How do birds keep their movements so orderly, so synchronized? Most people assume that birds play a game of follow-the-leader: the bird at the front of the flock leads, and the others follow. Indeed, people assume centralized control for almost all patterns they see in the world. But that's not necessarily so. In the case of bird flocks, most don't have leaders at all. Rather, each bird follows a set of simple rules, for example, matching its velocity to that of the other birds around it, and keeping a safe distance from the birds on either side.
A bird flock is one of many phenomena organized without an organizer, coordinated without a coordinator. In ant colonies, trail patterns are determined not by the dictates of the queen ant but by local interactions among the worker ants, such as following a scent that their fellow ants emit when they find a source of food. In human societies, macroeconomic patterns arise from the haggling between millions of buyers and sellers in marketplaces and stock markets around the world. And in immune systems, armies of antibodies seek out bacteria in a systematic, coordinated attack--without any "generals" organizing the overall battle plan.
The Era of Decentralization
A growing number of people are now choosing these kinds of decentralized models for the organizations and technologies they construct in the world, and for the theories they construct about the world. One such case began to unfold on December 7, 1991, when Russian President Boris Yeltsin met with the leaders of Ukraine and Belarus in a forest dacha outside the city of Brest. After two days of secret meetings, the leaders issued a declaration: "The Union of Soviet Socialist Republics, as a subject of international law and a geopolitical reality, is ceasing its existence." With that announcement, Yeltsin and his colleagues sounded the final death knell for a centralized power structure that had ruled for nearly 75 years. In its place, the leaders established a coalition of independent republics and promised a radical decentralization of both economic and political institutions.
The next day, halfway around the world, another powerful institution announced its own decentralization plans. IBM chairman John Akers publicly announced a sweeping reorganization of the computer giant, dividing the company into more than a dozen semi-autonomous business units, each with its own financial authority and its own board of directors. The goal was to make IBM more flexible and responsive to the needs of rapidly changing markets.
The coincident timing of the above two events actually symbolizes a broad decentralization trend that is sweeping through many different domains. For example:
* Organizations. All types of organizations--schools, companies, even countries--are pushing authority and power down from the top, distributing rights and responsibilities more widely. In U.S. education, for example, decentralization extends to several levels. School choice brings market-oriented thinking to the world of education, asserting that individual families--not state or local governments--should decide where their children go to school. Likewise, school-based management moves decision-making authority from state and district offices to individual schools. And child-centered learning, now adopted in many classrooms, transforms the teacher from a central authority into a catalyst, coach, and collaborator.
* Scientific models. For 300 years, researchers' thinking has been guided by Sir Isaac Newton's model of the physical world as a clocklike mechanism. Newton's world is ruled by a centralized notion of cause and effect--one gear turns, which makes another gear turn, and so on. Now, a new set of decentralized models and metaphors is spreading through the scientific community and gradually into the culture at large. Researchers now view a wide range of systems--everything from bird flocks to immune responses--less like clockwork mechanisms and more like complex ecosystems controlled by decentralized interactions and feedback loops.
* Psychology. Few concepts seem more obvious than the singular nature of the mind and self. Each of us experiences life as a single thread of consciousness, and each of us imagines our own mind as "I," not "we." But the idea of the unified, centralized mind, first challenged by Sigmund Freud, has eroded rapidly during the past decade. For example, Daniel Dennett, a professor of philosophy at Tufts University, proposes that there is no single stream of consciousness in the mind. He suggests instead that multiple narratives are simultaneously created and edited in different parts of the mind. Similarly, the field of artificial intelligence, once dominated by centralized models of the mind, now favors decentralization. Marvin Minsky, a professor at the MIT Media Lab, argues that the mind is a society of many simple agents that work together to accomplish complex tasks.
* Theories of knowledge. For centuries, philosophers strove for "objective knowledge." They put great faith in the power of logic to systematize all knowledge, to find ultimate meaning and truth. Today, philosophers are moving away from the notion of a single unifying conception of knowledge, arguing instead that knowledge speaks not with a single voice but with many. For example, traditional theories of literary criticism assumed that meaning was created by an author and conveyed through the author's writings. According to this view, reading is a search for inherent meaning in a document, an attempt to decipher the intention of the author. But modern schools of thought--such as poststructuralism, reader-response theory, and deconstructionism--all focus on readers as the main constructors of meaning. In this new view, texts have little or no inherent meaning. Rather, meanings are constantly reconstructed by communities of readers through their interactions with the text. Meaning itself has become decentralized.
Centralized Thinking
Even as the influence of decentralized ideas grows in many disciplines, a deep-seated resistance to such ideas remains. People seem to have strong attachments to centralized ways of thinking, assuming that every pattern must have a single cause, an ultimate controlling factor. The widespread resistance to evolutionary theories is an example: Many individuals still insist that someone or something must have explicitly designed the complex, orderly structures that exist in the biological world. They resist the idea that complexity can be formed through a decentralized process of variation and selection.
Similarly, many view the workings of the economy in centralized ways, assuming singular causes for complex, decentralized phenomena. In interviews with Israeli children between 8 and 15 years old, for example, David Leiser, a psychologist at Ben-Gurion University, discovered that nearly half of the children assumed that the government sets all prices and pays all salaries. Even children who said that employers pay salaries often believed that the government provides the money for the salaries. "The child finds it easier to refer unexplained phenomena to the deliberate actions of a clearly defined entity, such as the government," he wrote, "than to impersonal market forces."
The centralized mindset is not just a misconception of the scientifically naive. A similar bias toward centralized theories can be seen throughout the history of science, with scientists remaining committed to centralized explanations even in the face of discrediting evidence. The history of research on slime-mold cells, as told by Evelyn Fox Keller, a professor of science, technology, and society at MIT, provides a striking example. At certain stages of their life cycle, slime-mold cells gather into clusters. Scientists long believed that this aggregation process was coordinated by specialized slime-mold cells, known as "pacemaker" cells. According to this theory, each pacemaker sends out a chemical signal telling other slime-mold cells to gather around it.
In 1970, Keller and a colleague proposed an alternative model, showing how slime-mold clusters can form without any specialized cells. In this model, every individual slime-mold cell emits a chemical signal and follows signals produced by others. The result: aggregation without a leader. Nevertheless, for the following decade, other researchers continued to assume that pacemakers were required to initiate aggregation. As Keller writes, with an air of disbelief: "The pacemaker view was embraced with a degree of enthusiasm that suggests that this question was in some sense foreclosed."
It is not altogether surprising that people have strong commitments to centralized approaches. Many patterns and structures in the world are, in fact, organized by a central designer. When we see neat rows of corn in a field, we assume correctly that the corn was planted by a farmer. When we watch a ballet, we assume correctly that the movements of the dancers were planned by a choreographer. When we participate in social systems, such as families and school classrooms, we often find that power and authority are centralized, often excessively so. These phenomena reinforce the centralized mindset.
Another important factor is the way people think about themselves. Your mind (like all others) is composed of thousands of interacting parts. But you experience yourself as a singular being. This is a convenient, perhaps necessary, illusion for surviving in the world. When you do something like paint a picture or organize a party, you feel as if you are playing the role of the central actor. Only one entity seems to be in charge: you. So people naturally expect most systems to involve a central authority.
By clinging to this centralized mindset to explain all phenomena, politicians, managers, and scientists are working with blinders on, focusing on centralized solutions even when decentralized approaches might be more appropriate, robust, or reliable.
Decentralized Thinking Tools
To help people move beyond the centralized mindset and learn new ways of thinking about decentralized phenomena, I developed a new computer programming language called StarLogo. This language allows people to control the actions of thousands of graphic creatures on the computer screen. The user writes simple rules for the creatures and the environment in which they live, and then observes the group behaviors that emerge from their interactions. For example, a user might write simple rules for individual birds, then observe how the flock behaves.
In one StarLogo simulation, inspired by the controversy over slime-mold aggregation, the artificial creatures follow two simple rules: they emit a pheromone (chemical attractant) and, after "sniffing" the local area, move in the direction in which the pheromone is strongest. At the same time, the environment causes the pheromone to diffuse and evaporate. With this simple strategy, the creatures quickly assemble into clusters. The reason: When a few creatures get near one another just by chance, they create a pheromone "puddle," which attracts even more creatures, making the puddle even bigger and so on.
I have worked with several groups of high school students who have created decentralized "microworlds" using StarLogo. In one experiment, two students--Ari and Fadhil--wanted to study traffic jams. So they created a one-lane highway with a police radar trap to catch cars going above the speed limit. They then programmed each driver to follow three simple rules: If you come within two car lengths of the car in front of you, slow down. If no cars are within two car lengths ahead of you, speed up until you reach the speed limit. If you detect a radar trap (each car is equipped with a detector), slow down.
Both students expected that a traffic jam would form behind the radar trap, and indeed it did. As cars slowed down for the trap, the cars behind them were forced to slow down, creating a queue with roughly equal distances between the cars. When the cars moved beyond the trap, they accelerated smoothly until they reached the speed limit.
I asked the students what would happen if they removed the radar trap. The cars would be controlled by just two rules: if you see another car close ahead, slow down; if not, speed up. They predicted that the traffic flow would become uniform; cars would be evenly spaced, traveling at a constant speed. When we ran the program, however, a traffic jam formed. Along parts of the road, the cars were tightly packed and moving slowly. Elsewhere, they were spread out and moving at the speed limit.
At first, the students were shocked. Their comments revealed the workings of a centralized mindset: They argued that traffic jams need some sort of centralized "seed," like a radar trap or accident, in order to form. They couldn't believe that simple interactions among cars could create a jam. But as they continued to experiment with the simulation--modifying the speed and starting positions of the cars--they developed an understanding of how the traffic jams formed. When a few cars, by random chance, happened to get near one another, they slowed down, making it likely that even more cars behind them would have to slow down, leading to a jam.
Another student, Callie, chose to use StarLogo to simulate the behavior of termites. Termites are practically blind, yet they are considered the master architects of the insect world. In fact, on the plains of Africa, termites construct giant moundlike nests containing intricate networks of tunnels and chambers. Many people assume that the queen of the termite colony tells the blind workers what to do. But, as in ant colonies, the queen is more of a mother to the colony than a leader. On the termite construction site, no one is in charge of a master plan. Rather, each termite carries out a series of relatively simple tasks, relying on its sense of touch and smell.
Termites are thus well suited for StarLogo explorations. Callie started with the following goal for her StarLogo termite colony: Termites should gather randomly scattered wood chips and put them into a few orderly piles. As with real termites, she didn't want to put one termite in charge. Instead, she programmed each termite to walk around randomly, obeying two simple rules: If you are not carrying anything and you bump into a wood chip, pick it up. If you are carrying a wood chip and you bump into another one, drop the chip.
At first, we were both skeptical that this decentralized strategy would work. The strategy did not prevent termites from taking wood chips away from existing piles. So while termites were putting new wood chips on a pile, other termites might be taking wood chips away from it. It seemed like a good prescription for getting nowhere. But we ran the program with 1,000 termites and 1,500 wood chips.
Much to our surprise, the number of piles steadily declined and the number of wood chips in each pile grew. After several program iterations--in each iteration every termite took a step or picked up or dropped a chip--the wood chips had been gathered into hundreds of small piles. After 2,000 iterations, there were 100 piles with an average of 15 wood chips in each. After 10,000 iterations, there were fewer than 50 piles left, with an average of 30 wood chips in each pile. And after 20,000 iterations, only 34 piles remained, with an average of 44 wood chips in each pile.
The process was slow and frustrating to watch, as termites often carried wood chips away from well-established piles. But it worked. And as we watched the termites on the screen, it became obvious why this simple strategy is effective. Whenever the termites remove all the wood chips from a particular spot, the pile never restarts, since termites drop chips only where others already reside. The termites might drag chips back and forth between piles, but once a pile is gone, it is gone forever. So the total number of piles keeps shrinking.
Some Guiding Principles
People are usually fascinated by such decentralized phenomena. But when they try to understand or create their own decentralized systems, they often slip back into centralized ways of thinking. Through my work with high-school students, I have developed several guidelines to help people make sense of decentralized systems, highlighting some pitfalls to avoid and some possibilities not to overlook. For example:
* Positive feedback isn't always negative. Positive feedback is frequently symbolized by the screeching sound that results when a microphone is placed near a speaker. It is usually viewed as destructive because the situation often spirals out of control. By contrast, negative feedback is often symbolized by a thermostat that keeps room temperature at a desired level by turning the heater on and off as needed. It is thus considered useful because it keeps conditions under control. When I asked high-school students about positive feedback, most were unfamiliar with the term. But when I explained what it meant, the students quickly generated examples, most of which involved a loss of control, often with destructive consequences. One student talked about scratching a mosquito bite, which made the bite itch even more, so she scratched it some more, which made it itch even more. Another student talked about stock-market crashes: a few people start selling, which makes more people start selling, and so on.
Despite these negative connotations, positive feedback often plays a positive role in decentralized phenomena. Brian Arthur, an economist at the Santa Fe Institute, points to the geographic distribution of cities and industries as an example of a self-organizing process driven by positive feedback. After a small nucleus of high-technology electronics companies started in Santa Clara County south of San Francisco, an infrastructure developed to serve the needs of those companies. That infrastructure encouraged even more electronics companies to locate in Santa Clara County, which encouraged the development of an even more robust infrastructure. And thus Silicon Valley was born.
* Randomness can create order. Like positive feedback, randomness has a bad image. Most people think randomness simply makes things disorderly. They view randomness as annoying at best and destructive at worst. But randomness plays a crucial role in many self-organizing systems by creating fluctuations that act as natural seeds from which patterns and structures grow.
At concerts or sporting events, for example, spectators sometimes join together in seemingly spontaneous synchronized clapping. How do they coordinate their applause without a conductor? Initially, when everyone starts clapping, the applause is totally unorganized. Even people clapping at the same tempo are wildly out of phase with one another. But through some random fluctuation, a small subset of people happen to clap at the same tempo, in phase with one another. That rhythm stands out, just a little. People in the audience sense this emerging rhythm and adjust their own clapping to join it. The emerging rhythm thus grows stronger and even more people conform to it. Eventually, nearly everyone in the audience is clapping in a synchronized rhythm. Amazingly, the whole process takes just a few seconds, even with thousands of people participating.
* A traffic jam isn't just a collection of cars. It is fair to think of most objects as a collection of particular parts. For example, a particular chair might have four particular legs, a particular seat, and a particular back. But this is not so with objects like the termite wood-chip piles. The composition of the piles is always changing, as termites take away some wood chips and add other wood chips. After a while, few if any of the original wood chips might be in the pile, but the pile is still there. The wood-chip pile is thus an example of an "emergent object"--it emerges from interactions among lower-level objects. Similarly, a traffic jam is an emergent object, continuing to exist even though the composition of cars within it is always changing.
Students often have difficulty thinking about emergent objects. For example, two students, Frank and Ramesh, tried to use StarLogo to simulate an ant cemetery, in which ants gather their dead colleagues into neat piles. This problem was virtually identical to that of programming termites to create wood-chip piles. But Frank and Ramesh resisted the simple decentralized approach that Callie used for the termites. They were adamant that dead ants should never be taken from a cemetery once placed there. How can a cemetery grow, they argued, if the dead ants in it are continually being taken away? With this strategy, however, Frank and Ramesh ended up with lots of little cemeteries rather than a few big ones, simply because a cemetery, once started, could never disappear. If Frank and Ramesh had viewed the cemetery as an emergent object and allowed the composition of ant cemeteries to vary with time, they would have had much greater success in creating large ant cemeteries.
* The hills are alive. In his book Sciences of the Artificial, Herbert Simon, a Nobel laureate economist from Carnegie Mellon, describes a scene in which an ant is walking on a beach. Simon notes that the ant's path might be quite complex, but it does not necessarily reflect the complexity of the ant. Rather it might reflect the complexity of the beach. Simon's point: don't underestimate the role of the environment in influencing and constraining behavior.
Many people seem to resist the idea of an active and influential environment. For example, when I told a student about a StarLogo program in which ants find food by following pheromone trails, he was worried that the trails would continue to attract ants even after the food source at the end of the trail had been fully depleted. In his mind, the ants had to take some positive action to get rid of the pheromone. In fact, he proposed an elaborate scheme in which the ants, after collecting all of the food, deposited a second pheromone to neutralize the first pheromone. It didn't occur to him that the first pheromone would simply evaporate away.
Foundation for Discovery
A friend of mine has a daughter named Rachel. By the time she was three years old, Rachel had already developed a theory about why it rains on some days and not on others. "The clouds rain when the thunder tells them to rain," she explained. In her mind, some type of centralized decision making was necessary. Thunder commanded, and the clouds obeyed.
It is not surprising that Rachel came up with a centralized explanation for the rain. Most likely, she was unaware that other types of explanations even existed. But as Rachel grows up, will she continue to rely on centralized explanations? If she takes a physics course in high school, will she understand gravity as two objects pulling on one another with equal force, or will she think of gravity as a one-way force, with one large object pulling on a smaller one? If she takes an economics course in college, will she understand that interest rates and money supply can affect each other, or will she assume that one is the cause and the other is the effect? If the unemployment rate rises dramatically, will she search for explanations with multiple, interacting causes, or will she immediately assume some type of evil conspiracy?
An elementary or high-school course that teaches Ten Golden Rules of Decentralized Thinking probably would not have much effect on someone with a firmly entrenched centralized mindset. Young students are likely to become comfortable with decentralized ideas only if they get opportunities to design, create, explore, and play with decentralized systems.
What's needed are computer-based construction kits that let children like Rachel create their own decentralized microworlds. At school, Rachel might create an artificial environment with giraffes, elephants, and her other favorite animals, and program each to follow a few simple rules. She could then observe what patterns emerge from the interactions and how simple changes can affect the entire ecosystem. At home, she and her friends might simulate how people gather into groups at a party. By working on projects like these, Rachel could come to understand the importance of decentralized ideas in explaining the world around her.
By the time Rachel was four, she had developed a new theory about the rain. "The clouds get together at night, and they decide whether it should rain the next day," she explained. This new theory still involves some centralized planning, but there was no longer a central actor, the thunder, in charge of the whole process. If Rachel is surrounded by new types of computational tools and ideas as she grows, one can only wonder what new theories she'll develop to explain the rain.