A couple of notes on Hacking's chapters on Observation and Microscopes in Representing and Intervening, with possibly a mention or two of van Fraassen's 2000 APA paper "Constructive Empiricism Now."
Three questions and an observation:
1. Is there a distinction between observable and unobservable?
We've seen that Maxwell says "no": observable must mean observable in principle, and in principle we could devise instruments to let us observe pretty much anything.
van Fraassen of course says "yes." In his 2000 paper he gives the example of light: we can observe objects that are illuminated by light, but we can't see the light itself.
2. Do we see through microscopes?
Maxwell (and probably most people): "yes." van Fraassen: "no." Hacking points out that microscopes are much more complex than we usually think of them as being, and the way they cause us to see images is quite indirect (more so in some types of microscopes than others). Nevertheless he thinks we see through microscopes (as well as TV screens, monitors used in MRI imaging, etc.).
How far can we push this? Hacking thinks that I see an object when I look at a digitized image of it (photo, MRI, or whatever) provided that the route from object to image is sufficiently direct. On the other hand, I don't see the object if I look at a drawing of it (even a drawing indistinguishable from the digitized image). There's an interesting parallel here with an issue in aesthetics, namely whether we see objects "through" photographs. Kendall Walton, "Transparent Pictures," argues that when I look at a photo, I see the object photographed, but when I look at a photorealistic painting, I do not.
Notice that the exact location of the observable/unobservable divide is not essential to constructive empiricism, however. As van Fraassen and Hacking both observe, shifting the observable/unobservable distinction from where van Fraassen locates it so as to include things "seen through" microscopes would leave most of his view intact. Accepting a theory would require believing a little bit more than he thinks it does, but not very much more. The crucial questions for constructive empiricism are 1 and 3.
3. Is the observable/unobservable distinction epistemically significant?
(This of course presupposes that the answer to 1 is "yes.") van Fraassen: yes. Most of his opponents (except for antirealists of the positivist sort): no. Hacking: no -- whether we regard an entity as real has more to do with whether we can manipulate it than with whether we can observe it. (At least, I think this is his point, or one of them.)
Observation: Empirical vs. Pragmatic Virtues of Theories
Note how neatly van Fraassen's views on different topics fit together. There are two kinds of grounds for theory choice: empirical and non-empirical or pragmatic. The empirical basis is getting the observable part of the world right: what vF calls "empirical adequacy." Pragmatic virtues include things like explanatory power, unification, and simplicity. van Fraassen thinks that the pragmatic criteria are good reasons to favor one theory over another. But they do not give reason to think one theory rather than another is true. How can simplicity, e.g., be a good reason to accept one theory over another even though it is not a reason to think the favored theory is true? Because, for van Fraassen, accepting a theory does not require believing that it is true! It only requires believing that it is empirically adequate.