Models, Simulations, and Analogical Inference

Abstract

Models and simulations represent target systems by means of relations of similarity or analogy. Two objects or systems are similar if their attributes are close to each other or approximately equal. Two objects are analogous to each other if they are partly identical. From this perspective, it is useful to distinguish similarity models and analogy models as sources of learning about real targets. Similarity models include idealized models and their computer implementations which typically represent reality by deformation: while some irrelevant properties are excluded, some relevant properties are neglected by assigning them extreme values. Inferences from ideal similarity models are obtained either by approximation or by the concretization of counterfactual assumptions. Typical analogical models allow inference from the model to the target system by inductive inference from model data D to generalization C, and analogical reasoning from the model generalization C to the same generalization C about the real system.