Sum of the Parts

Intelligence is almost as baffling as stupidity. Only recently have scientists from wide-ranging disciplines started to piece together how complexity arises from tiny, unmistakably dumb parts. Steven Johnson, founder of Feed Magazine and author of Interface Culture, investigates the similarities between ant colonies, slime molds, and human intelligence in Emergence: The Connected Lives of Ants, Brains, Cities, and Software. Amazon.com Cyberculture Editor Rob Lightner connected with Johnson via e-mail.

Amazon.com: Early in Emergence , you note that the principle of complex behavior arising from simple components is nearly ubiquitous, yet we've only noticed it comparatively recently. Why did it take us so long?

Steven Johnson: I think what's happened recently is that we've begun to understand emergence as a kind of global phenomenon, one that shows up in a number of different fields and that seems to obey similar laws as you jump from field to field. But we've been observing the local renditions for a long time--that's why I spend so much time near the beginning of the book talking about Engels's visit to Manchester in the mid-19th century. He wrote a wonderful passage describing the way that city neighborhoods form orderly patterns without any master planners laying them out--which is the essence of urban emergence--but he didn't have the vocabulary yet to explain what was really happening, or to connect it to other disciplines.

The same goes for a number of the other fields I look at in the book. (For instance, we've understood that ant colonies have a kind of decentralized intelligence for centuries--if not longer.)

Why did we start seeing the global phenomenon now instead of a hundred years ago? That's always a hard question to answer, but part of the explanation has to be the invention of modern, graphical computers, which made it much easier for us to visualize, model, and experiment with these systems.

Amazon.com: Does emergence scale? If simple elements like ants and neurons can give rise to intelligent functions, are we likely to see the same kind of jump when we combine more complex elements (like humans)?

Johnson: I do believe that emergence scales, and you can see that scaling in action in any metropolitan center, where neighborhoods spontaneously form and sometimes keep their shape for hundreds of years. Or you can see it in something like Amazon's recommendation system. I don't want to sound like I'm preaching to the choir here, but the recommendation agent that we interact with at Amazon has gotten remarkably smart in a remarkably short time. If you've built up a long purchase history with Amazon, you'll tend to get pretty sophisticated recommendations. That's a kind of emergent behavior as well--a bottom-up intelligence that comes from looking for patterns in distributed, lower-level behavior.

For these systems to work, though, you need to eliminate a lot of individual complexity. In the city, the neighborhood clusters are formed through simple movement patterns: you visit this neighborhood on the weekend; you move out of another one after 10 years; you make a random visit to another neighborhood one afternoon. Your internal mental life and all its complexity is effectively irrelevant to the system of the neighborhood--it's all shaped by crude traffic patterns. Same goes for Amazon's recommendation system. The software doesn't know what it's like to read a book, or what you feel like when you read a particular book. All it knows is that people who bought this book also bought these other ones; or that people who rated these books highly also rated these books highly, etc. Out of that elemental data something more nuanced can emerge--if you set up the system correctly, and give it enough data.

Amazon.com: Given that we're participating in emergent phenomena, how can we recognize them as such, any more than an ant can recognize its place in the colony?

Johnson: For the same reason that we recognize many things that the ants don't recognize. We have giant neocortices that are incredibly good at rational thinking, observation, and language--while the individual ants have extremely limited conceptual tools and no complex language to convey whatever observations they might stumble across. Ants and humans also share a sense of smell, and in many ways the ant's smell system is far more nuanced than ours--it's the closest thing the ants have to a language--but even the most radical champion of ant intelligence will acknowledge that humans understand how smell works better than the ants do. The same goes for emergence.

Amazon.com: While comparing car traffic in Los Angeles to foot traffic in more traditional cities, you suggest that human interactions on freeways are too short-lived to give rise to higher-level order. Yet there are some freeway-related phenomena that seem inexplicable if one disregards emergence--traffic congestion is one example that you cite. What's going on here?

Johnson: That's a really interesting question, and I hadn't thought about it much since I wrote the line you refer to. I think the answer is the flip side of what I said about the need for eliminating individual complexity in emergent systems. I think the cars are too simple. Ants or humans have a relatively simple set of actions that contribute to the higher-level system of the colony or the neighborhood, but it's not too simple. It matters when someone decides to move to a neighborhood, but it also matters whether that person is an investment banker, a lesbian, a technology writer, or all of the above. The higher-level attributes of the neighborhood change based on those subtleties. The ants, similarly, have a sophisticated pheromone language that they use to communicate with one another. But the cars are far more crude in the flow of information between them. What do you care about when you're driving? You care about how fast you're going. Everything else is pretty much peripheral to that. So the primary piece of information that's being shared by the complex system of traffic is the number and position of cars and how fast they're going. All the other subtleties are pretty much useless.

Now, even with that limited supply of information, higher-level structures form, like traffic jams, or other weird periodic cycles. But they're nothing to write home about. Because the cars are too stupid.

Amazon.com: Will intelligence-generating tools like collaborative filtering be applied to politics and governance? What would that look like?

Johnson: I hope they will, if only because I think they'll lead to a fragmenting of the absurd red-vs.-blue political map that has appeared in the last few decades. Collaborative filtering--when it's done well--does a remarkably good job of finding smaller pockets of shared interests. Show our current political system 100 people, and it'll segment that population into two groups--Republicans and Democrats. Good collaborative filtering will find many different subcultures within the same 100 people. It's less rigged to segment the world off into two types of people.

What would that look like? European parliamentary systems, potentially. Lots of smaller parties, lots of coalitions between them, and shifting alliances. More diversity, no doubt, but also more power at the extremes.

Amazon.com: A dystopian or apocalyptic reader might fear that unstoppably destructive behaviors could emerge--or are emerging--from seemingly innocuous small-scale actions. Are we capable of slowing, stopping, or changing the course of our higher-order behavior?

Johnson: There are definitely some problems that are best not solved by emergent tools. There's too much random trial-and-error involved, too much experimentation. If there's a top-down approach that's clearly working, there's no need to try our hands at emergent approaches just for the fun of it. But where the existing top-down solutions aren't working--in inner city housing projects, to use the Jane Jacobs example--it's worth taking on the risk of the emergent approach.

Amazon.com: Many of the examples you give of guiding or tweaking emergence come from the world of programming. How can we direct our large-scale behavior in the real world?

Johnson: At the very end of the book, I talk about the anti-globalization movement, and its successes over the past two years. I think they've become so prominent partially because they've explicitly embraced the bottom-up approach to organization. Think about the popular imagery associated with this movement: there's no Jackson/Chavez/MLK/Ché figure pounding at the podium, energizing the masses into action. There are just these loose assemblages of different groups coming together for each big global finance event. They're self-organizing clusters, and that's precisely what makes them so powerful.