Thinking about the ultimate argument for realism

Abstract

Alan Musgrave has been one of the most passionate defenders of scientific realism.1 Most of his papers in this area are, by now, classics. The title of my paper alludes to Musgraves piece The Ultimate Argument for Realism, though the expression is Bas van Fraassens (1980, p. 39), and the argument is Hilary Putnams (1975, p. 73): realism is the only philosophy of science that does not make the success of science a miracle. Hence, the code-name no-miracles argument (henceforth, NMA). In fact, NMA has quite a history and a variety of formulations. I have documented all this in my (1999, chapter 4). But, no matter how exactly the argument is formulated, its thrust is that the success of scientific theories lends credence to the following two theses: a) that scientific theories should be interpreted realistically and b) that, so interpreted, these theories are approximately true. The original authors of the argument, however, did not put an extra stress on novel predictions, which, as Musgrave (1988) makes plain, is the litmus test for the ability of any approach to science to explain the success of science. Here is why reference to novel predictions is crucial. Realistically understood, theories entail too many novel claims, most of them about unobservables (e.g., that there are electrons, that light bends near massive bodies, etc.). It is no surprise that some of the novel theoretical facts a theory predicts may give rise to novel observable phenomena, or may reveal hitherto unforeseen connections between known phenomena. Indeed, it would be surprising if the causal powers of the entities posited by scientific theories were exhausted in the generation of the already known empirical phenomena that led to the introduction of the theory. So, on a realist understanding of theories, novel predictions and genuine empirical success is to be expected (given of course that the world co-operates). The aim of this paper is to rebut two major criticisms of NMA. The first comes from Musgrave (1988). The second comes from Colin Howson (2000). Interestingly enough, these criticisms are the mirror image of each other. Yet, they both point to the conclusion that NMA is fallacious. Musgraves misgiving against NMA is that if it is seen as an inference to the best explanation, it is deductively fallacious. Being a deductivist, he tries to correct it by turning it into a valid deductive argument. Howsons misgiving against NMA is that if it is seen as an inference to the best explanation, it is inductively fallacious. Being a subjective Bayesian, he tries to correct it by turning it into a sound subjective Bayesian argument. I will argue that both criticisms are unwarranted. Actually, I would have no problem with Musgraves version of NMA if deductivism were correct. But, as I will try to argue, the deductivist stance is both descriptively and normatively wrong. To avoid a possible misunderstanding, let me note that I have no problem with deductive logic (how could I?). My problem is with deductivism, that is the view that, as Musgrave (1999a, p. 395) puts it, the only valid arguments are deductively valid arguments, and that deductive logic is the only logic that we have or need. One could cite Bayesianism as a live example of why deductivism is wrong. But, I think, there are important problems with Bayesianism too.2 Put in a nutshell, the Bayesian critique of NMA is that it commits the base-rate fallacy. Howson tries to rectify this by arguing that a sounder version of NMA should rely explicitly on subjective prior probabilities. Against the Bayesian critique of NMA I will primarily argue that we should resist the temptation to cast the nomiracles argument in a subjective Bayesian form. However, I will also explore the possibility of accepting a more objective account of prior probabilities, if one is bent on casting NMA in a Bayesian form. Here is a brief summary of the menu. Section 2 defines scientific realism and investigates Musgraves own understanding of it. Section 3 explains, rather briefly, what I take the form and the aim of the no-miracles argument to be. Section 4 criticises Musgraves deductivism and his attempt to show that NMA is best understood as a deductive enthymeme. Section 5 explains how NMA (as an inductive argument) is supposed to commit the base-rate fallacy. Section 6 argues that there are ways to give a more objective account of the prior probabilities that are supposed to be necessary for NMA to be inductively sound. Section 7 explores some features of the base-rate fallacy and explains why it is reasonable to ignore the baserates (lets say the prior probabilities, though they are not the same) on certain occasions. Section 8 argues that if we look at case histories we can have strong reasons to be realists about several theories. Section 9 explores two ways to think of NMA that do not involve prior probabilities. © 2006 Springer.