Subscribe
Subscribe
MY AMERICAN SCIENTIST
LOG IN! REGISTER!
SEARCH
 
Logo IMG
HOME > PAST ISSUE > Article Detail

COMPUTING SCIENCE

Computation and the Human Predicament

The Limits to Growth and the limits to computer modeling

Brian Hayes

Model Models

Very small models can yield surprisingly rich behavior. One example is the predator-prey model developed (independently) by Alfred J. Lotka and Vito Volterra early in the 20th century. With just two equations and four parameters this model explains cyclic fluctuations in the abundance of predators and their prey, such as wolves and moose. Feedback and overshooting lead to prolonged oscillations rather than a direct approach to equilibrium.

The simplicity of the Lotka-Volterra model is part of its appeal, yet we cannot insist that everything of interest in the world be crammed into no more than two equations. If you want to describe the whole of human society and the planetary ecosystem, you probably need a few more parameters.

In this context climate models offer a useful point of reference. General circulation models for the atmosphere and the oceans, along with related models of ice sheets and atmospheric chemistry, have several points of similarity with World3. At a conceptual level the structure is much the same: There are flows of air, water, heat and other entities, which the model must sum or integrate. The time scales are similar: In both cases we want to know what’s going to happen several decades out. And feedback loops are essential mechanisms in both kinds of models. (There are even historical connections. The use of general circulation models to study global climate change began in earnest at MIT circa 1970. The instigator was Carroll Wilson of the Sloan School of Management, who was also the person who got Forrester involved with the Club of Rome.)

These similarities are outweighed by differences. Where the Limits team had a casual attitude to data gathering practices—and outright hostility to statistical methods—the climate science community is passionate about collecting data, verifying its provenance and quantifying its uncertainty. General circulation models are not based on rough estimates or guesses but on decades of meticulously curated measurements—what Paul Edwards, in A Vast Machine, calls a “climate knowledge infrastructure.”

The organizational scale of the two undertakings differs by orders of magnitude. The World3 model was put together by a dozen people working in isolation for a year or two. Climate modeling is Big Science, with contributions from several hundred workers, organized in groups that both compete and collaborate, with institutional and governmental oversight, not to mention a great deal of public scrutiny. The process has been ongoing for 40 years.

Another difference is that climate models focus mainly on physical and chemical processes where the underlying science is generally well understood. We know a lot about the absorption and emission spectra of molecules in the atmosphere, and we know how a volume of air will respond to heating or to a change in pressure. The social and economic systems modeled in World3 do not have natural laws of the same predictive power. In this sense the climate problem is easier.

In another respect, however, the task of climate models is more demanding. Where World3 promises only to “illustrate the basic dynamic tendencies” of the system, climate models are expected to produce precise quantitative predictions, such as a 1 percent change in global average temperature.








comments powered by Disqus
 

EMAIL TO A FRIEND :

Subscribe to American Scientist