Reliability and the Science Debate — Epoche B2
Beyond Seeing: Reliability and the Science Debate The dispute between scientific realists and their opponents is usually staged as a dispute about what exists [1] . Do electrons exist, or are they a useful fiction? Put that way, the question invites two entrenched answers and no procedure for settling between them. This essay argues that the staging is wrong, and that a piece of ordinary epistemology shows why. Once the debate is redescribed in terms of the reliability of belief-forming methods, it turns out that both sides accept that scientific method is reliable, and that their disagreement is about how narrowly the method should be described. That is a better question than the ontological one, because unlike the ontological one it can be argued about with cases. The claim is not that reliabilism decides the debate in favour of either party. It is that reliabilism identifies what the parties are actually disagreeing about, and that the identification is not neutral in its consequences: it puts a specific burden on each side. The two positions Begin with the standard opposition, since a reader should be able to state both before either is criticised. The realist holds that our successful theories are approximately true, including in what they say about things too small or too distant to observe. The argument, given its canonical phrasing by Hilary Putnam in Mathematics [2] , Matter and Method (1975), is that realism is the only philosophy that does not make the success of science a miracle. Its form is an inference to the best explanation: theories predict what nobody expected and the predictions hold; approximate truth explains this; nothing else does; so accept approximate truth. Bas van Fraassen's constructive empiricism [3] , set out in The Scientific Image (1980), accepts everything the realist says about the content of theories and almost nothing about our attitude to them. Theories are to be read literally: when a theory says there are electrons, it says there are electrons, and it is true only if there are. But the aim of science, on this view, is empirical adequacy, not truth — and accepting a theory involves believing only that it is empirically adequate, that is, that what it says about observable things is true. About the rest one may remain agnostic. This is not instrumentalism; van Fraassen does not say that talk of electrons is meaningless or that it is disguised talk about instrument readings. He says it may well be true and that belief is not required. He also has a direct answer to the no-miracles argument, and it is the part most often left out: I claim that the success of current scientific theories is no miracle. It is not even surprising to the scientific (Darwinist) mind. For any scientific theory is born into a life of fierce competition, a jungle red in tooth and claw. Only the successful theories survive — the ones which in fact latched on to actual regularities in nature. Success is explained by selection. Empirically inadequate theories are eliminated, so the survivors are empirically adequate, and no truth about unobservables is needed to account for that. Reliabilism Epistemology asks when a belief is justified. Reliabilism , developed by Alvin Goldman in 'What Is Justified Belief [4] ?' (1979), answers that a belief is justified when it is produced by a reliable process — a process that tends, across the range of cases in which it operates, to yield true beliefs rather than false ones. Vision in good light is reliable; wishful thinking is not. The theory is externalist: whether a process is reliable is a fact about the world, and a believer can be justified without being able to demonstrate that they are. Reliabilism has a well-known internal difficulty, and it is the difficulty that matters here. Every belief is produced by a process that can be described at many levels of generality. A belief formed on seeing a bird at dusk is produced by 'perception', by 'vision', by 'vision in poor light', and by 'identifying a small bird in poor light at forty metres'. These have very different reliability scores. Nothing in the theory says which description is the right one. This is the generality problem, pressed under that name by Earl Conee and Richard Feldman in a 1998 paper in Philosophical Studies [5] , which argues that no principled answer has been given. The redescription Now redescribe the debate. Instead of asking whether electrons exist, ask which belief-forming processes scientists employ and how reliable each is. On this description the parties agree about a great deal. Both accept that the process running from experiment and observation to claims about observable regularities is reliable; van Fraassen's Darwinian explanation is itself an argument that it is, since selection weeds out the empirically inadequate. The disagreement concerns one process only: inference to the best explanation, understood as the step from 'this hypothesis about unobservables would explain the data better than any rival considered' to 'this hypothesis is true'. And now the generality problem is not an embarrassment but a diagnostic. The realist and the empiricist are offering competing individuations of the same episode of reasoning. The realist describes the process broadly: explanatory inference in mature science . Described that way, it has a good record — it delivered atoms, genes, and the neutrino, each initially unobservable and each subsequently established by independent means. A process with that record is reliable, and beliefs formed by it are justified. The empiricist describes it narrowly, splitting it in two: explanatory inference to claims about observables , which has a good record, and explanatory inference to claims about unobservables , whose record cannot be assessed, because assessing it would require independent access to the unobservables and there is none. On this individuation, the second process has no established reliability, and belief formed by it