Author(s): Jiang Tianji Reviewed work(s): Source: The British Journal for the Philosophy of Science, Vol. 36, No. 4 (Dec., 1985), pp. 409- 423 Published by: Oxford University Press on behalf of The British Society for the Philosophy of Science Stable URL: http://www.jstor.org/stable/687400 . Accessed: 09/03/2012 11:55 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Oxford University Press and The British Society for the Philosophy of Science are collaborating with JSTOR to digitize, preserve and extend access to The British Journal for the Philosophy of Science. http://www.jstor.org Brit. J. Phil. Sci. 36 (1985), 409-423 Printed in Great Britain 409 Scientific Rationality, Formal or Informal?* by JIANG TIANJI x Introduction 2 Standard Conception of Scientific Rationality 3 Vindication of Kuhn's Conception of Informal Rationality 4 Other Models of Informal Rationality 5 Conclusion I INTRODUCTION About scientific rationality, which is the central problem, if not be identified with the whole field, of the philosophy of science, there are now two extreme positions, with quite a few intermediate ones in between. The standard conception of scientific rationality, as represented by logical positivism, critical rationalism and related views, claims that scientific rationality does not change over time, and it is implicit in all genuine science through different periods of its development. It is thus the aim of inductive logicians and normative methodologists to make explicit as fully as possible that permanent rationality of science. On the other hand, the revolutionaries of the sixties, as represented by Polanyi, Kuhn, Feyerabend et al., claim that standards of rationality change with the change of scientific beliefs, and there is no superstandard or supermethod that stands above scientific change and is immune to revision, that ensures continuity within science and demarcates genuine science from non-science and pseudo-science. Their heretical conceptions have been stigmatised as irrationalistic by the most orthodox of the other camp. In one respect the formal rationality of the standard conception is in striking contrast to the informal rationality of the heretical conception; in another respect the methodological monism of the former contrasts sharply with the methodological pluralism of the latter. In this paper I shall focus my discussion on the first topic, while lightly touching upon the second one. Section 2 will be devoted to a characteris- Received 2 December 1983 * This paper was first delivered as a lecture for the Course Philosophy of Science at the Inter- University Centre of Postgraduate Studies, Dubrovnik, Yugoslavia, and again read at the Institute of Philosophy, the University of Beograd, April 1983. I would like to thank all those who made helpful comments on both occasions. I am especially indebted to Bill Newton- Smith for help and encouragement. I am also grateful to a referce for helpful comments and suggestions. 410 Jiang Tianji ation of the standard conception of scientific rationality. Kuhn's heretical conception will be vindicated against charges of irrationalism in section 3, and other models of informal rationality will be briefly discussed in section 4. 2 STANDARD CONCEPTION OF SCIENTIFIC RATIONALITY Logical positivists and Popperians have believed that science proceeds by following a distinctive method, the scientific method, and rationality is nothing other than acting in conformity with the norms of this method. They have aspired after an explicit formalisation of rationality. In their opinion, rationality is identified with commitment to this formalisable method. From Mill to Carnap and Hempel, influential philosophers of science have continued to believe that some thing like a formal method ('inductive logic') underlies empirical science and that continued work might result in an explicit statement of this method, a formalisation of inductive logic comparable to the formalisation of deductive logic which was achieved by Frege and others. If such a method were discovered, then scientific rationality would be identified with the possession and employment of this method. This influential view was voiced by I. Scheffler as follows: Underlying historical changes of theory, there is ... a constancy of logic and method, which unifies each scientific age with that which preceded it and with that which is yet to follow. Such constancy comprises not merely the canons of formal deduction, but also those criteria by which hypotheses are confronted with the test of experience and subjected to comparative evaluation (Scheffler [19671, PP. 9-1o). Proponents of this view admit that 'we do not . .. have explicit and general formulations of such criteria at the present time', but they insist that criteria 'are embodied clearly enough in scientific practice to enable communication and agreement in a wide variety of specific cases (and such) communication and consensus indicate that there is a codifiable methodology underlying the scientific enterprise' (ibid.). It was assumed that there are universal principles of theory-acceptance implicit in scientific practice at least since the time modern science began; in other words, there is a rationality implicit in the work of natural scientists since about the time of Newton, this explains why this work is so successful. It is possible that these criteria of rationality are not well-understood at a given time and incorrectly applied on occasion. But it was further assumed that by i800 or 1850, this rationality was fully operative in the work of major natural scientists, so that we could speak of 'the' scientific method. Thus for logical positivists, it was assumed that scientific rationality takes the form of a logic. It is, therefore, formal. It is given once and for all; it never develops or changes. This was not to say that all scientists employ it Scientific Rationality, Formal or Informal? 41I equally well. But the assumption was that there is a norm of rational procedure which all scientists should endeavour to comply with and all philosophers should endeavour to disclose. Logical positivists have at- tempted at an explicit general formulation of the principles of rational acceptance. Take Hempel's logic of confirmation, for instance. In his 'Studies in the Logic of Confirmation', Hempel is engaged in a search for 'general objective criteria determining whether . .. a hypothesis H may be said to be corroborated by a given body of evidence E' (Hempel [19651, p. 6). His aim in the essay is to provide definitions of the concepts of confirmation and disconfirmation which characterise those relations in a purely formal way: the criteria of confirmation 'should contain no reference to the specific subject matter of the hypothesis' (ibid. p. Io). For Hempel, one ought to be able to tell simply by looking at the logical forms of a hypothesis-sentence and an evidence-sentence whether the confirmation relation holds between them, just as one can tell simply by inspecting the logical forms of premise- sentences and the conclusion-sentence whether the implication relation holds between them. The search, then, is for formal syntactic criteria of confirmation analogous to the formal criteria for the validity of deductive arguments. What would count as evidence for an hypothesis is determined by the form of hypothesis-sentences and evidence-sentences, not by their content. The situation seems from an epistemological point of view ideal: the justification of hypotheses becomes a very straightforward matter. But, alas, reality has a way of eluding the ideal, for actual evidential relations in science are not captured by such analysis of confirmation. The analysis Hempel provides is of a formal, syntactic relation between sentences. This relation holds only between sentences containing the same predicates. But most interesting scientific theories are not evidentially supported by statements containing the same terms as occur in the theory, but by statements containing quite different kinds of terms. Therefore the Hempelian approach is applicable only to a restricted set of evidential relations, the relation between observation statements and theories which are merely generalisations of observation reports. Generally speaking, the logical positivist account of theory-acceptance is oversimplified. In his inductive logic Carnap treats all logical consequences as having the same evidential relationships to the hypothesis they support: no distinction is drawn between those known in advance and the 'novel' ones. It would be very difficult for him to incorporate such a distinction into his scheme, because there is no way to know by a simple inspection of an observation statement whether it would count as 'novel' or not. Only the history of the theory in question can tell one that. But the positivists have adopted a very abstract model of rationality in which only the formal structure of a theory and the empirical evidences drawn from it are relevant to theory evaluation. They paid no attention to the prior historical development of a theory and the different degrees of support conferred by 412 Jiang Tianji different sorts of evidence upon it. According to their intuitive principles based on an a priori notion of evidence, all accepted consequences lend the same degree of support to the theory or the same degree of disconfirmation if they are refuted. What they emphasised is the mere fact that theory and evidences are deductively related. They can easily set up a postulational system of plausible principles of evidence, the trouble is that this does not apply to what actually goes on in science. It is now generally agreed that, in actual scientific practice, what counts most of all in favour of a theory is not just its success in prediction but its 'dynamical character' or its 'fertility', its ability to meet anomaly in a creative and fruitful way. This is something which manifests itself only gradually over the course of time; one cannot attest to it until the theory has survived many tests and been extended in illuminatingly new ways. Time plays an essential role here. Whether such actual evidential relations can be formalised or not, we have yet to see. Popper and Popperians have attempted to formulate a more complex model of rationality than that of logical positivists. But with them they shared the assumption that there are norms of the scientific method, an unchanging set of intuitive principles underlying validation in science. According to Popper, a scientist ought to act in accordance with the dictates of the method of falsificationism: propose testable theories; not investigate ad hoc theories; accept that theory for further theoretical and practical purposes, which is best corroborated; select a simpler hypothesis because it is more testable than a less simple one; and so on. Popper has claimed that his method will give us a better chance to achieve the goal of epistemic progress. In this model of rationality, three criteria are offered for rational theory- choice: first, the criterion of prior appraisal; second, the criterion of posterior appraisal, and third, the criterion of appraisal for theories which are known to be false. The notions of empirical content and explanatory power, etc., are defined in purely logical terms by Popper. In his explication of empirical content (or falsifiability), one important problem concerns the measurement or com- parison of contents. There are objections that content comparison is impossible because contents of theories are not subsets of each other. But Watkins still attempts to show that at least the idea of comparing empirical content-a pre-testing concept-can be rescued. As to theory testing, Popper thinks that only a novel fact can really support a theory. Thus he introduces the ideas of background knowledge and severity of a test to account for the distinction between 'evidence' known in advance and verified prediction of a novel fact, a distinction which cannot be made in Carnap's theory of confirmation. The logical relation of deducibility cannot of itself convey this difference; a temporal analysis of some kind is needed. But success in novel prediction is not fertility. Fertility is not a matter of prediction at all. He concedes the importance of time order in theory appraisal but never quite comes to viewing fertility as a criterion of appraisal. Scientific Rationality, Formal or Informal? 413 Finally, Popper's formal concept of verisimilitude will allow two false theories to be compared with respect to their truthlikeness, the theory with more truth-content or less falsity-content is to be preferred. However this concept of verisimilitude has come under criticism from D. Miller and P. Tichy, and it seems that a precise definition of verisimilitude couched in a formalised language will never be possible, because truth-contents and false-contents are not comparable. Therefore, both the problem of pre- testing measurement of content and the problem of post-testing compara- tive verisimilitude have met with severe technical difficulties, and the concept of degree of corroboration will not be very useful if all theories have been refuted or are bound to be refuted. The model of Popper's thus has its own unsolvable difficulties, not to speak of its inapplicability to theory- choice in actual scientific practice. Now we come to Lakatos's improvement of the Popperian methodology. Lakatos took one step further in insisting that the unit for appraisal is not the theory taken at a moment of time but the theory considered in terms of its entire historical career, what he called a 'research programme'. He had two main criteria for the appraisal of scientific research programmes. One is increase in empirical content, at least some of which has been experimentally confirmed. The other is the idea that a research programme is better than a competitor if it has more heuristic power, potential for creating 'evidential support', i.e., if it makes possible the construction of a sequence of theories with increasing 'evidential support'. Roughly, a research programme is fruitful if it can lead to a series of theories which produce novel knowledge. It appears that heuristic power is scarcely more than the capacity to generate increase in successful prediction. Urbach has attempted to show that the heuristic power of a research programme can be objectively determined. For Lakatos, a research programme is better than its rival if it has more 'evidential support' and more heuristic power, i.e., power to create theories with 'evidential support'. Thus everything hinges, again, upon the idea of 'evidential support'. But prediction failures are not relevant to the appraisal of research programmes. A major problem for Lakatosians is how to rule out ad hoc evidential support. Worrall suggests that those facts which have been used in the construction of the theory in question cannot 'support' the theory. Worrall maintains that it can be objectively determined whether a fact was used in constructing a theory. If a fact has been used once in a theory, then the same fact can only be explained ad hoc by that theory-a fact cannot be used twice. Lakatos comes near to the criterion of fertility with his emphasis on the historical development of a programme, but not quite; it seems what is sought in a 'good' development is mere increase in successful prediction. Of course Lakatos put emphasis on the importance of heuristics for the development of science. But this heuristic is said to set a 'pre-conceived' plan for the research programme. And heuristic power is reduced to power to generate increase in successful prediction. Thus the principal criterion 414 Jiang Tianji for rational theory-choice is excess evidential support, not different from what is dictated by falsificationism: chose the best corroborated theory. Just like falsificationism, its formal approach to rationality cuts itself off from the particularity and peculiarity of the historical situations where scientific theory-choices take place. Necessarily, what appears to be development of a theory (a scientific research programme) is really a sequence of classes of propositions (theories) logically related to one another as specified by method. The fertility of a theory, on the other hand, refers to the theory's proven ability to take up new forms which are not deducible from the previous versions, but which are suggested by them, according to Ernan McMullin's account (see his [1976], pp. 398-402). The important relation- ship then becomes that of suggestion, holding between a previous and a later version of the same theory. It is a very significant fact that, as McMullin points out, this suggestion relation is not a purely logical one, but necessarily dependent on the historical record. The fertility of a theory would then be conceived in terms of the developmental transformations it has undergone to solve new problems, where later formulations are not inferable from previous ones, but suggestible by them in the light of the historically available evidence. We cannot hope to state precise rules for the relation of suggestion. But Lakatos did believe that the methodology of scientific research programme can be characterised by general rules that would be binding and consciously or instinctively observed by scientists. Lakatos and Lakatosians claimed with the help of case studies that the evaluative criteria which scientists actually use are those described by the methodology of scientific research programmes. It was claimed that when scientists fail to conform to Lakatos' methodology it is because of external pressures. That is to say, when they are acting rationally, they act in accordance with the dictates of this methodology; when they are acting irrationally, then they deviate from them. Therefore this methodology is an explicit formulation of the universal rational method underlying empirical science. Lakatosians expressed views about the constancy of scientific method just like those of logical positivists as voiced by I. Scheffler. Elie Zahar, for instance, has this to say: MSRP entails that the presystematic methodology instinctively used by scientists in judging individual achievements did not change very much over the centuries, say from the time of ancient Greece until the present day. MSRP does not of course presuppose absolute stability, but it nonetheless implies that deviations from its norms have been in the nature of local fluctuations very different in magnitude from large scale scientific revolutions. This stability thesis, namely that there have been no methodological upheavals on a par with scientific revolutions, may prove too strong. Still the stability thesis is to my mind a very interesting working hypothesis which ought to be investigated. . . . Many people have a strong feeling that intuitive methodology, like intuitive logic, has been largely stable (Zahar [1982]). He seemed to imply that Lakatos' methodology is this instinctive meth- Scientific Rationality, Formal or Informal? 415 odology made explicit, that is why scientists would tacitly subscribe to it if they are not under external pressures. From what is described above, it follows that Popper no less than Carnap, and Lakatos no less than earlier Hempel, would equate rationality with formal rationality. Formal rationality was regarded as actual rationality in science, i.e., as explicit statement of the rationality implicit in scientific practice. Their models of rationality are based on the following assumptions: I. There is a unified scientific method or methodology which has been instinctively used by almost all scientists in all ages or at least from the time when real science began. This is the doctrine of unity, universality and stability of method. 2. The instictive methodology or implicit rationality can be formulated and formalised, even if we do not yet have such a formulation at the present time. This is the theses of formalisability of method. 3. The formal method or formalisable method can be used to achieve our objectives in science, be it the discovery of truth, epistemic progress, explanation and understanding, prediction and technological control, problem-solving, and what not. This is the thesis of rationality of method. 4. The formal or formalisable method also serves as a demarcation criterion between science and other human pursuits, between science and non- science or pseudo-science. It is the defining characteristic of science. This is the thesis of essentiality of method. From this it is concluded that the formal method can serve as the very definition of rationality. 3 VINDICATION OF KUHN'S CONCEPTION OF INFORMAL RATIONALITY This standard conception of rationality is rejected by Thomas Kuhn, the leading proponent of a more complex view of scientific rationality. Kuhn has challenged all the above assumptions of the standard conception. According to the standard conception, the crux of rationality is the acceptance or rejection of beliefs on the basis of evidence. Whether the inductive support of logical positivists, the corroboration of Popper or the evidential support of Lakatos, all these are more or less formalised accounts of evidential relations holding between the theory being tested and observations. Rationality of the decisions of scientists consists in the formal relationships they bear to logic, on the one side, and to empirical facts on the other. Experience and logic can determine uniquely the acceptance or rejection of a theory. For Kuhn, this is possible only in normal science for determining the validity of an articulation or an application of an existing theory. Whole theories or paradigms cannot be judged by the same sort of criteria that are employed when judging a theory's individual research applications, he said: EE 41 6 Jiang Tianji Observation and experiment can and must drastically restrict the range of admissible scientific beliefs, else there would be no science. But they cannot alone determine a particular body of scientific belief. An apparently arbitrary element, compounded of personal and historical accident, is always a formative ingredient of the beliefs espoused by a given scientific community in a given time (Kuhn [1972], p. 4). Because of this arbitrary element, the outcome of theory choice cannot be dictated by logic and experiment alone. Therefore any formal account of evidential relations would not apply to actual scientific practice. Objecting to the positivist and Popperian models of rationality Kuhn said, both of them assume that 'the problem of theory-choice can be resolved by techniques which are semantically neutral. The observational consequences of both theories are first stated in a shared basic vocabulary (not necessarily complete or permanent). Some comparative measure of their truth/falsity count then provides the basis for a choice between them. For Sir Karl Popper and his school, no less than for Carnap and Reichenbach, canons of rationality thus derive exclusively from those of logical and linguistic syntax' (Kuhn [1970], p. 243). Their assumption of the rationality and essentiality of a formal method for theory evaluation is radically rejected by Kuhn. Now come to Kuhn's positive views on theory choice. First, he affirms that theory choice is definitely rational, there being good reasons of the sort usually described for choosing between competitive theories. Second, these reasons constitute values to be used in making choices rather than rules of choice. There cannot be any shared algorithm by the use of which all members of a scientific group will reach the same decision. On the contrary, accuracy, simplicity, scope, fruitfulness, and so on, constitute a shared ideology for the group. Scientists who share the same ideology may nevertheless make different choices in the same concrete situation. Group behaviour will be affected decisively by the shared commitments, but individual choice will be a function also of personality, education, and prior pattern of professional research. Therefore individual scientists will not all apply the same values in the same way. Even if they agree about the importance of the various values, there would remain the problem of combining them all into one precise overall criterion which would de- termine which of two competing theories to accept. But there is no unanimity at all as to the relative weights that are to be assigned to different values in the overall comparison of theories. Third, the choice between competing theories lies in the hands of the group of specialists. These specialists share the same commitments. Even though they don't share precise criteria for theory choice, and even though their preference is determined in part also by idiosyncratic factors, there does in fact, in the process of a controversy over the relative merits of two competing theories, eventually emerge a consensus leading to the acceptance of one of the rivals. Though there are no explicit rules of acceptance for Kuhn, he neverthe- less offers what amounts to a justification for the group behaviour in theory Scientific Rationality, Formal or Informal? 417 choice. Here Kuhn argues that scientists' modes of behaviour as exemplified in theory choice have certain essential functions: 'In the absence of alternate mode that would serve similar functions, scientists should behave essen- tially as they do if their concern is to improve scientific knowledge' (Kuhn [1970], p. 237). Kuhn regards all behaviour essential to scientific develop- ment as rational; since scientific behaviour in theory choice conduces to the improvement of scientific knowledge, this is also an argument in favour of the claim that scientific activity is rational. Now the crucial question is: is theory choice as characterised by Kuhn really rational? There are at least several objections. First, many philosophers insist that an action ordecision in science cannot be considered rational unless it is based on logical or methodological rules which are objective in the sense that different scientists following the same rules and provided with the same information will arrive at the same decision. The basic assumption is that all rationality should be equated with formal rationality. Since Kuhn would not have anything to do with such rules, decision as to which of the competing theory is to be chosen cannot be based on logic or methodology, but rather based on a value system or ideology, the application of which in a concrete situation will not lead to unanimous decisions by different scientists. Personal factors have to be taken into account, and what actually determines a decision is individual judgement rather than general rules. But according to the objection, such personal factors dependent on historical contingencies are irrelevant to rational theory choice. Since for Kuhn, scientists actually rely on sub- jectivistic and psychologistic criteria for choosing between competing theories, such choice is not rational, it is irrational. Kuhn is accused of irrationalism. This objection is untenable. On Kuhn's view, both the shared values and personal factors are relevant to individual decisions. Because perception and judgement are involved in applying values in a concrete situation, each value can be judged quite differently, individual variability in decisions is naturally to be expected. But individual decisions are not final, they generally take risks and might be wrong. Still they are not irrational, because they are not made arbitrarily without reasons. And they do not violate any accepted rules because there are no such binding rules in scientific practice. What is important is the unanimous decision belatedly reached in a group. Neither individual choice nor group choice is based on what is called sub- jectivistic and psychologistic criteria, though psychology is not irrelevant because theory choice is a human activity, and occurs not in the 'world 3' of Popper, but in this real world of our scientists. Second, from the standpoint of the objectors, consensus eventually emerging within the group of scientists, leading to the acceptance of a paradigm, is not a rational decision either, for according to their conception of formal rationality, the consensus criterion is not objective, but sub- jectivistic. Only the logical positivists' degree of confirmation or inductive 418 Jiang Tianji support and the Popperian comparative degree of corroboration are objective criteria for theory choice. They seem to think that for Kuhn a group of scientists may choose any theory they like so long as they agree in their choice and therefore enforce it. This, of course, would be sub- jectivistic. But nothing of the sort is implied in Kuhn's concept of consensus. On Kuhn's view, when a group of specialists reach a unanimous decision, it is never arbitrary, but must be based on good reasons. It may be a conclusion reached at the end of a process of deliberation or argumentation, or it may be more or less easily arrived at as a result of the relative unanimity of the professional judgements of these specialists. Therefore the consensus among scientists is objectively arrived at, and is at least partly a product of objective social conditions and institutions. As Kuhn well said: Recognition of the existence of a uniquely competent professional group and acceptance of its role as the exclusive arbitor of professional achievement has further implications. The group's members, as individuals and by virtue of their shared training and experience, must be seen as the sole possessors of the rules of the game or of some equivalent basis for unequivocal judgements (Kuhn [1972], p. 168). Why should scientific communities be able to reach a firm consensus unattainable in other fields like art and philosophy? According to Kuhn, science is the only field that shows progress in increasing the number and precision of the solved problems. The nature of scientific communities 'provides a virtual guarantee that both the list of problems solved by science and the precision of individual problem-solutions will grow and grow'. At least, the nature of the community provides such a guarantee if there is any way at all in which it can be provided. What better criterion than the decision of the scientific group could there be? (Kuhn [1972], p. 170). If for Popper, 'there is nothing more "rational" than . . . the method of science' (Popper [1972], p. 27), then, Kuhn should have said, 'there is nothing more "rational" than the decision of a scientific group'. But unlike logical positivists and Popperians for whom inductive support or evidential support would be the sole or at least the principal criterion for theory acceptance, the ability to solve empirical problems is for Kuhn neither the unique nor an unequivocal basis for paradigm choice by a scientific group. There can be no criterion of that sort, and no criterion is needed. The group's decision is based on examples of what a scientific theory achieves and on judgements of value. That is why there is no formal rationality, informal rationality is all we have. Third, there are still some other philosophers who would neither denounce Kuhn's model of theory choice as irrational, nor concede that it is rational. According to them, 'any kind of action, including scientific choice, can be called rational only if it can be causally linked to deliberation or reasoning aimed at achieving specific ends. Scientific theory choice as characterized by Kuhn would not be rational in this sense but would rather be akin to... latently functional behaviour patterns, which serve a function Scientific Rationality, Formal or Informal? 419 they were never chosen to fulfill' (Hempel [I981]). I think what Hempel emphasises here is correct, the question whether an action is rational or non- rational hinges on this point: whether or not it involves a process of deliberation. A spontaneous, unconscious action is non-rational. But a deliberate and considered judgement, even when it is not the conclusion of a cogent reasoning, can be rational. We should not suppose that the judgements of a scientific group on the comparative merits of competing theories cannot be causally linked to any deliberation at all. The drux of the matter is: how extensive and detailed a process of deliberation is required before it can confer rationality upon the action that will ensue from it. If it were demanded that a scientist knows what his alternatives are, and he knows what the possible outcomes of these alternatives are, and he also knows what his preferences are among the possible outcomes (that is, if it were demanded that he must know what alternatives will maximise his expected utility, before his choice can be rational), then choices of an individual scientist are seldom rational. By the same token, group choices are very seldom rational, because we cannot expect any scientist to know all this. But why should we demand of a scientist or a group of scientists this knowledge of all alternatives and especially of all the possible outcomes of them? Actions of scientists, including theory choice, can be rational if a less extensive and less careful deliberation is required. The theory choice as characterised by Kuhn is not rational in the utility-maximisation sense of rationality. What is at issue is the problem of formal versus informal rationality. The standard conception of scientific rationality identifies rationality with formal rationality, commitment to a formal method of science. Logical positivists and Popperians have attempted to 'explicate' scientific pro- cedures as governed by explicit and precise rules. The acceptance or rejection of a theory is rational if it can be justified by some of these rules, otherwise it would be irrational or non-rational. This overly narrow and abstract conception of rationality would relegate most scientific decisions and beliefs to irrationality. The contributions of Thomas Kuhn consist in the liberalisation of rationality. Formal rationality becomes only a part, and may be an unimportant part, of rationality, because scientific method in an especially comprehensive and momentous context, namely in regard to scientific revolutions and theory change, cannot be characterised by explicit general rules to be observed by practising scientists. It is obvious that where formalisation of scientific procedures is impracticable or unsuccessful, there people have to rely on informal rationality. But is it true that where a formal method exists, no recourse to informal rationality is needed? Kuhn seemed to imply this in what he said: That part of his (Karl Popper's) writing which seek an algorithm for verisimilitude, if successful, would eliminate all need for recourse to group values, to judgements made by minds prepared in a particular way (Kuhn [1970], p. 238). 420 Jiang Tianji I think that this is incorrect. It is highly unlikely that a complete formalisation of all scientific procedures would be forthcoming, and even if we got such a formal method, in applying it we cannot dispense with human intelligence and judgement. 4 OTHER MODELS OF INFORMAL RATIONALITY While Thomas Kuhn has established, in a more or less descriptive way, that theory choice is under the sway of informal rationality, Hilary Putnam provides a powerful argument against the formalisability of scientific rationality. His central idea is that the need for an informal element is indispensable for any formal method. The formalised scientific method depends on inputs which are not formalisable. Take the Bayesian Approach for instance. In computing the degree of support of various hypotheses using the Bayesian theorem, as 'inputs' we need not only the computable likehoods, but also the prior probabilities of the various hypotheses, i.e., the 'subjective degrees of belief' assigned by scientists to those alternative hypotheses prior to examining the observational evidence. It turns out that difference in the prior probability function can lead to violent difference in the actual degrees of support assigned to theories. Putnam draws the conclusion that formal rationality does not guarantee real and actual rationality. If one starts out with a 'reasonable' prior probability function, then one will only assign degrees of support to 'reasonable' hypotheses, but if one's prior probability function is 'unreasonable', then one's assignment of degrees of support to various hypotheses will be 'irrational'. This prior probability metric represents the scientists' antecedent beliefs about the world. It seems that one of the inputs to the method itself is a set of substantive factual beliefs (or degrees of of belief). The need for an informal element is similar in other approaches. Therefore it is not possible to draw a sharp line between the content of science and the method of science. Both Putnam and Shapere lay stress on the dependence of scientific method upon scientific beliefs, and maintain that scientific method changes constantly as the content of science changes. If the method of science changes constantly, then the traditional belief in a universal stable method is untenable and the attempts to formalise it would be doomed to failure. Putnam has this to say, The hope for a formalisable method, capable of being isolated from actual human judgements about the contents of science (about the world), and capable of actually being stated seems to have evaporated (Putnam [1981]). If from the constancy of scientific method the ideal of formalisation is derived, and as a result the praises of formal rationality are sung, and formal rationality is identified with the very rationality itself, then from the changeability of scientific method we should be convinced that complete formalisation is hopeless and informal rationality must be restored to its Scientific Rationality, Formal or Informal? 421 legitimate status. During the period from fifteenth to seventeenth century, when new science was originated and developed, general methodological maxims instead of precise formal rules were put forward by scientists and philosophers for the guidance of scientific inquiry. In the application of general maxims to particular cases intelligence and judgement are required, thus it was informal rather than formal rationality that prevailed in scientific practice of that period. After a long tradition of attempts at formalisation from Mill to Carnap and Popper, it was Polany and Kuhn, in the years around i960, who first threw off the tyrannical yoke of formal rationality imposed by logical positivists and Popperians. Polanyi is sceptical of the degree to which the skill of theory-assessment which develops historically in the scientific community can be made explicit either by scientists or, post factum, by philosophers. In his view, scientific rationality can never be made fully explicit or even rendered unambiguous in outcome. Kuhn claimed that decision in theory-choice is based on ideology rather than on logic. There cannot be any algorithm for theory-choice which will dictate the same decision for all members of the group. The upshot of all this is to replace formal rationality by informal rationality. Other models of informal rationality include both the inferential and the noninferential ones. Shapere's discussion of principles of reasoning as having a central role in many scientific decisions between competing theories or lines of research is a clarification of the functioning of informal reasoning in scientific practice. These principles of reasonable scientific research are not formal rules, the reasons they provide for seeking a certain type of theory are of course not conclusive. Shapere said, 'that the reasons were not logically conclusive did not make them any the less reasons-and good ones, relative to the state of science at that time-nor did it make action in accordance with them any the less rational' (Shapere [19741, p. 409). Therefore in accounting for the rationality of scientific discovery as well as of justification our one recourse is to informal rationality. The non-inferential models of informal rationality are emphasised, besides Kuhn and Putnam, by Polanyi, Grene, Toulmin and Wartofsky. It is a form of rational judgement modelled on judgement in the law, in medical diagnosis and in aesthetic appraisal; this form of informed, skilled judge- ment involves the application of general maxims or directive to particular cases, but it is not rigorously rule-determined. It is rational in the sense that judgements may be defended by citing particular cases, previous appli- cations of the maxims and the like. Wartofsky wants to resurrect a historical alternative conception long overshadowed by the standard conception of rationality. This is the 'heuristic tradition' involving 'craftsmanlike skill in judgement' and 'tinkering'. The scientific theorists tinker with ideas (see his [1980]). This kind of down-to-earth tinkering found in crafts allows for creative innovation in science. Thus this model of informal rationality would make scientific behaviour both creative and rational. 422 Jiang Tianji 5 CONCLUSION We have seen that the standard conception of scientific rationality, the model of formal and abstract rationality, is divorced from scientific practice. This normative conception is in conflict with the historical fact that standards of rationality change over time. It is the assumption of fact-value dichotomy that renders the standard conception seemingly invulnerable. But starting from Kuhn and others, philosophers are increasingly coming to agree that scientific rationality includes an informal part whose importance is estimated differently by different philosophers. Both extreme positions- the identification of rationality with a formal method or the complete neglect of formal rationality-seem to be incorrect. Thomas Kuhn gives formal rationality its due in normal science. Hilary Putnam would think that each functioning of scientific rationality includes a formal part and an informal part. This 'liberation' of rationality from formal logic or normative methodology is a good thing. Let us quote Kuhn again: The man who embraces a new paradigm at an early stage must often do so in defiance of the evidence provided by problem-solving. ... There must also be a basis, though it need be neither rational nor ultimately correct, for faith in the particular candidate chosen. Something must make at least a few scientists feel that the new proposal is on the right track ... (Kuhn [1972], p. 158). For Kuhn, belief or faith in a theory's promise is a most important factor relevant to theory-choice, but the faith need not be based on a rational inference (and thus would be rejected by the standard conception of rationality as irrational), but based on a rational judgement or an insight: the scientists feel that the new proposal is on the right track. Only specialists can make such a judgement. Even if the judgement is sometimes inarticulate or turns out to be wrong, might we say that it is irrational? I think not. The liberalisation of rationality must be extended to the point of including such a judgement among the rational ones. Wuhan University People's Republic of China REFERENCES HEMPEL, C. G. [1965]: Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. HEMPEL, C. G. [1981]: 'Turns in the Evolution of Problem of Induction', Synthese, 46, p. 402. KUHN, T. [I970o]: 'Reflections on My Critics', in I. Lakatos and A. Musgrave (eds.), Criticism and the Growth of Knowledge. KUHN, T. [1972]: The Structure of Scientific Revolutions, 2nd ed. LAKATOS, I. and MUSGRAVE, A. (eds.) [I970o]: Criticism and the Growth of Knowledge. MCMULLIN, E. [1976]: 'The Fertility of Theory and the Unit of Appraisal in Science', in R. S. Cohen et al. (eds.), Essay in Memory of Imre Lakatos. POPPER, K. [1972]: Objective Knowledge. PUTNAM, H. [198 1]: 'The Impact of Science on Modern Conceptions of Rationality', Synthese, 46, p. 375. Scientific Rationality, Formal or Informal? 423 SCHEFFLER, I. [1976]: Science and Subjectivity. SHAPERE, D. [1974]: 'Discovery, Rationality and Progress in Science: A Perspective in the Philosophy of Science', in K. Schaffer and R. Cohen (eds.), PSA 1972. WARTOFSKY, MARX W. [I980]: 'Scientific Judgement: Creativity and Discovery in Scientific Thought', in T. Nickles (ed.), Scientific Discovery: Case Studies. ZAHAR, E. [1982]: 'Feyerabend on Observation and Empirical Content', British Journalfor the Philosophy of Science, 33, P. 407.