Poincaré, with all this virtue, made a serious philosophical mistake. In Poincaré’s own work, this error seems to me to have kept him from several fundamental discoveries in physics. (Howard Stein "Physics and Philosophy Meet: the Strange Case of Poincaré," 1.)
One way to describe this mistake--which I think is related to a wider philosophical tradition--is to say that Poincaré construed the function of theoretical work as essentially administrative--an activity of information-processing. The job of the theorist par excellence..-is to find the simplest and most systematic arrangement for storing scientific information--within the limits defined by those assumptions so entrenched as to function as a priori principles, i.e., as “definitions in disguise.” (Stein, 16)
This mistake is a characteristic one in the empiricist tradition. It involves what seems to me a very odd paralogism. One recognizes--and I cannot say too strongly how right I think this is--that experience is our only touchstone of true information. One observes, with Hume, that the data of experience do not logically entail knowledge of “objective existence” or of the future. In application to the results of science, one draws sophisticated conclusions about the merely hypothetical status of the objects and agencies of scientific theory--e.g., atoms or the ether. But one neglects thereby the earlier insight--the Berkeleyan insight, which precedes Hume--that in a fundamental empiricist epistemological analysis all the objects and agencies of common sense and ordinary life have that same status. In effect, this is to distinguish, whether tacitly or explicitly, between the “reality”--or just persuasiveness--of what the ordinary processes of common sense make of the empirical data, and the reality or persuasiveness of the conclusions reached by the sophisticated processes of science. (Stein 17)
[A]n inadequacy in what may be called Poincaré’s conception of the dialectic of scientific concepts and theories, and of their relation to experiment. And that inadequacy I have earlier associated with what I called a very odd paralogism that recurs in the empiricist tradition: in effect, a double standard of epistemological critique, which occurs with some frequency among empiricist and instrumentalist philosophers of science.....
There is no warrant at all in the instrumentalist view for grading the entities of a theory in degrees of reality or fictitiousness--regarding particles as more real than the ether. This is a double standard that recurs in the empiricist tradition, (Stein, 22, and Stein quoting himself.)
Some over-sensitive historians dislike using judgments like 'mistake' in treating of the past. They mistakenly believe that such judgments are a form of anachronism or hindsight bias. Howard Stein, the greatest living philosophical historian of science,* certainly is not afraid to introduce anachronism when it suits his purposes (including in the quoted passages above). But, that's not what is going on in his judgment that a mistake has been made. Agents, except, perhaps, a divine one, can make mistakes, and worse. Even Homer nods. For, one can specify circumstances in which an agent ought to have known better (given existing best practices, regulations, etc.). Malpractice lawyers make a fine living arguing the details.