Issue #01. April 1995. Pp. 81-95.
Epistemic Values in Science
Copyright (C) by SORITES and Valeriano Iranzo
Epistemic Values in Science
Valeriano IranzoFootNote 57
Laudan adopts a naturalistic stance, assuming that there is no cleavage between theories and methods, on one side, and goals, on the other. His reticulated model of science, developed in SV, emphasizes the interconnection among theories, methodological norms, and goals. Research about scientific aims has to take into account results in other levels because theories and methodological norms are basic to determine the legitimacy of the aim at issue.
Laudan points out that he is concerned with epistemic aims or values. Thus, his task focuses on a naturalistic account of epistemological normativity in science, setting aside the muddy question of ethical normativity. From now on, I will use the words «aim», «goal», and «value» in an epistemological sense. The distinctive feature of epistemic goals -- «explanatory power», «predictive accuracy», «truth», ... -- is their close relation to the goodness of our beliefs. No doubt, scientific practice is not isolated and it is externally controlled by social goals. The politicians' decisions that determine the research policies and the technological applications are embedded in ideological and moral values. But, on the other hand, I think that the increasing control over citizens or the use of military power over other countries, and the eradication of infectious diseases -- for instance -- are not epistemic aims, although they might be aims actually pursued by scientific research (perhaps through the previous achievement of epistemic goals -- think only of predictive accuracy concerning human behavior).
According to Laudan there are two main reasons on which to reject an aim: because it does not fit with current theories and practice, or because it is utopian, namely, because it is not realizable. I will call the former, the principle of coherence (PC), and the latter, the principle of realizability (PR). Let us begin with PC.
1. Coherence
To illustrate PC, Laudan offers two examples extracted from the historical record. The first is the shift, at the end of XVIII century and the beginning of XIX, from inductivism which refused to postulate unobservable entities to theories purporting the discovery of nature's deep structure. Against the inductivistic mainstream, Laudan refers to Hartley, Lesage and Boscovich, who were criticized by putting forward theories committed with inobservable entities. They had to develop a specific methodology (hypothetico-deductive), although its incompatibility with the aims widely acknowledged by the scientific community of that time somehow keeped them apart from it. However, confronted with a difficult choice, Hartley, Lesage and Boscovich did not modify their theoretical preferences. The empiricist qualms went by the board as they persisted in trying to understand the visible physical realm through an invisible one. Later, Herschel and Whewell claimed that «the axiology of empiricism was fundamentally at odds with the axiology implicit in scientists' theory preferences» (SV 59), and they gave strong and definite support to the postulation of unobservable entities.
Laudan offers another example to illustrate the feedback between theories, methods and aims. Now the aim in question is intelligibility, a goal strongly favoured by the cartesian way of doing science. From this point of view a good explanation involves some kind of reduction of the less intelligible to the more intelligible. Cartesian objections to newtonian physics estemmed from the notion of «action at a distance», a notion hardly intelligible for natural philosophers influenced by Descartes. Of course, the heart of the matter is the criterion of intelligibility, but by the 1740s -- Laudan continues -- Cartesians could not even convincingly show that the notion of action by contact (the only sort of action in a full universe such as that considered by Cartesians) was more intelligible than the notion of action at a distance. At this stage it became more reasonable to relinquish intelligibility as a desirable aim for science, since none of the physical theories had been entirely successful in eliminating all suspicious notions, notwithstanding serious efforts in that direction.
Both examples tend to show that the process of goal revision roughly consists in «an examination of what our best (or, here, all our available) theories seem capable of achieving» (SV 61). Notice that although PC denounces situations where there is a gap between explicitly deffended aims and current scientific practice, it does not force us to abandon an aim. It is highly desirable to increase the degree of conceptual coherence but changing aims is not the only choice. We can also modify theories and methodological rules keeping aims fixed, as Laudan himself acknowledges.FootNote 59 Nevertheless, there is no general way of knowing what to do in these situations. A reasonable choice has to take into account all contextual information that could be relevant and, surely, members of the scientific community are the best qualified to accomplish the task.
PC seems to be a reasonable condition. Something is wrong when our best theories do not have the properties we regard as legitimate aims of science. But, despite the fact that PC rightly stresses the feedback between theories and methods, on one side, and aims, on the other, it is too «soft». To claim that what we can achieve with our best theories and methodological rules occasionally can lead us to revise science's epistemological aims is most definitely a rather imprecise statement. In fact, it would probably even be accepted by those who do not get on with a naturalistic standpoint. However, taking for granted that we cannot fix in advance how the revision has to be carried out, we could reformulate PC as involving the claim that any goal whatsoever may be revised when the results -- e.g., theories and methodological rules -- clash with it, despite repeated attempts in that direction. PC thus reformulated is no more precise than before but, at least, it is fully in line with naturalism, since it is not yet possible to set up a trascendent goal for science, a goal unaffected by the workings of the two other levels.
2. Realizability
According to Laudan, a necessary condition for a rational -- or legitimate -- aim is its achievability: «... 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.» (SV 51) This is PR in its general formulation.
It is worth stating that we cannot infer the utopian character of an aim, properly speaking, from the fact that no theory is successful in achieving it. In that case there would be no difference between PC and PR. That fact is a necessary condition, not a sufficient one, to consider an aim as illegitimate. There is an outstanding difference as regards the resulting policies from PC and PR, since lack of conceptual coherence turns on the red lights and warns us that something is wrong -- although PC does not tell us where the shortcoming is --, while irrealizability of an aim discards it inmediately as a legitimate one.
Laudan thinks that PR is uncontroversial. We usually regard as irrational those actions aiming for unachievable aims as inmortality or perpetual motion machines; in the same way, if we have good reasons to think that a goal is beyond our faculties, then the most rational course of action is to discard it. In other words, an aim or a value is utopian when «we have no grounds for believing that it can be actualized or «operationalized», that is, we do not have the foggiest notion how to take any actions or adopt any strategies which would be apt to bring about the realization of the goal state in question.» (SV 51) In words alien to Laudan's viewpoint we could say that there is no cleavage between an instrumental rationality and a teleological one working separately, rather, instrumental rationality assesses the realizability of aims, and the realizability of aims determines, in its turn, the rationality of the aims. Axiological controversies are on a par with factual or methodological ones, consequently the same mechanisms are involved in settling any scientific disputes.
Laudan distinguishes three kinds of utopianism (demonstrable, semantic and epistemic). I will analyze them separately.
Demonstrable utopianism arises when we infer the impossibility to achieve the aim at issue from logical or physical laws. Laudan's instance is infallible knowledge. Physical laws are unrestricted generalizations but testability is radically limited to observational claims we have access to. Because of this, we can not be sure that our knowledge is infallible, at most we could say that up to now this piece of knowledge has not failed, but this is not enough for infallibility in its full sense. Infallibility could have been a goal for science during long periods of history but now there is a wide agreement about fallibilism, the opposite view. It claims that scientific knowledge is provisional, revisable. We can back it up not only by means of logical arguments, as Laudan does, but with information from neurophisiology or comparative biology. These sciences underline the crucial role played by the sensorial receptors and the nervous system of a species in shaping reality. Research in these fields casts serious doubts on the access to a rough reality independent of the knower, and stresses the changing character of the latter, subject to evolution processes that profoundly alter his appropiation of reality. Hence it is really a philosophical platitude -- both in science and in philosophy of science -- that infallibility is not a reasonable aim, at least in an absolute sense. The moral that can be drawn from Laudan's example is that we can infer grounds for or against a goal from the theories and methods we accept at some stage of scientific development.
Semantic utopianism arises when the aims are not unambiguously characterized: «If someone purports to suscribe to an aim, but can neither describe it in the abstract nor identify it in concrete examples, there is no objective way to ascertain when that aim has been realized and when it has not.» (SV 52) Laudan thinks simplicity and elegance are not legitimate scientific aims in this sense. According to him, most advocates of these goals have no clear ideas about what these aims consist of, they offer neither a coherent abstract definition nor good examples that supposedly instantiate it.
Unfortunately, Laudan is not clear enough about his intended sense of «clear». His remarks about semantic utopianism raise different issues. Firstly, the necessity of giving an accurate content to aims. It does not seem appropiate to use a goal as an emotive word, on pain of turning axiological debates into a confrontation of disguised subjective preferences. I agree with Laudan that those who subscribe to simplicity, for example, as a reasonable aim have a serious problem if it does not refer to an objective property of theories and is only a way to emotionally reinforce the acceptance of a theory -- they are defending a goal devoid of content. However, I think his picture could do justice to elegance but not to simplicity. Far from it, the problem with simplicity is that there is no agreement among its advocates because they have different interpretations about it.
When applied to scientific theories simplicity may demand a reduction in kinds of postulated entities, laws' parameters, basic principles, mathematical calculations, ... We are not bound to understand simplicity in the same way when working on different scientific subdomains, so perhaps there is no such a general property as simplicity, a property that all scientific theories possess in more or less degree. Besides, why do scientists prefer simpler theories? It seems that to equate simplicity to convenience is not enough. If simplicity is an epistemological value, in its full sense, it must be connected with more interesting epistemological properties as predictive accuracy, explanatory power, ... For Popper simplicity is related to falsability; Quine prefers linking it with high probability; and E. Sober is skeptical about the possibility of stating a general argument to justify our preference for the simpler hypothesis when confronted with two having the same score at observational accuracy.FootNote 60
Consequently, there is no general agreement among philosophers neither on how to define simplicity nor on how to justify it. In any case, a suitable account of scientific reasoning has to include simplicity insofar as the scientific judgment is under its influence. And, in relation to what we are mainly concerned with, the really important point is not whether everybody is talking about the same, but whether whatever each one of them is talking about may be a legitimate goal for science. Then, we have to isolate the reciprocal irreducible definitions (simplicity1, simplicity2, ... simplicityn) and treat them as different goals, instead of rejecting simplicity straightforwardly as Laudan does. Then, they have to be assesed making use of the coherence principle and the realizability principle. Therefore, the mere coexistence of different interpretations of `simplicity' is not a reason enough to exclude simplicity from the realm of legitimate aims.
These comments reveal Laudan's careless use of words like «goal», «aim» and «value». He treats them all as synonymous but it must be emphasized that, roughly speaking, a value is a worthy property and a goal -- or an aim -- is what we pursue by our actions. We regard scientific theories as good or bad insofar as they possess worthy properties. But not every worthy property should be properly considered as a goal. It sounds extremely odd to say that scientists look for simplicity, or compatibility with the body of accepted knowledge, although they are all worthy epistemological properties, i.e., epistemological values. They are rather means for other goals such as explanatory power, predictive accuracy, and, why not, truth. On this view, their legitimacy would be assessed not only by PC and PR, but through their historical success as reliable indicators of more interesting epistemic values -- the real goals of scientific theorization mentioned earlier -- as well. Nonetheless, Laudan does not distinguish between the epistemic values pursued by themselves -- the actual goals of science -- from the epistemic values which are means for ulterior ones. All this does not rule out the possibility that the very epistemic values may be turned into means for non-epistemic values (see above p. 1).
Lastly, epistemic utopianism, which is much more fully characterized by Laudan than the other species of utopianism. This version arises because there is no criterion to determine when the value is satisfied, despite having a clear definition and no demonstration that it is utopian. Truth is the only example Laudan offers at this point although in Science and Values he devotes an entire chapter to discussing the issue. The charge of epistemic utopianism is roughly stated in the next quotation:
Suppose, (...), someone claims to have the goal of building up a body of true theories. Moreover, let us suppose that he offers a coherent and straightforward characterization of what he means by a theory «being true» -- perhaps in the classic tarskian semantics of correspondence. Under such circumstances his goal is not open to the charge of semantic confusion. But suppose, as we further explore this person's goal structure, it emerges that, although we can define what it means for a theory to be true, he has no idea whatever how to determine whether any theory actually has the property of being true. Under such circumstances, such a value could evidently not be operationalized. ... In the absence of a criterion for detecting when a goal has been realized, or is coming closer to realization, the goal cannot be rationally propounded even if the goal itself is both clearly defined and otherwise highly desirable. (SV 53)
In fact, the chapter on truth goes beyond the delegitimation of an aim to turn into a refusal of realism. It must be conceded that if we had not have the remotest idea as to how to approach truth, we would have a conclusive reason to abandon it as a legitimate aim. But achievability may be relative and, radical inaccesibility excluded, the axiological status of truth depends on additional factors. In the same way as equality or freedom are legitimate aims in political theory or in morals, even though in practice it seems impossible to realize these values completely, truth may be a legitimate goal for science, although we know we will never develop a true -- in an absolute sense -- account of the world.
Therefore, I shall try to show -- contrary to Laudan -- that truth is a goal for science and that, being a genuine aim as it is, its rationality depends not only on how far it may be achieved, but on the explanatory role it plays in a fair account of science as well. Following his strategy, I shall deal with this issue separately.
3. Truth
Laudan distinguishes three varieties of realism. Semantic realism -- «to claim that all theories are either true or false and that some theories -- we know not which -- are true» is presupposed by epistemic realism -- to claim that one can know if theories are true or false by means of certain kinds of empirical support. The third modality is intentional realism: «the view that theories are generally intended by their proponents to assert the existence of entities corresponding to the terms in those theories.» (SV 105) Laudan is not interested in denying that theoretical claims have a determinate truth value. His is not a complaint against bivalence. And he is not interested in casting doubt on the intentions of scientists either: they usually propound theories as true claims about the world. Nonetheless, this realistic attitude is not Laudan's genuine target (the more interesting question is, of course, if those theories are really true). Laudan's concern is epistemological realism. In very brief compass, theories may be true or false, but we have good reasons to despair of ascertaining it. He does not discard the possibility that truth be a worthy property which scientific theories do possess, the problem is that we are unable to detect it. Besides, realism has a remarkable normative component; in fact, it is a doctrine about «what the aims or values of science ought to be.» (SV 106) According to realism, the main aim of science is «to find ever true theories about the natural world.» (id.) Laudan attacks the notion of truth because of its undetectability and, consequently, he eschewes it as a legitimate goal.
Before discussing Laudan's objections, it may be worth noting some well -- known remarks. Although research in the history of science shows us a non linear process, it is undeniable that recent theories, at least in mature sciences have a higher degree of empirical adequacy than their predecessors. And we must notice that improving predictive eficacy is closely related to improving instrumental success and technology. It is not difficult to find theories in present day science which encompass an impressive amount of empirical phenomena, much more than ancient generations of scientists would have ever imagined. Antirealists like Kuhn, van Fraasen and Laudan have no doubt about the high rate of empirical adequacy in science but they all warn us about seriously considering the ontological commitments of theories, especially the theoretical ones. And, as reference and truth are linked -- given that to devise a true theory with referentially empty central terms would be a rather complicated task --, suspicion over reference (theoretical entities) leads to suspicion over truth (theoretical claims).
Instead of stopping at the empirical level and remaining agnostic about the upper floors, realist-minded philosophers think that some theories are true, from which it follows that their theoretical claims are also true and that the referents of their theoretical terms do exist. The argument many realists (Boyd, Putnam, McMullin, Leplin, Newton-Smith, ...) make use to fill the gap between the empirical and the theoretical levels is based on the explanatory role of truth. For most of them, there is no better ground to affirm the existence of theoretical entities and, consequently, the truth -- at least the approximate truth -- of theories than their empirical success. Insofar as the success of later theories increases, we have a compelling reason to affirm their truth and the existence of the theoretical entities posited by them. Otherwise it would certainly be striking that the world behave as if these entities existed, whithout really being there. This argument is a version of a model of reasoning called «inference to the best explanation» (IBE) that recalls Peircean abductive inference and has the following form:
O (an account of a fact),
E1 is the best explanation of O (among the set of available and rival explanations E1, E2, ... En),
Therefore, E1 is highly probable.
If we apply the model to the case we are concerned with, O will be an account of the success of a theory and E1 will state that if a theory is successful, then it is true. A stronger -- and more general -- version of E1 could be that the growing success of theories is due to their truth.FootNote 61
Laudan claims, as do most anti-realists, that truth does not possess the explanatory power in which realists believe. If I have understood him, his rejection of truth is derived from two different contentions. The first has to do with devising truth as a property gradually instantiated; the second arises from the historical record and disputes the alleged connection between success and truth.
(a) Truth and Closeness to truth.
Even though Laudan does not accept a link from success to truth, as we will see later, he acknowledges that the converse entailment «if a theory is true, then it will be successful» is self-evident. (SV 117) The point is that there is no current scientific theory that could properly be considered fully true because realists are forced to weaken the notion of truth in response to Putnam's pessimistic meta-induction.FootNote 62 Since a great deal of past scientific theories have been falsified, we can't be sure that theories accepted now will survive every future test. It is quite likely that they will be eventually discarded. According to this, we would never be entitled to ascribe truth to a particular theory because those theories accepted now will be replaced by better ones in the future, just as they in turn replaced earlier theories. Indeed, in all probability every theory we have now is false.
Against this skeptical argument a minor change may be performed upon IBE so as to infer the approximate truth of successful theories. At most, all we have is more o less closeness to truth but not truth itself. Certainly, to adscribe complete truth to a scientific theory would prevent from revising it and that would not be the game we are playing -- science -- but a very different one. In practice, «true» is not an absolute parameter since there is nothing unsound in talking about more or less true, and scientific realists often prefer using expressions as «partial truth», «proximity to truth», «verisimilitude», «truth-content», and so on in order to avoid the commitment with ascriptions of truth in an absolute sense. There are theories closer to truth than other ones, and in mature sciences we have good grounds to consider later theories closer to truth than former ones. Therefore, truth is gradually instantiated. Even though we have no instances of a true theory in an absolute sense, the growing success of later theories enables us to consider them closer to truth than preceding ones. This is what scientific progress mainly consists in, from a realistic stance.
However, the notion realists employ to forego pessimistic meta-induction (approximate or partial truth) is unacceptable to Laudan. First, it has to be showed that a semantically adequate characterization of it is available; secondly, realists have not argued convincingly that approximately true theories are successful predictors; thirdly, an epistemical criterion for adscriptions of approximate truth is needed (SV 120).
The first requirement is difficult to fulfill because Laudan does not give us any clue about what a «semantically adequate characterization» would consist in. I suppose Laudan is not demanding a mathematical account of approximate truth. Do we need a technical definition like the Tarskian one? Or, is it enough with a notion that allow us to make comparative judgments between rival theories?
Measurement of closeness to truth is an awkward task. Some realists have tried to define closeness to truth in terms of truth-content. Nevertheless, this approach has to face great difficulties -- notice, for instance, that scientific theories have infinite observational consequences -- and there is a generalized skepticism among philosophers of science about the possibility of working out the relative truth content of two theories. Popper himself acknowledges the limits of analysis about verisimilitude.FootNote 63 Yet, he thinks that the lack of aplicability of verisimilitude is not a sufficient reason to discard the notion of truth. He reminds us that deducibility is not as clear a notion as some would like. Although a general procedure to decide in concrete examples if a formula is deducible from the axioms of a logical calculus -- and very often there is no time to work out the infinite number of valid deductions -- can not be offered, this fact does not lay aside notions as deducibility and formal validity. This is just what happens with closeness to truth.FootNote 64 To deffend the explanatory role of truth -- and, thus, its legitimacy as a goal for science -- it is not necessary to have a very exact notion of it. «Closeness» is a misguiding word here because it invites to measure the distance to the last stage; but talking about closeness -- or approximation -- is only a way of acknowledging that even our current best theories might eventually be rejected.
Here we may take into account recent developments which try to reconcile the relentless historical replacement of theories with the realist intuition that progress -- in mature sciences at least -- consists in a growing precision as to the identification of what there is. Causal theories of reference have to face great difficulties and it is dubious that scientific realism could ground upon them. Perhaps a «metaphorical» theory of reference like the ones developed by R. Boyd and E. McMullin would be more promising. Both of them appeal to the notion of metaphor and set up a more loose connection between theory and world than a pure causal theory whithout abandoning the realist viewpoint.FootNote 65 I shall not pursue the point here, but if it can be shown that recent theories -- through a refinement of an initial metaphor, for instance -- are better at identifying reference than earlier ones this would give support to our intuitive jugdments about approximate truth. In fact, scientists talk about true/false theories/hypothesis, at least for now, and it does not seem to create a perennial confusion among them. Judgments attributing truth or falsity are revisable but that is a different matter. In the selection of rival theories we may make errors but this does not undermine the global task of separating and excluding falsities.
Laudan himself develops an alternative to realist conception of scientific progress in Progress and Its Problems.FootNote 66 There he claims that the goal of science consists in solving problems -- both empirical and conceptual problems -- and avoiding anomalies. We have to choose theories -- or research traditions -- with a high rate of problem-solving efectiveness. I must confess that, from the point of view of precision, I see no advantage in replacing the rate of verisimilitude for the rate of problem-solving efectiveness. To start with, we have no clear criteria about what to count as a problem. Let us grant that we arrive at a precise definition of what counts as a problem and that we can neatly distinguish between two different problems and two different formulations of the same problem. Yet efectiveness in solving problems is not merely a matter of counting solved problems. The resolution of a certain problem may be crucial for ulterior developments in the discipline, or perhaps, for devising successful technology to face practical pressing needs. Since not all the problems have the same relevance, and their importance -- and not only the number -- is decisive for choosing one theory, we have to previously assess their relative weight. But, how can this be determined? If, in the end, we have to rely on scientific intuitions to assess the problem-solving efectiveness of two rival theories, then I don't think that Laudan is in a better position than the advocates of approximate truth. Problem-solving efectiveness is as fuzzy a notion as approximation to truth. Since we have no precise definition of both properties, Laudan's rejection of partial truth also forces him to abandon his own approach to scientific progress.
Laudan's second objection against approximate truth criticizes its purported link with predictive success: «No one of the proponents of realism has yet articulated a coherent account of approximate truth which entails that approximately true theories will, across the range where we can test them, be successful predictors».FootNote 67 He briefly discusses Popperian definition of approximate truth in terms of truth and falsity content and argues that it is possible we may not be able to ascertain that a theory T1 is more approximately true than T2 on the strenght of its predictive success (because its truth content is not the same as the truth content available to us: the former may be huge while the latter poor.) If the successful predictions of T1 are not available to us, we shall not consider it as a better approximation to truth than T2, even though it may be so indeed. It would be equally possible that T2 be more successful than T1, although it is further from truth because its falsity content unknown to us is greater than the falsity content of T1. Laudan's second objection must be understood as a concern with detection of approximate truth, and this conflates it with the third one -- the need of an epistemical criterion for adscriptions of approximate truth. He points out that success is not a reliable indicator of approximate truth insofar as the realists have not demonstrated a connection between approximate truth and success.
A few remarks are in order here. There are several ways to define approximate truth. The Popperian approach -- an algorithmic one -- is just an example and perhaps it is not on the right track. On the other hand, having granted, as Laudan does, that the connection between truth and success is self-evident, I see no problem in affirming a connection between approximate truth and success. Despite the fact that the Popperian attempt to define closeness to truth in terms of truth and falsity content is open to the logical objections raised by Laudan, realists are not bound to this definition. Approximation to truth could be understood as a consequence of a more exact determination of the entities with which we causally interact by means of sophisticated devices.FootNote 68 The plausibility of IBE as a general pattern of reasoning is untouched after replacing truth for approximate truth. The question now is: on which grounds may we infer partial truth from success? This takes us from logical to historical considerations.
(b) History of Science.
Realism affirms the existence of theoretical entities, while an antirealist like Laudan prefers to stay at the observational level. Obviously, if we took IBE as a conclusive argument from a logical point of view we would be committing a formal fallacy. Realism does not pursue such a kind of basis: IBE only claims that the antecedent is highly probable, not certain. However, Laudan finds evidence against the alleged connection between properties as «empirically adequate» and «true» in the history of science. In fact, some theories were once successful, well confirmed and widely accepted but now they are considered plainly false: the ploghiston theory, the caloric theory of heat, the humoral theory of medicine, among other examples. (SV 121) Accordingly, if success is not an indication of truth, we are not entitled to infer the truth of theoretical postulates from their empirical success and the very existence of the entities is seriously questioned.
There have been several attempts to meet the challenge,FootNote 69 and I think the most promisory defense of realist convictions consists in admitting that success by itself is not a sufficient condition for truth, while maintaining that truth is a basic notion in order to understand the workings -- and the success -- of science.
First of all, «realism is not a blanket approval for all the entities postulated of the past.»FootNote 70 A theory could be firmly believed by the scientific community and -- according to success standards of the age -- regarded as a successful theory, but this is not enough to infer its truth. Success has to be assessed during a significant period of time and it has to be accompanied by other important epistemic values that Laudan completely neglects.
Take, for instance, «ad hocity». Ad hoc explanations are not legitimate ones. They can be temporarily accepted, if there are no better alternatives. But it is commonly held that ad hocity is an undesirable feature, even though an ad hoc theory encompasses a large amount of empirical phenomena. A good and well-known example is Ptolemy's heliocentric system. It could have been successful, from a predictive point of view, during a large period of time but it is not true. Its truth can not be inferred from its predictive reliability; but its falsity can be inferred from its ad hocity. To save the phenomena is not enough and it is even a symptom of something going wrong.
Ad hoc theories go after observation, while scientific method somehow anticipates itself to phenomena. Of course, a successful prediction involves that we have anticipated what is going to happen, but I am thinking of a special kind of predictions: what has been called «novel predictions». The ability to make unexpected predictions is an epistemic value (fertility) that Laudan does not discuss. Ad hoc theories are not fertile theories, and in that sense they do not anticipate themselves to phenomena. Fertility is closely related to explanatory power, another value that is not in Laudan's agenda. Sometimes a theory works in a new field though, in principle, it was not thought out to handle it. This sort of success is not simple predictive success and, whatever the name we choose, it is more difficult to explain by antirealists. In such a case it seems that we are entitled to infer that the underlying mechanisms of the different kinds of phenomena are the same. The theory is anchored to solid rock by identification of theoretical entities, mechanisms and processes that really exist and it unifies previously separated realms increasing explanatory power. Thus, atomistic theory showed its explanatory power by dealing with heat, even though it was not primarily designed to apply there.
Therefore, the fact that scientists distinguish between ad hoc explanations and more natural ones gives reason to believe that there are other factors in addition to predictive success which function as reliable indicators of truth. These factors are worthy properties -- values -- as fertility and explanatory power. The lack of any allusion to them reveals an important neglect in Laudan's axiological discussion. Observational success by itself may not be sufficient for truth, but there are other values that give some grounds to believe that truth is not so a blurry notion as Laudan suggests. Approximate truth of theories, and the existence of referents partially similar to the theoretical posits, can be inferred when predictive success plus fertility plus explanatory power go together.
How to measure values as fertility or explanatory power? Certainly, they are more difficult to assess than predictive accuracy or instrumental success but we can recall historical examples to show that they can be discerned. It is commonly assumed that Newton's explanation of free-fall is better than Galileo's one, even though both are false. But the superiority of the former is not simply a question of predictive success, rather it suceeds in offering a more comprehensive and accurate picture of physical phenomena. This is not surprising. One of the most peculiar features of scientific methodology is self-correctness. The criteria of what counts as a good explanation have changed for centuries and the scientific community modifies them in order to make them more powerful and effective in representation and manipulation of phenomena. In D. Shapere's words, it is not only a matter of coming to know about the world, but of learning how to learn, to think and to talk about nature as well.FootNote 71 However, to deffend the legitimacy of truth as a scientific aim it is not necessary to be commited with a perfect theory as the result of the iterated application of scientific methodology. I think we can hardly make sense of that notion indeed.
In his first book Laudan provocatively compared pursuit of truth with pursuit of immortality, of the philosopher's stone, .... as if it were a completely misguided enterprise, if not a chimerical dream.FootNote 72 But to affirm that predictive success -- which is relatively easy to assess empirically -- is not enough for truth does not involve that truth is a misterious and undetectable property. On the other hand, our judgments concerning theoretical truth are historical. They are determined by the amount of knowledge we have at certain time and we may fail in our ascriptions of truth. Again, this does not mean to equate «T is true» with «T is widely accepted» or «T is justifiyingly believed» because truth is not a purely epistemic concept. It has to do not only with the way we represent the world but with the way the world is, and a historical approachment to mature sciences shows certain referential stability and an increasing detail of the internal mechanisms postulated.
4. Conclusions
It is now time to look into the merits of Laudan's attempt to set up objective criteria in order to settle axiological controversies in science.
PR is more disputable than PC. Notice that PC is closely related to a version of PR: the demonstrable utopianism. To refuse a goal for being demonstrably utopian means to call for some kind of coherence. At the bottom we have to deal with coherence because in both cases the rejection of an aim is made in order to avoid incoherence between what we claim and what we do, on one side, and what we try to achieve, on the other side. Why does not Laudan subsume demonstrable utopianism under PC as a particular version of incoherence? There is a subtlety. Demonstrable utopianism allows us to infer conclusively the impossibility of the aim in question from the accepted theories and to reject it outright, whereas the revision of the aim according to PC is the result of repeated failures in achieving it. We could say the former points at a theoretical incoherence; the latter at a practical one.
In relation to semantic utopianism, I have already pointed out why Laudan's argument is not sufficiently powerful. Now, I would like to make a more general remark. Laudan discards simplicity and elegance implicitly assuming another value which he does not argue for: precision. Precision is the value that supports the charge of semantic utopianism but then we have to adress some questions: why is precision a more fundamental value than simplicity or elegance? what sort of justification could we offer for precision? could it not be that precision was also a utopian goal according to some of the three modalities suggested by Laudan? The point is that PC and PR are themselves grounded on values. Stating the problem in a more general form: are we not forced to show that the values involved in the analysis of science are justified from the very science (in accordance with a naturalized conception of knowledge)? I am not sure this is a severe requirement for Laudan's reticulated model. Perhaps Laudan could reply that answering this general question goes beyond an analysis of scientific rationality, but I think it would be desirable to fill the gap. Meanwhile the legitimacy of the principles that legitimate scientific aims is in question.
Regarding the last sort of utopianism -- epistemic utopianism --, truth is not so utopian as Laudan claims. Problem-solving efectiveness is not clearer than partial truth. On the other side, there are basic distinctions in the appraisal of theories that could not be grounded if we do not assume an ability to identify actual constituents of the world on the part of some scientific theories. The way to truth is neither straight nor conclusive but taking into account epistemic values as fertility and explanatory power is neccesary to sustain the realist cause.
For Laudan, axiological choices are on the same footing as the theoretical and methodological ones: all of them may be objectively grounded. The generality of the principles and their naturalistic flavour are the most remarkable merits of Laudan's account but the results are rather meagre. PC may be, in the end, a mere a posteriori justification of changes in axiological direction carried out by the scientific community. The rejection of a demonstrable utopian goal, granting naturalistic assumptions, is completely sound but it has a very limited scope. I am afraid science could not demonstrate much about goals. From the rejection of semantic utopianism we can draw a need for a previous clarification rather than sustantive criticisms and, finally, Laudan's charge of epistemic utopianism is very controversial, as I have tried to show.
Valeriano Iranzo
Valencia, Spain