A recent study looking at how colonies of ants regulate their foraging behaviour has caused a bit of a buzz online. A lot of the coverage has focused on a similarity highlighted in the press release, which says that the ants “determine how many foragers to send out of the nest in much the same way that Internet protocols discover how much bandwidth is available for the transfer of data”. While it’s wonderful that the study has received so much attention, I can’t help but feel that the really interesting aspect of this study has been overlooked in the excitement about the “anternet”. While the similarity between the two systems is striking, I’m more fascinated by a basic difference: unlike our computer networks, the regulation system in ants isn’t purposefully designed but emerges from uncoordinated decisions made by individuals.
Ants usually live in nests that can contain anywhere from a few dozen individuals to a few million. Food for the colony is provided by older ants who leave the nest to forage. Some species make foraging trails: long, tiny highways full of ants shuffling back and forth to bring food to the colony. These form when an ant returning with food lays down a special chemical (called a pheromone) on her way back to the nest; since the ants have a slight tendency to move towards the pheromone, they follow the chemical path and reinforce it, soon forming a trail. This shows how decisions made by individual ants (to follow the pheromone) can sum together to generate an overall behaviour that seems purposeful (forming a foraging trail). This kind of behaviour makes it possible for colonies to solve challenging problems, like finding the shortest path to food, without any kind of centralized planning or decision making.
Finding food isn’t the only challenge facing an ant colony; they also have to deal with a variety of tasks ranging from maintaining the nest to moving to a new home. Just like everything else in the colony, the allocation of workers to different tasks has to be coordinated without any central planning. Understanding how this kind of decision making system works and evolves is an important and exciting field of research. A team of scientists at Stanford University addressed this question by investigating how colonies of the red harvester ant Pogonomyrmex barbatus regulate their foraging behaviour. Since P. barbatus feed on seeds which are scattered by winds and flood rather than being collected in patches, they forage individually rather than making trails. An ant returning with a seed drops it in the narrow entrance tunnel to the nest; she is then ready to go foraging again while another ant carries the seed deeper into the nest. Searching for food in the dry deserts of Arizona is an expensive proposition, so waiting ants decide when to leave the nest based on brief interactions with returning foragers, optimizing the colony’s resource usage.
To understand how this regulation works, the scientists measured how frequently ants went out foraging during a 20 minute observation period; for a few minutes out of this period, they prevented returning foragers from reaching the nest. By comparing the rate before, during and after the interruption, they discovered that the foraging rate depends on how often foragers return with food. This is very effective approach to regulation. If food is plentiful, the foragers will return quickly and recruit more ants to help; on the other hand, a scarce food supply will mean that foragers return more slowly, so fewer ants start foraging. The team used a mathematical model to test the idea that this behaviour was the result of decisions made by individual ants. In their model, ants leave the nest to forage at a certain rate which increases when other workers return and decreases as time passes, though it never drops below a fixed minimum. Since the results from the simulations were reasonably similar to the observations, it seems likely that the ants use something like this mechanism to decide when to forage.
Most of the media coverage about this study has focused on the similarity with the way traffic is regulated on computer networks. Professor Balaji Prabhakar, one of the authors of the paper, noticed the similarity with the Transmission Control Protocol (TCP) used by computers to find out how much bandwidth is available. Data sent using TCP has to be acknowledged by the receiving computer; the acknowledgement rate is used to estimate the quality of the connection. A low acknowledgement rate means there isn’t much bandwidth available, so the sender slows down the transmission rate. While this is an excellent example of ants and humans converging on the same solution to a problem, I think the really exciting part is the difference between the two systems.
While TCP was explicitly designed to solve a problem, the mechanism used by ants is emergent. Emergence, a feature of so-called “complex systems”, is when interactions between individuals lead to the appearance of a new behaviour or property at higher levels. Individual ants aren’t trying to find the shortest path to a food source or to optimize their colony’s foraging behaviour; they’re simply following a scent they like or going out to forage shortly after a nestmate returns with food. Nevertheless, the overall effect of these decisions is colony-level behaviour that looks just like a purposeful attempt to solve certain problems. This sort of emergence is an endless source of fascination to me. Complex systems tend to be both robust and flexible, which makes them ideal for everything from decision making to regulating development. Complexity theory provides a way to understand the behaviour of a wide range of dynamic systems, ranging from traffic flow and economics to weather. The plentiful computing resources available today have made it far easier to study these sorts of systems and to begin to understand the rich patterns and dynamics that emerge from this exhilarating and profoundly beautiful approach.
Balaji Prabhakar, Katherine N. Dektar, & Deborah M. Gordon (2012). The Regulation of Ant Colony Foraging Activity without Spatial Information PLoS Computational Biology, 8 (8) DOI: 10.1371/journal.pcbi.1002670
(Since it’s published in the PLoS journals, this article is open-access, meaning anyone can read it)