+ All Categories
Home > Documents > Invention and Induction Laudan, Simon and the Logic of Discovery

Invention and Induction Laudan, Simon and the Logic of Discovery

Date post: 22-Dec-2016
Category:
Upload: robert-mclaughlin
View: 213 times
Download: 1 times
Share this document with a friend
15
Invention and Induction Laudan, Simon and the Logic of Discovery Author(s): Robert McLaughlin Source: Philosophy of Science, Vol. 49, No. 2 (Jun., 1982), pp. 198-211 Published by: The University of Chicago Press on behalf of the Philosophy of Science Association Stable URL: http://www.jstor.org/stable/186918 . Accessed: 04/02/2011 04:30 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=ucpress. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. 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 [email protected]. The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR to digitize, preserve and extend access to Philosophy of Science. http://www.jstor.org
Transcript

Invention and Induction Laudan, Simon and the Logic of DiscoveryAuthor(s): Robert McLaughlinSource: Philosophy of Science, Vol. 49, No. 2 (Jun., 1982), pp. 198-211Published by: The University of Chicago Press on behalf of the Philosophy of Science AssociationStable URL: http://www.jstor.org/stable/186918 .Accessed: 04/02/2011 04:30

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=ucpress. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR todigitize, preserve and extend access to Philosophy of Science.

http://www.jstor.org

INVENTION AND INDUCTION LAUDAN, SIMON AND THE LOGIC OF DISCOVERY*

ROBERT McLAUGHLIN

University of Melbourne and Macquarie University

Although on opposite sides of the logic of discovery debate, Laudan and Si- mon share a thesis of divorce between discovery (invention) and justification (appraisal); but unlike some other authors, they do not base their respective versions of the divorce-thesis on the empirical/logical distinction. Laudan argues that, in contemporary science, invention is irrelevant to appraisal, and that this irrelevance renders epistemically pointless the inventionist program. Simon uses his divorce-thesis to defend his account of invention, which he claims to be non- inductive-so evading the problem of induction. Underlying both authors' po- sitions are inadequate conceptions of inductive inference. Laudan here ignores the role in contemporary science of plausibility arguments, which provide a cru- cial link between invention and appraisal, and thence an epistemic rationale for inventionism. Simon's account of invention does covertly call upon inductive principles from the context of appraisal, and this is what gives his program epistemic import; otherwise he would be vulnerable to Laudan's "no rationale" critique. The tensions in both authors reveal the falsity of the divorce-thesis, and the essential function of induction in both appraisal and invention of hy- potheses.

1. The Divorce Between Discovery and Justification. That the "logic" of scientific discovery, if such there is, differs crucially from the logic of justification is not a new idea, but novel arguments relying on this thesis have been advanced recently by Laudan (1980) and by Simon (1977). Ironically, these authors stand on opposite sides of the debate about the propriety of attempting a philosophical account of discovery-Laudan urging the traditional view (associated with logical empiricism) that philosophy has no particular contribution to make to the

study of discovery, and Simon insisting that there are indeed important and interesting questions here of a philosophical character, apart from the

psychological and other empirical problems that evidently are involved. Yet they share one thesis-namely, that the "logics" or inference pro- cedures involved in discovery and in justification are different and in- dependent. This thesis permits Laudan to argue that there is no rationale for the philosophical study of discovery-in contrast to justification, which is of central interest to epistemology. The same thesis enables Simon to pretend that he can sidestep the problem of induction: justifi-

*Received June 1981; revised November 1981.

Philosophy of Science, 49 (1982) pp. 198-211. Copyright ? 1982 by the Philosophy of Science Association.

198

INVENTION AND INDUCTION

cation involves generalization, predicting and testing, and this procedure is inescapably inductive;' whereas discovery, according to Simon, is not concerned with generalizing or predicting, but simply with detecting pat- terns (1977, p. 331). Pattern-detecting, he claims, is non-inductive, and so the logic of discovery can avoid the notorious epistemic problem of induction.

The thesis that the logic of discovery can be divorced from the logic of justification is false, as I hope to show. His commitment to this di- vorce-thesis exposes Simon to the full thrust of Laudan's critique of the philosophical study of discovery: Simon's account of a non-inductive logic (or set of algorithms) for hypothesis-discovery turns out to be philo- sophically pointless if there is no logical connection between discovery and justification. Closer scrutiny of Simon's proposals reveals that he does tacitly rely on inductive plausibility considerations in his reconstruc- tion of discovery, and this is what makes his analysis intuitively appeal- ing. Indeed, it can be argued much more generally that the divorce-thesis is false; and its falsity renders otiose Laudan's critique of the philosoph- ical study of discovery. With rejection of the divorce-thesis, and thereby with acknowledgment of the close logical connection between discovery and justification, it will emerge that, contra Laudan, there is a philo- sophically interesting logic of discovery, which is the same as part of the logic of justification (i.e., the logical contexts of discovery and justifi- cation may intersect); and, contra Simon, this philosophically interesting logic of discovery is mainly inductive.

At the outset, let me note a quite different view (from those of Laudan and of Simon) concerning the thesis of divorce between discovery and justification. Reichenbach (1938, pp. 6-7)2, Braithwaite (1953, pp. 20-21), Feigl (1970, p. 4) and other logical empiricists, as well as Popper (1959, pp. 31-2), have sought to base the discovery/justification dis- tinction on the categorial distinction between empirical and logical mat- ters. This construal has been echoed very recently by Siegel (1980), in a penetrating contribution to this Journal. On this popular view, discovery

1In this paper I adopt Salmon's (1967) view of the inevitability of induction in scientific inference. Here 'induction' means any correct form of ampliative inference; it is not con- fined to "straight" enumerative induction or extrapolation, but additionally may embrace ampliative arguments employing principles of analogy and simplicity, for example. The problem of induction I take to involve showing that there are any such correct ampliative inferences.

2As Nickles (1980) and Curd (1980) separately have noted, Reichenbach's view of the discovery/justification distinction is equivocal. While the primary basis of his distinction seems to be a categorial dichotomy between logical and empirical accounts, this tends to be conflated in his writing with other distinctions, such as: descriptive vs normative; sub- jective vs objective; before vs after "hitting upon" hypothesis. These ambiguities are par- ticularly unfortunate, since Reichenbach is usually regarded as the primary source of the discovery/justification distinction.

199

ROBERT McLAUGHLIN

is taken to be (or is defined as) an empirical matter, the province of such empirical disciplines as psychology and history; while justification is taken to be (or is defined as) logical and epistemological in character, and is the province of philosophy. It follows that to speak of a logic of discovery is to commit a category mistake-is to conflate the distinct categories of logical and empirical matters. This sort of category mistake has been dubbed 'psychologism' by Popper (1959), for one.

To define the discovery/justification distinction in these categorial terms is merely to make logical reconstructions of discovery guilty by definition of psychologism! Surely the propriety of such reconstruction programs deserves to be judged on less whimsical grounds than that. Elsewhere I have suggested the replacement of 'discovery' and 'justifi- cation' by the terms 'invention' and 'appraisal', in order to avoid such a question-begging construal (McLaughlin, 1982). These latter terms are to be understood as category-neutral: the context of invention embraces all items (logical and empirical) which are conceptually or causally rel- evant to the "hitting-upon" of a hypothesis; and the context of appraisal comprises all items conceptually or causally relevant to the acceptance/ rejection of a hypothesis.3 On this usage, there is no question of attempts at logical reconstruction of invention involving a category mistake, and charges of "psychologism" levelled against such attempts miss their mark; the possibility of a logic of invention cannot be dismissed on such a priori grounds. Hereafter, to avoid confusion, I shall use the terms 'invention' and 'appraisal' where appropriate, instead of 'discovery' and 'justification'; and I shall refer to the program of logically reconstructing the context of invention as 'inventionism'. Objections to, and defences of, inventionism then can be examined on epistemological rather than semantic grounds.

To the credit of each, neither Laudan nor Simon rests his respective divorce-thesis on the empirical/logical category distinction. Each accepts the propriety of speaking of a logic of discovery (invention); but each seeks to divorce this from, and to contrast it with, the logic of justification (appraisal). Thus both authors are saddled with two different "logics". For Simon, the contrast is between the logic of confirmation and 'retro- duction', a term coined by Peirce (1931) and employed more recently by Hanson (1961) and others. For Laudan, the distinction is between modern confirmation logic and enumerative induction. I propose to examine these

3Sequences of psychological events, such as thought-processes, sometimes may involve

reflecting-upon logical items, such as arguments. This would provide a basis for a cor-

respondence between some psychological and logical items (in both invention and ap- praisal). Appropriate correspondence of this sort (where the arguments reflected-upon are

correct/valid ones) would be one necessary condition for the rationality of the psycho- logical activities concerned. See ?4 below.

200

INVENTION AND INDUCTION

attempted separations, which threaten the fertile marriage of invention and appraisal.

2. Inventionism Lacks a Rationale (Laudan). 2.1 Laudan's Argu- ment. The essence of Laudan's (1980) paper is that the logical recon- struction of hypothesis-appraisal has a clear epistemic rationale, but that logical reconstruction of hypothesis-invention, which is divorced from

hypothesis-appraisal, does not. From his divorce-thesis it follows that there is no point in epistemologists, or philosophers generally, studying invention-at least qua philosophers. This was not always so. There was a time, Laudan remarks, especially during the 17th and 18th centuries, when the invention and the appraisal of hypotheses were conceived as closely connected, or even identical, procedures. According to this meth- odology, the same observations and inferences which provided a basis for the generation or invention of a hypothesis typically would serve also to verify or authenticate that hypothesis. In this older program, I take it, the observation of 100 white, and no non-white, swans would both gen- erate the hypothesis that all swans are white and also serve to verify that

hypothesis; enumerative induction or extrapolation being the inference

procedure in both the invention and the appraisal of the hypothesis in this

type of case.4 (Some scientific practice in that period was certainly more

complex than this conception suggests; but Laudan's concern here is with the latter.) This early era of modem science, in which invention and ap- praisal were viewed as intimately related, was characterized, according to Laudan, by an infallibilist epistemology and also by the circumstance that most interesting scientific hypotheses were closely linked with ob- servation-i.e., were "low-order" or "empirical" generalizations. On the methodology prevailing at this stage of science, the logic of invention was, or constituted, the logic of appraisal. Ordinarily this logic was enu- merative induction. Laudan mentions Bacon and Newton, among others, as representative of this older program.

All this changed, according to Laudan, in the early 19th century, with the development of "high-order" hypotheses or theories, linked only in-

directly with observation, and the abandonment of an infallibilist view of scientific knowledge. Under this newer program, the way in which a hypothesis was invented had no bearing on its epistemic status. Enu- merative induction largely gave way to consequentialist procedures of

appraisal (e.g., the H-D method), which involved the examination of the

4This is deliberately simplified. The inductive invention method would require correct initial description and classification of the data, before generalization or amplification comes into play. This is analogous to Simon's (1977) "pattern-discovery" (see ?3.1 be- low), which is, I argue, itself a further inductive procedure.

201

ROBERT McLAUGHLIN

observational consequences of high-order theories. The divorce of inven- tion from appraisal was a direct result of this new conception of science, as exponents whereof Laudan names Herschel and Whewell. It became, and remains, the case that invention has nothing to do with epistemology, and hence the study of invention is philosophically pointless. This, for what it is worth, is Laudan's argument.

2.2 Critique of Laudan. The credibility of Laudan's divorce-thesis re- lies on the over-simple logic of appraisal which he invokes. It is over-

simple even according to his own conception of appraisal, as sketched

by him early in his essay (1980, pp. 173-174). There he acknowledges the existence of a "nether region", which he calls the context of pursuit, lying "[b]etween the context of discovery and the context of ultimate justification." This pursuit phase seems to constitute the initial stage of

hypothesis-appraisal, as I would put it: pursuit involves plausibility judg- ment, in contrast to "ultimate justification", which ordinarily would in- volve testing (via H-D or other consequentialist schemata). The idea of an initial plausibility-assessment stage of appraisal is not new; it was sug- gested by Salmon (1967, p. 114), for example.5 In "Invention and Ap- praisal" I urge that 'enhancement' is preferable to 'pursuit' as the title of this stage. Terminology aside, the important point for the present dis- cussion is that there is more to hypothesis-appraisal than testing, and Laudan himself has acknowledged this. The "something more" is plau- sibility-assessment, which normally precedes testing, and which I now want to argue is intimately linked with invention.

Against this background, and against Laudan, I suggest that, in the newer program of 19th century science, neither the demise of infallibil- ism, nor the shift of interest to high-order theories, rendered invention irrelevant to appraisal. On the contrary: in the newer program, invention and appraisal remain logically connected-but this connection can only be recognized if one adopts a more sophisticated view of the logic of both invention and appraisal than Laudan entertains.

As to his first claim, it seems clear that the end of infallibilism and the

development of a consequentialist logic of appraisal did not require, nor result in, the divorce of invention from appraisal. Enumerative induction can continue to play a role in both contexts, as may be seen from the case of the 100 white swans, now treated in a fallibilist fashion. As be- fore, observations of 100 white swans generate (by enumerative induc-

tion) the hypothesis H, 'All swans are white'; however, on a fallibilist

epistemology, these observations do not suffice to verify (infallibly) this H. But the observations are not simply irrelevant to the appraisal of H,

5Essentially similar trichotomies have been mentioned recently by a number of authors, including Kordig (1978), Nickles (1980), Curd (1980).

202

INVENTION AND INDUCTION

either. Many contemporary epistemologists-certainly Salmon, and per- haps even Laudan in a kindly mood-would allow that the observations of 100 white swans provide some "initial" or "prior-to-test" support for H; they give H a degree of plausibility.6 If this initial plausibility-value is high enough (taking account of the utilities in the research situation), the appraisal of H will proceed into the testing stage; otherwise H will be discarded as unworthy of further appraisal. In the case of the swans, the newer program would then prescribe a consequentialist schema for testing, requiring the inference from H to observational predictions, such as that the next n (n = 1 . . . oo) swans observed will be white.

This reveals the continuing dual role of an inductive argument in both invention and appraisal, even on a fallibilist epistemology. Despite Lau- dan, it seems that the decline of an infallibilist view of appraisal did not of itself require the divorce of invention from appraisal. There is no rea- son why a given argument may not function as what I call an advance- ment argument in the context of invention, and also continue to function as an enhancement argument in (the initial or "pursuit" stage of) the context of appraisal. Advancement/enhancement arguments are the keys to the logic of invention. Typically they are inductive arguments-but they are by no means confined to inductions by enumeration or extrap- olation.

To develop this idea, I proceed to my second criticism of Laudan: the shift of interest to high-order theories in the 19th century did not, as he claims it did, render invention irrelevant to appraisal. What it did was to open the way for a recognition that more sophisticated inference pro- cedures (than enumerative induction) may function in both contexts. (In practice, more sophisticated inferences than enumerative induction were discernible in the older program; presumably Laudan would argue that more explicit metascientific attention to these procedures characterized the transition to the newer program.) Salmon (1967), in his treatment of plausibility considerations which function in the prior-to-test stage of appraisal, mentions as examples of these such principles as analogy, sym- metry and other forms of simplicity; in his Bayesian schema of appraisal (1967, p. 115ff.), their essential role is to estimate a prior probability

6This plausibility value can play two important roles in the context of appraisal (cf. Curd 1980, p. 203). First, as noted, it permits a decision-theoretic judgment on whether to proceed with further appraisal, including testing. Second, if one adopts a Bayesian ap- praisal-schema (Salmon 1967, p. 115ff.), the plausibility estimate yields a prior probability for insertion therein. By the way, Kordig (1978, p. 113) has got this point quite back-to- front. He says, "Plausibility arguments might hinge on factors independent ... of prob- ability." In fact, as Salmon (1967, p. 126ff.) insists, it is prior probability estimates that hinge on plausibility arguments! This is because there often are no frequencies available to determine prior probabilities; in such cases, the prior probability estimate must be based upon other plausibility considerations, such as simplicity and analogy.

203

ROBERT McLAUGHLIN

for a hypothesis, in the absence of relevant frequencies (cf. footnote 6). They thus contribute significantly to the determination of a posterior (to test) probability for the hypothesis-i.e., they are indispensable com- ponents of the context of appraisal. Similar principles were discussed by Hanson (1961); and Laudan himself would perhaps concede the role of such plausibility considerations in his stage of pursuit. What I want now to suggest is that arguments employing such considerations and principles may function equally well in the context of invention. That is, the same analogy argument, symmetry argument, etc. may serve as both an ad- vancement (invention) and an enhancement (initial appraisal) argument for a hypothesis. Which title we give it will depend solely on extra-logical factors-such as whether it is taken account of by the scientist before or after or both before and after he/she hits upon the hypothesis.

Examples of plausibility (enhancement or pursuit) arguments are not hard to find; nor is it difficult to recognize that, in many cases, such arguments may play dual roles-may function equally well as advance- ment arguments in the context of invention; and conversely. Hanson (1961) gave several examples of analogy and symmetry based arguments which functioned in discovery: e.g., analogical reasoning by Kepler to the conclusion that Jupiter's orbit would be non-circular, and by Dirac to a Lorentz-invariant electron theory. Ironically, Hanson noted as an embarrassment for his divorce-thesis that such invention arguments could function also in the context of appraisal. This, of course, is a good reason to discard the divorce-thesis (see my further discussion of Hanson below, ?3.1). Salmon (1967, p. 127), in the course of noting that Hanson's dis- covery arguments in fact function as plausibility arguments, mentions a quite ancient simplicity based advancement/enhancement argument for the inverse-square law of gravitation, relying on principles of isotropy and homogeneity of space. Einstein, in his revealing "Autobiographical Notes" (1959), gives great weight to what he terms the "naturalness" or "logical simplicity" of a theory, not only as affecting its plausibility, but also as a consideration influencing the direction of inventive thinking. As he remarks: "This point of view, an exact formulation of which meets with great difficulties, has played an important role in the selection and evaluation of theories since time immemorial." (1959, p. 23)

In molecular biology, to take another branch of science, a rich source of examples of dual invention/appraisal arguments is Watson's The Dou- ble Helix (1968). For instance, one of his arguments for the helical struc- ture of the DNA molecule relies on its analogy with the chemically sim- ilar TMV (tobacco mosaic virus) molecule, which is helical. He derives the same conclusion from a simplicity argument employing the principle that the simplest form for any regular polymeric molecule is a helix. These and similar advancement arguments for the hypothesis about the structure of DNA serve equally well as enhancement arguments for that

204

INVENTION AND INDUCTION

hypothesis, giving it a degree of initial plausibility, as Watson's account makes clear. More generally throughout the biological, and also the be- havioral, sciences, impressive numbers of advancement/enhancement arguments rely upon analogies, such as obtain between chemically or structurally similar substances or organisms, including those between various animals or human beings. In his encyclopedic The Act of Creation (1964), Koestler gives a huge array of examples of "bisociative" think- ing, which he takes to be involved in creativity, in science as well as in other inventive spheres; it could be argued that a great deal of such think- ing is analogical in character, and has a dual advancement/enhancement role. For my present purposes, the point seems clear enough: in a sig- nificant proportion of cases of scientific invention, arguments are in- volved (explicitly or implicitly) in the advancement of a hypothesis (cf. footnote 3); and where this happens, the same argument normally will serve also to enhance (make plausible) the hypothesis. In short: a sci- entist's reasons for hitting upon a hypothesis often will be reasons for him/her to be confident of its truth, or to assign to it a particular prior probability. This point, once made, seems sufficiently evident not to need laboring.

I have been at pains to show that the contexts of invention and appraisal often may intersect in order to exhibit the philosophical rationale of in- ventionism, which Laudan has denied. In his older program, enumerative induction played the dual role of advancing and enhancing (verifying, in that program) low-order hypotheses. In the newer or contemporary pro- gram, I maintain that more sophisticated inductive principles (e.g., anal-

ogy, simplicity) likewise play the dual role of advancing and enhancing (making plausible, in this program) high-order hypotheses. Contrary to Laudan's view, the present situation is not very different in principle from the older one. The rationale of inventionism, which he admits applied then, also applies now. Philosophers of science rightly are concerned with the epistemic status of scientific hypotheses, and I suggest that insistence on a logical divorce of appraisal from invention wrongly truncates this

epistemic task. Typically, a rationally invented hypothesis enters its con- text of appraisal bearing a plausibility-value from its context of invention. Adequate logical reconstruction of its appraisal demands attention to the invention (advancement) arguments which provide it with this plausibil- ity-value (in the role of enhancement arguments). Thus to ignore, with Laudan, the logic of its invention would be to overlook an important element in the logic of its appraisal, and so to do a defective job of episte- mological reconstruction.

3. Invention is Non-Inductive (Simon). Simon, like Laudan, insists that the logic of invention is independent of the logic of appraisal. He thinks this is the way to evade the problem of induction, which he imag-

205

ROBERT McLAUGHLIN

ines to have discouraged others' attempts to reconstruct a logic of inven- tion (1977, p. 329). What he fails to recognize is that his avowal of this logical divorce exposes him directly to Laudan's type of critique-that is, to the charge that his inventionist program is pointless-a critique that I have avoided above by showing that inventive inferences typically are both inductive and identical with certain arguments in the context of ap- praisal. I shall sketch the essentials of Simon's account, which-apart from its mistaken divorce-thesis-has much to recommend it.

3.1 Simon's Argument. Simon (1977) begins by contrasting Popper's (1959) and Hanson's (1961) views on the possibility of a logic of dis- covery (invention). I have already (?1) touched on Popper's anti-inven- tionist stance, as expressed in his question-begging charge of "psychol- ogism" (1959, p. 31). Here it is noteworthy that Hanson, perhaps the most famous recent exponent of inventionism, committed himself to the same false thesis as (later) did Laudan and Simon: the thesis, namely, that discovery (invention) and justification (appraisal) involve different "logics". For Hanson, the logic of discovery-which, following Peirce (1931) he called retroduction-was to be viewed as quite distinct from the logic of justification (e.g., the H-D schema). This commitment to the logical separation of invention and appraisal nevertheless troubled Hanson somewhat (see ?2.2 above): for it seemed that, while retroduction was supposed to function exclusively in invention (a claim which Salmon (1967, pp. 113-4) refuted), there were other types of inferences which appeared to operate in both invention and appraisal. As Hanson remarked: "Analogical reasons, and those based on symmetry ... are reasons both for proposing that H will be of a certain type and for accepting H." (1961, p. 27) I have already noted, as grist to my mill, this dual role of analogical and symmetry based arguments. Simon, in any case, ignores this embarrassment for the divorce-thesis, and does not challenge the basic idea of Hanson, Peirce and others that invention and appraisal have different logics; he even retains the Peirce-Hanson term 'retroduction' for the envisaged logic of invention. But although he takes himself also to be defending retroduction, Simon in fact proposes a quite different view of the logic of invention from that of Hanson.

Simon's key contention is that what he calls a 'law discovery process' is not inductive. In his account, "pattern discovery"-which serves as a model for "law discovery"-is a matter of recoding parsimoniously a finite portion of a data-sequence. In simple cases, at least, automatic algorithms are available for pattern-search, amounting essentially to se- lective trial-and-error strategies (see his 1966a and 1966b for more de- tailed treatment of this approach). According to Simon, this recoding process does not of itself involve any extrapolation or generalization-it

206

INVENTION AND INDUCTION

is not inductive. The problem of induction "arises only if we wish to predict and test whether this same pattern will continue to govern the sequence when it is extrapolated." (1977, p. 331) Simon proceeds to offer the following two definitions.

"A law-discovery process is a process for recoding, in parsimonious fashion, sets of empirical data."

"A normative theory of scientific discovery is a set of criteria for evaluating law-discovery processes."

(1977, p. 331)

The inventionist enterprise then is seen by Simon as the construction of a normative theory of scientific discovery.

3.2 Critique of Simon. What is "discovered" by means of Simon's "law-discovery process" is not, strictly speaking, a law at all; for a law, as the term is ordinarily understood, is a general proposition-inferred, if it is inferred at all, by extrapolation and/or some more sophisticated ampliative (inductive) procedure, from a finite set of data.7 The product of Simon's recoding process is simply a more economical re-description of the same data. His idiosyncratic use of 'law' blurs this point. What makes his "law-discovery process" interesting-what gives it an epis- temic point or rationale-is that it yields a pattern which may be gen- eralized into, or from which may be induced, a law in the more usual sense of the term, or a hypothesis as I should prefer to call it. 'Law- discovery', in common parlance, is a process combining pattern-discov- ery with generalization or some other form of ampliative inference. It is thus straightforwardly inductive. Simon, has, in effect, shifted the de- marcation-point between invention and appraisal. This point is usually defined by the "hitting upon" of the general hypothesis (cf. ?1 above): the generalizing or other ampliative process is part of invention. Simon has moved the demarcation-point back into the context of invention, as this usually is conceived, and has defined (his) invention as terminating with pattern-discovery-i.e., as preceding extrapolation or other amplia- tive inference. But the problem of induction cannot be avoided merely by definition, as I shall now show: pattern-discovery itself is an inductive procedure.

It is not just any pattern that is of interest to scientists, but rather those patterns which, when generalized or amplified, will constitute plausible

7In a full-scale account of inductive invention, the data may include anomalies or other scientific problems (see, e.g., Laudan (1977), Nickles (1978), Burian (1980)); the more sophisticated inductive procedures may include analogy or simplicity based advancement arguments (cf. ?2.2).

207

ROBERT McLAUGHLIN

hypotheses. This is tacitly recognized by Simon in his inclusion of the qualifying phrase 'in parsimonious fashion' in his definition of a law- discovery process. Parsimony is recognized by him (and by most scien- tists) as a desideratum in plausible hypotheses. Now a normative theory or set of criteria is goal-oriented, as Simon himself remarks (1977, p. 328). The goal of a law-discovery process surely is the discovery of hy- potheses which will turn out to be highly-confirmed when tested (or to have a high posterior probability-cf. Salmon (1967))-that is, plausible hypotheses. So a normative theory of scientific discovery must be a set of criteria for evaluating the efficacy of various law-discovery processes with respect to the goal of discovering plausible hypotheses. Some pat- terns yielded by such processes will be better than others, according to criteria specified by the normative theory. That is, some patterns will be more parsimonious, symmetric, elegant and so forth than others, since these tend to be characteristics which, in the context of appraisal, render

hypotheses more plausible. As soon as the notion of better or worse patterns enters the story, so

does induction. 'Better or worse' means things like 'more, or less, par- simonious', and this in turn means 'more, or less, plausible when gen- eralized or amplified'; for parsimony is a plausibility-consideration in the context of appraisal, warranted by inductive (e.g., analogical) inference from past successes of other parsimonious hypotheses. There is no other reason for preferring more parsimonious encodings of data (if we are concerned with a discovery, and not merely a convenient data-handling, procedure). In choosing constraints like parsimony to apply to his law-

discovery processes (or invention algorithms), Simon is tacitly invoking inductive plausibility principles from the context of appraisal; principles such as that parsimonious hypotheses tend to be successful, or that hy- potheses analogous in some significant respect (e.g., formal symmetry) to already well-confirmed ones have a good prior probability. Inferences

employing such principles are instances of what I have called advance- ment/enhancement arguments; they reside in the intersection of the log- ical contexts of invention and appraisal.

It is only by covertly employing plausibility principles that Simon avoids Laudan's line of criticism. For the divorce of invention from ap- praisal, which Simon purports to have achieved, would indeed render

philosophically pointless his program of reconstructing a logic of inven- tion. Just as Laudan has urged, such a divorce would rob Simon's in- ventionist enterprise of any epistemic rationale. Thus one might ask Si- mon: Why choose parsimonious patterns? Why not seek complex, cumbersome, ugly, asymmetric patterns? If appraisal considerations were indeed quite separate from invention, he could offer no answer to these

challenges. In short, if he had succeeded in divorcing invention from

208

INVENTION AND INDUCTION

appraisal, he would have defeated his own purpose. I have argued that he does not succeed in this divorce, and the philosophical merit of his proposals results from this failure: he does invoke appraisal considera- tions in his invention strategies, and this is what makes his procedures both interesting and inductive!

4. Concluding Reflections. It is the concept of plausibility which pro- vides the epistemic link between invention and appraisal, and which is central to the rational status of each. A rationally-invented hypothesis enters the context of appraisal bearing an initial plausibility-value from the context of invention; and rational appraisal of the hypothesis must take account of this plausibility-value. Two linked objections immediately come to mind. First, why should invention be viewed as rational at all? Second, why should plausibility play a role in final appraisal?

Nowhere have I maintained the wild thesis that all inventings are ra- tional. Clearly there are cases where the immediate genesis of a scientist's hitting upon a new idea has little to do with rational reflection-that is, with thought-sequences which correspond to good (advancement) argu- ments. One need not be goaded with such famous cases as Newton's apple or Kekule's dream of snakes to concede this.8 Yet I do want to insist that an interestingly high proportion of cases of scientific invention are rational, in the sense I have sketched (cf. footnote 3). Typically, in- ventings of hypotheses occur, not in a cognitive vacuum, but rather in a setting where a scientist has been grappling intellectually with a prob- lem. To suggest that all such intellection is either irrational or irrelevant to the eventual invention of the hypothesis is too far-fetched to be cred- ible. Simply put, my theme is that science is rational in character, which is not to exclude non-rational factors; and that the same or similar rational procedures typically function in invention and in appraisal, viewed as intimately linked phases of the one epistemic enterprise.

As to the role of plausibility in appraisal, it seems clear that to try to divorce the two would be to make nonsense of the concept of plausibility. If plausibility has nothing to do with final appraisal-that is, acceptance or rejection of a hypothesis H-then why prefer plausible to implausible H's? Why test some H's and discard others without testing? Why seek

8Two features of such "Eureka!" cases are noteworthy, nevertheless. First, while the triggering or precipitating event may have nothing to do with rational reflection, it remains the case that having a body of scientific knowledge, or at least belief, together with a

history of intellectual grappling with the problem, typically is a necessary condition for the inventing to occur. Second, there seems good reason to think that much unconscious reasoning is involved in inventing (cf. Koestler 1964)-especially in the "bolt from the blue" cases; such unconscious processes may be nonetheless rational, in the sense that they correspond to good argument sequences.

209

ROBERT McLAUGHLIN

plausible rather than implausible H's? The collective answer is that plau- sibility is a measure of inductive support for a hypothesis, just as positive test results are a measure of inductive support for it. The two factors cannot licitly be divorced, nor is either properly dispensable in hypothesis appraisal. As Salmon emphasises, plausibility arguments ". . . are not only admissible into the logic of justification; they are an indispensable part of it." (1967, p. 118) The same point is stressed by Kordig, when he says: ". . there is no fundamental difference between reasons for plausibility and reasons for acceptability." (1978, p. 116) Indeed, any attempt to divorce plausibility from appraisal would be vulnerable to a Laudan-style critique: such a divorce would deprive the reconstruction of the "logic of pursuit" (or plausibility, or enhancement) of any epis- temic rationale. Few epistemologists-apart, perhaps, from Popper-would be comfortable with this.

Due attention to plausibility arguments reveals Laudan's (1980) ac- count of the character of inductive inference in science as too limited. In that essay, at least, he appears to confine such inference to enumerative or extrapolative induction (older program) and consequentialist induction from predicted observations to high-order or deep-structure hypotheses and theories (newer program). This is to ignore a large and interesting class of inductive inferences, of which analogical arguments, and ones relying on various kinds of simplicity considerations (e.g., parsimony, symmetry, isotropy) are good examples. It is these kinds of inductive inferences which play a key role in both invention and (initial) appraisal of high-order hypotheses, in the form of dual advancement/enhancement (plausibility) arguments. Simon (1977), too, has overlooked them. Yet they are indispensable to his otherwise laudable inventionist enterprise: they give it a rationale. Their dual function in what Laudan calls the newer program discloses that in contemporary science, as in fact it was in the older program of the 17th and 18th centuries, induction in various forms is essential to both invention and appraisal.

REFERENCES

Braithwaite, R. (1953), Scientific Explanation. London: Cambridge University Press. Burian, R. (1980), "Why Philosophers Should Not Despair of Understanding Scientific

Discovery" in Nickles, T. (ed.), Scientific Discovery, Logic, and Rationality: 317-336. Dordrecht: Reidel.

Curd, M. (1980), "The Logic of Discovery: An Analysis of Three Approaches" in Nic- kles, T. (ed.), Scientific Discovery, Logic, and Rationality: 201-219. Dordrecht: Reidel.

Einstein, A. (1959), "Autobiographical Notes" in Schilpp, P. (ed.), Albert Einstein: Phi- losopher-Scientist: 1-95. New York: Harper.

Feigl, H. (1970), "The 'Orthodox' View of Theories: Remarks in Defense as well as Critique" in Radner, M. & Winokur, S. (eds.), Analyses of Theories and Methods of Physics and Psychology (Minnesota Studies in the Philosophy of Science, Vol. IV): 3-16. Minneapolis: University of Minnesota Press.

210

INVENTION AND INDUCTION 211

Hanson, N. (1961), "Is There a Logic of Scientific Discovery?", in Feigl, H. & Maxwell, G. (eds.), Current Issues in the Philosophy of Science: 20-35. New York: Holt, Rinehart and Winston.

Koestler, A. (1964), The Act of Creation. London: Hutchinson. Kordig, C. (1978), "Discovery and Justification", Philosophy of Science 45: 110-117. Laudan, L. (1977), Progress and Its Problems. Berkeley: University of California Press. Laudan, L. (1980), "Why Was the Logic of Discovery Abandoned?" in Nickles, T. (ed.),

Scientific Discovery, Logic, and Rationality: 173-183. Dordrecht: Reidel. McLaughlin, R. (1982), "Invention and Appraisal" in McLaughlin, R. (ed.), What?

Where? When? Why?: 69-100. Dordrecht: Reidel. Nickles, T. (1978), "Scientific Problems and Constraints" in Hacking, I. & Asquith, P.

(eds.), PSA 1978, Vol. 1: 134-148. East Lansing, Mich.: Philosophy of Science Association.

Nickles, T. (1980), "Scientific Discovery and the Future of Philosophy of Science" in Nickles, T. (ed.), Scientific Discovery, Logic, and Rationality: 1-59. Dordrecht: Reidel.

Peirce, C. (1931), Collected Papers. Cambridge, Mass.: Harvard University Press. Popper, K. (1959), The Logic of Scientific Discovery. London: Hutchinson. Reichenbach, H. (1938), Experience and Prediction. Chicago: University of Chicago

Press. Salmon, W. (1967), The Foundations of Scientific Inference. Pittsburgh: University of

Pittsburgh Press. Siegel, H. (1980), "Justification, Discovery and the Naturalizing of Epistemology", Phi-

losophy of Science 47: 297-321. Simon, H. (1966a), "Thinking by Computers", in Colodny, R. (ed.), Mind and Cosmos:

3-21. Pittsburgh: University of Pittsburgh Press. Simon, H. (1966b), "Scientific Discovery and the Psychology of Problem Solving" in

Colodny, R. (ed.), Mind and Cosmos: 22-40. Pittsburgh: University of Pittsburgh Press.

Simon, H. (1977), "Does Scientific Discovery Have a Logic?" in Simon, H., Models of Discovery: 326-337. Boston: Reidel. Reprinted from Philosophy of Science 40: 471-480.

Watson, J. (1968), The Double Helix. Harmondsworth, Middlesex: Penguin.


Recommended