On Trails (12 page)

By regarding ant colonies as intelligent systems composed of individuals following simple rules, Deneubourg was able to make another leap forward: He found that he could describe their movements with mathematical formulas, which could then be used to create computer models. Ant colony algorithms—in which myriad initial routes are explored, the best ones being amplified while the others fade—have since been used to improve British telecommunications networks, to design more efficient shipping routes, to sort financial data, to better deliver supplies during disaster relief operations, and to schedule tasks in a factory. Scientists chose to model their algorithms after ants (as opposed to, say, tent caterpillars) because ants are constantly tweaking their designs and probing for new solutions; they tend to find not only the most efficient solution, but also a slew of effective backup plans.

I spoke with Deneubourg one winter morning at his home in Brussels. He greeted me at the door: a compact, spritely, gray-haired man with big ears and a wide smile. If wrinkles are a graph of all past expressions, his pointed decidedly gleeward.

From the outset of his career, when he studied under the famed systems theorist Ilya Prigogine, Deneubourg had sought to reveal the invisible systems that underlie animal behavior. He realized early on that collective intelligence extends well beyond insect colonies: indeed, historically the notion referred first to humans, and only much later to insects. The term “collective intelligence” appears as early as the 1840s, when the democracy activist Giuseppe Mazzini used it to critique Thomas Carlyle's belief that history was nothing more than the record of the actions of “Great Men.” Mazzini argued that the
greater goal of history was to discern the “collective thought . . . in the social organism”; for too long, he wrote, historians had focused on the petals rather than the whole flower. A fervent Catholic, he believed that the “collective intelligence” of humans ultimately stemmed from an almighty God, of whom humans were mere “instruments.”

Deneubourg sought to dispense with divine explanations and instead show how collective intelligence can emerge (among insects
and
people) from the interactions of individuals. In one early paper he argued that people tend to build their settlements stigmergically, just like ants: they unconsciously modify the environment, which sends a signal instructing other people how and where to build. For example, if you build a home in an unpopulated area, other people may start to perceive that area as a nice place to build a home; build enough homes and someone may build a shop; build enough shops and someone may build a factory or a shipping port. No top-down oversight is necessary; cities can arise from the ground up.

The week before I met Deneubourg, I had talked with one of his disciples, a professor in Toulouse named Guy Theraulaz, who showed me a video of how
Messor sancta
ants dig a network of branching tunnels in a disc of dirt. Next, he showed me aerial photographs of unplanned cities—
villes spontanées—
like Benares, Goslar, and Homs. The similarity was striking. He and his colleagues had found that both these systems found a near-optimal mathematical balance between efficiency (a minimum of paths) and robustness (the good kind of redundancy, whereby the collapse of a single avenue does not lead to a systematic collapse).

“The interesting thing is that in Rome, originally that was a grid system,” he said. “The whole system was destroyed by time, and then it converged into a medieval organic system.” Likewise, many cities across Europe that were built on the Roman grid—Damascus, Mérida, Caerleon, Trier, Aosta, Barcelona—later collapsed back into an organic layout, as people began taking shortcuts across empty
quadrants, filling in extravagant plazas, and altering the imperial road network to their needs. Left to their own devices, people unwittingly redesigned their cities precisely as ants would.

Sitting in Deneubourg's office, I thought back to this experiment. I wondered how a veteran collective intelligence researcher, knowing what he knows, would use that knowledge to design a better city. So I asked Deneubourg: If he were the mayor of a new city being built ex nihilo, like Brasília, how would he organize it?

“I would like to see the
emergence
of the town,” he said. “If I was the mayor—and the probability of that happening is quite low—my attitude would be very liberal. My objective would be to offer different types of material to help the citizens find the solution that they prefer.”

I found this answer somewhat surprising. By all accounts, he was an expert in the design of efficient systems. And yet he would withhold his expertise and allow the town's residents to plan their own town?

“Yes,” he replied, with a look of impish mirth. “To believe that you have the solution for another person is a form of stupidity.”

+

As the human population continues to swell and gather in ever more densely crowded, hive-like cities, the collective intelligence of ants begins to look all the more astonishing by comparison. Much of ants' inventiveness arises from their almost utopically (or dystopically, depending on your outlook) high degree of selflessness, which we notably lack. For example, when traversing a V-shaped ramp in a lab, army ants will construct a bridge out of their own bodies to create a shortcut across the crook. In human terms, this would be as if a businessman, while rushing off to work, decided to speed the passage of his fellow commuters by laying his body down over a gap in the sidewalk. I do not foresee us developing this kind of altruism any time
in the near future. Nevertheless, there are many lessons humans can glean from the wisdom of ants.

When walking among large crowds, for example, both humans and ants naturally form lanes of traffic. However, among human crowds, those lanes break up and then slow down to reformulate every thirty seconds or so, whereas ant lanes remain in a constant, steady, orderly flow. To find out why, a crowd theorist named Mehdi Moussaid set up video cameras on balconies overlooking some of the busiest pedestrian areas in the French city of Toulouse. What he found was that a single impatient person tends to weave through the crowd, disrupting the smooth flow and slowing everyone else down. (When Moussaid told me this, I laughed in uncomfortable recognition. While commuting across Manhattan to my old job, I saw this phenomenon every morning in the crowded subway tunnel between Sixth and Seventh Avenues. Perpetually late for work, I usually was that jerk.) Ultimately, moving with the flow, rather than racing through it, gets everyone in the swarm to their various destinations more quickly.

In another surprising study on ant traffic dynamics, a former colleague of Moussaid's named Audrey Dussutour showed that ants never get stuck in traffic jams. One advantage ants have is that their highways have flexible boundaries, so they are able to effortlessly widen them as traffic increases. Even in artificially constricted conditions, however, ants still adapt better than we do. Dussutour proved this by pouring Argentine ants into a basic bottleneck-shaped maze; from above, the ants resembled a dense crowd of people trying to exit a theater through a narrow set of doors. But no matter how many ants she poured into the bottleneck, she could not induce them to grind to a halt the way people inevitably would.

She told me she had recently stumbled upon a likely explanation: She had noticed that when the crowd reaches a certain density, a small number of ants—about ten percent—will stop cold in the middle of the flow, “like stones.” Remaining frozen for up to twenty minutes,
the stationary ants split those moving around them into lanes, which prevents jams. By slowing down, certain self-sacrificing individuals allow the colony to move faster. This finding meshes with similar research on human crowds, which has shown that placing an obstacle like a pillar directly in front of a doorway will cleave crowds into neat rows and quicken their flow.

Talking with Dussutour, I began to envision a future where swarms of driverless cars would use ant-based algorithms to forever eradicate traffic jams. In the past, such techno-utopic schemes had always seemed far-fetched to me, because I imagined the cars would require a centralized supercomputer to coordinate their movements. (Think of the hellacious traffic jams that would ensue if that supercomputer were to malfunction.) But a growing body of research—especially in the nascent field of swarm robotics—has proven the cars could effectively coordinate themselves without a godlike hand steering them; highly sophisticated coordination can arise from the bottom up, through individual machines following simple rules.

However, Dussutour stressed that it would be a mistake to think that just because ants behave selflessly and cooperatively, they are all identical and predictable, like robots. Her work has led her to believe that the next big paradigm shift in collective intelligence research will stem from the realization that there are notable individual differences
between
members of a swarm. “People always say ants are the same,” she said. “Bullshit.” For example, she noted, researchers have found that fourteen percent of common black garden ants never lay a trail during their various foraging trips. Another study found that at least ten percent of foraging green-headed ants will eat whatever they find without ever bringing anything back to the nest. A third study found that as many as twenty-five percent of
Temnothorax rugatulus
do no work at all. No one knows why these selfish ants exist—whether they provide some hidden evolutionary advantage to the colony, or
whether they merely demonstrate that no species is without its share of rebels and slackers.
III

Systems built on universal trust are universally easy to exploit. This is why, among humans, members of utopian communes must expend an enormous amount of energy policing against shirkers and charlatans. (“Communes,” wrote the social psychologist Jonathan Haidt, “can survive only to the extent that they can bind a group together, suppress self-interest, and solve the free rider problem.”) Because of the premium placed on social cohesion, cooperative communities—­from hives to nations—are also prone to being swayed by charismatic leaders. Experiments among shoals of golden shiner fish have shown that a single emphatic individual can alter the trajectory of an entire school of fish regardless of whether it is in the best interest of the group. Likewise, it has been found that among humans, the most confident, talkative member of a group often becomes the group's leader, more or less regardless of the quality of his or her input (a phenomenon called the “babble effect”).

“The wisdom of crowds doesn't work all the time,” said Simon Garnier, who runs a research laboratory at the New Jersey Institute of Technology called the Swarm Lab. “If you play it right, you can make crowds go wherever you want.”

Garnier was referring to the 2004 book
The Wisdom of Crowds
,
by
James Surowiecki, which described the ways that crowds of perfectly average people can collectively make judgments that rival those of the most highly regarded experts. The canonical example of this phenomenon is an experiment run by the British scientist Francis Galton. In 1906, Galton collected data from a group of people at a country fair who were trying to guess the weight of a fat ox. Of the roughly eight hundred people who wagered a guess, most were wide of the mark. However, the average of all their guesses was nearly perfect.

This experiment would later be repeated many times. Oddly, researchers learned that the key to the experiment was that each person needed to judge the weight of the ox independently, without sharing their guesses with one another. In similar experiments where people were given access to one another's answers, the collective intelligence of the group worsened. Often, the early guesses provoked a false consensus to form, a vicious circle that caused the later guesses to hurtle toward ever-greater error. “The more influence a group's members exert on each other,” wrote Surowiecki, “the less likely it is that the group's decisions will be wise ones.”

I was startled when I first read this finding, because it appeared to contradict everything scientists have learned over the past three hundred years about how trails form. When trails are taking shape, every member of a crowd has access to every previous walker's guess, because their choices are written right there in the dirt. And yet trails nevertheless tend to find optimal routes across the landscape, rather than veering off on wild, mistaken detours. How could this be?

I recently ran across an answer in a paper by a bioscientist named Andrew J. King, who conducted a clever update on the famous Galton ox-weighing experiment. In it, he asked a group of 429 people to guess the number of sweets in a jar. But this time, he made a few tweaks: He gave each member of one group access to the guess of the previous guesser. He gave another group the mean of all previous guesses. And he gave the members of a third group access to a random previous
guess. All these pieces of additional information, as predicted, skewed the group's answers for the worse. However, when he gave members of a fourth group access to the “current best guess”—the previous guess which was currently closest to the mark—he found that that group not only outperformed the three others, but in certain respects it also outperformed the classic Galtonian, private-information-only group.
IV
Among crowds, sharing more pieces of random information is generally unhelpful—like rumors swirling through a school, they amplify toward ever-greater falsehood as they go. But more
reliable
information—even if it is not perfectly correct—kicks off a process of fine-tuning, until the answer is revealed.

Every trail is, in essence, a best guess: An ant does not leave a strong pheromone trail unless it has found food, which means that it has already made a correct calculation of where the food is. The same rule applies to humans—we generally don't make trails unless there is something on the other end worth reaching. It's only once an initial best guess is made, and others follow it, that a trace begins to evolve into a trail.