Modeling and Validation Challenges for Complex Systems

Abstract

Many important systems, both natural and artificial, may be classified as complex, and the study of complex systems is ongoing. Such systems have special defining characteristics, including sensitivity to initial conditions, emergent behavior, and composition of components. Complex systems are increasingly prevalent as the subject of modeling efforts. There are at least two reasons for this; first, the systems that are of the greatest practical interest and thus most likely to be modeled tend to be complex, and second, because complex systems resist closed form analysis modeling is often the only way to study them. Unfortunately, the special characteristics of complex systems lead to additional challenges in both effectively modeling them and in validating the models. This paper, which takes the form of an introductory tutorial and literature survey, first defines complex systems in terms of their key characteristics and describes how validation risk applies to models of them. It then identifies a series of modeling and validation challenges that follow from the defining characteristics and suggests mitigation approaches for those challenges.