All models are wrong

British statistician George Box rather notoriously stated in 1976 that, “All models are wrong, some are useful.” For almost 50 years now these words have both haunted and challenged model builders. While it is factually true that it is almost impossible for a model to be 100% correct all the time, especially complex ones such as those endeavoring to capture the nuances of aerospace or drug delivery, there is a much more important point to be made about his usefulness comment. Models can be wonderful tools to explore choices, aid decisions and even avoiding disasters. Early in a modeler’s career they are exposed to the power of a model’s ability to perform a sensitivity analysis. As opposed to a simple equation where inputs and outputs are rather well inferred, in complex systems one does not know what a small change to an input might mean to an output. Often more importantly, models tell us where a system is most reactionary to small inputs and where even large inputs potentially cause little or no output change. This zone of critical operability is essential to things like systems design, urban planning and health management. Sensitivity analysis is extremely useful and even in models that might not exactly hit the mark in terms of predicting the future, they can help us navigate it much more effectively.