We are sometimes misled by false analogies in the words we use. 'Natural selection' is an example of this, I think. The phrase 'natural selection' was chosen in order to parallel 'artificial selection'. This is great, since natural selection does parallel artificial selection; but the choice of the phrase has misled people into thinking that what natural selection and artificial selection have in common is selection, when in fact what they have in common is sorting. Artificial selection is sorting that involves selection; natural selection is sorting that is selection-like in some of its products, but does not involve selection.
This, incidentally, is why I have grave doubts that we can have any successful etiological account of function, i.e., an explanation of functions that involves natural selection picking out the function of something. Natural selection doesn't pick out anything; it doesn't even sort anything, because it is a sorting of a population (of whatever) by environmental factors that affect differential survival and differential reproduction. Systemic or 'cybernetic' accounts of function are much more promising; although they have more than their share of difficulties to face, as well.