In Defense of Cities

Dan Gillmor, in his San Jose Mercury News column of last Saturday, makes the case that in light of the attack on New York City, "...the logic of decentralization has never been more clear." He notes that the Internet's decentralized architecture performed well during and after the attack. He goes on to call for a similar Internet-style decentralization of populations away from cities, predicting that the attacks should or will precipitate a migration away from cities into more decentralized patterns of living.

He is right about data infrastructure, but he is wrong about cities, because cities are not cause but effect. (Full disclosure: I am a long-time resident of New York, and have spent the better part of two careers working in cities of various flavors.)

Cities are not isolated things so much as the large-scale intersection of countless small forces, forces which in aggregate give cities the kind of homeostasis and adaptability that have made them such surprisingly long-lived features of human life.

Cities are decentralized systems

In fact, cities exist because of decentralization, not in spite of it. You need only read Jane Jacob's marvellous The Death and Life of Great American Cities to see the difference between the organic growth and regeneration of urban areas that allow for decentralized decision-making by the populace, versus the disaster that befell neighborhoods and even whole cities that adopted central planning.

Clay Shirky is a featured speaker at the O'Reilly Conference on Peer to Peer and Web Services, Nov. 5-8 in Washington, D.C. Related Articles: The Great Rewiring The End of Innovation? Next Step for P2P? Open Services Convergence of Peer and Web Services O'Reilly P2P and Web Services Conference: Program Chair's Best Bets

Is America marked by decentralization or by power centralized in urban centers? Does NY's massive decentralization mean it will persevere or is the geographic centralization of financial markets a stunning vulnerability? Post your comments

More generally, the sort of uneven distribution of people in cities is true of all decentralized systems, where decentralization of some lower-level technology allows its users higher degrees of freedom. In pursuing those freedoms, the users create emergent patterns which do not have smooth or flat distributions, but rather highly uneven ones. Buildings are technologies, their residents are users, and cities are an emergent property of myriad overlapping choices about the placement and use of those buildings.

This uneven distribution is usually described as a "power law," where there are a small number of big things, a medium number of medium-sized things, and a large number of small things. In the case of American cities, size is very unevenly distributed -- a large percentage of people who live in cities live in the 100 most populous ones, and of those hundred a large percentage live in the 10 most populous, and of those ten a large percentage live in New York.

New York City happened not because the Bureau of Centralized Cities decreed that New York City should be the largest. Indeed, at the founding of the United States, either Philadelphia or Boston would have seemed liklier candidates for that sort of pre-eminence. New York is big because over time more people came than left, because millions of uncoordinated actors decided independently to move to New York. The population is not a single variable, it is the sum of these countless distrbuted decisions.

Michael, Jacob, and Matthew

In a power law, the nth item generally has a size or frequency of 1/n of the largest -- the tenth largest city is about one-tenth the size of the largest, and this emergent higher order organization is a feature of many human systems. English word frequency is the canonical example of a power law distribution, with speech and writing including many occurences of few words, like I and the, some occurences of some words, like possible and important, and few occurences of many words, like dilate and efficacious.

The Internet is decentralized, but Web site traffic observes power law distributions. Childbirth is decentralized, but baby's names observe a power law distribution. Market participation is decentralized, but stock ownership observes a power law distribution, and on and on. Power law distributions are not only inevitable, they are a sign that decentralization is working, because the freedom that decentralization provides allows for the kind of emergent properties power laws describe.

The politics of the attempted terrorist disruption of city life are best addressed in less technical forums, but the possibility of decentralized systems leading to smooth distributions, of city populations or anything else, is nil. Higher-order organization will always arise out of sufficiently complex systems of independent actors.

As long as we have a system where individuals are free to make their own choices, we will always have a most popular song, we will always have a most widely held stock, we will always have a favorite TV show, we will always have a largest city. It will not always be New York, but it will always be somewhere, and we need to plan for a world where such big cities exist, because cities are a stable effect of the decentralization Gillmor rightly prizes.

Clay Shirky writes about the Internet and teaches at NYU's Interactive Telecommunications Program. He publishes a mailing list on Networks, Economics, and Culture at shirky.com/nec.html.