In 1984, the developer Will Wright had just finished work on his first video game, a shoot-em-up called Raid on Bungeling Bay. In it, the player controls a helicopter dropping bombs on enemy targets on a series of islands. Wright was happy with the game, which was a commercial and critical success, but even after it was released, he continued tinkering with the terrain editor he had used to design Raid’s levels. “I found out,” Wright later told the Onion AV Club, “that I was having a lot more fun doing that part than just playing the game and going around bombing stuff.” Enthralled by the islands he was making, Wright kept adding features to his level editor, adding complex elements like cars, people, and houses. He became fascinated with the idea of making these islands behave more like cities, and kept tinkering with ways to make the world “come alive and be more dynamic.”
Looking to understand how real cities worked, Wright came across a 1969 book by Jay Forrester called Urban Dynamics. Forrester was an electrical engineer who had launched a second career as an expert on computer simulation; Urban Dynamics deployed his simulation methodology to offer a controversial theory of how cities grew and declined. Wright used Forrester’s theories to transform the cities he was designing in his level editor from static maps of buildings and roads into vibrant models of a growing metropolis. Eventually, Wright became convinced that his “guinea-pig city” was an entertaining, open-ended video game. Released in 1989, the game became wildly popular, selling millions of copies, winning dozens of awards, and spawning an entire franchise of successors and dozens of imitators. It was called SimCity.
Almost as soon as SimCity came out, journalists, academics, and other critics began to speculate on the effects that the game might have on real-world planning and politics. Within a few years of its release, instructors at universities across the country began to integrate SimCity into their urban planning and political science curriculums. Commentators like the sociologist Paul Starr worried that the game’s underlying code was an “unreachable black box” which could “seduce” players into accepting its assumptions, like the fact that low taxes promoted growth in this virtual world. “I became a total Republican playing this game,” one SimCity fan told the Los Angeles Times in 1992. “All I wanted was for my city to grow, grow, grow.”
Despite all this attention, few writers looked closely at the work which sparked Wright’s interest in urban simulation in the first place. Largely forgotten now, Jay Forrester’s Urban Dynamics put forth the controversial claim that the overwhelming majority of American urban policy was not only misguided but that these policies aggravated the very problems that they were intended to solve. In place of Great Society-style welfare programs, Forrester argued that cities should take a less interventionist approach to the problems of urban poverty and blight, and instead encourage revitalization indirectly through incentives for businesses and for the professional class. Forrester’s message proved popular among conservative and libertarian writers, Nixon Administration officials, and other critics of the Great Society for its hands-off approach to urban policy. This outlook, supposedly backed up by computer models, remains highly influential among establishment pundits and policymakers today.
150 Equations, 200 Parameters
Jay Wright Forrester was one of the most important figures in the history of computing, but he is also one of the least understood. He trained at Gordon Brown’s Servomechanisms Laboratory at MIT, spending World War II designing automatic stabilizers for the U.S. Navy’s radars. After the war, he led the development of the Whirlwind computer, arguably the most important computer project of the early postwar period. This machine, after humble beginnings as a flight simulator, morphed into a general-purpose computer which stood at the heart of the Semi-Automatic Ground Environment (SAGE), a multibillion-dollar network of computers and radars that promised to computerize the U.S. Air Force’s response to a Soviet nuclear attack by streamlining the detection of incoming bombers and automatically deploying fighters to intercept them.
In 1956, with the SAGE system not yet finished, Forrester abruptly changed careers, shifting his gaze from electronic systems to human ones. From the unlikely setting of the MIT Sloan School of Management, he founded a discipline called “industrial dynamics” (later rechristened “system dynamics”). At first, this field focused on creating computer simulations of production and distribution problems in industrial firms. But Forrester and his cadre of graduate students later expanded it into a general methodology for understanding social, economic, and environmental systems. The most famous example of this group’s work was the “doomsday” World 3 model that stood at the center of the landmark environmental text The Limits to Growth, a book that warned of a potential collapse of industrial civilization by 2050.
Urban Dynamics was Forrester’s first attempt to apply his methodology outside of the corporate boardroom. He came up with the idea of tackling the problems of cities after meeting John F. Collins, a conservative Democratic politician and the outgoing mayor of Boston, who had recently taken a position at MIT. Listening to Collins’s stories about his time as mayor in the 1960s, Forrester became convinced industrial dynamics could be used to study the poverty and capital flight associated with the US’s ongoing “urban crisis.” Even though Forrester had no expertise in urban affairs—or in social science more generally—Collins agreed that a collaboration could prove fruitful.
Over the course of 1968, Forrester devoted about twenty-five hours a week to his project with Collins. During that time, he met with the former mayor and his team of advisors and constructed a sprawling flowchart representing the relationships between different aspects of the city’s structure. Forrester translated this flowchart into the group’s house simulation language, DYNAMO. After a staff secretary or a graduate student punched the DYNAMO equations onto cards, it could be loaded on a machine. From there, the computer would generate a running version of the model, and output line-plots and tabular data representing the decade-by-decade evolution of the simulated city.
Forrester spent months tinkering with this model, tested and corrected it for errors, and ran a “hundred or more system experiments to explore the effects of various policies on the revival of a city that has aged into economic decline.” Six months after beginning the project, and over 2000 pages of teletype printouts later, Forrester declared that he had reduced the problems of the city to a series of 150 equations and 200 parameters.
At the start of a standard 250-year run in Forrester’s model, the simulated city is empty. No land is occupied, there is no economic activity, and there is little incentive for construction. As the city gradually develops, increases in housing, population, and industry all serve to reinforce one another, and the city enters a period of sustained economic and population growth. During this period, people are attracted to the city, and new housing and businesses are constructed at a brisk pace.
But as the city ages and its land area reaches full occupancy, growth slows. Areas which were considered “attractive and useful” have already been occupied. New construction takes place on more marginal land, and because this land is less attractive, the pace of construction slows. When no greenfield parcels of land are left for development, new construction becomes impossible, and new housing and firms can only be built when old ones are demolished. New migrants, once a boon to the city’s industry, continue to flow into the metropolis, causing overcrowding and underemployment, dampening economic vitality, and sending the city into a death spiral of blight and decay.
The arc of this story reflected the simplified, and sometimes entirely fictional, assumptions of Forrester’s model. On its most basic level, Urban Dynamics modeled the relationship between population, housing stock, and industrial buildings against a background of government policies. The city inside Forrester’s model was a highly abstracted one. There were no neighborhoods, no parks, no roads, no suburbs, and no racial or ethnic conflicts. (In fact, the people inside the model didn’t belong to racial,ethnic, or gender categories at all.) Economic and political life in the outside world had no effect on the simulated city. To the extent that the world outside the model existed, it served only as a source for migrants into the city, and a place for them to flee to if the city became inhospitable.
The residents of Forrester’s simulated city belonged to one of three class categories, “managerial-professional,” “worker,” and “underemployed.” As one moved down the class ladder in the urban dynamics model, classist assumptions about the urban poor piled up: birth rates were higher, tax contributions were lower, and the use of public expenditures increased. This meant that the urban poor served as a massive drag on the health of the simulated city: they did not add to its economic life, they had large families which strained public services, and they contributed only paltry amounts to the city’s coffers.
Forrester was cagey about how much this mapped onto real life. He cautioned that his model was a “method of analysis” and that it was unwise to take its conclusions as applicable without first ensuring that the model's assumptions fit a specific city’s situation. At the same time, Forrester used the simulation as a stand-in for cities in general, making sweeping claims about the failure of what he regarded as “counterproductive” urban policies.
To Forrester, low-income housing was an especially egregious example of a “counterproductive” urban program. According to the model, these programs increased the local tax burden, attracted underemployed people into the city, and occupied land which might otherwise have been put to more economically healthy uses. Housing programs aimed at improving the condition of the underemployed, Forrester warned, “increased unemployment and reduced upward economic mobility” and would condemn the underemployed to lifelong poverty. This idea wouldn’t have seemed new to anyone steeped in the conservative or libertarian tradition, but Forrester’s technical approach helped update it for the digital age.
The Perversity Thesis
When we consider the social effects of computers in political and social life, we usually think in terms of expanded power and new possibilities. This perspective on computation permeates even our critical visions of technology. But we should also be attentive to the power that computers and the accompanying language of “systems” and “complexity” have to narrow our conception of the politically possible.
Forrester thought that the basic problem of urban planning—and making social policy in general—was that “the human mind is not adapted to interpreting how social systems behave.” In a paper serialized in two early issues of Reason, the libertarian magazine founded in 1968, Forrester argued that for most of human history, people have only needed to understand basic cause-and-effect relationships, but that our social systems are governed by complex processes that unfold over long periods of time. He claimed that our “mental models,” the cognitive maps we have of the world, are ill-suited to help us navigate the web of interrelationships that make up the structure of our society.
For him, this complexity meant that policy interventions could, and usually would, have very different social effects than those imagined by policymakers. This led him to make the stark assertion that “the intuitive solutions to the problems of complex social systems” are “wrong most of the time.” In essence, anything we do to try to improve society will backfire and make things even worse.
In this respect, Forrester’s approach to the problems of American cities mirrored the “benign neglect” outlook of influential Nixon adviser Daniel Patrick Moynihan and the rest of the administration. Indeed, Moynihan was an enthusiastic proponent of Forrester’s work and recommended Urban Dynamics to his fellow White House officials. Forrester’s arguments enabled the Nixon Administration to claim that its plans to slash programs created to help the urban poor and people of color would actually, counterintuitively, help these people.
Forrester’s central claim about complexity wasn’t a new one; it has a long history on the political right. In a 1991 book, Rhetoric of Reaction, the development economist and economic historian Albert O. Hirschman identified this style of argument as an example of what he called the “perversity thesis.” This kind of attack, which Hirschman traced back to Edmund Burke’s writings on the French Revolution, amounts to a kind of concern trolling. Using this rhetorical tactic, the conservative speaker can claim that they share your social goal, but simultaneously argue that the means you are using to achieve it will only make matters worse. When commentators claim “no-platforming will only make more Nazis,” that welfare programs lock recipients into a “cycle of dependency,” or that economic planning will lead a society down a “road to serfdom,” they’re making this kind of perversity argument.
What Forrester did was give the perversity thesis a patina of scientific and computational respectability. Hirschman himself makes specific reference to Urban Dynamics and argues that the “special, sophisticated attire” of Forrester’s models helped reintroduce this kind of argument “into polite company.” In the nearly fifty years since it has come out, Forrester’s “counterintuitive” style of thinking has become the default way of analyzing policy for mainstream wonks. For many, “counterintuitivity” is the new intuition.
Expert knowledge, of course, has an important place in democratic deliberation, but it can also cut people out of the policy process, dampen the urgency of moral claims, and program a sense of powerlessness into our public discourse. Appeals to a social system’s “complexity” and the potential for “perverse outcomes” can be enough to sink transformative social programs that are still on the drawing board. This might not matter in the context of a virtual environment like that of Urban Dynamics or SimCity, but we have decades of real-world evidence that demonstrates the disastrous costs of the “counterintuitive” anti-welfare agenda. Straightforward solutions to poverty and economic misery—redistribution and the provision of public services—have both empirical backing and moral force. Maybe it’s time we start listening to our intuition again.