Introduction. Collingwood (1966, p. 11) differentiates between metaphysics as the science of pure being and metaphysics as the science, which deals with the presuppositions underlying ordinary science. My use of metaphysical will be in conformity with the latter throughout this paper. But it also includesin contradistinction to Collingwoodontological presuppositions. In the logical positivist tradition of philosophy of science there were no place for metaphysicsor so it was claimedin either sense of the term Collingwood has distinguished. On the other hand, in the post-Kuhnian philosophy of science it has been claimed that science is not free of metaphysical presuppositions. This has led to three different strands. (i) Science with metaphysical presuppositions cannot be acceptable, since it makes science irrational. (ii) Science is actually irrational, and thus cannot be differentiated from other kinds of irrational activities in this respect. (iii) Science does have metaphysical presuppositions, but this does not make science irrational. In this paper I shall defend the third position, surprisingly enough, relying upon Carnaps distinction between external and internal questions. It may sound surprising because, as everybody knows, Carnap was one of the leading figures of logical positivism. There is a growing literature, however, that documents that Carnap anticipated many ideas of Kuhn, and Kuhn did not try to show that science is an irrational activity, as many supposed, and he is quite close to Carnap as far as rationality of science is concerned. (See, for example, Irzik and Grnberg 1995.) In what follows, I shall first consider a rereading of Carnap that will reveal the metaphysical involvement of science, and, second, a case study which is an attempt of showing the rationality of a method of confirmation as an answer to an external question in Carnaps sense, and, as such, when properly interpreted, involves metaphysical presuppositions.
Carnaps distinction of external and internal questions. Metaphysical presuppositions of science are typically ontological. For example, There are sense-data (such as patches of color), There are physical entities (such as tables), There are theoretical entities (such as electrons), and There are abstract entities (such as numbers). Carnap considers such sentences as expressing only pseudo-propositions when taken literally. However, they become meaningful when reformulated as the expression of a decision to use a particular linguistic framework. For example, to say that physical objects exist is tantamount to accepting the so-called thing-language (i.e., a language in which the bound variables range over physical things). Now, acceptance of a linguistic framework for science involves the use of formation and transformation rules, semantical rules, and methodological rules of confirmation, such as the rules of a system of inductive logic. Carnap (1956, p. 207) calls external questions, questions concerning the choice of a linguistic framework among alternatives. And since the result is based upon a decision, and not on verification of a statement or a system of statements, these questions are 2 practical. 1 Once a particular linguistic framework concerning a particular category of entities is accepted, the statement that there are entities of the given category becomes analytically true relative to the accepted framework. For example, the statement that there are physical objects is analytic relative to the thing language. On the other hand, There are black swans in Australia is a synthetic existential statement. Questions concerning the truth-value of statements formulated in a given linguistic framework are internal questions. (Carnap 1956, p. 207.) The two examples we mentioned above are answers to internal questions. Internal questions are theoretical in the sense that their answers are propositions having truth-value. Note that the analyticity of the statement that there are physical entities does not commit us to the theoretical acceptance of reality of physical things, but only to the practical acceptance of the thing- language, which is a matter of decision. But, such a decision is not arbitrary. It can be evaluated by criteria of rationality, which are permanent and ultimate. Carnap gives as examples of such criteria, efficiency, fruitfulness, and simplicity. Kuhn (1977, pp. 320-329) also endorses similar ones, adding accuracy, consistency, and scope to the list. The rationality of a decision of choosing a linguistic framework for the realization of a purpose consists in the conduciveness to the successful realization of that purpose. The question whether a linguistic framework realizes its purpose is itself a theoretical, or rather meta-theoretical, question in the sense that the realization of a purpose is an objective fact. Science might have diverse purposes, ranging from most practical to most theoretical ones. Of course, the most desirable purpose of science is truth seeking. Questions of truth and falsity, however, cannot arise at this stage, because scientific hypotheses can be formulated and interpreted only by means of the constitutive rules of a given linguistic framework. This is quite the same with Collingwood. For him what he calls absolute presuppositions in a given epoch, like Every event has a cause, are not propositions proper, and thus they do not have truth-value. Truth or falsity of scientific statements is only relative to the accepted metaphysical presuppositions, quite in the same way as an answer to an internal question is based upon an answer given to an external question in Carnap. However, Collingwood differs both from Carnap and Kuhn in that he does not have a rational explanation regarding how a metaphysical presupposition (or a linguistic framework) is chosen among alternatives. He just takes them as historical contingencies Thus, turning back to the choice of a linguistic framework (or metaphysical presupposition in Collingwoods sense), the criteria of rationality are the ultimate norms for evaluating the proper ontological presupposition corresponding to a linguistic framework. In this sense we can consider them as meta-philosophical presuppositions by means of which we choose a linguistic framework among the alternatives. They can be formulated as normative rules
1 Carnap (1963, pp. 941-942), however, was inclined to accept Bar-Hillels proposal to interpret (rather than replace) a metaphysical proposition asserting the existence of a category of objects as expressing the expediency of the linguistic framework referring to the objects in question. Thus, metaphysical presuppositions, reinterpreted in accordance with Bar-Hillels proposal, may become genuine theoretical theses for Carnap. 3 like One ought to choose the most efficient linguistic framework or One ought to choose the most simple linguistic framework and the like. Now, if the aim of science is efficiency then one cannot further question why one ought to choose the most efficient framework, simply because this is the very aim science is trying to achieve at. (You cannot ask why you want to realize your purpose!) And it is in this sense why these criteria of rationality are ultimate, i.e., they do not need any further justification and, as such, they play the role of absolute presuppositions. On the other hand, the thesis that these criteria are permanent is justifiable in a historico-empirical way. Throughout the history of science, scientists tried to discover the most efficient, simple etc. linguistic framework, or scientific theory, so that the aims of science have never changed, and there is no reason to assume that they will change in the future. Note that Collingwoods absolute presuppositions are ultimate and permanent only in a given epoch, i.e., they are historical and, as such, strictly speaking, they are neither permanent nor ultimate, and therefore cannot properly be called absolute. I shall turn now to the case study.
A case study: Rationality of the method of confirmation constitutive of the linguistic framework of a scientific theory. A method of confirmation for Carnap consists in a particular credibility function (or class of credibility functions), i.e., a function assigning to scientific hypotheses formulated in a linguistic framework, a degree of confirmation. The credibility function (or class of credibility functions) constitutes a system of inductive logic, which is determined by a set of axioms. Therefore, the choice of a linguistic framework requires a choice of a credibility function among alternatives. Carnap contends that this choice is based upon a priori intuitive judgments. I shall defend the view that these a priori intuitions are nothing but the result of Carnap and Kuhns criteria of rationality (such as efficiency, fruitfulness and the like) for the choice of linguistic frameworks. Carnap (1963, p. 978) writes, The reasons [for accepting any axiom of inductive logic] are based upon our intuitive judgments concerning inductive validity, i.e., concerning inductive rationality of practical decisions (e.g., about bets). Now, the rationality of practical decisions requires the expectation of the successful realization of the purpose of the decided action. The efficiency of the action performed for realizing such a purpose consists in the quickness and the easiness of that realization at least cost. The basic axioms of inductive logic, namely, Kolmogorovs axioms of probability calculus, have been deduced from the fairness of the betting systems. (See Kemeny 1963, p. 720, n. 13.) This means that in case a betting system is fair, the betting ratios satisfy Kolmogorovs axioms. And it is also proved that the fairness of the betting system is deducible from the Kolmogorovs axioms, namely, if the betting ratios characterizing a betting system satisfy Kolmogorovs axioms, the betting system is fair. (See ibid., p. 720, n. 13.) A betting system is fair just in case it is not possible that the bettor always looses. It is strictly fair just in case it is fair, and, furthermore, the bettor can sometimes win. (See ibid., 719-720.) An agent who decides to enter a betting system performs an action whose purpose is gaining. Therefore, if an agent looses without gaining at all she would not realize her purpose, and a fortiori her acceptance of a bet would not be efficient. It follows that both fairness and strict 4 fairness are necessary conditions of the efficiency of the action of betting. In general the rationality of scientific activity results from the rationality of the decision to use a suitable linguistic framework involving, in particular, a method of confirmation. The adoption of a particular set of axioms of inductive logic constitutes a particular method of confirmation. Let us remember that the choice of a linguistic framework is an answer to an external question. Internal questions refer to the activity of the scientists within a particular linguistic framework, which has already been chosen. For Carnap internal questions refer to the confirmation of hypotheses formulated within the framework. More precisely, an internal question is of the form What is the degree of confirmation of scientific hypothesis h given evidence e? The degree of confirmation is computed on the basis of the credibility function characterizing the linguistic framework. Hence, a hypothesis cannot be confirmed and, in general, an internal question cannot be answered independently of a linguistic framework.
Conclusion. We see that science starts with ultimate and permanent criteria of rationality that can be formulated as universal metaphysical norms of purposeful actions. They express a priori intuitions and that is why we may call them metaphysical. For each particular purpose these norms will serve to adopt a particular linguistic framework. The rationality of the answers to internal questions within a linguistic framework results from the rationality of the choice of the framework, hence of the rationality of the answer to an external question. For example, the rationality of the assignment of a degree of confirmation to a hypothesis formulated in a given linguistic framework derives from the rationality of the choice of the axioms of inductive logic for that framework.
David Grnberg Philosophy Department, Middle East Technical University david@metu.edu.tr
References
Carnap, R. (1956), Empiricism, Semantics, and Ontology, in Meaning and Necessity (2 nd ed. enlarged), Chicago: The University of Chicago Press. Carnap, R. (1963), Replies and Systematic Expositions, in P. A. Schilpp (ed.), The Philosophy of Rudolf Carnap, La Salle: Open Court. Collingwood, R. G. (1966), An Essay on Metaphysics (first edition, 1940). Oxford: The Clarendon Press. Irzik, G. and Grnberg, T. (1995), Carnap and Kuhn: Arch Enemies or Close Allies?, British Journal of the Philosophy of Science, 46, pp. 285-307. Kemeny, J. G. (1963), Carnaps Theory of Probability and Induction, in P. A. Schilpp (ed.), The Philosophy of Rudolf Carnap, La Salle: Open Court. Kuhn, T. S. (1977), The Essential Tension. Chicago: The University of Chicago Press.