[PHI 3800] The theory-ladenness of observation and the observable/unobservable distinction: a match made in hell?

According to Kuhn, observation is theory-laden. That is to say, observational evidence is not "purely objective." Rather, observational data is infused with the theoretical commitments of those who perform the observations.

The notion that observation is theory-laden can figure in an argument against the "objectivity" of theory choice in science as follows:

1. Scientists can objectively choose between competing theories only if observational evidence is theory-free.
2. Observational evidence is not theory-free (it is theory-laden).
3. (Therefore) Scientists cannot choose objectively between competing theories.

Now, it might seem as if this argument from the theory-ladenness of observation is another weapon that anti-realists can use in order to argue against scientific realism. But is it?

The most prominent form of anti-realism today is probably constructive empiricism. Constructive empiricists claim that the most appropriate attitude to take with respect to what scientific theories say about unobservable entities, events, and processes is agnosticism. In other words, constructive empiricists recommend agnosticism with respect to theoretical entities, such as atoms, genes, and the like. Clearly, the observable/unobservable distinction is a key tenet of constructive empiricism.

Now, are the notion that observation is theory-laden and the observable/unobservable distinction complementary? Or are they in tension? If observation is not theory-free, but is rather laden with theoretical assumptions, can a real observable/unobservable distinction still be drawn?