If you’re building a system designed to learn from the ground level, a system where macrointelligence and adaptability derive from local knowledge, there are five fundamental principles you need to follow. Gordon’s harvester ants showcase all of them at work:
More is different. This old slogan of complexity theory actually has two meanings that are relevant to our ant colonies. First, the statistical nature of ant interaction demands that there be a critical mass of ants for the colony to make intelligent assessments of its global state. Ten ants roaming across the desert floor will not be able to accurately judge the overall need for foragers or nest-builders, but two thousand will do the job admirably. “More is different” also applies to the distinction between micromotives and macrobehavior: individual ants don’t “know” that they’re prioritizing pathways between different food sources when they lay down a pheromone gradient near a pile of nutritious seeds. In fact, if we only studied individual ants in isolation, we’d have no way of knowing that those chemical secretions were part of an overall effort to create a mass distribution line, carrying comparatively huge quantities of food back to the nest. It’s only by observing the entire system at work that the global behavior becomes apparent.
Ignorance is useful. The simplicity of the ant language—and the relative stupidity of the individual ants—is, as the computer programmers say, a feature not a bug. Emergent systems can grow unwieldy when their component parts become excessively complicated. Better to build a densely interconnected system with simple elements, and let the more sophisticated behavior trickle up. (That’s one reason why computer chips traffic in the streamlined language of zeros and ones.) Having individual agents capable of directly assessing the overall state of the system can be a real liability in swarm logic, for the same reason that you don’t want one of the neurons in your brain to suddenly become sentient.
Encourage random encounters. Decentralized systems such as ant colonies rely heavily on the random interactions of ants exploring a given space without any predefined orders. Their encounters with other ants are individually arbitrary, but because there are so many individuals in the system, those encounters eventually allow the individuals to gauge and alter the macrostate of the system itself. Without those haphazard encounters, the colony wouldn’t be capable of stumbling across new food sources or of adapting to new environmental conditions.
Look for patterns in the signs. While the ants don’t need an extensive vocabulary and are incapable of syntactical formulations, they do rely heavily on patterns in the semiochemicals they detect. A gradient in a pheromone trail leads them toward a food source, while encountering a high ratio of nest-builders to foragers encourages them to switch tasks. This knack for pattern detection allows meta-information to circulate through the colony mind: signs about signs. Smelling the pheromones of a single forager ant means little, but smelling the pheromones of fifty foragers in the space of an hour imparts information about the global state of the colony.
Pay attention to your neighbors. This may well be the most important lesson that the ants have to give us, and the one with the most far-reaching consequences. You can restate it as “Local information can lead to global wisdom.” The primary mechanism of swarm logic is the interaction between neighboring ants in the field: ants stumbling across each other, or each other’s pheromone trails, while patrolling the area around the nest. Adding ants to the overall system will generate more interactions between neighbors and will consequently enable the colony itself to solve problems and regulate itself more effectively. Without neighboring ants stumbling across one another, colonies would be just a senseless assemblage of individual organisms—a swarm without logic.
*
Gordon’s harvester ant colonies contain another mystery. If we understand how local interactions can lead to global problem-solving, we still don’t have an answer to the question of how colonies develop over time. This is one of those scientific questions that nobody thought to ask, because the phenomenon had gone unobserved. And that phenomenon had gone unobserved because people had been thinking about ants—and watching ants—using the wrong scale. Until recently, entomologists studied colony behavior in snapshots, surveying a given nest for days or months at a time, then moving on to other nests or back to the lab. But successful colonies can live as long as fifteen years—the life span of the egg-laying queen ant, whose demise signals the final death of the colony itself. Entomologists had been looking at individual colonies in the scale of weeks or months. But to understand how colonies develop, you needed to work on the scale of decades.
In the mideighties, when she first began doing fieldwork in Arizona, Gordon made a bold research gamble that turned out, in hindsight, to be brilliant: she decided to track individual colonies year to year, following them through their birth at the end of a successful mating flight all the way to their fifteen-year-old senescence. After a half decade or so in this time-consuming project, the results began to come in, and they were fascinating. Like a stop-motion film of a vine winding its way around a branch, Gordon’s research transformed the way that we think about ants by transforming the temporal scale with which we perceived them. The colonies cycled through a clearly defined infancy, adolescence, and mature phase over their fifteen-year existence. “I had never thought about it, or read anything about it, because without long-term data, nobody really knows the ages of their colonies,” she says now. “So it wasn’t until I had been watching the same colonies year after year, and began to be able to count how old the colonies were, that I could start to see that young colonies were more active.” As she continued her observations, a number of differences emerged between colonies of varying ages, differences that were eerily reminiscent of other developmental cycles in the animal kingdom.
For one, younger colonies are more fickle. “I’ve done experiments that mimic the kinds of changes in environment that a colony usually experiences—say, a change in the availability of food,” Gordon tells me. “If I do the same experiment week after week with older colonies, I get the same results: they respond the same way over and over. If we do the same experiment week after week with a younger colony, they’ll respond one way this week, and another way next week, so the younger colonies are more sensitive to whatever’s different about this week than last week.”
“Typical teenagers,” I say, laughing.
“Maybe.” She smiles. “And the other thing that might be more typical of teenagers would be the difference between older and younger colonies in the ways that they respond to their neighbors. Neighboring harvester ant colonies meet when foragers from the two colonies overlap and search the same places for food. If older colonies meet a neighbor one day, the next day they’re more likely to turn and go in the other direction to avoid each other. The younger colonies are much more persistent and aggressive, even though they’re smaller. So they meet one day and they’ll go right back the next day—even if they have to fight.”
The developmental cycles of colonies may be intriguing enough at face value, but consider this additional fact: while the overall colony evolves and adapts over fifteen years, the ants that make up the colony live no longer than twelve months. Indeed, the hapless male ants—who only show up once a year for the mating flight—only live for a single day. (Their life span is so abbreviated that natural selection didn’t bother to endow them with jaws to eat, since they don’t live long enough to get hungry.) Only the queen ant lasts for more than a year, and yet she does nothing but lay eggs and is entirely uninvolved with the behavior of worker ants out in the field. The colony grows more stable and less impetuous as it develops, and yet the population of the colony starts over from scratch each year. How does the whole develop a life cycle when the parts are so short-lived?
*
It would not be wrong to say that understanding emergence begins with unraveling this puzzle. The persistence of the whole over time—the global behavior that outlasts any of its component parts—is one of the defining chara
cteristics of complex systems. Generations of ants come and go, and yet the colony itself matures, grows more stable, more organized. The mind naturally boggles at this mix of permanence and instability. We can understand it when we stumble across, say, a Tudor house in the Cotswolds whose every plank and beam and brick has been replaced at least once in its lifetime, because those bricks are being replaced by “master planners”: craftsmen or residents who know what the house itself is supposed to look like, and who deliberately follow the original blueprints. Gordon’s ant colonies are more like a house that automatically replaces its skin once a year, without anybody helping out. Or better yet, given that ant colonies grow more durable over time, it’s like a house that spontaneously develops a sturdier insulation system after five years and sprouts a new garage after ten.
The ant colony may amaze us with its capacity to grow and evolve while discarding entire generations of worker ants, but as it turns out, we’re not all that different from social insects like ants, termites, or bees. As the science writer Matt Ridley observes, “The relationship between body cells is indeed very much like that between bees in a hive. The ancestors of your cells were once individual entities, and their evolutionary ‘decision’ to cooperate, some six hundred million years ago, is almost exactly equivalent to the same decision, taken perhaps fifty million years ago by the social insects, to cooperate on the level of the body; close genetic relatives discovered they could reproduce more effectively if they did so vicariously, delegating the task to germ cells in the cells’ case, or to a queen, in the case of bees.”
The human body is made up of several hundred different types of cells—muscle, blood, nervous, and so on. At any given time, approximately 75 trillion of these cells are working away in your body. In a very real sense, you are the sum of their actions; there is no you without them. And yet those cells are dying all the time! Thousands probably died in the time it took you to read the last sentence, and by next week, you will be composed of billions of new cells that weren’t there to enjoy the reading of that sentence, much less enjoy your first step or your high school prom. Cells are dying all the time in your body—and most of them are being replaced at a tremendous clip. (Even brain cells turn out to regenerate themselves far into adulthood.) And yet somehow, despite that enormous cellular turnover, you still feel like yourself week to week and year to year. How is this possible?
Some readers might be inclined to object at this point that humans are in fact closer to that endlessly rebuilt Tudor house than an ant colony, because in the case of human development we do have a master planner and a blueprint that we can follow: those coils of DNA wrapped neatly in every cell in our body. Our cells know how to build our bodies because natural selection has endowed them with a meticulously detailed plan, and has seen to it that 75 trillion copies are distributed throughout our bodies at any given time. The tyranny of DNA would seem to run counter to the principles of emergence: if all the cells are reading from the same playbook, it’s not a bottom-up system at all; it’s the ultimate in centralization. It would be like an ant colony where each ant started the day with a carefully planned agenda: forage from six to ten; midden duty until noon; lunch; and then cleanup in the afternoon. That’s a command economy, not a bottom-up system.
So does this mean our genes are secret Stalins, doling out the fixed plan for growth to the Stakhanovites of our cells? Are we more like a socialist housing complex than an ant colony? No one questions that DNA exerts an extraordinary influence over the development of our cells, and that each cell in our body contains the same genetic blueprint. If each cell were simply reading from the chromosomal playbook and behaving accordingly, you could indeed make the argument that our bodies don’t function like ant colonies. But cells do more than just follow the dictates of DNA. They also learn from their neighbors. And without that local interaction, the master plan of our genetic code would be utterly useless.
Cells draw selectively upon the blueprint of DNA: each cell nucleus contains the entire genome for the organism, but only a tiny segment of that data is read by each individual cell: muscle cells read from the lines of code that concern muscle cells, while blood cells consult the passages that relate to blood cells. This seems simple enough, until you ask the question, how did a muscle cell get to be a muscle cell in the first place? And that question underlies one of the most fundamental mysteries of emergence, which is how complicated organisms, with a wide variety of building blocks, can develop out of such simple beginnings. We all start life as a single-celled organism, and yet by the end of our development cycle, we’re somehow composed of two hundred variations, all intricately connected to one another, and all performing stunningly complex tasks. How does an egg somehow know how to build a chicken?
The answer is not all that different from the solution that ant colonies rely on. Cells self-organize into more complicated structures by learning from their neighbors. Each cell in your body contains an intricate set of tools for detecting the state of surrounding cells, and for communicating to those cells using various chemical messengers. Where ants used pheromones to inform each other of their activities, cells communicate via salts, sugars, amino acids—even larger molecules such as proteins and nucleic acids. The messages are partially transmitted through cell “junctions,” small passageways that admit molecules from one cell’s cytoplasm to another. This communication plays an essential role in all cellular activity, but it is particularly critical for embryonic development during which a single-celled organism self-organizes into a mouse or a roundworm or a human being.
We all begin life as a single-celled embryo, but seconds after conception, the embryo divides itself into two compartments: a “head” and a “tail.” At that point, the organism has joined the ranks of multicellular life, being composed now of two distinct cells. And those two cells—the head and the tail—have separate instructions for growth encoded in their DNA: one cell turns to the “head cell” chapter, the other to the “tail cell” chapter. At this early stage of development, the instructions follow a predictable pattern: divide into another “head” and “tail.” Thus, in the second round of embryonic development, there are four cells: the head of the head, the tail of the head, the head of the tail, and the tail of the tail. Those four units may not sound like much, but this cycle of cell division continues at a blistering clip. A frog embryo self-divides into nearly ten thousand cells in a matter of hours. The runaway power of geometric progression is not just a mathematical oddity—it is also essential to the very origins of life.
Once the embryo reaches a certain size, cell “collectives” start to form, and here matters get more complicated. One group of cells may be the beginning of an arm, while another group may be the first stirrings of the brain’s gray matter. Each cell has somehow to figure out where it is in the larger scheme of things—and yet, like the ants, cells have no way of seeing the whole, and they have no fixed address stamped upon them when they come into the world, no factory serial number. But while cells lack a bird’s-eye view of the organism that contains them, they can make street-level assessments via the molecular signals transmitted through the cell junctions. This is the secret of self-assembly: cell collectives emerge because each cell looks to its neighbors for cues about how to behave. Those cues directly control what biologists call “gene expression”; they’re the cheat sheet that enables each cell to figure out which segment of DNA to consult for its instructions. It’s a kind of microscopic herd mentality: a cell looks around to its neighbors and finds that they’re all working away steadily at creating an eardrum or a heart valve, which in turn causes the cell to start laboring away at the same task.
The key here is that life does not simply reduce down to transcribing static passages from our genetic scripture. Cells figure out which passages to pay attention to by observing signals from the cells around them: only with that local interaction can complex “neighborhoods” of cell types come into being. The Nobel laureate Gerald Edelman calls this process
topobiology, from the Greek word for “place,” topos. Cells rely heavily on the code of DNA for development, but they also need a sense of place to do their work. Indeed, the code is utterly worthless without the cell’s ability to determine its place in the overall organism, a feat that is accomplished by the elegant strategy of paying attention to one’s neighbors. As Ridley writes, “The great beauty of embryo development, the bit that human beings find so hard to grasp, is that it is a totally decentralized process. Since every cell in the body carries a complete copy of the genome, no cell need wait for instructions from authority; every cell can act on its own information and the signals it receives from its neighbors.” And so we have come full circle back to Gordon’s ants, and their uncanny ability to generate coordinated global behavior out of local interactions.
*
Neighbors and neighborhoods. The words seem more attached to the communities of human settlements than the microscopic domains of muscle cells or harvester ants. But how do we extend our vision up one more level on the chain of life to the cultural “superorganism” of the city? Certainly it is possible to model the behavior of cities by using the tools of swarm logic. Computer-based simulations can teach us a tremendous amount about complex systems: if a picture is worth a thousand words, an interactive model must be valued in the millions. But a quick look at the software bestseller lists will tell you that city simulations are more than just an educational device. Will Wright’s SimCity franchise has now sold millions of copies; it’s likely that the number of virtual towns created using Wright’s tools exceeds the number of real towns formed in modern human history. Some games attract our attention by appealing to our appetite for storytelling, following a linear progression of move and countermove, with clearly defined beginnings and endings; other games catch the eye by blowing things up. SimCity was one of the first games to exploit the uncanny, bottom-up powers of emergence. Wright’s genius was not simply in recognizing the fun of simulating an entire metropolis on your screen. He also hit upon a brilliant programming trick that enabled the city to evolve in a more lifelike way—a trick that closely resembles the behavior of ant colonies and embryos.
Emergence Page 7