A Kuhnian Defence of Inference to the Best Explanation

of 10
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Documents

Published:

Views: 4 | Pages: 10

Extension: PDF | Download: 0

Share
Description
Description:
Tags
Transcript
  A Kuhnian defence of inference to the best explanation David Walker Department of Philosophy, University of St. Andrews, Edgecliffe, The Scores, St. Andrews, Fife KY16 9AL, UK  a r t i c l e i n f o  Article history: Received 16 August 2010Received in revised form 25 May 2011Available online 28 September 2011 Keywords: ExemplarExplanatory virtueIBEKuhnLoveliness a b s t r a c t According to inference to the best explanation (IBE), scientists infer the loveliest of competing hypothe-ses, ‘loveliness’ being explanatory virtue. This generates two key objections: that loveliness is too subjec-tive to guide inference, and that it is no guide to truth. I defend IBE using Thomas Kuhn’s notion of  exemplars : the scientific theories, or applications thereof, that define Kuhnian normal science and facili-tate puzzle-solving. I claim that scientists infer the explanatory puzzle-solution that best meets the stan-dard set by the relevant  exemplar of loveliness . Exemplars are the subject of consensus, eliminatingsubjectivity; divorced from Kuhnian relativism, they give loveliness the context-sensitivity required tobe truth-tropic. The resulting account, ‘Kuhnian IBE’, is independently plausible and offers a partial rap-prochement between IBE and Kuhn’s account of science.   2011 Elsevier Ltd. All rights reserved. When citing this paper, please use the full journal title  Studies in History and Philosophy of Science 1. Introduction It is generallyagreed that inferenceto the bestexplanation (IBE)is widely used in science. According to IBE, we infer what would, if true, be the best explanation of the evidence. 1 Scientists formulatea pool of competing potential explanations, identify the best, and in-fer on those grounds that it is the actual explanation of the evidence(or  an  actual explanation—some evidence may be explained in sev-eral compatible ways). But which is the  best   potential explanation in any pool? PeterLipton, whose account of IBE (Lipton, 2004) is definitive, arguesthat the best explanation is the  loveliest   explanation, ‘loveliness’being explanatory virtue: ‘‘the explanation that would, if correct,be the most explanatory or provide the most understanding [is]the ‘loveliest’ explanation’’ (Lipton, 2004, p. 59). As Lipton (2004, p. 60) notes, a good account of inductive inference must tell uson what basis we judge one proposed conclusion likelier thananother, and IBE claims to do this in terms of   explanatory  consider-ations. Loveliness is thus essential to IBE; failure to take lovelinessseriously is failure to acknowledge IBE’s identity as an account of induction. Consequently, IBE is correctly defined as inference tothe loveliest potential explanation (hereafter, I take ‘IBE’ to be syn-onymous with this definition). 2 It should be emphasised at once that ‘loveliness’ is  not   a bywordfor aesthetic value. If one endorses IBE one does not thereby en-dorse the controversial thesis that the most beautiful hypothesesare likeliest to be true (whatever that amounts to). As just noted,loveliness is a matter of   explanatory virtue ; what makes forloveliness is just what makes for understanding (of course,lovely-making factors may also be aesthetically appealing).Consider the ‘dormitive virtue’ explanation of opium’s causingdrowsiness. This explanation provides almost no understanding:‘opium has a dormitive virtue’ says little more than ‘opium causes(has the power to cause) drowsiness’, and it is precisely this powerwe seek to explain. Hence the explanation has few explanatoryvirtues, and none to any great extent—it is only minimally lovely.We may infer it, since  some  understanding is given (it does morethan paraphrase the evidence, avoids ad hoc clauses, and so on),but it is clear we can do better. Empirical investigation of opium’s‘dormitive virtue’ yields more informative hypotheses, enabling usto move away from this banal explanation towards a thoroughunderstanding of the phenomenon. If true, these later hypotheses 0039-3681/$ - see front matter    2011 Elsevier Ltd. All rights reserved.doi:10.1016/j.shpsa.2011.09.007 E-mail address:  djw21@st-andrews.ac.uk 1 In common with most discussions of IBE, I construe explanation factively: for something to be an explanation, it must be true. 2 More strictly, IBE is the inference that the loveliest of a pool of competing potential explanations is an actual explanation,  where the loveliest is lovely enough to be inferred andsufficiently lovelier than its competitors . If the italicised condition is not met, inference may be postponed. Studies in History and Philosophy of Science 43 (2012) 64–73 Contents lists available at SciVerse ScienceDirect Studies in History and Philosophy of Science journal homepage: www.elsevier.com/locate/shpsa  will display greater explanatory virtue—they will be lovelier. Inthis context then, ‘loveliness’ is merely shorthand for ‘degree of explanatory virtue’.In this paper, I offer neither an analysis of explanatory virtuenor an account of any candidate virtue, let alone an exhaustive list.Rather, my aim is to defend IBE against two crucial objections, viz.that loveliness is too subjective to guide inference,and that it is nota guide to truth (an account of loveliness does emerge, but it isbroadly functional in character).Perhaps surprisingly, my main tool in answering these objec-tions is Thomas Kuhn’s notion of   exemplars , the scientific theories,or applications thereof, that define periods of Kuhnian normalscience. In Section 2 I outline Kuhn’s account of science, focussingon the role of exemplars. In Section 3 I present the two objectionsto IBE—the subjectivity objection (SO) and the truth objection(TO)—in more detail and bring out their significance. In Section 4I answer SO by claiming that exemplars instruct scientists inhow to solve problems by providing standards of loveliness againstwhich to assess potential puzzle-solutions. These standards areshared by all relevant scientists; hence loveliness is not subjectiveas SO claims. On my account, loveliness  is  relative to puzzle-solving context,but this context-sensitivity is to be expected, giventhe ways the various sciences approach their problems. It is alsobeneficial, since it promotes problem-solving.In Section 5 I answer TO. Kuhn states that paradigms aretailored for effective puzzle-solving and that, across paradigms,science makes puzzle-solving progress. I argue that epistemologi-cal reliabilism explains these phenomena: science progressesbecause its puzzle-solving method gets better at tracking the truth.Since exemplars are crucial to that method, the standards of loveliness they generate are truth-tropic.This response to TO makes clear that my approach to Kuhn ispiecemeal. I claim that we may divorce Kuhn’s historically-in-formed insights about the structure of science from his philosoph-ically-motivated relativism. Rejecting the latter and keeping theformer, we may illuminate both IBE and Kuhnian science. The pro- ject is begun with exemplars, but in Sections 6 and 7 I explore theextent to which IBE is compatible with other key aspects of Kuh-nian science. Having shown ‘Kuhnian IBE’ to be plausible, I con-clude by noting the modest rapprochement between IBE andKuhnian science thereby achieved. 2. Kuhn’s account of science Kuhn (1996) identifies a pattern in the histories of maturesciences: extended periods of normal science governed by aparadigm or tradition of scientific work, punctuated by occa-sional scientific revolutions in which old paradigms are replacedby new ones. Crucial to this pattern is the theoretical  exemplar  (Kuhn, 1996, pp. 187–198). According to Kuhn, in any period of normal science, exemplars are ‘‘the concrete problem-solutionsthat students encounter from the start of their scientific educa-tion [and] at least some of the technical problem-solutionsfound in the periodical literature that scientists encounter dur-ing their post-educational research careers and that also showthem by example how their job is to be done’’ (Kuhn, 1996,p. 187). For Kuhn, that job is puzzle-solving, an enterprise hedistinguishes from problem-solving in a way that does notconcern us here (reasons for ignoring the distinction are givenin Section 5; Kuhn confuses his position by using the terms‘problem’ and ‘puzzle’ interchangeably, as the above quotationattests). Exemplars earn their name because ‘‘scientists solvepuzzles by modelling them on previous puzzle-solutions’’ (Kuhn,1996, p. 189).Exemplars are crucial to productive normal scientific work, butthis is not their only function. Exemplars are at the root of para-digms’ ability to define—both conceptually and in the sense of demarcation—the subject matter of a particular science during aperiod of normal science; further, they tell the relevant scientificcommunity how to investigate it. This is because they are the focusof the consensus that constitutes normal science. Exemplars revealthe all-important similarities between hitherto recalcitrant phe-nomena and those already understood within the paradigm. Expo-sure to exemplars (typically received as a student) thus habituatesscientists to see certain extant and emerging puzzles as relevant,certain ways of solving them as appropriate, and certain solutionsas better than others (Kuhn, 1996, pp. 37–40, 45–47). During nor-mal science, scientific communities accept exemplars unreflective-ly. On Kuhn’s account, it is this widespread dogmatism that allowsproductive puzzle-solving to take place; normal scientists do notconstantly question the basis of the science in which they work(Kuhn, 1996, p. 164).As these brief remarks indicate, Kuhn is unclear on whetherexemplars are correctly seen as applications of theories, i.e. specificpuzzle-solutions, or theories themselves, i.e., puzzle-solving tools.puzzle-solving tools. The quotations above, for instance, suggesttheformer,butsomefunctionsKuhnascribestoexemplars—thedef-initionofscientifictermsandeducationwithrespecttosymbolism,for example (Kuhn, 1996, pp. 188–191)—are better served by theo-ries. Further, if exemplars are to help explain the appearance anddisappearance of the consensus they generate in accordance withKuhn’s historicalclaims, thenthey are better seen as theories, sincetheory-change isthekeyfeatureofscientificrevolutions.IreturntothisambiguityanditsrelevancetothepresentprojectinSection4.1.Kuhn’s account of science is much more complex than this, butmore detail here is unnecessary. In Sections 4 and 5 I use exem-plars to defend IBE against SO and TO. I now take a closer look atthese objections. 3. The objections  3.1. The subjectivity objection The subjectivity objection (SO) is this: SO : loveliness is too subjective to be a guide to inference. Con-ceivably, different groups of scientists may fail to converge on asingle explanation when asked to select the loveliest from agiven range of competitors; thus inference will be impossible. 3 The charge is plausible, for at least two reasons. Firstly, despitewhat was said above, ‘loveliness’ has connotations of aestheticpreference, and good inductive inference is not a matter of taste.Secondly, it is notoriously difficult to determine what the explana-tory virtues are and how they are weighted in cases of conflict.Thus SO identifies a deep-seated worry about IBE: I find one expla-nation loveliest, you favour another, and we may never reachagreement.Defending IBE, one might appeal to its descriptive merits. Forexample, inductive inference can be audience-relative: partieswho evaluate evidence differently or have different backgroundbeliefs may legitimately make different inferences about the sameevidence. Loveliness’ flexibility allows IBE to accommodate this(Lipton, 2004, p. 143). That is right, but SO worries about subjectiv-ity, which audience-relativity does not exhaust. The allegation is 3 Lipton (2004, p. 70) calls SO ‘Hungerford’s objection’, after Margaret Wolfe Hungerford, whose novel  Molly Bawn  (1878) contains the line ‘beauty is in the eye of the beholder’.I reject his terminology since the reference to beauty is misleading. D. Walker/Studies in History and Philosophy of Science 43 (2012) 64–73  65  that scientists would disagree about loveliness in much the sameway as music enthusiasts disagree about their favourite tune: theymay talk all night trying to convert each other, yet in the morningstill prefer the melody that satisfies their peculiar tastes. If loveli-ness is radically subjective in principle, then IBE cannot be a cor-rect account of induction, either descriptively (loveliness cannotground a correct description of an activity displaying widespreadconsensus) or normatively (loveliness can only impede good infer-ence). A proper answer to SO must be descriptively plausible andgive loveliness normative support by showing that, whatever formit takes, it will not be too subjective to guide inference. I give suchan answer in Section 4.  3.2. The truth objection The truth objection (TO) is this:  TO : loveliness is not indicative of truth. IBE’s claim that, inchoosing the loveliest potential explanation, scientists therebychoose a true explanation, is unwarranted. 4 More generally,for the loveliest explanations of all phenomena to be true wewould have to live in the loveliest of all possible worlds, andwe have no reason to think we do. 5 Again, the objection has great intuitive force: suppose scientistsgenerally agree on the loveliest explanation (answering SO); whyshould their preference tell them anything about its truth-value?The lack of any obvious connection between loveliness and truthis, for many, the biggest problem with IBE, but IBE’s supportersmight think the objection unfair. As Hume showed, the searchfor independent reasons to think that  any  inductive method willgenerate true conclusions is futile; arguably, TO is just Hume’sproblem reformulated for IBE (Lipton, 2004, pp. 144–147). It stillraises a serious challenge—inductive success is mysterious and de-mands explanation—but an equivalent challenge awaits all IBE’scompetitors.This strategy fails, for two reasons. Firstly, although TO shouldnot ask for  independent   reasons to think that loveliness indicatestruth, it may still ask for IBE’s connection with inductive successto be illuminated. Justifications of inductivemethods may be circu-lar and unpersuasive to sceptics, but may explain and clarify thosemethods to those already disposed to use them (Psillos, 1999, pp.86–89). Even a circular argument for a reliable link between love-liness and truth would dramatically strengthen the case for IBE.Secondly, TO is in fact more serious than Hume’s problem. Onedoes not need to be an inductive sceptic to think that the link be-tween loveliness and truth needs clarificationin a way that the linkbetween a generalisation and its instances does not.To see this, consider reliabilism in epistemology. Reliabilists re-solve Hume’s problem by saying that what matters for knowledgeis that our beliefs are, as a matter of fact, formed by an inductivemethod that reliably delivers truth. We need know nothing aboutthat method or its reliability since justification does not dependon having independent reason to trust our inductive practices.Modally speaking, in any law-governed world our inductive meth-ods are reliable, and in the set of all possible worlds, the number of law-governed worlds is large (if the laws of nature are necessary,then all possible worlds are law-governed). It is a priori plausiblethat we live in one of these worlds; thus, plausibly, our inductivemethods are reliable. Reliabilists hold that our inductive beliefsare justified not because we know that our methods conduce totruth, but because those methods (whatever they are) form beliefsin (what is very likely) a law-governed world.Thus reliabilism answers Hume’s problem; but reliabilism doesnot thus answer TO. TO accuses IBE of placing us in the loveliest of all possible worlds, and that world is unique—it is  not   a priori plau-sible that we live  there . The defender of IBE must show that, of allthe lovely law-governed possible worlds, the actual world has theloveliest laws. Thus TO poses a challenge to IBE that goes beyondHume’s challenge to rival accounts of induction.To answer TO, the IBE supporter must argue that whenever sci-entists infer the loveliest explanation, they thereby tend to inferthe truth (a necessary connection between loveliness and truth isnot required—IBE is an account of   inductive  inference). As withSO, only a normative answer will do. I give such an answer in Sec-tion 5. Presently, I defend IBE against SO. 4. A Kuhnian response to SO SO claims that loveliness is too subjective to guide inference. Aproper answer must show that judgments of loveliness are nothighly contingent and that scientists will typically converge on asingle explanation when selecting the loveliest from a range of competitors.Kuhnian exemplars have an important feature with respect toSO: within the relevant community, they are the source of al-most all scientific understanding, as described in Section 2. Fur-ther, any puzzle-solving success achieved under exemplars’guidance (and this should be considerable, given that paradigmsprevail only if they solve puzzles) may be indirectly attributed tothem. Thus exemplars hold an elevated status within the rele-vant community. Now recall that according to IBE the loveliestpotential explanation is that which would offer most under-standing. My claim against SO is that exemplars function by pro-viding scientists with standards of loveliness against which toassess potential explanations. Normal-scientific inference is amatter of evaluating candidate puzzle-solutions against exem-plars, so if puzzle-solving means explanation-giving, as it oftendoes, then that inference is governed by what we might call exemplars of loveliness . If exemplars did not generate standardsof loveliness, they could not play the puzzle-solving role thatKuhn describes whenever a puzzle demands an explanatorysolution. The potential explanation/puzzle-solution that bestmeets the standard set by the relevant exemplar provides theloveliest explanation of the evidence, and should thus beinferred (see Day & Kincaid, 1994, especially p. 289).Scientists may be unaware that their loveliness judgments areguided by exemplars, and unaware of the value of so judging. Sim-ilarly, the concept of loveliness that scientists form throughacquaintance with exemplars may never be fully articulated. Nev-ertheless, they will have such a concept—it is simply a matter of resemblance (it is difficult to imagine exemplars promoting a kindof loveliness they did not themselves instantiate). It is well moti-vated too: exemplars are the source of most theoretical and prac-tical understanding in normal science and the key to puzzle-solving success; thus they are held in high esteem. A concept of loveliness is an entirely natural, perhaps inevitable, consequenceof training and working within a normal science tradition (seeThagard, 1978).The following example illustrates these claims. Copernicus’ the-ory of planetary motion was an exemplar of loveliness for 16thcentury astronomers in that, among other things, it committedthem to inferring only explanations that saw planetary orbits ascircular. At that time, explanatory puzzle-solutions that repre-sented orbits by any other geometric shape could not have been 4 Despite construing explanation factively, I say ‘true explanation’ here since it draws attention to the thrust of the objection. 5 Lipton (2004, p. 70) calls TO ‘Voltaire’s objection’. In Voltaire’s play  Candide  (1759), the character of Dr. Pangloss uncritically endorses the Leibnizian thesis that we live in thebest of all possible worlds. Voltaire makes him the subject of ridicule. 66  D. Walker/Studies in History and Philosophy of Science 43 (2012) 64–73  considered lovely. This was a sensible constraint: theories assum-ing circular orbits had scored considerable empirical success sinceAristotle proposed his astronomical views in the 4th century B.C.(the most significant intermediate theory, that of Ptolemy, as-sumed circular orbits even though, via the equant point, it allowedthat planets generally deviate from a circular path).But within the context of 16th century ‘mathematical astron-omy’, in which the typical problem was to calculate the apparentposition of a planet from the earth, the solutions offered by Coper-nicus’ theory remained frustratingly inaccurate. Attempting toremedy this, Kepler proposed that the planets orbit in ellipses. Thisallowed Kepler’s theory to accommodate a greater amount of dataof greater accuracy than Copernicus’ theory; it even predicted thepositions of Mercury, the planet hitherto most resistant to astro-nomical modelling. This empirical success led to the adoption of Kepler’s theory over Copernicus’, despite the fact that great meta-physical and aesthetic significance was attached to the circle andthe ellipse was seen as inferior (McAllister, 1996, pp. 178–181).Kepler’s theory thus formed a new exemplar of loveliness inastronomy. Demands of predictive accuracy revealed the Coperni-can standard of loveliness to be flawed; hypotheses lovely byCopernican lights could not explain enough evidence, while thosethat accorded with Keplerian standards had greater scope andwere more precise. This is important: the change in lovelinesswas driven by empirical demands (see Section 5). Astronomersdid not replace a standard that recommends circular orbits withone that recommends elliptical orbits because they began to preferellipses to circles; as noted above, they did not. Rather, astrono-mers changed their standard because that which happens to rec-ommend elliptical orbits allows more precise calculation of planetary positions. Plausibly, mathematical precision brings withit the satisfaction of understanding and thus falls within the scopeof loveliness just as invoking circular orbits does. Thus the changeis better described as a move from the Copernican ‘loveliness asgeometrical orbit’ to the Keplerian ‘loveliness as accordance withmathematical laws’. Elliptical orbits  were  part of the new standardof loveliness—and sure enough, astronomers’ distaste for them dis-appeared over time—but only because explanations that assumedelliptical orbits solved more problems. For their respective astro-nomical communities, Copernicus’ and Kepler’s theories wereexemplars of loveliness, providing clear and universally endorsedstandards against which explanatory puzzle-solutions were as-sessed. This eliminated the subjectivity that SO claims would bepresent wherever loveliness is employed. 6 Astronomers may have been unsettled by the shift between thetwo standards—a phenomenon that Kuhn (1996, pp. 150–152)recognised—but it is not difficult to imagine scientists growingattached to a standard of loveliness that unlocks empirical success,however different it may be from its predecessor. Such attachmentshould be part of an account of loveliness. Although lovely-makingfeatures are not aesthetic properties, there is an important sense inwhich standards of loveliness become ‘internalised’, motivating akind of adherence that may be phenomenologically similar to aes-thetic preference. Scientists naturally find theories that conform totheir standard of loveliness desirable and wish to maintain thatstandard when it comes under threat. The fact that scientistscleave to standards in this way means that, rather than makingany kind of inference that some potential explanation is loveliest,scientists ‘see’ the loveliness of that candidate, preferring it to itscompetitors without consciously applying the relevant standardto the pool. Difficult cases may force scientists to scrutinise thatstandard and apply it more deliberately, but otherwise it may re-main tacit, and exert its influence over inferential judgments muchas aesthetic preference exerts its influence over aesthetic judg-ments. However, perhaps a better analogy is with ethical judg-ments, since it is more widely understood that these are thesubject of (at least) community-wide consensus. In the majorityof cases, ethical judgments achieve this uniformity without moralagents applying shared rules or making conscious inferences aboutthe moral situations they encounter. I say more about scientists’attachment to standards of loveliness in Section 4.2. 4.1. Loveliness and the role of exemplars Having sketched the idea of exemplars of loveliness, I want tostress something important, and note an advantage of my view.What I want to stress is that it is a concept of   loveliness  that scien-tists form in response to exemplars. Exemplars govern explanatorypuzzle-solving by prompting a clear idea of the features a puzzle-solution should have in order to create understanding of a phe-nomenon. Exemplars do not inculcate a concept of likeliness orwarrant, though they may influence scientists’ ideas about the fea-tures a puzzle-solution should have in order to achieve a high de-gree of likeliness, and rightly so. 7 One would expect the empiricalsuccess that exemplars enjoy to be related in some way to theirlovely-making features. But it is  these  features scientists seek toreproduce, not exemplars’ likeliness directly; likeliness cannot befaithfully reproduced in this way. The extent to which puzzle-solu-tions share exemplars’ likeliness is an empirical matter, separatefrom the issue of loveliness. Exemplars do not decide the likelinessof the puzzle-solutions they inspire. The advantage of my view I wish to note is that it need not com-mit itself to a particular account of loveliness. For my purposes,loveliness can be defined functionally as ‘those lovely-makingproperties exhibited by the relevant exemplar’, lovely-makingproperties being those which enable a hypothesis, if true, to en-hance scientific understanding. The notoriously vexed question of which properties make for lovely explanations can thus be left toone side. Judgments about the plausibility of my account dependneither on loveliness being analysed in some particular way, noreven on loveliness being analysable (in any illuminating way) atall. Accordingly, I offer no analysis of any lovely-making propertyand do not consider whether any such property is necessary or suf-ficient for lovely explanation, although I do suggest some candi-date explanatory virtues and indicate how further work on thisissuemightproceed.That workwill involveempirical investigationof relevant scientific episodes and consideration of psychologicalresearch on understanding. Another reason I avoid analysing love-liness is that such tasks lie beyond the scope of the present paper. 8 This freedom extends to the definition of exemplars. Kuhn is va-gue as to whether exemplars are scientific theories or applicationsthereof. The balance of his work suggests he saw them more asspecific applications of theories in problem-contexts, but someexemplary functions are better discharged by theories themselves.For instance, Kuhn claims that later paradigms solve importantoutstanding puzzles and retain (much of) the puzzle-solving powerof their predecessors (Kuhn, 1996, pp. 169–170). If exemplars,which change during revolutions, are to help explain this 6 I am not committed to the specific criteria of loveliness suggested by this or any historical example presented here. I deliberately avoid offering a constitutive account of loveliness (see Section 4.1). 7 Lipton (2004, pp. 59–62) contrasts likeliness and loveliness, but the distinction is unclear and potentially troublesome. Although it plays an important role in his developmentof IBE, it has been omitted from this discussion. 8 Lipton offers an account of loveliness and understanding which, though tentative and incomplete, goes beyond that assumed here (Lipton, 2004, pp. 30, 66, 122–123, 138–139). D. Walker/Studies in History and Philosophy of Science 43 (2012) 64–73  67
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x