The Limits of Simplicity: Toward Geocomputational Honesty in Urban Modeling

Occam's razor informs us that simpler models are better. Yet models of geographical systems must often be complex to capture the richness of the human and natural environments. When critical decisions, such as those affecting human safety, depend on models, the model must be sufficient for the task, yet can become impossible for all but the modeler to understand holistically. This paper outlines five model components, and examines how a move toward simplicity impacts each, focussing on the critical stage of model calibration. Three model applications to the same data are examined in terms of how their calibrations took place. A comparison of the results shows that model outcomes are subjective, and more shaped by their role in decision-making than by scientific criteria. Model reduction, a goal as yet lacking methods, may therefore not be a useful way forward for geocomputational modeling until the full context of models and the ethics of modeling are better understood