Nautilus has an interview with Bas van Fraassen which is worth the read.
For me, it underscores my sense that the realism/anti-realism debate as it has been framed in the last few decades poses a false dichotomy. As Van Fraassen frames it, the choice is to be an empiricist who believes only the observable predictions of science or a realist who thinks that science is after the singular underlying truth of everything. In images that the interview plays with, observations are just fragments of the real world: the tips of icebergs above the surface, in contrast to the vast totalities beneath the waves which are postulated by theory. The empiricist just believes in the fragments while the metaphysical realist believes in the vast totality.
The alternative is an obvious middle ground which believes in lots of the fragments and in some of the contours beneath the waves, without any pretense that science should or could reveal the totality of everything. For example, I am more confident in the existence of electrons than I am in some poorly confirmed empirical phenomena. But that doesn’t mean I believe in superstrings, nor that I am sceptical about observables in general.
In the interview, van Fraassen pulls on similar intutions. He says:
My main point is that it is practically impossible to describe the chaos of what actually happens in the world. We can construct useful theories or models that are empirically adequate—that tell us something, for instance, about the behavior of what we call electrons, without having to say what an electron is.
The thing to note about this is that there can only be behaviour of electrons if there are electrons. Even though electrons are unobservable, we learn about what they do. This is compatible with denying that we learn anything about their fundamental nature or about deep metaphysics. Of course, van Fraassen goes further by arguing that we just ought to believe in the observable results of electrons.*
We can disagree over whether my position should be called “realism” or not, but there is precedent in usage of the term. For example, so-called “entity realists” argue that we should believe in entities (like electrons) insofar as we can manipulate them.**
* Insofar as this is motivated by the humble thought that we can’t describe the chaos of what actually happens, it seems like an odd move. It’s in one way too humble, because it stops short of even the most well-confirmed unobservable entities. It’s also not humble enough in imagining that science aims to capture all the observables. There are lots of trivia which are observable only at great pain and expense, which are outside the purview of science.
** I differ from the entity realist by thinking that the contour between what merits belief and what’s mere construction doesn’t always cut between entities and natures. But entity realism sets the precedent that partial realism which eschews deep metaphysics still counts as “realism”.