to give orders to others.

All ants have a developed sense of smell and with their tentacles they are able to distinguish between hundreds of different chemicals. They are also covered with an oily coating of hydrocarbons that protects them from drying out. When one ant touches its prey with its tentacles, it assesses the scent carried by this oily coating. Scientists know that there are species in which the chemical composition of these hydrocarbons, called cuticular hydrocarbons, varies according to environmental conditions. A harvester ant wandering in the hot desert sun smells different than an ant that spends most of its time in the nest. As a result, the ant's scent reflects its duties.

Michael Green from the University of Colorado at Denver and I tried to understand how ants make connections with each other using their tentacles. To this end, we conducted an experiment in which we coated tiny glass beads with extracts of cuticular hydrocarbons from ants performing certain tasks, and placed them in ant nests. We discovered that when one ant touches its prey with its tentacles, it simply receives a message that it has encountered an ant with a given odor and nothing more. It turns out that the rate of interactions is the factor that determines how the insect reacts. In our experiments we were able to change the behavior of the colony by changing the frequency with which the ants encounter the glass beads.

How do ant colonies organize their work using simple scent interactions? During the last 30 years I have been researching harvest ports in the Southwest USA. In this species, the need to conserve water appears to have been the driving force in the evolution of the process that uses interactions to regulate foraging. The harvest ant eats seeds of grasses and annual plants, which provide the colony with both food and water. But the colony must waste water to get water. Forager ants lose water simply because they are outside looking for seeds. A forager ant therefore does not leave the nest before it has had enough encounters with foragers returning to the nest with food. Since each collecting ant continues to search until it finds food, the feedback it receives from the foragers returning to the nest creates a correlation between the gathering activity and the amount of food: the more food there is, the shorter the search time: the foragers return to the nest faster and more foragers leave the nest in search of food.

The long-term study I conducted on a population of harvester ant colonies allowed us to understand how evolution shapes their collective behavior. To understand how natural selection works now, we needed to know if the way a colony regulates its foraging activity affects its ability to produce daughter colonies. The first step was to determine which colonies are the offspring of which mother colonies. No one has tested this before in ant colonies. But since 1985 I have been monitoring a population of about 300 colonies at a site in southeastern Arizona. Every year I locate all the colonies that were there the year before, say goodbye to those that died and map the newly founded colonies. These long-term data show that a colony lasts 25 to 30 years. Every year there is a mating meeting where the males, who only live a short time until mating, meet with unmated queens from all the colonies in the population. After mating, the males die, and the newly mated queens fly away and found new colonies. Each queen lays a new set of sterile workers, and once the colony is large enough, she also lays fertile females and males. Thus, every year, for the rest of her life, with the help of the sperm from that original mating session. Krista Ingram from Colgate University, Anna Pilko From the University of California in San Diego and I were able to link daughter colonies with their mother colonies with the help of DNA collected from about 250 colonies and to learn how food gathering activity is related to successful reproduction.

We found that colonies that have daughter colonies tend to be the ones that conserve water by limiting food gathering on hot, dry days, and sacrificing food consumption to conserve water. This result surprised us because many studies on animals assume that the more food there is, the better. But it turned out that the colonies that for years I saw as unreliable and weak colonies, because they rarely gathered food on hot and dry days, were the great-grandmothers, while the most active colonies, which went out to gather every day, failed to reproduce. Because colonies can store seeds for long periods of time, there is no survival cost to avoiding foraging on certain days.

Natural selection only acts on traits that are passed from parents to offspring, and there is intriguing evidence of the heredity of collective behavior in the harvester ant: daughter colonies are similar to their parent colonies in choosing the days when they reduce gathering activity. Thus, our findings provided the first demonstration, to my knowledge, of contemporary evolution of collective behavior in a wild animal population.

Ecological solutions

Different species of ants demonstrate how the system of interactions used by the species is related to its ecology. I also study gliding ants (Cephalotes) the animals in the trees in the tropical forests of western Mexico. In the tropics, the air is very humid, and food is plentiful, so the operating costs of food gathering are low compared to the desert. On the other hand, the competition in these areas is great because many other ant species utilize the same resources. I discovered that colonies of gliding ants create collecting paths in the trees along which the ants regularly move from one food source or nest to another. Unlike the harvester ant, gleaning glider ants continue to forage as long as there is no particular interaction that causes them to slow down or stop. For example, interactions with ants of other species inhibit activity. A gliding ant leaving the nest is expected to go further and further along the track, unless it meets an ant of another species. One ant of the kind is enough pseudomyrmex, which patrols back and forth on a branch, flashed and designed like a racing car, to completely block a certain branch in the path of the gliding ants, who are robust but more timid, if it encounters a large number of them. Colonies are so consistent in maintaining a smooth flow of ants along the route when the area is clear, and resuming the flow when the threat is gone, that it may be easiest to avoid confrontations.

Simple interactions between ants thus create the network of routes of the ants that glide through the dense vegetation in the forest tops. These interactions make the network both durable and flexible. Each ant marks its route while walking with the help of a chemical pheromone, and it follows the scent left by ants in front of it. Saket Naolahha from the Salk Institute for Biological Studies and I are trying to understand the algorithm that ants use to maintain and repair their tracks. When an ant reaches a junction between one branch and another, it tends to follow the path most ants have taken recently. Often, a rickety bridge between one stem and another collapses due to a gust of wind, a passing lizard, a rotting branch or, sometimes, the experimental intervention of my scissors. Ants recover quickly. It seems that when they reach the first broken end, they go back to the previous bifurcation point and look for a pheromone trail from there, until they create and eventually prune a new trail that will connect with the other side of the track.

Collective behavior of ants has evolved in response to the way resources, such as food, are distributed in the environment, as well as in response to the costs of gathering and the behavior of other species the ants encounter. There are resources grouped together in one place and others scattered randomly. Ants of many species excel at exploiting clustered resources, such as picnic sites. They use interactions based on pheromones that cause one ant to follow the other and create recruitment routes. Recruitment is a logical process when resources are pooled. After all, where there are sandwiches, there are likely to be cookies as well. On the other hand, ants searching for scattered resources, such as seeds, do not use recruitment routes, because finding one seed does not guarantee finding another seed nearby.

Even to find food in the first place, collective behavior is necessary. Since ants work mainly with the sense of smell, an ant has to get close to the food to find it. As the food is spread over a wider area, the ants have to scan a larger area. But as the food is hidden in different places, the ants have to scan the surface of the ground more thoroughly. I discovered that the species known as Argentine ant deals with this contradiction very effectively by adjusting the search paths according to density. When there are several ants in a small area, each ant follows a winding path, which allows it to scan the area thoroughly. But when there are several ants in a large area, they walk in straight tracks, allowing a whole group to scan a larger area. Individuals can sense crowding through a simple signal: the pace of interactions with others. The more contacts there are between tentacles, the more tortuous the ants' walking path. The Argentine ant an invasion to Mediterranean climate zones around the world. The efficiency with which it first reaches new food sources may explain why this invasive species tends to outcompete native species wherever it invades.

Go to the ant, learn its ways and be wise

The ways in which ants use simple interactions to thrive in a given environment can provide solutions to problems that arise in other systems. The computer scientist Balaji Prabkar Stanford University and I noticed that the harvest ports regulate harvesting activity using an algorithm similar to the TCP/IP protocol used to regulate data traffic on the Internet. We called the analogy "Internet". This protocol was originally designed for an environment with high operating costs: in the early days of the Internet, the network was small and did not have much redundancy. Therefore, it was essential to ensure that no packet of information was lost. Just as pickers won't go on tour if they don't have enough interactions with pickers that have returned with food, so a packet of information won't leave the source computer if it doesn't get confirmation from the router that the previous packet had enough bandwidth to continue moving towards the destination. Presumably 130 million years of ant evolution has produced many other useful algorithms that humans have not yet thought of and that could help us find ways to organize information networks with simple interactions that use a minimal amount of information.

I believe that we will see a similar correlation between an algorithm and an ecological situation in many other types of collective behavior. For example, cancer develops in response to conditions in its microscopic environment. It must be assumed that a type of cancer that tends to send metastases to a certain type of tissue undergoes an evolution that will allow it to use the resources concentrated in this particular tissue. These cancers, like the ant species that have evolved to use resources concentrated in certain places, may be the ones that tend to send cells back to the original tumor to recruit more cells, as breast cancer cells tend to do. In such a case, cells that recruit other cells to food resources would be the best target for toxic bait.

In biology and engineering there is great interest in how collective behavior relies on simple interactions. It is increasingly becoming clear that such interactions respond to changing conditions. Systems biology, a field built on a hundred years of research that showed in detail what happens inside the cell, is now turning its focus to interactions between cells, and is aided by the amazing developments in the field of imaging. In neuroscience, new methods make it possible to locate patterns in the timing of the electrical activity of thousands of nerve cells. We humans can see certain types of movements and hear certain types of sounds because the neural pathways in our brains have evolved to respond collectively to certain environmental features such as the rate at which important objects, such as parents or predators, often move, and the frequency range that is most important to us to hear. Engineering systems have also evolved. The enormous increase in the scope of the Internet and the number of devices connected to it, as well as the speed of interactions, require new, distributed solutions.

Scientists can now look for trends in how different natural systems have evolved similar collective behavior to face similar ecological challenges. We may be able to apply this knowledge to intervene in processes that operate without central control, and perhaps also solve some of the problems of human society at the same time.