In this short essay I point out several important aspects of Ian Hacking's realism in his book, and consider its implication on the scientific status of astronomy.
1. Realism about what?
Hacking makes a basic distinction between two kinds of realism (27):
(a) realism about theories -- theories in mature science are approximately true.
(b) realism about entities -- good many theoretical entities really exist.
Hacking's emphasis is apparently on (b). This is apparent from his slogan, "if you can spray them, then they are real" (22). But we should note that he is not saying that "if you cannot spray them, then they are not real" (see, e.g., 275).
How about (a)? His comments on Cartwright (215-219) are suggestive. Cartwright is an anti-realist about theories, and Hacking seems to agree. But we should be careful. Instead of supporting theories, they support local causal models. Models are "approximate representations of the universe" (217). It sounds to me that Hacking is arguing for realism about models, i.e., models in mature science are approximately true. If this interpretation is correct, this is a kind of variation of (a).
2. Cosmic accidents and manipulation
To know what makes him a realist, a good way is to compare his refutation of the cosmic accidents argument (54-55) with his argument for the reality of microscopic phenomena (200-205). Despite his own comments, the differences between two arguments are not so significant, except for one point -- through microscopes, we see grids manufactured by us. Because we made them, it cannot be a cosmic coincidence that we see the grids through (several different kinds of) microscopes. Anti-realists' argument against the cosmic accidents goes like this: such a remarkable coincidence shows that the tested theory is empirically very adequate, but it tells no more than that. But where we can intervene the process, there seems to be something more than empirical adequacy, namely causal relationship behind the phenomenon.
This is an argument for the reality of causal relationship (so somewhat supports realism about causal models), but it is not enough as an argument for the reality of theoretical entities (a causal relationship can be hold without any entity involved in it). I wonder if he can really bridge the gap.
3. Is astronomy a science?
Now we can turn to the question of the scientific status of astronomy. Does Hacking say that discipline in which "intervention" is impossible cannot deal with a real world and thus is not a science? First of all, Hacking says nothing about the demarcation between science and non-science in this book. Moreover, he admits the possibility of the existence of black hole and things like that (274-275). These points are enough to see Hacking has no reason to deny the scientific status of astronomy and other purely observational disciplines. But he has more positive argument for observation. He accepts Shapere's criteria for direct observation (183):
(1) information is received by an appropriate receptor
(2) information is transmitted directly to the receptor from the entity x.
Hacking adds another criterion (185):
(3) the theories upon which the observation relies are not intertwined with the facts about the subject matter under investigation.
If something satisfies these three criteria, it "counts as observing rather than inferring" (185).
It is not very clear if he proposes them as criteria for a scientific observation or criteria for the reality of observed entities. If he means the latter, then this sounds again a cosmic coincidence argument, and he needs some defense. Anyway, if astronomy uses this kind of observation, Hacking should be happy to call astronomy a science, even though he denied the scientific status of astronomy in a later paper.