Understanding World Models

Computer models of the world system produce very different results, ranging from economic collapse and massive starvation in the 21st century to universal prosperity for double or triple the present world population. The strikingly different conclusions that arise make it urgent to compare them effectively with one another, and see what it is about them that produces such diverse policies. And even insofar as the policies are similar, one would like to know more about how they arise from the models. This paper suggests a line of analysis that permits comparison of properties among such models. It takes up two ways of seeing what is in a model in addition to examining its documentation: first, making alternative transparent models that check the partial results of the complex model; and, second, 'black-box' experiments leading to a truncated linear form of the complex model. These two methods of assessment are designed to replace most of the documentation, and to allow the user to understand more effectively what assumptions he commits himself to in using the model