Scientific realists generally think we should believe in the observable and unobservable entities posited by our best scientific theories. Anti-realists oppose realism in various ways. One motivation to be an antirealist comes from empiricist worries about unobservables. Past mistakes in positing unobservable things like aether, phlogiston and vital force were lessons learned for these empiricists. So, from here on out, unobservables were not to enjoy the same status as observables for these antirealists.
There’s much, much more to describing realism and anti-realism, but that will be beyond the scope of this post. See the SEP.
Two popular forms of antirealism are instrumentalism and constructive empiricism. Traditionally, instrumentalists like the logical positivists take talk of unobservables to be useful fictions. Constructive empiricists take talk of unobservables to be meaningful (like the realist), but say that we cannot know whether they exist. On this view, science aims to be empirically adequate, and a theory is empirically adequate if and only if what it says about observables is true.
As Andre Kukla says:
Both types of anti-realist wish to ascribe a philosophically superior status to the observational – for instrumentalists, the superiority is metaphysical; for constructive empiricists, it is epistemological. Evidently, the coherence of these positions depends on there being a coherent way to distinguish the observational from the nonobservational. On the standard view, the distinction rests on a difference between two parts of the scientific vocabulary: observation language and theoretical language. Kuhn argued that this linguistic distinction could not be made. Anti-realism would seem to be a non-starter if Kuhn is right.
Theoretical language, as used above, just means language about unobservables, e.g. ‘electron’. For Kuhn, the distinction between observation and theory could be not made because all language and even perception is theory-laden.
Bas Van Fraasen agrees with the Kuhnian view of the theory-ladenness of language, but he thinks that we can maintain the observable/unobservable distinction. It’s intuitive enough that rocks are observable. Electrons, by contrast, are unobservables, because we don’t actually see the electron; rather, we see their effects, like a streak in the cloud chamber.
According to Andre Kukla, this raises some problems: anti-realists want to draw a hard line between observables and unobservables; but what counts as observable changes as our instruments develop.
For Van Fraassen, something counts as observable if we can detect it without the aid of instruments. But this still doesn’t draw a hard line between observables and unobservables. People with bad eyesight, or the blind will have a smaller class of observables.
Maybe we should talk about scientific communities, rather than individuals; because it makes sense to say that a blind person can adopt the beliefs of the expert scientists as his own beliefs–most of us don’t do science, we get our science from scientists. But what if we have a mutation that allows us to see better; or, what if we allow a highly advanced alien with super-sight into our scientific community? We are back to the ambiguous line between observables and unobservables.
Kukla takes these considerations to be problematic for anti-realists, but I don’t see why it should phase the anti-realist. I don’t see why Van Fraassen et al. need to hold to a hard, fixed line between observables and unobservables. Constructive empiricists want to place epistemic superiority on observables, but why should there be a problem if our epistemic situation changes? What counts as epistemically superior will change with time.
Kukla, Andre. Observation in The Routeledge Companion to Philosophy of Science.