By the end of this chapter you will:
A little ant is walking on the sandy soil of Arizona looking for food. The food is meant for feeding the brood that is being taken care of by other ants in the colony. Different ants have different tasks in the colony, and the ant we are following is charged with leaving the nest every day in search of food. The amazing aspect of ant colonies is that such a complex organizational structure exists that is not controlled by the queen or a small group of bureaucratic ants. There is also no plan or to-do list that the ants are following. No, the complexity of the ant colony emerges out of the local interactions among the ants. The ant we observe is following a trail of pheromones, indicating that other ants of the colony have found food nearby and dropped the pheromones on their way back to the nest. The use of pheromones is therefore a way in which ants communicate to others “follow my trail and you may find food.” The pheromone trail will evaporate at a certain speed, and will therefore only be of limited use. If other ants do not follow this trail, bring back food and add their own pheromones to the trail, the trail will disappear. But when the trail is enhanced by successful use by others, a highway of ants may emerge. On such a highway, we see one lane of ants empty handed following the pheromone signal, and the other lane of ants bringing food back to the nest.
There is an enormous diversity of ant-species, all of which have some variation in their social organization. Some ant-species produce ants with different physical characteristics that distinguish their role in the colony: workers, foragers, soldiers, etc. In other ant-species, all of the colony’s ants are physically similar and can switch roles when needed. For example, when foraging ants are killed by an ant-eater, the colony will experience a reduction of food being delivered, which then signals other ants to exchange their caring-for-the-brood role for a foraging role.
Ants change the environment, which subsequently changes the behavior of other ants. This indirect influence of agents via the change of the environment is called stigmergy. Another example of stigmergy is the digital trails we develop when we interact with websites. By buying books, renting dvds or listening to music files we leave information “pheromones.” We get recommendations of other books, dvds or music that people “like you” also bought, rented or listen to. These stigmergic interactions can lead to a reinforcement of choices. Popular movies featured on YouTube tend to get even more viewers. If we want to understand how certain books, movies or songs become so popular, we need to look into the various ways choices are influenced and reinforced by others.
Emergence is a macro-level phenomenon, like the functioning of an ant colony, as a result of local-level interactions. Like the cartoon in Figure 2, understanding of macro level phenomena can only be derived by studying the consequences of the behavior of micro level agents.
Other examples of emergence are:
Have you ever been stuck in a traffic jam only to reach the front of the slow-down to discover that there was no obvious reason for it? Sometimes it happens that traffic jams, defined by slow moving cars, develop rapidly in the opposite direction of the flow of traffic. As a consequence you may enter in a traffic jam that was formed many miles ahead. Traffic jams can also form without direct causes like accidents. When drivers have different preferred speed levels, they need to slow down as they approach a slower car in front of them. As a consequence other cars behind this car will need to slow down too. As we will see in later chapters, we can simulate the emergence of traffic jams by simply assuming differences in the desired speed of the drivers.
Mexican Waves and Standing Ovations
In a Mexican wave people stand up and sit down from their seats in succession, creating a human wave. In a standing ovation people stand up to applaud the performance at the end of an excellent play. Both are examples of emergence in stadiums and theaters. No single individual in the audience is able to create a Mexican wave or a standing ovation. So, how do they emerge? When a few people start others may follow. How many need to start to trigger a wave and does it matter where they are sitting?
These emergent phenomena can be studied by complex adaptive systems. With complex adaptive systems we refer to a group of (locally) interacting agents, who constant act and react to actions of other agents. The coherent emergent behavior that might occur in a system arises from the local interactions of the agents. In the case of a school of fish, we can explain the flocking behavior as a result of a few simple rules such as avoiding local crowding, steering toward the average heading of local fishes, and steering toward the average position of local fishes.
Other examples of complex adaptive systems are:
Complex adaptive systems and emergence are determined by what we call bottom up processes. The opposite would be a system of top-down control. In a top-down controlled system all individual components obey the rules of a central commander or a blue print. Examples are armies and bureaucratic organizations, which are trained to execute orders from higher echelons. They have precise instructions on what to do in order to meet the goals of the organization as a whole. Taking individual initiatives are not appreciated since it may disrupt the well-trained order in the system.
An interesting example of emergence and complex adaptive systems in social systems is irrigation in Bali, Indonesia. Farmers who irrigate their fields in Bali have to solve a complex coordination problem that emerges at the intersection of two separate issues: pest outbreaks and water shortage. On one hand, control of pests is most effective when all rice fields in a watershed have the same schedule of planting rice. The reason is that rice is harvested at the same time and no food for the pests exists in a larger area to survive. If harvest of rice was in small fields only, pest can survive after a harvest by moving to a neighboring field where rice is still on the field. On the other hand, the terraces are hydrologically interdependent, with long and fragile systems of tunnels, canals, and aqueducts. Therefore, to avoid water shortage, the irrigators should not plant rice all at the same time.
To balance the need for coordinated fallow periods and use of water, a complex ritual system has been developed that details what rituals and actions should be done on each specific date in each organized group of farmers—called a subak. These actions are related to offerings at temples, ranging from the little temples at the rice terrace level to the temples at the regional level and all the way up to the temple of the high priest Jero Gde, the human representative of the Goddess of the Temple of the Crater Lake. Crater Lake feeds the groundwater system, which is the main source of water for irrigating in the entire watershed. These offerings were collected as a counter gift for the use of water that belonged to the gods.
There is no person who has control over the whole irrigation system. At the temple level, subak leaders come together regularly to exchange information on their irrigation experiences in their subaks. When a particular subak makes a decision that has unfavorable effects on neighboring subaks, leaders of the local temple community come together to discuss these matters. This may lead to various actions, including the threat of cutting-off the water supply to the disrupting subak. When seasonal rainfall is different than expected, subaks leaders may come together to discuss alternative cropping pattern to avoid unfavorable circumstances.
The function and power of the water temples were invisible to the planners involved in promoting the Green Revolution during the 1960s. They regarded agriculture as a purely technical process. Farmers were forced to switch to the miracle rice varieties, which were predicted to lead to three harvests a year, instead of the two harvests of traditional varieties. Farmers were motivated by governmental programs that subsidized the use of fertilizers and pesticides. After the governmental incentive program was started, the farmers continued performing their water temple rituals, but they no longer coincided with the timing of rice-farming activities. Soon after the introduction of the miracle rice, a plague of plant-hoppers caused huge damage to the rice crop. A new variety was introduced, but then a new pest plague hit the farmers. Furthermore, there were problems of water shortage.
During the 1980s, an increasing number of farmers wanted to switch back to the old system, but the engineers interpreted this as religious conservatism and resistance to change. It was anthropologist Steve Lansing who unraveled the function of the water temples, and was able to convince the financers of the Green Revolution project on Bali that the irrigation was best coordinated at the level of the subaks with their water temples. Lansing built an agent-based model of the interactions of subak management strategies and the ecosystem, and the local adaptation of subaks to strategies of neighboring subaks, and showed that for different levels of coordination, from farmer level up to central control, the temple level was the level of scale where decisions could be made to maximize the production of rice. He also showed how the coordination might have been evolved as a result of local interactions. In his agent-based model, agents, representing subaks, make a decision at the end of the year which cropping pattern to use for the next year. They will look at their neighboring subaks to determine whether one of them is doing better than themselves. If this is the case, the subak copies the strategy of the better performing subak. Within a few years, the average simulated yield increases to a level that is close to the maximum yield. The result is that there are pockets of subaks who have the same cropping pattern, solving the pest problem. These pockets of subaks are small enough to avoid serious water shortages. Steve Lansing therefore shows that a complex irrigation network is able to organize themselves by pure local interactions.
In this book we will use models to study complex adaptive systems. You will learn how to develop such models in Netlogo. Netlogo is a modeling platform developed by http://ccl.northwestern.edu/uri/">Uri Wilensky and his team at Northwestern University, and can be downloaded for free at http://ccl.northwestern.edu/netlogo/. We will use Netlogo version 5.0 for the examples in this book.
To get an idea of the types of models we are looking at during the course you can explore the large Netlogo Model library. There are many examples of various topics. Before you learn the programming, you can play with the existing models. We discuss now two examples in more detail.
In Netlogo Model library under the heading Biology there is a model titled Ants that simulates the foraging of a colony of ants. The ant colony in the center has 3 lumps of resources in the neighborhood. In Figure 7 the purple circle is the colony, and the blue circles are food piles. The ants initially move randomly around until they find a food resource. If an ant finds a food resource, they return back to their colony leaving a trail of pheromones behind them. Ants walking randomly around may encounter pheromone trails, and they follow this in the direction away from the colony. The three food piles are at different distances from the colony. You will experience that the ants consume first the closest food pile and end with the most distant food pile. One can explore the dynamics of the foraging behavior by changing the number of ants, the rate at which an ant leave pheromones behind them, and the rate the pheromones evaporate.
Another example of emergence is flocking behavior. In the biology section of the model library of Netlogo you will find the model titled Flocking. The agents monitor the movements of other agents in their neighborhood. If they are too close (minimum separation) to their neighbors they turn their direction (max-separate turn) to increase the separation. If an agent is not very close to its neighbors it turns its direction in line with the neighbors (max-align turn) to decrease the separation (max-cohere turn). The result is that groups of agents emerge who move around in a group. The values of the sliders determine whether a large group of agents or smaller groups of agents emerge.
 Note for OS X users: The Google Chrome browser is unable to run the Java applets that enable NetLogo models in the browser. You must use the Safari or Firefox browser. Otherwise, you may download the model code and run it using the NetLogo application installed on your computer.
Complex adaptive systems are systems that are defined by a group of (locally) interacting agents, who constant act and react to actions of other agents. Examples of such systems are ant-colonies, immune systems, ecosystems, financial markets, etc.
One of the phenomena of complex adaptive systems is the occurrence of emergence, which is a macro-level phenomenon as a result of local-level interactions. Examples are ant-trails, immune system response to a virus, flocking behavior of birds and bubbles in financial markets.
One of the mechanisms in complex adaptive systems that can cause emergent phenomena is stigmergy. Individuals change the environment, which, in turn, affects the behavior of other individuals.
In this course we will use agent-based models to study complex adaptive systems. We will also use computer simulation to find out how simple rules of local interactions of agents lead to emergent phenomena.
Gordon, D. (1999) Ants at work: How an insect society is organized. New York/London: W.W. Norton & Company.
Holland, J.H. (1998) Emergence from chaos to order. New York: Oxford University Press.
Lansing, J.S. (2006) Perfect order: Recognizing complexity in Bali. Princeton, NJ: Princeton University Press.
Lansing, J.S. and J.N. Kremer (1994) Emergent properties of Balinese water temples. American Anthropologist 95 (1): 97-114.
Mitchell, M. (2009) Complexity: A guided tour. New York: Oxford University Press.
Waldrop, M.M. (1992) Complexity: The emerging science at the edge of order and chaos. New York: Simon & Schuster.