Scientific Realism and the Stratagema de Divide et Impera

In response to historical challenges, advocates of a sophisticated variant of scientific realism emphasize that theoretical systems can be divided into numerous constituents. Setting aside any epistemic commitment to the systems themselves, they maintain that we can justifiably believe those specific constituents that are deployed in key successful predictions. Stathis Psillos articulates an explicit criterion for discerning exactly which theoretical constituents qualify. I critique Psillos’s criterion in detail. I then test the more general deployment realist intuition against a set of well-known historical cases, whose significance has, I contend, been overlooked. I conclude that this sophisticated form of realism remains threatened by the historical argument that prompted it. 1 A criterion for scientific realism 2 Assessing the criterion 3 A return to the crucial insight: responsibility 4 A few case studies 5 Assessing deployment realism 1 A criterion for scientific realism Scientific realists offer a hypothesis about scientific theories, which, they propose, should itself be treated as a scientific theory. In its basic formulation, the hypothesis is that our successful scientific theories are (approximately) true. In this hypothesis (approximate) truth is being attributed not only to a theory’s claims about observables, but also, and quite crucially for the scientific realist, to the theory’s claims about unobservables. Readily granting that we cannot conclusively establish the hypothesis, the realist claims that we are at least justified in believing it, given what is commonly called ‘the no-miracles argument’: It would be a miracle were our successful theories not (at least, approximately) true; the (approximate) truth of T provides the only (or at least the best) explanation for T’s success. However, testing the realist hypothesis against the historical record, it looks as though numerous theories have been successful but cannot, as wholes, be approximately true (Laudan, The Author (2006). Published by Oxford University Press on behalf of British Society for the Philosophy of Science. All rights reserved. doi:10.1093/bjps/axl021 For Permissions, please email: [email protected] Advance Access published on August 17, 2006 [1981]). The basic realist hypothesis appears significantly falsified. And, unable to appeal to approximate truth to explain these successes, the realist appears to be faced with a list of inexplicable successes—or, taking her argument literally, ‘miracles’. In defense, realists have modified their hypothesis. Here, I focus on the recent defense against this historical argument that appears to be most strongly favored by scientific realists. The realist notes that theoretical systems can be divided into numerous constituents. She also notes that, while certain past successful systems, taken as wholes, may fail to be approximately true (by present lights), certain postulates contained within them have nonetheless been retained. Accordingly, the realist revises her hypothesis to focus on particular theoretical constituents. Of course, to be a realist about any actual constituents, one must specify particular conditions for identifying those constituents that are deserving of belief. Philip Kitcher ([1993]) and Stathis Psillos ([1999]) contend that we can justifiably believe those, and only those, constituents that are deserving of credit for the significant successes of the theory. Kitcher says, we must ‘distinguish between those parts of theory that are genuinely used in the success and those that are idle wheels’ (p. 143, footnote 22). Proposing that we ‘really focus on the specific successes of certain theories’ (p. 109), Psillos wants to identify those constituents that are ‘responsible’ (p. 108) for a given successful prediction, ‘those which ‘‘really fuel the derivation’’’ (p. 110). He characterizes this step as the ‘divide et impera move’ (p. 108). ‘Realists’, he contends, ‘need care only about those constituents which contribute to successes and which can, therefore, be used to account for successes’ (p. 110). The no-miracles argument then takes the following form: a theory’s success would be miraculous if those constituents that had a genuine bearing on its successful predictions were not (at least approximately) true. That given, says the realist, we are justified in believing the hypothesis that those constituents are (approximately) true. And if, say, the proposition ‘the ether exists’ (Kitcher [1993]) was not ‘deployed in’—that is, if it were not among the constituents responsible for—successful predictions, it is not deserving of credit for those predictions. And despite its falsity, it would not stand as a counterexample (or an apparent ‘miracle’). The historical argument against realism is thought to be deflected. Taking a cue from Kitcher, we can call this sophisticated position deployment realism. Other philosophers who advocate this position are Jarrett Leplin ([1997]), Ilkka Niiniluoto ([1999]), and Howard Sankey ([2001]). However, no one of these philosophers has articulated and attempted to apply the deployment realist’s intuition to the extent that Psillos has ([1999]). For this reason it is crucial to address his articulation (references to his [1999]). Psillos proposes that our criterion for deployment realism be the ‘essentiality’ of a constituent in bringing about a given successful prediction. ‘Theoretical constituents 538 Timothy D. Lyons