Written for my Honors College science colloquium course: Bio Breakthroughs. Based largely on this really cool TED talk, as well as the RadioLab episode “Emergence.”
Taking a few minutes to watch a population of ants might lead one to conclude that they are a very unintelligent species. They move around seemingly at random, appear to have no goal in mind, and sometimes even push one another around. Even though as individuals ants may seem really unintelligent, looking at the colony as a whole reveals just what they are capable of. Ant colonies can efficiently find and transport food, build expansive underground nests, protect those nests from flooding conditions, and more. The intelligence of the group appears to be greater than the sum of its part, a phenomenon known as swarm intelligence.
Generally, “swarm intelligence” is used to refer to the collective behavior of any large system. The term was originally coined by Gerdo Beni and Jing Wang to describe artificial intelligence in robotic systems. However, they note that it can occur in both natural and artificial systems, and modeled many of their ideas off of what has been observed in the natural world. (705) Swarm intelligence arises from de-centralized groups of individuals. That there is no one specific leader giving instructions to the whole and instead, order arises from the decisions and interactions of individual members. Swarm intelligence occurs in a wide array of species including the flocking patterns of birds, schools of fish, and even in bacterial growth patterns.
One place that swarm intelligence can be readily observed is in the foraging activity of ant colonies. For populations of ants, finding sources of food and transporting it back to the nest is crucial to their survival. However an individual ant is not very good at this task; individuals are often seen “wandering” or moving randomly about with no goal in sight. However, when observing the larger population as a whole, the entire process becomes much more organized. The ant colony quickly and efficiently locates a food source and determines the most efficient path from the nest. Soon, a trail of ants leads from the nest to the food, and back—with every individual ant following along in line. Individual ants interacting with one another have no “global view,” that is they do not know the actions of every other individual in the colony nor the location of the food source. However, these individual interactions tend to lead the entire colony toward efficient choices and behaviors. (Prabhakar, Dektar, and Gordon) Videos of ant colonies in laboratory settings consistently show groups of ants that move seemingly at random, but rapidly form, and follow, efficient paths. (LeBlanc and Couzin) This apparent creation of order out of disorder is what makes swarm intelligence such a unique occurrence.
Some scientists have suggested that there are evolutionary pressures that promote swarm intelligence in certain populations. Swarm intelligence allows the population to be very efficient in expending energy in order to gather enough resources to support its individuals. (Gordon) For instance, colonies of ants exhibit natural feedback loops in their behavior. If there is much food available and many individual ants are returning to the nest with food, more ants will follow them to gather more food. However if food is scarce and few ants return with food, fewer ants will expend energy following them to gather more food. (Gordon 2) Initially, swarm behavior may seem counterintuitive from an evolutionary standpoint, in which Darwin’s theory is generally thought of as the “survival of the fittest.” Richard Dawkin’s The Selfish Gene offers another context to consider evolution in which swarm intelligence might make more sense. He contends that natural selection simply favors “survival machines” that will most likely allow the DNA “replicator molecule” to survive. (Dawkins 26) Individual intelligence is generally considered to increase the fitness of a species, but with swarm intelligence the fitness of an individual is diminished but vastly increases the survival likelihood of the colony. Arguably, ant species are incredibly successful evolutionarily because of swarm intelligence.
Like many other discoveries in biology, swarm intelligence was not discovered by a single scientist in a single breakthrough moment. Instead, it slowly developed over several decades through the work of many different scientists. The very notion that order could arise out of disorder is closely related to religion and creationism. For this reason, it was essentially ignored by scientists for several years as a topic not worth studying. (Abumrad and Krulwich) Swarm intelligence could be observed in many different species, but for hundreds of years it simply was not considered an area worth of scientific study. This changed in the 20th century when one scientist became determined to figure out how ant populations communicated with one another to follow the same paths. Edward Wilson of Harvard University cultured fire ants, and in a very crude process, dissected the ants into individual organs. He then smeared each organ in a straight line to see which one the ants would follow. Eventually he determined that the ants were secreting a chemical everywhere they went, which other ants could then smell and know to follow. (Wilson) These eventually became known as pheromones and are critical to understanding how swarm intelligence functions in ant populations.
There are a few different general methodologies to study swarm intelligence in different species. One such method is lab-based culturing and observation of specimens. One example of this is Deborah Gordon, a scientist at Stanford University, who cultures colonies of ants within her lab. The laboratory setting allows for the careful control of several variables for experimentation. Gordon put the ants into different environments, changed the amount of food available in each environment, and observed the ants’ behavior in each. By doing so, she determined that the colony as a whole was very efficient in its usage of energy and only sent several individuals to forage for food when a large amount was available. (Gordon) Another example of scientists studying swarm intelligence in a lab setting is the work of Iain Couzin. He used cameras to record ants as they moved from their nest to a food source, and then made computer models and animations of their paths. His findings showed initial random movement that quickly organized into a few distinct paths. (LeBlanc and Couzin) Control of experimental variables, such as food availability, and the ability to closely record behaviors are really only possible within the lab setting, which is why it is an important methodology in the study of swarm intelligence.
Another way that scientists study swarm intelligence is through in situ field observation. In this setting it is less possible for scientists to control experimental variables, but it does offer the ability to observe many more elements that might contribute to the behavior of the species. For example, Balaji Prabhakar, Katherine Dektar, and Deborah Gordon observed populations of harvester ants in their natural habitat. They observed that these ants made brief antennae contact as they entered or left the nest in order to detect how much food each ant was returning with. This detection was associated with the colony’s self-regulation of foraging behavior. (Prabhakar, Dektar, and Gordon) By observing the species in their actual environments, scientists were able to discover another aspect of the behaviors associated with swarm intelligence.
A third method that is used by scientists in the study of swarm intelligence is the use of mathematic models. These methods are more abstract than the previous two, and help provide a lot of the technical groundwork to understand the effects—but not necessarily the causes—of swarm intelligence. One example of this is a paper that was published in the SIAM Journal on Applied Mathematics about firefly populations in Southeast Asia. Fireflies in this region are unique because large groups of individuals will “sync up” their flashing at night. Entire sections will light up and turn off, all in perfect synchronization, and without any one particular leader. The paper considered the fireflies as “common oscillators,” and outlined mathematic models which demonstrated that any decentralized network of oscillators will eventually sync up if they are able to respond to feedback from nearby individuals. (Mirollo and Strogatz) Interestingly, the paper also suggested that such synchrony would either never occur if they were coupled equally to one another through a common resistor—such as if there was a single “leader” firefly. Steve Strogatz, one of the authors of the paper, wrote another paper that attempted to model the self-organizing networks. Not only does he offer mathematical models of how networks of fireflies organize, he also suggests that the same models can apply to areas outside of biology such as social networks and other man-made technologies. These theoretical modeling methods help apply the biology studies of swarm intelligence and bring those discoveries into other fields. For this reason, mathematic models are another important method used when studying swarm intelligence.
The discovery of swarm intelligence in various species has implications both within biology, and in other fields entirely. As one example, scientists relied on the basic ideas of swarm intelligence to crowdsource the discovery certain protein structures. An online game called “FoldIt” presents protein folding problems as a game that anyone can play and take part in. Higher scores are given for configurations that lead to lower energy levels, and thus represent a likely model of how the protein is structured. Even players who have no experience with biology at all are able to take part and contribute to the study. (Cooper et al.) It took the FoldIt system only 10 days to discover the structure of an HIV-causing protein, something that had evaded scientists for nearly 15 years. (Khatib et al.) Much like the colonies of ants, each individual player may not have known much about protein folding, but as a collective group they were able to discover something previously unknown to science.
Outside of biology, the ideas of swarm intelligence can be applied to areas such as crowd behavior and pathfinding. For example, an article in Scientific American described the behavior of individuals trying to escape from the World Trade Center on 9/11. It showed that rather than acting in an uncontrolled panic, people “show a remarkable ability to organize themselves and support one another.” (Drury and Reicher) This has potential implications to safely design buildings and structures that consider likely flow paths of large groups. These same ideas can be used in non-living applications as well. In a highly generalized and over-simplified sense, the Internet works by connecting different computers to one another and sharing information between them. However, finding the most optimal path between systems and between networks can be difficult and time-consuming. However, some scientists have used swarm intelligence and a method called Ant Colony Optimization (ACO) to program systems that quickly and efficiently find such paths. In ACO, each “hop” along the path chooses the next destination, but never seeing the whole path. This saves computational power and eventually leads to efficient routing paths. (Sharma and Khurana) These cases represent just a few examples of the applications that swarm intelligence science has outside of the field of biology.
It is very easy to quickly look at a species such as ants, fireflies, or bees and write them off as being simple and unworthy of study. Just because an individual may seem unintelligent, its colony as a whole may exhibit incredibly behaviors and tendencies. Swarm intelligence is not just an exciting new discovery, but underscores the importance of looking closely at the world around us and studying the way it works.
Abumrad, Jad, and Robert Krulwich. Emergence. N.p. Audio Recording.
Beni, Gerardo, and Jing Wang. “Swarm Intelligence in Cellular Robotic Systems.” Robots and Biological Systems: Towards a New Bionics?. Ed. Paolo Dario, Giulio Sandini, and Patrick Aebischer. Springer Berlin Heidelberg, 1993. 703–712. link.springer.com. Web. 28 Feb. 2017. NATO ASI Series 102.
Cooper, Seth et al. “Predicting Protein Structures with a Multiplayer Online Game.” Nature 466.7307 (2010): 756–760. www.nature.com. Web.
Dawkins, Richard. The Selfish Gene: 40th Anniversary Edition. New York, NY: Oxford University Press, 2016. Print.
Drury, John, and Stephen D. Reicher. “Crowd Control.” Scientific American Mind 21.5 (2010): 58–65. Print.
Gordon, Deborah M. “The Ecology of Collective Behavior.” PLOS Biology 12.3 (2014): e1001805. PLoS Journals. Web.
Khatib, Firas et al. “Crystal Structure of a Monomeric Retroviral Protease Solved by Protein Folding Game Players.” Nature Structural & Molecular Biology 18.10 (2011): 1175–1177. EBSCOhost. Web.
LeBlanc, Simon, and Iain Couzin. “Pheromone Trail Networks in Ants.” Couzin Lab – Collective Animal Behavior. N.p., 22 Sept. 2009. Web. 28 Feb. 2017.
Mirollo, Renato E., and Steven H. Strogatz. “Synchronization of Pulse-Coupled Biological Oscillators.” SIAM Journal on Applied Mathematics 50.6 (1990): 1645–1662. Print.
Prabhakar, Balaji, Katherine N. Dektar, and Deborah M. Gordon. “The Regulation of Ant Colony Foraging Activity without Spatial Information.” PLOS Computational Biology 8.8 (2012): e1002670. PLoS Journals. Web.
Sharma, Parveen, and Neha Khurana. “Study of Optimal Path Finding Techniques.” International Journal of Advancements in Technology 4.2 (2013): 124–130. Print.
Wilson, Edward O. “Chemical Communication among Workers of the Fire Ant Solenopsis Saevissima.” Animal Behaviour 10.1 (1962): 159–164. alliance-primo.com. Web.