Models and reality : how did models divorced from reality become epistemologically acceptable?

Economic models translate real problems to an artificial world, and calculate outcomes. The match between artificial worlds populated by rational robots, and the real world, is never assessed. Instead, models are judged on aesthetic grounds, involving conformity to preconceived principles of optimization and equilibrium. Despite methodological proclamations to the contrary, models are not judged by predictive performance. Economics models are formulated axiomatically, and never cross-checked against reality. Taking this (controversial) characterization of economic methodology for granted, this paper sketches trends in philosophy of science which led to a methodology which permits creation of mental models disconnected from reality. A key development was the separation of the observable phenomena from the underlying reality (noumena) which eventually allowed empiricist philosophers to jettison the underlying reality as part of what good models attempt to describe. The paper discusses contemporary methodologies for assessing models in economics and econometrics, and explains why these lead to models disconnected from reality.