Pruebas para alcanzar los objetivos de investigación

We have seen in the preceding section that different ways of measuring the success of science ultimately reduce to different views about the aim of science. The task of studying this aim belongs to the axiology of science.

Historical and sociological naturalism (cf. Section 1.4) suggests that axiological issues about science should be resolved by studying the actual behaviour and opinions of the scientists: the aim of science is what the so-called scientists really aim at. This proposal does not work unless we already have a criterion for picking out ‘the scientists’ from the world. Are medicine men, engineers, priests, artists, or astrologists scientists? Answers to such questions depend partly on the characteristic activities of these professions, but also on their aims. Any workable definition of science and scientist makes some reference to axiological concepts. Therefore, the axiology of science cannot be entirely ‘naturalized’, but it remains a genuine part of philosophy.

Larry Laudan's important article ‘Progress or Rationality? The Prospects for Normative Naturalism’ (1987a) helps a lot to clarify the debate about naturalism. Laudan suggests that methodological rules can be understood as conditional norms of the form

• (12)

•

Such statements, which express connections between ends and means, are true (or warranted) if y really promotes the goal x, i.e. if

• (13)

•

is true. As (13) is a contingent empirical statement, Laudan argues, it follows that the

‘naturalist meta-methodologist’ will rely—instead of on pre-analytic intuitions or choices of the scientific elite—on historical data concerning means–ends relationships.¹⁴

One may point out that the scientist's own ‘methodological norms’, like social norms in general, are unconditional (‘Avoid ad hoc explanations!’, ‘Make sure that your

experiments are repeatable!’). Laudan's point is that methodology, as a systematic study of such norms, should construe them as conditional rules (see Kaiser 1991).

Statements of the form (12) are called technical norms by G. H. von Wright (1963a). I have earlier suggested that technical norms define the typical form of knowledge that is sought in applied research (see (1) above). Laudan's thesis can thus be expressed by saying that methodology is an applied science.

While I agree with this thesis, it seems to me that conditional norms of the form (12) are not typically justified inductively by historical data, but rather by studying theoretical models of knowledge acquisition. Such models study the effectiveness of cognitive strategies relative to epistemic goals and factual assumptions about the world. Examples can be found from disciplines like applied mathematics, mathematical statistics, game theory, decision theory, and operations research:

• (14)

•

• (15)

•

• (16)

•

An important feature of conditionally normative results of this type is the possibility of proving them a priori by mathematical demonstration

end p.171

(for (15), see Section 4.6). At the same time, their application in some particular

situation—i.e. deriving from them non-conditional norms or recommendations—requires factual, hypothetical, or empirically warranted knowledge (about the antecedents of the conditionals and the situation of their application).

Some methodological rules may be based upon conceptual connections between precisely explicated notions, so that again they can be justified a priori. For example, a scientific realist may formulate methodological rules of type (12) by means of the following results (Niiniluoto 1987a) which are analytically true (given Hintikka's measure of corroboration and my measure of truthlikeness):

• (17)

•

• (18)

•

(Cf. Section 6.5 for a more precise account.)

We may thus conclude that Laudan's conception of methodology as consisting of conditional norms legitimizes, besides a role for historical data as empirical evidence, also the possibility of a formal philosophy of science, so that the approach of logical empiricism is partially rehabilitated (cf. Section 1.4; Niiniluoto 1991b).

Laudan's conception of methodology is limited to strategic rules which express means–

ends relationships. They are comparable to principles which state how one plays chess effectively—how to attack, defend, build strong positions, and eventually beat the opponent. However, an institutionalized activity also has constitutive rules (to use Searle's term) which characterize its legitimate ‘moves’. In the case of chess, such constitutive rules state how the different chessmen may be moved on the table. Violation of these rules does not lead to ineffective playing, but to not playing chess at all.

Many debates in the philosophy of science concern the constitutive rules of science—

rules about both the characteristic methods and aims of science. As such rules define what science is, they have a ‘conventional’ element, as Popper (1959) says. But a demarcation between science and non-science, or the explication of the concept of science, should also be ‘close’ to the accepted use of the terms ‘science’ and ‘scientific’

(cf. Kuhn 1983). The method of justifying constitutive rules cannot be purely logical or empirical, but consists in an attempt to reach a ‘reflective equilibrium’ (to use Rawls's term) between our normative demands for science and its actual practice (cf. Thagard 1988; Kaiser 1991).

end p.172

Laudan (1983) rejects the demarcation problem: creationism is for him bad science rather than non-science. Therefore, it can be understood that he does not formulate constitutive rules for science. But how could one propose strategic rules for any game without knowing its constitutive rules?

However, Laudan (1987a) makes the important addition that ‘methodology gets nowhere without axiology’. He acknowledges the ‘need to supplement methodology with an investigation into the legitimate or permissible ends of inquiry’. And in Science and Values (1984a) he has proposed a ‘reticulational model’ for showing how questions about scientific values can be resolved by appeal to (temporarily) shared theories and methods.

One might object to Laudan's model that it is too restricted: disputes about scientific values may refer or appeal not only to scientific practices, theories, and methods but also to philosophical principles from fields like logic, epistemology, aesthetics, and ethics (cf.

Niiniluoto 1987b). This would not be quite fair, however, since Laudan may include naturalized versions of such principles among ‘theories’ in his model. But still it seems problematic how a network of descriptive statements and conditional norms could ever give a positive justification for pursuing some value in science. In this respect, Laudan's assessment of values remains within the framework of instrumental means–ends

rationality (i.e. Max Weber's Zweckrationalitet), and resembles Giere (1988), whose

version of naturalism admits only instrumental rationality: goals are evaluated in terms of their accessibility relative to the available means, not by their having intrinsic value or being ‘reasonable’ (cf. Aarnio 1987; Putnam 1981; Siegel 1996).

Laudan's reticulational model thus proposes a negative way of eliminating utopian or unrealizable goals: ‘the rational adoption of a goal or an aim requires the prior

specification of grounds for belief that the goal state can possibly be achieved’ (Laudan 1984a: 51). In my view, it would be too strong to understand this as requiring that any rational goal can actually be reached. But it is a serious matter, if a goal cannot be approached. For example, epistemological arguments against the infallibility of factual knowledge have led philosophers to reject the traditional quest for complete certainty as a general value in science. But these arguments gain their strongest support from the

observation that in many (though not in all) situations even the gradual approach to certainty is excluded, since the relevant hypotheses contain counterfactual or idealizing assumptions and, therefore, are known to be false.

Thus, scientific values should be regarded as respectable if there are reasonable criteria for claiming that we have made progress in realizing them.

end p.173

(Similarly, it is a reasonable goal to be an excellent, and even perfect, piano player, even if this goal will not be reached.) For this reason, it seems to me, Laudan's (1984a)

argument that truth is a utopian aim for science is not convincing (cf. also Rorty 1998: 3).

Peirce's fallibilism (cf. Section 4.1) admits that, even if there are no infallible criteria for recognizing whether truth has been realized, it is probable that this goal has been realized in many particular cases. And even in cases where truth is at best the asymptotic limit of enquiry, the measures of verisimilitude help us to assert with a fallible empirical warrant that this goal has been approached. For example, given the problem described above in (15), it can be shown that the infinite sequence of point estimates converges with probability one to the true value of the unknown parameter (see Section 4.6).

The axiology of chess is dominated by a single supreme rule: the aim of the game is to beat the opponent. A secondary rule states that, if you have lost your chance of winning, you should try to save a tie. It is only when chess becomes an art that the style of playing becomes an end in itself: we try to win with a new, short, and beautiful combination of moves.

In my view, the axiology of science should likewise be governed by a primary rule: try to find the complete true answer to your cognitive problem, i.e. try to reach or approach this goal. Truthlikeness measures how close we come to this goal. As secondary rules, we may then require that our answer is justified, simple, consilient, etc. Further, if it is known that the available answers do not include a true one, then our rule is to search for the least false among them.