The science of emergent behavior - how large chaos orders itself into sense (kind of how your brain works)
Caught in the swarm
How Hanoi's chaotic traffic system fascinated, and almost killed, MIT computer-science guru Seymour Papert
By Matt Steinglass/Boston Globe
THE FIELDS OF computer science and education suffered a blow on Dec. 5, when Seymour Papert, the 78-year-old cofounder of MIT's Artificial Intelligence Lab, was struck by a motorbike in Hanoi. Papert, who had come to Hanoi for a conference on teaching math with computers, remained in a coma as of Friday.
Strangely, shortly before the accident, Papert had been discussing how to build a computer model of Hanoi's notoriously chaotic traffic. He found it an interesting instance of a theme closely associated with his work: "emergent behavior," or the way that large groups of agents following simple rules, with no central leader, can spontaneously create sophisticated systems and activities. Examples include schools of fish, anthills, bee swarms, and, apparently, Vietnamese motorbike drivers.
Papert's involvement in emergent behavior grew out of Logo, a child-friendly programming language he invented in the 1970s for teaching math. Kids initially used Logo to drive an electric toy, which Papert called a "turtle." To program the turtle to drive in various patterns, the kids had to figure out how to do plane geometry. Logo turtles exemplified Papert's interactive educational philosophy, which he calls "constructionism."
Like all the best inventions, Logo turned out to have applications far beyond those it was built for. Once modified to run multiple on-screen, virtual turtles, it turned out to be a perfect way to model emergent behaviors. Programmers could create thousands of turtles, then program them to hunt for "food" and follow the trails of other turtles that had found it. Pretty soon, they had an anthill. Other turtles might be programmed to stay close together, move quickly, and avoid hitting each other or any obstacles -- and become a flock of birds.
As the inventor of Logo, Papert helped set off an explosion of interest in emergent behaviors in the 1990s, when many of the Internet's early enthusiasts began to see the global online network as a kind of self-organizing swarm composed of human beings. In his 1994 book "Out of Control," Kevin Kelly, then executive editor of Wired magazine, praised the superiority of distributed intelligence over central control. He was particularly enamored of bee swarms, coining the term "hive mind," which he claimed encapsulated "the true nature of democracy and of all distributed governance."
Like bees, Hanoi motorbikes move in swarms, unrestrained by laws, lanes, or traffic signals. Somehow, the swarm self-organizes to keep people moving and, mostly, not crashing into one another. Papert was fascinated, and spent his first days in Hanoi talking with his former student, Northwestern computer scientist Uri Wilensky, about how to use NetLogo (Wilensky's modification of Logo) to model the city's traffic flow. As the two were crossing a six-lane road separating their hotel from the university, Papert was hit.
That Papert was struck by Hanoi's traffic while thinking about how to model it is ironic, in the proper sense of the word. It's as if he had been hit by one of his own turtles.
One of the first people to use Logo turtles to model traffic was Papert's MIT colleague Mitchel Resnick, author of the 1994 book "Turtles, Termites, and Traffic Jams." Resnick, like Kelly, was an evangelist: he saw recognition of emergent behaviors as part of an epochal shift away from Newtonian mechanical models and toward biological ones.
The thrill of such spontaneous organization is familiar to anyone who has ridden a motorbike in Hanoi. Traffic in Hanoi is a self-defining flow of merging groups. Lights are few and often ignored, and divisions between lanes are determined less by lines on the ground than by a shifting, implicit group consensus. "Hanoi is one of the first places I've been, even more than in Delhi, where traffic really is organic," says Wilensky. "It really is more like a herd of buffalo."
The system would be impossible without a large reserve of tolerance and informality. Newcomers to Hanoi, who are often unable to figure out how to cross the street, are told to step into traffic at a steady pace; the motorbikes will part around you. When the city began introducing traffic lights some years ago, a Vietnamese performance artist went to one of the busiest intersections in town and videotaped himself repeatedly crossing against the light to see whether traffic would make way for him; it did. "People were still nice," as he put it -- they hadn't yet embraced the tyranny of traffic lights.
So far, so emergent. But there's a problem: as Vietnam grows richer, the number of motorbikes and cars on the street is rising furiously.
From 2002 to 2004, the percentage of Vietnamese households owning motor vehicles went from 22 percent to 33 percent. These new vehicles are pouring into a hopelessly inadequate grid of winding alleyways, ancient dike roads, and Soviet-style highways. And as density rises, drivers behave more aggressively: 9,400 Vietnamese died in traffic accidents in the first nine months of this year, up 8 percent from the same period in 2005. Worse, Wilensky says that when Papert was hit, he was obeying the Hanoi rule he'd been taught: walk across at a steady pace. But the driver failed to make way for him.
Clearly, the rules are changing. And when rules change, the swarm no longer functions very well. "If you don't have uniform rules, then you don't have as much predictability," Wilensky says. Density also increases the risk of ripple effects from faraway events, which create sudden unpredictable moves by other drivers. All of this, Wilensky says, argues for more formal regulation -- traffic lights, crosswalks, speeding tickets. "If there were any good that could come out of this, it would be that they pay some attention to the traffic problem. It's out of control."
One thing about emergent phenomena that the pioneers of the field tended not to emphasize is that they are often unkind to their constituent agents: Ant colonies are not very solicitous of the lives of individual ants. Hanoi traffic is a fascinating emergent phenomenon, but it didn't take good care of Seymour Papert when he became one of its constituent agents. As a result, the world risks losing one of its greatest thinkers about emergent phenomena.
"Seymour is still, at 78, extremely actively contributing to so many things, intellectually," says Wilensky. "He was not at all done." Let's hope he still isn't.
(Matt Steinglass lives in Hanoi, where he reports for the Globe and other publications.)
0 Comments:
Post a Comment
<< Home