Prediction and Prescription in Systems Modeling

This chapter discusses how to conserve computational power in modeling, by using the hierarchical properties of systems to aggregate and thereby simplify them, and by substituting symbolic modeling, where appropriate, for numerical modeling.

Abstract

Modeling is a principal tool for studying complex systems. Since models may be used for predictions, for analysis, or for prescription, we must ask what our goals are before we build our models. Historically, predictive numerical models have dominated our practice. Since the world we are modeling is orders of magnitude more complex than even the largest models our computers can handle, we must conserve computational power, first, by asking how much temporal detail we need and how much can be supported by available data and theories, second, by asking whether knowledge of steady states may not be more important than knowledge of temporal paths, third, by using the hierarchical properties of systems to aggregate and thereby simplify them, and, fourth, by substituting symbolic modeling, where appropriate, for numerical modeling.