Model Evaluation

The aim of model evaluation is to learn about the quality of one or more scientific models. Quality may be conceptualized as accurate and comprehensive representation, relevant similarity, fitness-for-purpose, or in some other way. The conception of quality that is adopted carries implications for the practice of model evaluation, including whether it must attend to factors other than model-target relationships. Numerous obstacles and challenges arise in model evaluation, such as limited observations of the target system, model opacity, holism in assessment, and uncertainty about how to quantify model quality. As a consequence, it is sometimes difficult to reach confident conclusions about model quality.