The audience of academics and journalists gathered at the Brookings Institution had every reason to be excited: the venerable liberal-leaning think tank was announcing the publication of a new book with, in the words of Robert E. Litan, director of economic studies, "revolutionary" implications. It was, Litan declared, the "most innovative, potentially pathbreaking" book ever to bear the Brookings name. The work, Growing Artificial Societies, by Joshua M. Epstein and Robert Axtell, describes the two scholars' investigations of their computer-based world called Sugarscape. After the lights went down in the auditorium, Epstein and Axtell provided commentary while a giant display showed how "agents"--simpleminded red and blue blobs representing people--scurry around a grid, competing for yellow resources whimsically named "sugar" and "spice." The agents--more elaborate versions of the dots in John H. Conway's game called Life--eat, mate, trade and fight according to rules set down by their human "gods."
The agents' antics invite anthropomorphism. (Epstein referred to one as practicing "subsistence farming" far from the sugar "mountain.") The model's purpose, he explained, is, by employing "radical simplification," to study interacting factors that affect real societies. When agents are let loose in Sugarscape, trends emerge that would be hard to predict. A population may oscillate in size, and often a few individuals garner much of the "wealth," a well-known phenomenon in human societies. The researchers hope to discover which rules, such as inheritance laws, generate which effects. The pair is now using a similar approach to model the population decline of the Anasazi of the American Southwest.
A reporter (this one, truth to tell) asked how the Brookings scholars would know that the critical rules identified in Sugarscape are actually important in the real world. "You don't," Epstein admitted, because many different real-life rules could produce the same outcome. That is a problem with all science, he said, noting that Sugarscape nonetheless provided an improvement on existing economic theory ("mainly just a lot of talk"). But Thomas C. Schelling of the University of Maryland, an early pioneer of agent-based modeling, reinforced Scientific American's lingering doubt: "You still want to ask, If there are many different ways of producing [a] phenomenon, how do we know we have captured what's really there?" Schelling observed.
Another reporter rained on Brookings's parade by asking whether Sugarscape has implications for, say, tax policy. That prompted a reminder that the model is still in early development and that Sugarscape would first have to model governments. Epstein earned notoriety in 1991, when he used other computer simulations to calculate that between 600 and 3,750 U.S. soldiers would die in the Gulf War. In fact, only 244 died. But his confidence is apparently undented. On a roll, he started speculating at the Sugarscape book launch that agents could shed light on whether free markets will survive in Russia. John D. Steinbruner, a senior foreign policy fellow at Brookings, interjected that no computer model is ready to answer that question. Soon after, he wound down the discussion, noting that although "well short of a complete account," Sugarscape offers tools "that do appear to be very useful in the process of conceptualization."
This article was originally published with the title Which Came First?.