Dawkins’ claim is based on this false dichotomy: either reality is completely arbitrary, or scientific exists independently of all of our concepts. Obviously reality is not completely arbitrary, but he ignores the fact that all the scientific truth we have ever had was either corrected by present science or will be corrected by future science. it’s logically possible that someday the skies will part, the perfect theory will manifest itself to us, and no one will ever make any errors again. But it’s far more likely that our theories will always be models which approximate reality rather than capture it perfectly. consequently, the mind-independent “scientific truth“ that he talks about will be something that no scientist will ever actually see. What we have instead is a variety of theories, which are true to varying degrees, and none of which are true in the sense Dawkins is talking about. This is not as scary as it first appears. Here’s a quote from my book “neither brain or ghost“ which discusses this issue.

“Surely we don’t want to say that truth always admits of degrees. After all, statements like “Paris is the capital of France” and “2 plus 2 equals 4" are 100 percent grade A true; there are no degrees of truth there. This objection, however, blurs a subtle but important distinction. The claim that an entire theory cannot be false in a binary true-or-false sense does not apply to statements made within the context of a theory. Suppose that a medieval merchant and his customers sincerely believed that various exotic items for sale in the merchant’s shop (such as narwhal horns and kudu antlers) were unicorn horns. A customer asks, “Are there any unicorn horns in this shop?” The clerk goes back to the stockroom, and after a casual search that fails to reveal a small box with a few of these items on the third shelf from the left, comes back and tells the customer, “No, there aren’t.” The clerk’s statement is false, in the traditional binary sense of true or false. Within the context of discourse for that statement, there are unicorn horns in the shop, and the clerk has made an error. But to conflate the observation statement “There are no unicorn horns in the stockroom” with the ontological assertion “There are no unicorn horns” is to make a serious category mistake. After all, the discovery of narwhals would not have justified the clerk’s saying to the customer, “See, I told you we didn’t have any unicorn horns.”

A continuum theory of truth does not require us to refrain from criticizing or comparing theories. But just because a theory is muddle-headed, or prone to errors, or in need of replacement, does not mean that what it is talking about exists only in the head. Astrology is an illegitimate and useless conceptual system, but it is still a real objective fact that there are a certain number of Sagittarians in Kansas, and this is a fact about Kansas and Sagittarians, not a fact about our mind-brains. Even if almost everything that astrology says about Sagittarians is false, it still provides us with an observation language that divides up the world in a distinctive way, however crude and misleading. Once we have divided up the world with such an observation language, it becomes possible to make binary judgments. This is a good thing, for it is impossible to act in the world without making binary judgments. Every activity presupposes judgments like “Is this cake done or not?” or “Is this a predator or not?” The brittle silicon-inspired metaphors that are the basis for cognitive science models of mind are surely right about this, whatever their other limitations. (If the cake is done, take it out. Otherwise leave it in and repeat step one at a later time.) But before we can act, we must first embrace an ontology. The choices between ontologies, unlike the choices made within any given ontology, are not binary. No ontology is perfect, and which ontology is preferable varies depending on one’s goals and purposes.”