Scientific Realism

Scientific realism is the philosophical position that the theoretical entities posited by successful scientific theories — electrons, quarks, genes, tectonic plates — exist independently of our theories about them, and that our best theories provide approximately true descriptions of these entities. Scientific realism is not a scientific claim but a philosophical one: it concerns the relationship between our representations and the mind-independent world they purport to describe. The position is contested by instrumentalism (theories are tools for prediction, not descriptions of reality), constructive empiricism (we can only be confident about claims concerning observable entities), and social constructivism (scientific knowledge is socially constructed rather than discovered). The realism debate is foundational in philosophy of science because it determines what we take to be the point of scientific inquiry: is science trying to describe reality, or to generate successful predictions?

The Miracle Argument

The strongest argument for scientific realism is Hilary Putnam's 'no miracles argument': if our best scientific theories were not even approximately true descriptions of the world, it would be a miracle that they are so successful at prediction and manipulation. The theory of electrons allows us to build semiconductors; the theory of DNA allows us to engineer organisms. This predictive and technological success is not explained if theories are merely useful fictions that happen to generate correct predictions — that would be a miraculous coincidence. The realist concludes: the success of science is best explained by the approximate truth of its theoretical claims about unobservable entities.

The force of this argument depends on what counts as explanation. The anti-realist response — associated with Bas van Fraassen's constructive empiricism — is that the success of science is explained by natural selection among theories: theories that are not empirically adequate are discarded; only those that generate successful predictions survive. This selection process does not require that the surviving theories are true. It requires only that they are empirically adequate — that they save the phenomena. Whether they are also true of the unobservable world is a further question that the selection process does not settle.

The Pessimistic Meta-Induction

The strongest argument against scientific realism is Larry Laudan's pessimistic meta-induction: the history of science is a graveyard of successful theories whose central theoretical terms failed to refer. Phlogiston was the theoretical entity that explained combustion — and phlogiston does not exist. The caloric fluid was the theoretical entity that explained heat conduction — and caloric fluid does not exist. The luminiferous ether was posited to carry electromagnetic waves — and the ether does not exist. Each of these theories was successful by the standards of their time; each posited theoretical entities that were later abandoned. If past successful theories failed to refer, by what principle do we conclude that current successful theories succeed in referring?

The standard realist response is selective realism: not all theoretical posits are equal. The entities that earn realist commitment are those that play an indispensable explanatory role across multiple independent theories and experimental contexts — those that, in Ian Hacking's formulation, we can manipulate to produce new phenomena. Electrons earned realist commitment not merely because electron theory predicted spectral lines but because we can spray them, exploit their spin, and build computers using their behavior. An entity we can use as a tool in multiple independently corroborated ways is not easily dismissed as a useful fiction.

The Inference to the Best Explanation

Scientific realism typically relies on inference to the best explanation (IBE) as its core epistemic principle: when a theory explains the available evidence better than any rival, we are entitled to believe it is approximately true. The principle licenses the move from 'this theory explains the evidence' to 'this theory is true.'

IBE faces a structural problem: the 'best' among available explanations may still be quite distant from the truth. The principle tells us to believe the best available hypothesis, but gives no account of whether the best available hypothesis is good enough. In a domain where all available hypotheses are inadequate, IBE directs us to believe the least inadequate one — which may be substantially false. This is not a refutation of IBE but a constraint on it: the principle is only as reliable as the quality of the hypotheses in the pool from which 'best' is selected.

The Rationalist conclusion is blunt: scientific realism is the correct default position, but the arguments for it are weaker than its defenders typically acknowledge. The miracle argument assumes that only truth explains success — which is not established. IBE assumes that the best available explanation is good enough — which is not guaranteed. The pessimistic meta-induction establishes that historical track record cannot support naive confidence in current theoretical entities. What scientific realism requires — and what philosophy of science has not yet provided — is an account of the specific conditions under which theoretical posits earn the confidence the realist extends to them. Until that account exists, realism is a philosophical disposition, not an argument.