Formal methods in the philosophy of science (Q941743)

Traditionally, the merit of first-order logical formalizations of philosophical problems -- in particular, in the philosophy of science -- has been regarded as twofold: First, logical formalization forces philosophical investigators to define all concepts precisely, thereby sometimes discovering that a concept can be ``constructed'' from more primitive notions. Second, the insistence on formal logical derivation requires philosophical investigators to state all assumptions explicitly. In the first half of the twentieth century, great hopes were attached to this methodology. As the authors put it, the general expectation among analytic philosophers was that ``by using logical methods all philosophical problems can be solved or dissolved'' (p.~151). Whilst first-order logical formalization is still leading to remarkable findings in the philosophy of science (cf., e.g., \textit{H.~Andréka, J. X. Madarász, I. Németi} and \textit{G. Székely} [Stud. Log. 89, No. 2, 163--186 (2008; Zbl 1148.03028)]), the authors argue in this article that it would be naïve to assume that ``logic is the right framework for approaching every problem in the philosophy of science'' (p.~157). Rather, a host of other formal methods has been employed successfully in recent years in many branches of the philosophy of science. The authors mention the application of probability theory in the analysis of causation [cf. \textit{E.~Eells}, Probabilistic causality. Cambridge: Cambridge University Press (1991; Zbl 0785.60003); \textit{J. Pearl}, Causality. Models, reasoning, and inference. Cambridge: Cambridge University Press (2000; Zbl 0959.68116)], the use of game theory to study philosophical questions regarding common knowledge [cf. \textit{D.~Lewis}, Convention. Cambridge, MA: Harvard University Press (1969)], and the utilisation of graph theory in the theory of causality and events [cf. \textit{L.~Horsten}, ``Impredicative identity criteria'', Philos. Phenomenolog. Res. (to appear)]. In isolated cases, even unconventional mathematical techniques such as nonstandard analysis have been applied to problems in the philosophy of science [cf. a recent article by the reviewer on D.~Lewis' ``zero-fit problem'' in Br. J. Philos. Sci. 58, No. 1, 25--43 (2007; Zbl 1155.00304)]. Moreover, the emerging theory of judgment aggregation [cf., e.g., \textit{F.~Dietrich} and \textit{C.~List}, Soc. Choice Welfare 29, No. 1, 19--33 (2007; Zbl 1138.91019)], which has some overlap with the philosophy of science, draws upon both logical and probabilistic methods. It is argued that whenever one of these formal methods is sufficiently widely used, it can be viewed as a paradigm in the Kuhnian sense. Adopting this terminology, the philosophy of science has seen a Kuhnian paradigm shift in recent decades: First-order logic was abandoned as the exclusive means of formalization, and other formal tools, most notably probabilistic methods and concepts, were embraced. (As an example, the authors cite the Bayesian theory of confirmation, which ended a decades-long history of many unsuccessful attempts -- based solely on logical methods -- at analyzing scientific confirmation.) Thus, as far as the Kuhnian insight into revolutionary scientific progress is concerned, philosophy of science is no different from other sciences.