When non-scientist friends and family ask me the dreaded question "What's the aim of your work?", I'm never quite sure how to frame my answer. This is particularly true when I'm asked "Aren't scientists looking for the ultimate truth?" (no, we're looking for grant money and pizza). Lately, I've come up with a reply that I think helps me explain my aims while remaining relatively agnostic about the realist vs. instrumentalist debate:
Okay, maybe I don't sit around all day drawing maps (or programming for Google Maps), but the underlying concept is the same. While I would imagine that the average cartographer accepts the existence of the Earth, I doubt they spend much time concerning themselves with the existence of some sort of Platonic ideal of a "nation" or a "desert." If you need to know where you'll need a passport, you use a map showing national borders. If you're interested in the local climate and ecology, you'll use a map showing the Earth's biomes. If you need to drive somewhere, you'll use a road map. Roads may objectively exist, while deserts may be merely a classification scheme, but neither map is objectively "right" or "wrong." Creating a map inevitably involves ignoring plenty of things that are really there, while simultaneously putting in some things that aren't. But this doesn't bother anyone, because we're looking for a useful guide to the world, not an exact replica of it.
By the same token, any theoretical model inevitably excludes effects that are "really there" (spin-orbit coupling, the gravitational pull of the moon, whatever), while introducing effects that aren't, often as a "fudge-factor" for ignoring the effects that are there. While I accept that there probably is such a thing as an electron (though its "reality" may be drastically different than anything we've imagined), I'm far more skeptical as to whether molecular-orbitals or density-matrices actually exist. But the point is that the actual existence of the concepts under discuss is irrelevant. If the molecular-orbital "map" guides me where I'm going, I'll use that; if the density-matrix "map" is more convenient, I'll use that instead. When I design models, my ultimate goal is to create a map that strips away anything irrelevant to that model's purpose—regardless of whether it exists—and to put in things that will further that purpose—again, regardless of whether they correspond to "reality." Ultimately, a map/model is not judged by it's "correctness," but rather by its usefulness.
Okay, so maybe at the end of the day I'm something of an instrumentalist after all. What do others (in particular experimentalists) feel of my response?