Ian Garrick Mason

Book Reviews

San Francisco Chronicle -- June 27, 2004

Critical Mass: How One Thing Leads to Another

Philip Ball
Farrar, Straus & Giroux, 520pp, $27.00

by Ian Garrick Mason

The application of precise scientific methods to the messy world of human interactions might sound like just another exercise in ultramodern overconfidence. But ultramodern it's not: One of the first things that Philip Ball points out in Critical Mass: How One Thing Leads to Another is that philosophers have been attempting this feat ever since 1651, when Thomas Hobbes in "Leviathan" tried to deduce from first principles the shape that society should take. For the rationalist philosophers who succeeded him, the overarching aim was to uncover the "laws" of society, as scientists like Sir Isaac Newton had uncovered the laws of physics.

It's not necessarily an exercise in overconfidence, either, because the wall between physics and social science has rarely been impermeable. When 19th century Scottish physicist James Clerk Maxwell, for example, was faced with the insurmountable task of predicting the behavior of gases made up of trillions of rapidly moving molecules, he borrowed from social scientists the mathematical tools of statistics and successfully applied them to the problem.

Now knowledge is crossing the border again. Ball, a prolific science writer for publications such as Nature, surveys a whole series of recent attempts to apply to society the tools developed for physics, including the latest techniques of computer modeling. These models -- in which diverse individuals are radically simplified, "smoothed down to billiard balls," and described by only a handful of attributes and motivations -- can demonstrate remarkable things. One shows how realistic flocking behavior can spontaneously emerge from a crowd of independent "boids" (simulated birds), each of which is programmed only to avoid obstacles and to always seek the center of gravity of the nearest group of boids. Another model demonstrates how clusters of objects moving down corridors in opposite directions will naturally coalesce into the right-left streams familiar to pedestrians in order to avoid collisions. Ball presents such research convincingly and identifies ways in which it might be used to shorten traffic jams or improve building safety.

Other research, however, seems on weaker ground. Ball enthusiastically discusses the work of political scientists Robert Axelrod and Scott Bennett in modeling the international situation before World War II. In their computer program, each European belligerent is represented by a particle with several attributes -- religion, ideology, economy and so on -- that determine its propensity to ally with other nations. When the program runs, the particles attempt to combine to find the lowest "energy level," just as real particles do when finding a stable state such as solid, liquid or gas. In the event, the model predicted two "stable" arrangements of European powers: one in which the historical lineup of Axis versus Allies is reflected fairly accurately, and a second in which all of Europe is aligned with Nazi Germany against the Soviet Union. Ball sees in this program a useful tool: "We can talk in (somewhat) quantitative terms about worlds that might have been, and identify the factors that helped things turn out this way and not that."

The problem is that instead of staying true to a "physics" of many simple particles interacting with each other, the model uses only 17 particles, each of which has six or seven attributes -- and therefore its results may well be determined more by the attributes than by the interactions. Is it really that surprising, after all, that a model that includes "ideology" as a national attribute should end up reflecting Europe's historical divisions? Indeed, why shouldn't we make the model even more accurate by adding attributes like the conflicting personalities of national leaders? Or less accurate by adding attributes nations have in common, like post offices, railways or poor people?

It all seems rather arbitrary: Axelrod's model isn't bottom-up physics applied to international relations, but international relations abstracted and crammed into a physics program. If you want to understand the interactions of small numbers of complex nations, you're better off just writing history.

This is not to dismiss the book as a whole, which is chock-full of stimulating hypotheses and well-told stories. To his credit, Ball often expresses an awareness that the work he depicts is at an early stage of development and needs to be treated carefully. "Just because you get what looks like the right kind of collective motion," he warns about crowd modeling, "that doesn't mean you have the right rules."

Such cautions, however, are generally overshadowed by his open enthusiasm for the science's potential. While this certainly makes the book a pleasure to read, it also means that a reader needs to approach it with a skeptical mind, keeping a sharp eye out for those times when one thing doesn't, in fact, lead to another.