I'm teaching my course on scientific realism again. I especially enjoy it as an opportunity to re-trace the historical evolution of the realist and antirealist positions. In particular, I'm struck by the following: an earlier version of a realist argument might have been quite cognizant that it blunted a certain line of criticism (call it X), but then that argument is revised in response to something else in a manner that makes realists vulnerable to X. Here's one that's really struck me re-reading early versions of the "No Miracles Argument.":
Originally, scientific realists claimed that they, but not empiricists, could explain science's empirical success. Very roughly, the thought was that the realist could claim that the approximate truth of a theory best explains its empirical adequacy, but the empiricist was stuck claiming that a theory's empirical adequacy explains its empirical adequacy, which is no explanation at all.
However, owing to the pessimistic induction, the realist's explanandum became restricted to high-grade empirical success, such as novel prediction. Hence, the revised realist claim is that a theory's approximate truth explains its high-grade empirical success.
However, to my knowledge, few have noted that this means that empiricists thereby have an explanation of high-grade success. Empiricists can claim that a theory's empirical adequacy explains its high-grade empirical success. This is no longer circular, and indeed seems to be part of a general class of "success explanations," X's general reliability in domain D explains why X achieved a high-grade success in D. Ex. LeBron James' athleticism and basketball prowess explains why he was able to run down the shooter and block the latter's shot.
At this point, however, one may note that empirical adequacy, at least as defined by van Fraassen, is unlikely to be the true of any theory. However, so long as the empirical success in the explanans is more encompassing than the high-grade empirical success, empiricists can explain the high-grade empirical success. Let's call this general kind of empirical success which explains high-grade success empirical reliability, which is roughly akin to "approximate empirical adequacy."
What I can't see quite clearly is whether this empiricist explanation of high-grade success is better or worse than the realist explanation. But, of course, this has always been a difficulty with using Inference to the Best Explanation to adjudicate anything.