https://ebookmass.com/product/causation-in-science-and-the-
Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...
Research Methods in Practice: Strategies for Description and Causation – Ebook PDF Version
https://ebookmass.com/product/research-methods-in-practice-strategiesfor-description-and-causation-ebook-pdf-version/
ebookmass.com
Scientific Testimony: Its roles in science and society Mikkel Gerken
https://ebookmass.com/product/scientific-testimony-its-roles-inscience-and-society-mikkel-gerken/
ebookmass.com
Forensic Science: An Introduction to Scientific and Investigative
https://ebookmass.com/product/forensic-science-an-introduction-toscientific-and-investigative/
ebookmass.com
Race, Popular Culture, and Far-right Extremism in the United States 1st Edition Priya Dixit
https://ebookmass.com/product/race-popular-culture-and-far-rightextremism-in-the-united-states-1st-edition-priya-dixit/
ebookmass.com
NNA Notary Signing Agent National Loan Documents
Sourcebook 2nd Edition National Notary Association
https://ebookmass.com/product/nna-notary-signing-agent-national-loandocuments-sourcebook-2nd-edition-national-notary-association/ ebookmass.com
System Architecture and Complexity Vol. 2: Contribution of Systems of Systems to Systems Thinking Printz
https://ebookmass.com/product/system-architecture-and-complexityvol-2-contribution-of-systems-of-systems-to-systems-thinking-printz/ ebookmass.com
No Good Tea Goes Unpunished Bree Baker
https://ebookmass.com/product/no-good-tea-goes-unpunished-bree-baker/ ebookmass.com
Prosecution of the President of the United States: The Constitution, Executive Power, and the Rule of Law 1st Edition H. Lowell Brown
https://ebookmass.com/product/prosecution-of-the-president-of-theunited-states-the-constitution-executive-power-and-the-rule-oflaw-1st-edition-h-lowell-brown/ ebookmass.com
Age of Emergency Erik Linstrum
https://ebookmass.com/product/age-of-emergency-erik-linstrum/ ebookmass.com
https://ebookmass.com/product/five-first-chances-sarah-jost/
ebookmass.com
CausationinScienceandtheMethods ofScientificDiscovery
CausationinScienceand theMethodsofScienti fic Discovery
RaniLillAnjumandStephenMumford
GreatClarendonStreet,Oxford,OX26DP, UnitedKingdom
OxfordUniversityPressisadepartmentoftheUniversityofOxford. ItfurtherstheUniversity’sobjectiveofexcellenceinresearch,scholarship, andeducationbypublishingworldwide.Oxfordisaregisteredtrademarkof OxfordUniversityPressintheUKandincertainothercountries
©RaniLillAnjumandStephenMumford2018
Themoralrightsoftheauthorshavebeenasserted
FirstEditionpublishedin2018
Impression:1
Allrightsreserved.Nopartofthispublicationmaybereproduced,storedin aretrievalsystem,ortransmitted,inanyformorbyanymeans,withoutthe priorpermissioninwritingofOxfordUniversityPress,orasexpresslypermitted bylaw,bylicenceorundertermsagreedwiththeappropriatereprographics rightsorganization.Enquiriesconcerningreproductionoutsidethescopeofthe aboveshouldbesenttotheRightsDepartment,OxfordUniversityPress,atthe addressabove
Youmustnotcirculatethisworkinanyotherform andyoumustimposethissameconditiononanyacquirer
PublishedintheUnitedStatesofAmericabyOxfordUniversityPress 198MadisonAvenue,NewYork,NY10016,UnitedStatesofAmerica
BritishLibraryCataloguinginPublicationData
Dataavailable
LibraryofCongressControlNumber:2018939482
ISBN978–0–19–873366–9
Printedandboundby CPIGroup(UK)Ltd,Croydon,CR04YY
LinkstothirdpartywebsitesareprovidedbyOxfordingoodfaithand forinformationonly.Oxforddisclaimsanyresponsibilityforthematerials containedinanythirdpartywebsitereferencedinthiswork.
Preface
Thereisnodoubtthatthereiscausationinscience.Someofthechiefgoalsof scienceareunderstanding,explanation,prediction,andtechnicalapplication.Only iftheworldhassomesignifi cantdegreeofconstancy,inwhatfollowsfromwhat, canthesescientificactivitiesbeconductedwithanypurpose.Thisregularityin natureisthebasisofprediction,forexample,whichwouldotherwisebeentirely unreliable.Butwhatisthesourceofsuchconstancyandhowdoesitoperate?How isitensuredthatonesortofeventtendstofollowanother?Thisisaquestionthat goesbeyondscienceitself beyondtheempiricaldata andinevitablyrequiresa philosophicalapproach:ametaphysicalone.Weargueinthisbookthatcausation isthemainfoundationuponwhichscienceisbased.Itiscausalconnectionsthat groundregularity.Theyarethereasonthatonekindofthingtypicallyfollows anotherwithenoughreliabilitytomakepredictionandexplanationworthwhile pursuits.Itiscausationthatgroundsthepossibilityofknowledge.
Shouldscientistsconcernthemselveswithwhatphilosophershavetosay?Onthis issueweargueforaresoundingaffirmative.Philosophytellsusaboutthenatureof causationitselfandthuswhatweshouldbelookingforwhenweinvestigatetheworld. Allscientistsmakepresuppositions.Theyresttheirendeavoursuponassumptionsthat arenotthemselvesempiricallydemonstrablethroughdata.Theseassumptionsare metascientific.Scientistsconductexperiments,forinstance:trialsandinterventions thatchangeonethinginordertoseewhatchangeswithit.Withoutabeliefincausal connections,orinsomethingbeingcausallyresponsibleforthosechanges,such experimentationwouldhavenopointbecausetheinterventionwouldnotbeproductiveoftheresult;alsotherewouldbenopossiblefutureapplication.Furthermore,the methodsadoptedorpreferredbypractisingscientistsmustrevealsomethingofwhat theytakecausationtobe.Attheveryleast,theymustbelookingforsomethingthat theyacceptasevidenceofsuchcausalconnections.
Forsometime,philosophershavepaidaninterestinscience,attemptingtomake theirphilosophicaltheoriesbetterinformedempirically.Toooftenneglectedisthe oppositegoal:ofmakingscientificpracticebetterinformedphilosophically.Inthe caseofcausationthisisabsolutelycrucialforitcanmakeasignificantdifferenceto howscienceisconducted.Ifoneispersuadedthatcausesaredifference-makers,for instance,thisiswhatonewillseekto find.Ifonethinkscausesareregularities,one willinsteadlookforthem.
Nowwhyshouldanyofthismattertoanything?Actually,itmattersalot.For example,oneofthegreatestchallengesfacinghumanityisclimatechange.Forallwe know,thisisathreattoourveryexistence,oratleastthatofasignificantnumberof theworld’spopulation.Itishardtothinkofagreaterproblemthanthat.Rightly,we
looktosciencetohelpussortoutthefacts,manyofwhichremaincontroversial. Thereareanumberofquestionswecanpose,allofwhichshouldhaveascientific answer.The firstissimplywhetherclimatechangeisreal,andalmostallimpartial scientistsnowagreethatitis.Otherquestionshavethenbecomemorepressing.Did humanactivityproducethisclimatechange?Whatwerethemechanisms?And, lookingforward,whatarelikelytobethemedium-andlonger-termeffectsofsuch change?Willthepolaricecapsallmelt?Willsealevelsrise?WilltheGulfStream switchoff?Willtherebecropfailures?Willwestarve?Canitbestopped?Canitbe mitigated?Allofthesearecausalquestions.Theyconcernwhat-will-cause-whatand wewantscientificwaysofuncoveringthesecausalfacts.Oneofourbiggestever problemshascausationatitsheart,therefore:causalmattersthatwelooktoscience toexplainand,hopefully,control.
Theaimofthisbookisnottoofferanoverviewofthephilosophyofcausation. Otherbooksalreadydothis.Therecent Causality byIllariandRusso(2014)doesan excellentjobofpresentinganimpartialsurveyoftheavailabletheoriesofcausation andhowtheyrelatetoscience.Thishasfreedusupforamoreopinionated investigationofthetopic.Wethinkthattherearecertaindominantphilosophical viewsofthenatureofcausationthatcontinuetoinformandinfluencemethods,orat leastinfluencewhatwethinkthenormsofscientificdiscoveryshouldbe.Aswehope toshow,however,someofthoseperceivednormscreateaconflictwhenthey confrontthescientistineverydaypractice.Scientistsknowthattheycannotperfectly preserveanormofobjectivity,thattheycannot findexceptionlessregularitiesthat areindicativeofcausalconnections,thattheycannotattainabsolutecertainty,that causalknowledgeisdefeasible,andsoon.Scientificpracticethusprovidesahardtest forourreceivednormsandphilosophicaltheoriesofcausationand,weargue,gives usgoodgroundstooverthrowandreplacesomeofthosenorms.Westrive,therefore, forascientificallyinformedphilosophyaswellasaphilosophicallyinformedscience.
Withtwenty-eightchapters,itwouldcertainlybepossibleforabusyreadertodip inandoutofthebook.Thechapterscanbereadinanyorderandinisolation.Buta thesisisdevelopedinthewholethatwouldrewardasequentialreading.Ifyoulike, youcouldproceedstraighttotheconclusion.There,welistninenewproposed normsofsciencethataresupportedbytheargumentsinthebook.Undereachnorm, welistwhichchaptersarerelevant:thoseinwhichsupportforadoptionofanew normisjustified.Youcouldthenlookattherelevantchapters.Anotherwaytoread thebookistoskimthekeywordsthatwehavegivenforeachchapter,insteadofan analytictableofcontents,andthenconsultthetopicsthatinterestyouthemost. Althoughwehavemadeiteasytoreadthebookintheseways,westillhopethatsome readerswillchoosetofollowthewholetextfromstartto finish.
Examplesareusedtoillustratethephilosophicalpoints.Theyaredrawnfroma rangeofsciencesandarekeptassimpleastheycanbe.Forus,itisimportantthat readerswithdifferentbackgroundscanunderstandandrelatethediscussionstotheir owndisciplineandperhapsconsidermoredetailedandrelevantexamplesthemselves.
Thebookconsistsofeightparts.PartIcontainsthreechaptersaboutthephilosophicalmotivationforthebook.Chapter1, MetascienceandBetterScience,urges thatphilosophersandscientistsneedtoaddresstheissueofcausationtogether, aimingtowardsareflectiveequilibriumsatisfactoryboththeoreticallyandempirically.Sincethescientificmethodsfordiscoveringcausesrevealontologicalcommitmentsastothenatureofcausation,thesecommitmentsmustbemadeexplicit andsubjecttocriticalreflection.InChapter2, DoWeNeedCausationinScience? , wearguethatcausationoccupiesafoundationalplace,underpinningthepurposesof sciencetoexplain,understand,predict,andintervene.Chapter3, EvidenceofCausationIsNotCausation,drawsadistinctionbetweenhowwelearnaboutcausation (epistemology)andwhatwetakecausationtobe(ontology).Therestofthebookis anattempttoshowthatthesetwo,althoughtheymustnotbeconfused,aretightly andinevitablylinked.Howweseektoestablishcausationmustbeinformedbywhat wethinkcausationis.
Thenextfourchapters,inPartII,discussanorthodoxviewthatcausationis conceptuallyandepistemologicallylinkedtoperfectcorrelations.Chapter4, What’ s inaCorrelation?,concernshowweseparatecausalfromaccidentalcorrelations, whileneo-Humeanismstrugglestomakeontologicalsenseofsuchadistinction. Chapter5, SameCause,SameEffect,questionsthetheoreticalassumptionthat causationshouldberobustacrossallcontexts,sincethisisnotsupportedempirically. Chapter6, UnderIdealConditions,takesthisdiscussionfurther,showinghowour theoreticalexpectationofcausalnecessitationisphilosophicallysalvagedbystipulatingasetofidealconditions.InChapter7, OneEffect,OneCause? ,wewarnagainst oversimplifyingcauses.Typicallytherearemultiplecausesofaneffectandtreating theminisolationcanletusmisstheimportanceoftheirinteraction.
PartIIIconsistsofthreechaptersthatpresentanalternativeapproachtothe Humeanorthodoxy.Chapter8, HaveYourCauseandBeatIt,introducesthenotion ofadditiveinterference,whereaneffectiscounteractedbytheaddingofsomething. Thisexplainswhycausationisessentiallysensitivetocontext.InChapter9, From RegularitiestoTendencies ,acaseismadeforunderstandingcausesastendencies ratherthanconstantconjunctions,conceptuallydetachingthenotionofcausation fromperfectregularity.InChapter10, TheModalityofCausation,wearguethat causationinvolvesaprimitiveanddispositionalmodality,weakerthannecessitybut strongerthanpurecontingency.
PartIVpromotescausaltheoriesandmechanismsasanalternativeto finding causationinregularityandrepetition.InChapter11, IstheBusinessofScienceto ConstructTheories?,wearguethatacausaltheoryisneededinadditiontothedataif wewantsomethingmorethanmerelymappingthefacts.Chapter12, AreMoreData Better?,makesacaseforcausalsingularism,wherecausationhappensintheconcrete particular,overHumeanism,wherethesingleinstanceisderivedfromauniversal claim.InChapter13, TheExplanatoryPowerofMechanisms,wepointtohowwe needqualitativeandmechanisticknowledgefordeepcausalunderstandingand
explanation.Weherepresentourdistinctiveaccountofmechanisms.However,in Chapter14, DiggingDeepertoFindtheRealCauses?,wescrutinizethereductive projectof findingcausalmechanismsalwaysatlowerlevelsofnature.Weconclude thatmechanismscanbehigherlevel,andwegiveaccountsofholismandemergence.
PartVconcernshowcausesarelinkedtoeffects.InChapter15, Makinga Difference,wediscussthecounterfactualtheoryandarguethatitfailstoaccount forsomeinstancesofcausation.Thisshowsthatcausationisnotthesameas difference-making.Chapter16, MakingNothingHappen,developsthepointthat therearecasesofcausationwherenochangeoreventfollows,butthattheseare someofthemostimportantcausalsituations.Wethenlookatthematterofcausal chainsinChapter17, ItAllStartedwithaBigBang,andconsiderwhethertheir existenceshowsthatcausationistransitive.ThepartendswithChapter18, Does ScienceNeedLawsofNature?,wherewequestiontheontologicalneedforuniversal lawsinadditiontointrinsicpropensitiesandtheircausalinteractions.
PartVIisaboutprobability.Chapter19, Uncertainty,Certainty,andBeyond, focusesondegreesofbelief,suchasdoubtandcertainty.Notably,wemakeroom foraclassofcasesinwhichyoucanstillbecertainofsomethingevenwithless evidencethanbefore.Chapter20, WhatProbabilisticCausationShouldBe,iswhere weofferourtheoryofworldlychance,basedonadistinctiveaccountofpropensities. InChapter21, CalculatingConditionalProbability?,weshowthatourintuitive notionofconditionalprobabilitymustbeseparatedfromthestandardratioanalysis ofP(AjB),forepistemologicalandontologicalreasons.
PartVIIisprimarilyontheproblemofexternalvalidityandcontainstwochapters: Chapter22, RiskyPredictions,andChapter23, WhatRCTsDoNotShow.The firstof thesearguesthatourcausalpredictionsareessentiallyfalliblebutusefulnevertheless. Indeed,anytheoryofpredictionthatdidnotaccountforitsfallibilitywouldbe flawedbecauseofthat.Thesecondpointstosomesignificantshortcomingsofthese large-scalepopulationstudiesindealingwithcausalfactorssuchasindividual variation,heterogeneity,complexity,andmarginalgroups.
PartVIIIcontains fivechaptersoncausaldiscovery.Chapter24, GettingInvolved, developsthemanipulationistinsight,butmainlytoarguethatourcausalknowledge happensincloseinteractionwiththeworld,notbydistancedobservation. Chapter25, UncoveringCausalPowers,presentsouraccountoftechnologicalinnovation,whichwearguerestsmainlyinteasingoutthehiddenpowersofthings. Chapter26, LearningfromCausalFailure isaconsiderationoftheopportunitiesof newknowledgethatarisefromunsuccessfulexperimentsanddiscrepancies.Given thatthereisadiminishingreturninconfirmingevidence,afterapoint,bigbreakthroughsaremorelikelytofollowfromnegativeresults.InChapter27, Plural Methods,OneCausation,weargueforacombinationofepistemicpluralismand anontologicalmonism.Thefailureofanalysishasledotherstoassumethatcausationinrealitymustbemanydifferentthings.Butthisoverlookstheotherpossibility: causationisonethingbutprimitive.Wemusttheninvestigateitthroughits
symptomsandourmethodsmustdetectthose.InChapter28, GettingRealaboutthe IdealsofScience,wechallengethemotivationforidealizationandabstractionin sciencewhendealingpracticallywithamoremessyreality.Thereisacrisisin reproducibility,whichshowshowsomeofourexpectationsofscienceareunrealistic andbasedonamistakennotionofcausation.
Whatemergesfromalltheseconsiderationsisthat,ifweunderstandcausation right,weshouldapproachitinacertainwaythroughourscientificmethods.We finishbyofferingninenewnormsforcausalscience,drawingtogetherthemany themesofthebook.
Thisbookistheculminationoftheresearchcarriedoutonthe Causationin Science (CauSci)projectattheNorwegianUniversityofLifeSciences(NMBU)from 2011to2015,fundedbytheFRIPROschemeoftheResearchCouncilofNorway.We thanktheCauSciteam(originalandsubstitute),aswellasthemanyprojectcollaborators,forengagingwithourwork.Thematerialwasfurtherdevelopedthrougha teachingcourseatNMBUandwearegratefulforthediscussionsandinputfromthe studentsofPHI302andPHI403.Wehavealsoreceivedinvaluablecontributions fromtheCauseHealthteamandcollaborators.Inparticular,wewanttothank FredrikAndersenforraisinganddiscussingwithusmanyoftheissuesofthe book.WearegratefultoJohanArntMyrstad,the sinequanon ofChapter21. AspecialthankyoutoElenaRocca,ourpractisingscientistinresidence,forsuggestingandresearchingmanyoftheexamplesweuse.
Contents
ListofFigures xv
PartI.ScienceandPhilosophy
1.MetascienceandBetterScience3 science – philosophy – norms
2.DoWeNeedCausationinScience?11 scepticism – observation – empiricism
3.EvidenceofCausationIsNotCausation19 ontology – epistemology – operationalism
PartII.PerfectCorrelation
4.What’sinaCorrelation?29 regularities – statistics – coincidences
5.SameCause,SameEffect37 robustness – irregularities – falsification
6.UnderIdealConditions46 idealization – generality – exception
7.OneEffect,OneCause?53 complexity – backgroundconditions – triggers
PartIII.InterferenceandPrevention
8.HaveYourCauseandBeatIt63 interference – intervention – causalexpansion
9.FromRegularitiestoTendencies71 incidence – tendencies – vectormodels
10.TheModalityofCausation79 dispositionality – necessity – powers
PartIV.CausalMechanisms
11.IstheBusinessofSciencetoConstructTheories?89 rawdata – theorydevelopment – underdetermination
12.AreMoreDataBetter?97 induction – causaluniqueness – dispositionalism – evidencebase
13.TheExplanatoryPowerofMechanisms104 mechanisms – quantitative – qualitative
14.DiggingDeepertoFindtheRealCauses?112 reductionism – levels – emergence – demergence
PartV.LinkingCausestoEffects
15.MakingaDifference123 comparison – counterfactuals – overdetermination
16.MakingNothingHappen131 stability – homeostasis – absences
17.ItAllStartedwithaBigBang139 causalchains – determinism – transitivity
18.DoesScienceNeedLawsofNature?146 governance – properties – coveringlaw
PartVI.Probability
19.Uncertainty,Certainty,andBeyond157 probability – Bayesianism – credence
20.WhatProbabilisticCausationShouldBe165 chance – frequentism – propensities – overdisposing
21.CalculatingConditionalProbability?174 ratioformula – conditionals – compositionality
PartVII.ExternalValidity
22.RiskyPredictions185 uncertainty – fallibility – externalvalidity
23.WhatRCTsDoNotShow194 variation – exclusion – bias – policy
PartVIII.DiscoveringCausesandUnderstandingThem
24.GettingInvolved205 manipulation – presuppositions – processeddata
25.UncoveringCausalPowers212 technology – hiddenpowers – sideeffects
26.LearningfromCausalFailure220 diminishingreturns – negativeresults – discrepancies
27.PluralMethods,OneCausation227 pluralism – primitivism – evidencehierarchies
28.GettingRealabouttheIdealsofScience237 reproducibility – replicability – messiness
ListofFigures
5.1.Modelversusdata40
5.2.Anoutlier42
9.1.Avectormodeloftwotendencies74
9.2.Additiveinterference74
9.3.Subtractiveinterference74
20.1.Asymptoticrelationbetweenextentofpropensityandprobability171
22.1.AvectorwiththeresultanttendencyR189
27.1.Causationanditssymptoms232
PARTI ScienceandPhilosophy
1
MetascienceandBetterScience
1.1WhatScienceIsandShouldBe
Scienceismanythings.Wemightsaythatitisanactivityaimedatdiscoveringaclass ofimportanttruthsabouttheworld;andthenshowingwhatcanbedonewiththat knowledge.Wecanthinkofscienceasatool,amethod,or,moregenerally,a philosophy.Centraltoscience,perhapsevenconstitutiveofit,isasetofnormsfor thecorrect,systematicacquisitionofempiricalknowledge.Byanorm,wemeana standardofwhatought,oroughtnot,tobedone.Forscientificnorms,theconcernis whatweoughttodooroughtnottodoinordertobestgatherandutilizeknowledge abouttheworld.Empiricalknowledgeisknowledgederivedfromexperience,as opposedtomathematical,logical,andconceptualtruths,whichcanbediscovered throughuseofreason.
Forexample,itisarguablethatforknowledgetocountasscientific,itneedsto beobjective.Itcannotbejustoneperson’sview,forinstance.Hence,ifyouclaimto havefoundsomethingremarkableunderyourbedthatnooneelseisallowedtosee, thereport,inthatstate,willnotcountasscientificandwillthusnotbeanadmissible datumforanyscientifictheory.Beingjustoneperson’sexperience,thisreportwill beclassedassubjectiveandunverified.However,ifknowledgeisacquiredaccording tothenormsofscience,includingthenormthatitbeobjective,thenitmayqualify forthestatusofscientific.Thenormsarethusimportanttomanyformsofenquiry becauseitoftenmattersthataclaimisscientific.Itmeansitshouldbetrustworthy, forinstance.Ajustificationofsciencecouldbethatitisknowledgeacquiredthrough thesenorms,whichhavebeenacceptedastherightwaytoknowtheworld.Scienceis valuable,itmightbesaid,becauseitisobjective,amongotherthings.
Thereareanumberofnormsforcorrectempiricaldiscovery.Itneedstobe admitted,however,thatthesenormscontinuetobecontested.Thehistoryofscience showsmultiple,ongoingdebatesaboutwhatisthecorrectscientificmethod.Should sciencestartwiththerecordingofdata,forinstance,orisiteverpermissible indeed inevitable tobeginwithahypothesisthatyoucanthensubjecttotesting?Thereis stillworthwhilediscussiontobehadonthisquestion.Itmaynowsoundsurprisingto hearsomeoftheolderdebates,especiallyonsubjectswethinkareclear-cut.Galileo (1632)offeredanexperimentalmethodforscience,forexample;butexperimentation wascontroversialatthetime(Gower1997:23).Wenowthinkitobviousthatscience
shouldconductexperiments.InatraditiondatingbacktoPlato(seetheallegoryof thecave, Republic 514a–520a),however,thesenseswereregardedasunreliablesowe wereurgedtotrustreasoninginstead,whichwaspreciseandconclusive.Evenin Galileo’stime,evidencefromthelargelyunaidedsenseswasimpreciseanduncertain, sothoughttohaveonlylimiteduseforscience.Galileocompromised.Mostofthe experimentsheproposedwerethoughtexperiments.Hereasonedthroughwhathe thoughthadtohappen,including,itseems,inthefamousexperimentinwhichtwo objectsofthesamesizebutdifferentweightswerethrownfromatallbuilding.Itwas theuseofreasonthatledGalileotoconcludethattheywouldfallatthesamerate. Theexperimentcouldinprinciplebeperformed ‘forreal’ butthereseemednocause todoso,thoughtGalileo,sinceanyresultotherthanhisownwouldbecontraryto reasonandthereforecouldnothappen(Gower1997:30ff).
Wemightnowthinkthatwepossessaccuratemeasuringdeviceswithinexperimentalset-upsandthatwecantrustourobservations,whichareaspreciseaspure reasoning.Butthisisstillcontestable.Sometimesthereissuchconfidenceinthe reasoningwithinatheorythatithardlyneedsanyempiricalconfirmation.Or,asin theoreticalphysics,wearenotyetabletoempiricallyconfirmthetheoryandthus havetoresttheargumentonpurelyrationalconsiderations.Opposingthisviewisthe thoughtthatreasonitselfcannolongerbetrusted,inthattheworldneedn’tworkina particularlyrationalorintuitiveway.Someresultsincontemporaryphysicsseemto offendcommonsense,forinstance.Inonewell-knowninstance,weareaskedto acceptthatSchrödinger’scatisbothdeadandaliveatthesametimeandthatthis explainshowsomeaspectsoftheworldreallywork.Wemighthavetoreviseour viewofwhatislogical.
Wesaidthatthenormsofscienceweremany.Itmayhelpifweofferupsome initialcandidatenorms.Allofthemhaveaninitialplausibilityandtheymight collectivelybetakenasareasonabledefaultorstartingposition.Wecouldsay, then,thatthefollowinglistcapturesmuchofwhatscienceoughttodo:
• Beobjective.Forexample,resultsshouldbereproduciblesothatotherscan observethesameexperimentandrecordthesameoutcome.
• Atheoryshould ‘preservethedata’,i.e.beconsistentwithiteventually,moreor less.
• Themoreempiricalevidencethereisinfavourofatheory,themorelikelyitisto betrueandthusthemoreinclinedoneshouldbetoacceptit.
• Favourtheoriesthathavegreaterexplanatorypowerorscope.Ifanewtheory canexplaineverythingthattheoldtheoryexplained,andmore,thatisareason tofavourthenewtheoryovertheoldone.
• Predictivesuccesscountsinfavourofatheoryinthatonesignofagoodortrue theoryisthatitgetsconfirmedbysubsequentresults.
Afurtherreasonforcallingthesenorms ‘default’ isthatwewanttoleaveopenthe questionofwhethertheyarestrict, ‘hardandfast’,orwhethertheyshouldbetreated
asheuristic ‘rulesofthumb’,admittingexceptions.Afterall,perhapstherearesome caseswhereyoushouldfavouratheorywithnarrowerscopeoveronewithbroad scope,suchasifthebroadertheoryrestsonimplausibleassumptions.Weshould allowthatthedefaultnormsmayberevisedorgivenmoreprecision.Predictive successisnotastraightforwardmatter,forinstance.Thereareproblemssuchas confirmationbiasandtheory-dependenceofobservationwhereitcouldlooklikea theoryhasachievedpredictivesuccessbuthasn’treally.Onewouldhavetoruleout suchcasesinordertopreservetheintegrityofthenorm.Ifonecan’tdothat,one mighthavetoabandonthenorm.
1.2WhyPhilosophy?
Thisaccountshowsthatthereareatleasttwojobsforthephilosophyofscience.First, wehavetodecidewhatthecorrectnormsofscienceareandjustifythem.Isitrightto favourthetheorywiththemostempiricalevidence,forinstance;andifso,why? Second,onceallthenormsareinplace,wemustalsoconsiderwhatthebestwaysare ofsatisfyingthem.If,forexample,weagreethatscienceshouldbeobjective,onethen hastoconsiderwhetherthisdemandsreproducibilityofresults(seeChapter28).Does itmeanthatanyone,anywhere,shouldbeabletofollowthemethodandgetthesame outcomeastheonereported?Arethereotherwaysofsatisfyingtherequirementof objectivity?Areanyofthemmoreimportantthanthenormofreproducibility? Why,itmightbewondered,dowesaythatthesequestionsareforphilosophyto answer,ratherthanscienceitself?Can’tscienceworkitoutunaided?Isn’titpresumptuoustoclaimthatscienceneedsphilosophy?Thereisindeedaviewthat philosophyisoflittleornorelevancetothisdiscussion,norisanythingelseoutside ofscience.Scientism,sometimescallednaturalism,istheviewthateverymeaningful questioncanbeansweredscientifically.Insofarasphilosophyisattemptingatask whichisnon-empirical,hencenon-scientific,scientismquestionsphilosophy’sright topronounceaboutthenatureoftheworld.Someproponentsofthisviewarewithin science.StephenHawking(2011:5)hasdeclaredthatphilosophyisdeadbecause physicshasansweredallitssubstantialquestions.Butsomedefencesofscientismalso comefromwithinphilosophy,suchasfromLadymanetal.(2007),whoarefollowing anaturalisttraditionpromotedbyQuine(1995,foroneinstance).Wewishto overcomeanysuchconflictbetweenphilosophyandscience.Beingpro-philosophy doesnotmeanoneisanti-science.Similarly,beingpro-scienceshouldnotmean beinganti-philosophy.Bothareneeded,asweaimnowtoexplain.
First,ifourtaskistounderstandwhatthecorrectnormsofscienceare,theanswer clearlycannotcomefromwithinscienceitself,foranysuchanswerwouldbequestion begging.Onemightbetemptedbytheoppositeview.Whenaskedhowscienceshould bedone,someonecouldsaythatoneshouldlookatscienceandseehowitisdone,or thatoneshouldconsultatrainedscientistwhoknowsexactly.Butthisisnogood.We wanttoknowhowscienceoughttobeconducted,whichisn’tthesameashowitis
actuallyconducted.Ifwetakethenormstobedefinitiveorconstitutiveofscience, thenclearlyonecanonlyconsiderthevalidityofthosenormsbysteppingoutsideof thescientificpracticeitself.Considerationofthenormativeaspectofscienceisthus inherentlyaphilosophicalandabstractenterpriseratherthananempiricalone.The keyissueiswhetherthecurrentnormsofapracticecouldbewrongorincomplete. Onecannotmakethisjudgementfromwithinthatpractice.
Second,ithastobeacknowledgedthatthereisphilosophyinsciencewhetherone likesitornot.Sciencerestsonphilosophicalassumptions,includingmetaphysical ones.Theseassumptionscannotbeprovenbyscienceitself,butonlyassumed,and thisshowsusthatscientismisuntenable.Thisiseffectivelyadmittedduringan attempteddefenceofscientismbyLadymanetal.: ‘WithrespecttoLowe’s...claim [Lowe(2002:6),that “Naturalismdependsuponmetaphysicalassumptions”],itis enoughtopointoutthatevenifnaturalismdependsonmetaphysicalassumptions, thenaturalistcanarguethatthemetaphysicalassumptionsinquestionarevindicated bythesuccessofscience’ (Ladymanetal.2007:7).Thisconcessionisdecisive.We canseeanumberofsignificantpointsinitthatareworthdetailing.
First,theconcessionacceptsthepossibilitythatsciencedoesn’ttelluseverything. Itcanrestuponnon-scientificassumptions.Itisclearthattheseassumptionscannot beevaluatedbyscienceitselfbecausetheymustbeassumedinordertohavescience inthe firstplace.
Next,theconcessionadmitsthebasisofastrongargumentinfavourofacertain metaphysicalassumptionorsetofassumptions.Ittellsusthatsuchassumptions wouldbejustifiedwhensciencerestsuponthemandscienceisasuccess.Wewillsee inChapter2thatthisgivesusagoodargumentfortherealityofcausation.Science assumesthevalidityofbothobservationandinterventionandtherecouldbeneither withoutcausation.Wehavegoodgrounds,then,forsayingthatcausationisreal becausesciencerestsuponitandissuccessful.Sothisisaneffectiveargumentin favourofafoundationalmetaphysics,groundingasuccessfulpursuitsuchasscience.
AnothersignificantpointisthattheLadymanetal.concessionassumesatypeof justificationforsciencethatcannotcomefromscienceitselfbut,asarguedabove,isa normthatcanonlybeevaluatedfromoutside.Thisistheclaimthatsuccess for instance,explanatory,predictive,andtechnological iswhatvindicatesscience. Scienceitselfcannotshowthatyououghttofollowthepracticethatproducessuccess ofthiskind.Itdependsonwhatyouwant.Differentactivitiesaresuccessfulon differentgrounds:groundsthathavetobechosenthroughrationalenquiry.Weare entitledtoaskspecificallywhatthemarkersofsuccessareinscience.Whatexactlyis successinthiscase?
Finally,itisworthconsideringhowthisnormofsciencewouldbeassessed. Supposethenaturalistweretoclaim,withoutresortingtophilosophy,thatit’ssimply obviousthatpredictivesuccessistheyardstickagainstwhichweshouldmeasure science.Thatmaybeso.Isn’titjustintuitivethatweshouldjustifytheworthof science,andthereforeitsmetaphysicalbasis,onthegroundsofitssuccess?What
morethanexplanatory,predictive,andtechnologicalsuccesscouldonesensibly want?However,thisisadefencethatLadymanetal.havealreadyruledout.They arguethat ‘asnaturalistswearenotconcernedwithpreservingintuitionsatall’ (Ladymanetal.2007:12)andtheygoontopointouthowmanythingsinscience aretruebutcounterintuitive.Oneshouldnotaccepttheargument(thatscienceis validatedbyitssuccess)onintuitivegrounds,therefore;atleastnotaccordingto defendersofnaturalism.
Yet,thistypeofargumentwouldnotbequitesobadotherwise.Ladymanetal. cannotacceptitbecauseitunderminestheirdefenceofscientism.Forsomeonewho isnotseekingtodefendscientism,itcouldhoweverbeaccepted.Indeed,socompellingistheintuition,thattheauthorsuseitveryquicklyasabasisforaccepting anothernormofscience.Tojustifythatscienceisseekinga ‘relativelyunifiedpicture ’ (p.27),theyarguethattheoppositewouldbea ‘mystery’.Thatsoundsfairenough, exceptforthefactthattheauthorshadjustclaimedthatbeingintuitiveisnovirtuein science.What,then,cantheymaintainisbadaboutsomethingbeingamystery?
1.3PhilosophicalAssumptionsinScience
Philosophicalassumptionsareinevitableinscience.Therearethentwotypesof scientists:thosewhoareawareofscience’sphilosophicalunderpinningsandthose whoarenot.Onlyifweareawareofwhatwehaveassumedareweabletoreflect criticallyuponitsvariousaspectsandaskwhetheritreallygivesusasoundfoundation.Withoutsuchawareness,ourscientificconclusionsmightcontainerroneous presuppositionsthatdonotserveuswellorwithstandrationalscrutiny.
Letuslookatonewayinwhichthismighthappenand,indoingso,movetowards thecentraltopicthatoccupiesusinthisbook:thescientificdiscoveryofcausation. Weshouldalsounderstandsomeofthedifferencesbetweenphilosophicaland scientificapproachestoknowledgeofcausation.Whilescienceisempiricaland dealswiththeconcrete,suchasparticularmattersoffact,recordedasdata,philosophydealswiththegeneralandabstract.Thus,inscience,includingsocialscience,we mightwantto findoutwhatcauseswhat:whetheraparticulargenotypemakesa personsusceptibletocancer;whetherheatincreasesordecreasesthetensilestrength ofacertainmetal,orwhetherareadingprogrammeimprovestherateofsocial mobility.Theseareallparticularcausalhypotheses,eventhoughnoneofthemuses theterm ‘ cause ’ explicitly.Theycouldeasilyhavedoneso(‘doesreadingcausean increaseinsocialmobility’,andsoon).Philosophy inparticularmetaphysics, whichstudieswhatthereisinthemostgeneralterms(Mumford2012) willlook atwhatiscommontothesethreescientificinstances.TheyallclaimthatsomeAis causingsomeB.Thisshowsthatphilosophyisabstractinthesensethatitsconcernis withthenatureofcausation,irrespectiveofwhatthecausesandeffectsare.Philosophy thusstudieswhatitisforonethingtocauseanother,abstractedawayfromthe particularthingsthatarecausingandbeingcaused.
Anaturalistmightclaimthatwecanjustletsciencetelluswhatitisforonething tocauseanother,butthatwouldbeamistake.Sciencecannotdothiswithoutceasing tobescience.Todosoistogobeyondthedataandenterintophilosophy.Science canproposetheoriesofwhatcauseswhat,basedonthedata,butitexceedsthenorms ofsciencetocommentonthelargelynon-empiricalmatterofwhatcausationisitself.
Naturalists findthemselvesinatrap,then.OnecanonlysaythatAcausesBifone hasanaccountofwhatitisforonethingtocauseanother;butscienceunaidedby philosophycannotjudgeonthatmatter.Forthisreason,somescientistsdeclineever tomakecausalclaims.Causalclaimsarerarewithinacertaintraditioninepidemiology,forinstance,eventhoughsuchasciencewouldseempointlessunlessitis discoveringcauses(Broadbent2013).Itmightbethoughtthatthescientifictaskis completeoncethedataarerecorded:perhapsaraisedincidenceofsomediseaseupon increaseofanothervariable.Anaturalistreactiontothetrapis,then,toproclaimthat thereisnocausationinscience,ornoneinoneparticularscience.Thisquestionisthe topicofChapter2.Wecansayhere,however,thatthissortofclaimmustbebasedon anassumptionofwhatcausationisandisnot(Kerryetal.2012,Andersenetal.2018). Basedonthatassumedviewofcausation,onemayjustconcludeagainstitspresence. Butwhatifthenatureofcausationisnotasthenaturalistassumed?Perhapsno regularitiesordeterminaciesarefoundinphysics,butthatonlycountsagainstthe presenceofcausationifcausationhastobedeterministicandissueinregularities.The problemhereisthatwecanonly findtherightorbestaccountofthenatureof causationandallitsfeaturesbydoingatleastsomephilosophy.Thenaturalistis trappedbyarefusaltodoanysubstantialmetaphysics,whichhindersprogress.
Thereareofcoursemanyothernotionsinsciencethatbenefitfromahigher-level understanding.Whatisanexplanation,forinstance,alawofnature,oranatural kind?Onlywithamatureunderstanding,ofwhatthesethingsaresupposedtobe, canoneacceptorrejecttheminscience.Onecouldnotsay,forinstance,thatthere arenonaturalkindsinrealitybecausenothinghasfeatureN,iffeatureNisnotreally apartofwhatitistobeanaturalkind.Andthislatterquestionisatleastinpart and perhapsprimarily aphilosophicalmatter.Similarly,therearemanyotherassumptionsatplayinsciencethatarenotempiricallygrounded,concerning,forinstance, thenotionofcomplexity(seeRoccaandAndersen2017).Someacceptreductionism, forexample.Perhapsonescienceisreducibletoanother,suchasifwecanexplain psychologyintermsofevolutionarybiologyoreconomicsintermsofneuroscience (seeChapter14).Perhapseverythingisreducibletophysics.Again,though,weneed tounderstandwhatconditionswouldqualifyasareduction,whichthengivesusa griponwhat,ifanything,wouldallowustorecognizeacaseofitinreality.
Ourjob,however,concernsspecificallycausationinscience.Onetaskistotease outtheassumptionsthatarebeingmadeaboutthenatureofcausation:articulate themsothatwecanscrutinizethemandjudgewhethertheyareright.Butthisisnot ouronlytask;indeeditisnotourmaintask.Thisbookisprimarilyaboutthenature ofcausalscienceanditsownphilosophicalpresuppositions,asmanifestedinthe
normsofcausalscience.Whatwemeanbycausalscienceisthescienceof finding causesinthenaturalworldandusingthatknowledge.Webelievethatsmoking tobaccocausescancer,forexample,andthattheheatingofironrodscausesthemto expand.Dowehavegoodgroundsforthesebeliefs?Wecanseethatthiscomesdown tothequestionofwhetherthecorrectnormsofcausaldiscoveryhavebeenfollowed andgivenapositivejudgement.Butdoweknowthatthosecorrectnormshavebeen yetfound?Furthermore,isitpossiblethatsomenormsofcausaldiscoveryare inappropriate?Thiscouldagainbeapartiallyphilosophicalmatter.Supposeone hasaregularitytheoryofcausation:thatitconsistsinanabsoluteregularityor constantconjunction(Hume1739:I,iii,2).Onemaythenadoptasanormofcausal sciencethatoneshouldrecordalltheconstantconjunctionsanddrawcausal inferencesfromthem.Butthisnormisinappropriate,weallege,becauseitwillfail toreliablyidentifythecausalfacts.Itcouldpronouncethatsmokingisnotacauseof cancer,becausenotallsmokersgetcancer;or,contrariwise,thathumanityisthe causeofmortality,whichwewouldsayitisnot,eventhoughallhumansareindeed mortal.Whattheseexamplesillustrateissimplythatourphilosophicalviewofthe natureofcausationshapesthenormsthatweadoptforcausalscience;thatis,it determineswhatwethinkweshouldlookforwhenseekingcausalconnections.
1.4TheCriticalFriend
Thedefenceofaphilosophicalcontributiontoscienceshouldinnowaybeunderstoodasanattackonscience.Althoughsciencecannotanswereveryquestion,itstill givesusthebestwaytoanswerquestionsofacertainkind,includingtheempirical matterofwhatiscausallyconnectedtowhat.Philosophyshouldbeseenasscience’ s criticalfriend,pointingoutthosenormativeaspectsthatinpartconstitutescience andexaminingthecredibilityoftheunderlyingmetaphysicsuponwhichitrests.In return,sciencecanbephilosophy’scriticalfriend,too,allowingtheabstracttheoryto beinformedbyabetterunderstandingoftheconcretereality.Itmightbeassumedin thephilosophicalaccount,forinstance,thattherecannotbeinstantaneousaction overadistance.TheEinstein–Podolsky–Rosenexperimentinphysicsshowsan allegedcasewhereperhapstherecouldbesuchaction,underoneinterpretation. Thephilosopherthenhasanoptionofdroppingthestricturefromhermetaphysics ofcausation.Butistheinterpretationasoundenoughbasisfordroppingthe stricture?Therecanbe,thus,atoingandfroingbetweentheconcreteandtheabstract asourunderstandingdevelops.Thegoalisto findatheorythatissatisfactoryboth scientificallyandphilosophically.Thisiswhatwemeanbyareflectiveequilibrium. Inthisbook,weofferasampleofphilosophicallyorienteddiscussionswithwhich scientistsshouldwanttoengagebecausetheyrelatetothescientificdiscoveryof causalconnections.Shouldscienceconstructcausaltheories,oronlydealwithdata? Howcanourscientificstudieshaveexternalvalidity,beyondthedataset?Andcan weassumethatthemoredata,thebetteristheevidenceofcausation?Whatdowe
