"We may consider the horse as not having a determinate color, or else we may consider the horse not as having a determinate color. To consider the horse as not having a determinate color is to hold, or attempt to hold, as the object of our thought a horse that simply has no determinate color—a creature never encountered in physical reality... This sort of abstraction falsifies and contradicts the concretes on which it is based. But to consider the horse not as having a determinate color is simply to consider the horse as a horse without considering its color one way or the other; and here no falsification is involved.
In short, a precisive abstraction is one in which certain actual characteristics are specified as absent, while a nonprecisive abstraction is one in which certain actual characteristics are absent from specification. [The latter is to be preferred in models of human behavior because it permits the useful distillation of complex phenomena for study without succumbing to falsification.]...
[Milton] Friedman…thinks that a worthwhile economic theory ‘must be descriptively false in its assumptions,’ since it ‘takes account of, and accounts for, none of the many other attendant circumstances’ but instead ‘abstracts the common and crucial elements from the mass of complex and detailed circumstances.’ Friedman is of course quite right that an economic theory needs to leave aside a mass of complex details; but so long as it leaves them aside by failing to specify them, rather than by specifying their absence, it does not need to be descriptively false.
A perfect-competition model, for example, does not merely fail to specify the existence of entrepreneurial error; if it did, it would fail to explain much about the workings of the market, but at least it would not say anything false. Rather, a perfect-competition model, by positing that all economic actors possess complete (and completely similar) information, explicitly specifies the absence of entrepreneurial error—and it is to this falsification that Austrians object. George Reisman (1968) was exactly on target when he characterized the perfect-competition model as ‘Platonic Competition’; and Friedman is making precisely the Platonic mistake of treating all abstraction as a form of idealization."
--Roderick T. Long "Realism and Abstraction in Economics: Aristotle and Mises versus Friedman" (2006)