The validation of models of complex systems

This paper begins with a short review of various approaches to validation proposed in the modeling literature. It then discusses three alternative views of models: (1) the concept of model in mathematical logic, (2) the conventional (empiricist) view of models in the social sciences, and (3) an alternative view that sees models as the products of metaphor construction and interpretation. This third view leads to the identification of four different areas in model validation: (1) evaluation of adequacy of metaphors, (2) observation, measurement, and experimentation, (3) internal validation, and (4) external validation. The process of external validation proposed here recognizes that models are inherently partial, incomplete, and somewhat idiosyncratic and need to be confronted and compared with each other to produce new insights collectively.