Conduct, Misconduct, and Cargo Cult Science 1407

DOCTORAL COLLOQUIUM KEYNOTE ADDRESS

CONDUCT, MISCONDUCT, AND CARGO CULT SCIENCE

James R. Wilson

Department of Industrial Engineering North Carolina State University Raleigh, North Carolina 27695, U.S.A.

In the South Seas there is a cargo cult of people. During the war they saw airplanes land with lots of good materials, and they want the same thing to happen now. So they’ve arranged to make things like runways, to put fires along the sides of the runways, to make a wooden hut for a man to sit in, with two wooden pieces on his head like headphones and bars of bamboo sticking out like antennas—he’s the controller—and they wait for the airplanes to land. They’re doing everything right. The form is perfect. It looks exactly the way it looked before. But it doesn’t work. No airplanes land. So I call these things cargo cult science, because they follow all the apparent precepts and forms of scientific investigation, but they’re missing something essential, because the planes don’t land. Now it behooves me, of course, to tell you what they’re missing. . . . It’s a kind of scientific integrity, a principle of scientific thought that corresponds to a kind of utter honesty—a kind of leaning over backwards. For example, if you’re doing an experiment, you should report everything that you think might make it invalid—not only what you think is right about it: other causes that could possibly explain your results; and things you thought of that you’ve eliminated by some other experiment, and how they worked—to make sure the other fellow can tell they have been eliminated. . . . In summary, the idea is to try to give all of the information to help others to judge the value of your contribution; not just the information that leads to judgment in one particular direction or another.

—Richard P. Feynman, “Surely You’re Joking, Mr. Feynman!” (1985)

ABSTRACT ity and unimpeachable analysis of experiments per- formed with those models. Finally I will discuss the ethical dilemmas inherent in the peer review system, I will elaborate some principles of ethical conduct in and I will offer some concrete suggestions for improv- science that correspond to Richard Feynman’s well- ing the process of refereeing primary journal articles. known precepts of “utter honesty” and “leaning over backwards” in all aspects of scientific work. These principles have recently been called into question by 1. INTRODUCTION certain individuals who allege that such rules are based on a misunderstanding of “how science actu- Much has been written recently about what consti- ally works” and are therefore potentially “damag- tutes scientific misconduct, and public esteem for sci- ing to the scientific enterprise.” In addition to ex- ence has been damaged by high-profile episodes such amining critically the general basis for these allega- as the “cold fusion case” at the University of Utah tions, I will discuss the particular relevance of Feyn- (Huizenga 1993) and the “David Baltimore case” at man’s ideals to the field of computer simulation; and MIT (Elliott and Stern 1997). Against this back- I will emphasize the need for meticulous validation of drop I will examine several claims about principles of simulation models together with exact reproducibil- ethical conduct in science that were made by James 1406 Wilson

WoodwardandDavidGoodsteinoftheCaliforniaIn- aboutthediscoveryofNrays,hewenttoFranceto stituteofTechnologyinanarticleentitled“Conduct, observeBlondlot’sexperimentalprocedure.Atthat MisconductandtheStructureofScience,”whichap- timeBlondlotwasusingaspectroscopefittedwithan pearedintheSeptember1996issueoftheAmerican aluminumprismtomeasuretherefractiveindicesof Scientist.Thegistoftheprinciplesinquestionis Nrays.AlthoughBlondlot’sexperimentswereper- summarizedinthequotationbyRichardFeynman formedinadarkenedroom,asmallred(darkroom) givenabove.Iwillarguethattheseprinciplesare lanternenabledBlondlottoseeagraduatedscalefor especiallyrelevanttothefieldofcomputersimula- measuringtothreesignificantfiguresthepositionof tion,andIwillelaboratemyviewthatFeynman’s averticalthreadcoatedwithluminouspaint.The idealsof“utterhonesty”and“leaningoverback- threadwassupposedtobrightenasitcrossedthe wards”constituteamandateformeticulousvalida- invisiblelinesoftheN-rayspectrum.Accordingto tionofsimulationmodelstogetherwithexactrepro- LangmuirandHall(1989),WoodaskedBlondlotthe ducibilityandunimpeachableanalysisofexperiments followingquestion: performedwiththosemodels.Severalkeyreferences ...fromjusttheopticsofthething,with arehighlightedinthisdiscussion—inparticular,see slits2mmwide,howcanyougetabeam thepamphletsentitledOnBeingaScientist(1995) sofinethatyoucandetectitspositionto andHonorinScience(1986).Interestedindividu- withinatenthofamillimeter? alsareinvitedtoexaminetherelevantliteratureand tojudgeforthemselvesthevalidityofthearguments Blondlotisreportedtohavegiventhisreply: givenhere. That’soneofthefascinatingthingsabout Nrays.Theydon’tfollowtheordinarylaws 2.“THESCIENCEOFTHINGS ofscience...Youhavetoconsiderthese thingsbythemselves.Theyareveryinter- THATAREN’TSO” estingbutyouhavetodiscoverthelawsthat governthem. InadditiontoperformingNobelPrize–winningre- search,theAmericanphysicistIrvingLangmuirex- Hissuspicionsarousedatthispoint,Woodused ploredextensivelyasubjecthecalled“pathological thecoverofthedarkenedroomtoremovetheprism science,”definingthisas“thescienceofthingsthat andputitinhispocket.WoodthenaskedBlondlot aren’tso.”Althoughheneverpublishedhisinves- torepeatsomeofhismeasurements.Withthecrit- tigationsonthissubject,hepresentedacolloquium icalcomponentoftheexperimentalapparatusmiss- onpathologicalscienceatGeneralElectric’sKnolls ing,Blondlotobtainedexactlythesameresults.In AtomicPowerLaboratoryonDecember18,1953. aletterthatwaspublishedinNature,Wood(1904) SubsequentlyRobertN.Hall,oneofLangmuir’sfor- exposedBlondlot’sexperimentsonNraysasacase mercolleaguesatGeneralElectric,transcribedand ofself-deception.AlthoughWood’sletterkilledre- editedarecordingofLangmuir’spresentationsothat searchonNraysoutsideFrance,itisinteresting itcouldbepublishedintheOctober1989issueof tonotethattheFrenchAcademyofScienceschose PhysicsToday.LangmuirandHall(1989)shouldbe Blondlottoreceivethe1904LeContePrize—even requiredreadingforeveryonewhopursuesacareerin thoughtheotherleadingcandidatewasPierreCurie, scientificresearch. whotogetherwithMarieCurieandHenriBecquerel Thisarticleisafascinatingaccountoffamous hadsharedthe1903NobelPrizeinphysicsforpio- casesofself-deceptionbyscientistsworkinginabroad neeringworkonradioactivity. diversityofdisciplines.Perhapsthemostremarkable LangmuirandHall(1989)alsodiscussanumber ofthesecasesconcernsthediscoveryofNraysby ofotheranomalousphenomena,andtheyanalyzethe theFrenchphysicistRenéBlondlotin1903.This mainsymptomsofpathologicalscience(orcargocult exoticformofradiationwasclaimedtopenetrate science,touseFeynman’smorecolorfulexpression). inchesofaluminumwhilebeingstoppedbythinfoils ThesesymptomsaresummarizedinTable1.The ofiron.WhenNraysimpingedonanobject,Blond- caseofNraysexhibitsallofthesesymptoms.It lotclaimedaslightincreaseinthebrightnessofthe isimportanttobearthesesymptomsinmindwhen object;butheadmittedthatgreatexperimentalskill consideringthevalidityofcertainclaimsmadeby wasneededtodetecttheeffectoftheserays. WoodwardandGoodstein(1996)aboutethicalcon- Duringtheperiodfrom1903to1906,over300pa- ductinscience.Numerouscasesofpathologicalsci- perswerepublishedonNraysby100scientistsand enceinvolvingpseudoscientificcranksarediscussed medicaldoctorsaroundtheworld(Nye1980).When inthebookFadsandFallaciesintheNameofSci- theAmericanphysicistRobertW.Woodlearned encebyMartinGardner(1957).Somefamouscases Conduct, Misconduct, and Cargo Cult Science 1407

Table1:Langmuir’sSymptomsofPathologicalScience

1.Themaximumeffectthatisobservedisproducedbyacausativeagentofbarelydetectableintensity, andthemagnitudeoftheeffectissubstantiallyindependentoftheintensityofthecause. 2.Theeffectisofamagnitudethatremainsclosetothelimitofdetectabilityor,manymeasurementsare necessarybecauseoftheverylowstatisticalsignificanceoftheresults. 3.Thereareclaimsofgreataccuracy. 4.Fantastictheoriescontrarytoexperiencearesuggested. 5.Criticismsaremetbyadhocexcusesthoughtuponthespurofthemoment. 6.Theratioofsupporterstocriticsrisesuptosomewherenear50%andthenfallsgraduallytooblivion.

ofself-deceptionbylegitimatescientistsaredetailed reticallyneutral”languageforrecordingob- onpages107–125ofthebookBetrayersoftheTruth servationsisimpossible. byWilliamBroadandNicholasWade(1982). Iclaimthatinthecontextofcomputersimulation experiments,thisstatementissimplyuntrue.Byus- 3.THELOGICALSTRUCTUREOF ingportablesimulationsoftware,wecanachieveex- SCIENCE actreproducibilityofsimulationexperimentsacross computerplatforms—thatis,thesameresultscanbe 3.1BaconianInductivismvs.DataSelection obtainedwhetherthesimulationmodelisexecuted Asabasisfortheirdiscussionofhowscienceactually onanotebookcomputerwitha16-bitoperatingsys- works,WoodwardandGoodsteinexaminecritically temoronasupercomputerwitha64-bitoperating thetheoriesofthescientificmethodthataredueto system.Moreover,theaccumulationofrelevantper- FrancisBacon([1620]1994)andKarlPopper(1972). formancemeasureswithinthesimulationmodelcan Baconianinductivismprescribesthatscientificinves- bepreciselyspecifiedinawaythatiscompletelyin- tigationshouldbeginwiththecarefulrecordingof dependentofanytheoryunderinvestigation.Thus observations;andasfaraspossible,theseobserva- wecanattainFeynman’sidealof“akindofutter tionsshouldbeuninfluencedbyanytheoreticalpre- honesty”inwhicheverysimulationanalysthasavail- conceptions.Whenasufficientlylargebodyofsuch ablethesameinformationwithwhichtoevaluatethe observationshasbeenaccumulated,thescientistuses performanceofproposedtheoreticalormethodolog- theprocessofinductiontogeneralizefromtheseob- icalcontributionstothefield.Inmyview,itisim- servationsahypothesisortheorythatdescribesthe possibletooverstatethefundamentalimportanceof systematiceffectsseeninthedata. thisadvantageofsimulatedexperimentation;andwe Onthecontrary,WoodwardandGoodsteinassert aredeeplyindebtedtothedevelopersandvendors that“Historians,philosophers,andthosescientists ofsimulationsoftwarewhohavetakenthetrouble whocarearevirtuallyunanimousinrejectingBaco- andexpensetoprovideuswiththetoolsnecessaryto nianinductivismasageneralcharacterizationofgood achievethereproducibilitythatisanessentialfeature scientificmethod.”WoodwardandGoodsteinargue ofalllegitimatescientificstudies. thatitisimpracticaltorecordalloneobservesand AccordingtoWoodwardandGoodstein,Baconian thatsomeselectivityisrequired.Theymakethefol- inductivismleadstothepotentiallyerroneousand lowingstatement: harmfulconclusionthatdataselectionandoverinter- pretationofdataareformsofscientificmisconduct, Butdecisionsaboutwhatisrelevantin- whilealessrestrictiveviewofhowscienceactually evitablywillbeinfluencedheavilybyback- workswouldleadtoadifferentsetofconclusions. groundassumptions,andthese...areoften Inmanyprominentcasesofpathologicalscience,the highlytheoreticalincharacter.Thevocab- rootoftheproblemwasdataselection(“cooking”) ularyweusetodescribetheresultsofmea- surements,andeventheinstrumentsweuse thatmayhavebeensubconsciousbutwasnonetheless tomakethemeasurements,arehighlyde- grosslymisleading.InadditiontothecaseofBlond- pendentontheory.Thispointissometimes lot’snonexistentNrays,LangmuirandHall(1989) expressedbysayingthatallobservationin andBroadandWade(1982)detailseveralothernote- scienceis“theory-laden”andthata“theo- worthycasesofsuchcookingandoverinterpretation 1408 Wilson ofexperimentaldatainthefieldsofarchaeology,as- dationinthefieldofcomputersimulation,weneed tronomy,geology,parapsychology,physics,andpsy- fundamentaladvancesinboththepracticeandthe- chology.Iclaimthatwhateverthetheoreticaldefi- oryofmodelvalidation.SofarasIknow,thesimu- cienciesofBaconianinductivismmaybe,theyhave lationliteraturecontainsverylittledocumentationof nobearingonthefieldofcomputersimulation;more- real-worldapplicationsinwhichasimulationmodel over,therearesoundpracticalreasonsforinsisting wascarefullyvalidated.Acomprehensivemethod- thatresearchersinallfieldsshouldavoidselectionor ologyforvalidatingsimulationmodelsisdetailedin overinterpretationofdatathathaseventheappear- KnepellandArangno(1993)andSargent(1996),but anceofpathologicalscience. itnotclearthatmanypractitionersandresearchers havegivendueconsiderationtoeithertheimplemen- tationortheextensionofthismethodology.Ibelieve 3.2Validatingvs.“Cooking”Simulation thatweneedtopaymuchgreaterattentiontosim- Models ulationmodelvalidationinteachingandresearchas Becausesimulationistsworkfarmorecloselywiththe wellasinpracticalapplications. endusersoftheirtechnologythanspecialistsinmany otherscientificdisciplines,wearesometimesexposed 3.3PopperianFalsificationism togreaterpressurefromclientsorsponsorstofudge or“cook”ourmodelstoyieldanticipatedordesired NextweturntothefalsificationistideasofKarlPop- results. Withtheadventofpowerfulspecial-and per.Accordingtothistheoryofthescientificmethod, general-purposesimulationenvironmentsincluding wetestahypothesisbydeducingfromitaprediction extensiveanimationcapabilities,suchmodel-cooking thatcanbetestedinanexperiment.Iftheprediction isfareasierforsimulationiststocarryoutthanitis failstoholdintheexperiment,thentheassociated for,say,atmosphericphysicists. hypothesisissaidtobefalsifiedandmustberejected. Inadditiontointentionalmodel-cooking,thereis ThusPopperianfalsificationismrequiresascientistto thedangerofunintentionalself-deceptionresulting holdahypothesistentatively,toexploreandhigh- fromfaultyoutputanalysis.Inmanyofthecases lightthewaysinwhichthehypothesismightbreak ofself-deceptiondocumentedinLangmuirandHall down,touncoverandscrutinizeevidencecontraryto (1989)andBroadandWade(1982),themostnotable thehypothesisratherthandiscardingorsuppressing commonfeaturewastheexperimenter’sattemptto suchevidence,andingeneraltoavoidexaggeration detectvisuallyanextremelyfaintsignalinsituations oroverstatementoftheevidencesupportingthehy- whereauxiliarycluesenabledtheexperimenterto pothesis.Perhapsthemostforcefulstatementofthis knowforeachtrialobservationwhetherornotthe viewofsciencewasgivenbyRichardFeynmaninthe signalwassupposedtobepresent.Forexamplein quotationatthebeginningofthisarticle. theN-rayexperimentsdescribedpreviously,Blondlot AccordingtoWoodwardandGoodstein,thereare couldseethescalemeasuringthecurrentpositionof alsoseriousdeficienciesinPopperianfalsificationism thethreadcoatedwithluminouspaint.Witheach asageneraltheoryofgoodscientificmethod: changeinthethread’sposition,Blondlotknewifhe Oneofthemostimportantoftheseissome- wassupposedtoseeabrighteningofthethread— timescalledtheDuhem-Quineproblem.We andthushewasabletodeceivehimselfinto“see- claimedabovethattestingahypothesisH ing”effectsthatotherexperimenterscouldnotre- involvedderivingfromitsomeobservational produce.Inthecontextofsimulationexperiments, consequenceO.Butinmostrealisticcases animationcanbeoneoftheprimaryvisualmeans suchobservationalconsequenceswillnotbe forself-deception.Equallydangerousisfaultyoutput derivablefromHalone,butonlyfromH analysisbasedonvisualinspectionofcorrelograms, inconjunctionwithagreatmanyother assumptionsA(auxiliaryassumptions,as histograms,confidenceintervals,etc.,computedfrom philosopherssometimescallthem)....It aninadequatevolumeofsimulation-generateddata. ispossiblethatHistrueandthatthereason Withallofthesesimulationtools,thereistheever- thatOisfalseisthatAisfalse. presentdangerofseeingthingsthatsimplydonot ...Itmaybetrue,asPopperclaims,that existorofnotseeingthingsthatdoexist. wecannotconclusivelyverifyahypothesis, Toguardagainstcookingasimulationmodelor butwecannotconclusivelyfalsifyiteither. itsoutputs,simulationistsshouldplacemuchgreater emphasisonmeaningful,honestvalidationoftheir Themostdistinctivefeatureofcomputersim- modelsasaccuraterepresentationsofthecorrespond- ulationexperimentsisthatthesimulationisthas ingtargetsystems.Toreemphasizetheroleofvali- completecontrolovertheexperimentalconditions Conduct, Misconduct, and Cargo Cult Science 1409 via(a)therandomnumberstreamsdrivingthe 4.THESOCIALSTRUCTUREOF simulationmodel’sstochasticinputprocesses,and SCIENCE (b)thedeterministicinputsgoverningmodelopera- tion.Thusinsimulatedexperimentationitispossible WoodwardandGoodsteinclaimthatultimatelyin- toisolatetheeffectsofauxiliaryassumptions,sothat ductivismandfalsificationismareinadequateastheo- theDuhem-Quineproblemcanbeeffectivelyresolved. riesofsciencebecausetheyfailtoaccountforthepsy- Howeverasseveralcolleagueshavepointedout,often chologyofindividualscientistsandthesocialstruc- practitionersfailtoevaluatetheeffectsofauxiliary tureofscience.FirstWoodwardandGoodsteincon- assumptionsinlarge-scalesimulationprojects.This sidertheroleofsocialinteractionsinscientificinves- failuremaybeduetothelackofawell-documented, tigation: widelyrecognizedmethodologyforaddressingthe Duhem-Quineprobleminthecontextofsimulation Supposeascientistwhohasinvestedagreat studies.Futuresimulationresearchshouldfocuson dealoftimeandeffortindevelopingathe- thedevelopmentofsuchmethodologytogetherwith oryisfacedwithadecisionaboutwhetherto continuetoholdontoitgivensomebodyof acomprehensiveinvestigationoftheconnectionsbe- evidence....Supposethatourscientisthas tweenmethodsforsolvingtheDuhem-Quineproblem arivalwhohasinvestedtimeandresources andmethodsforvalidatingasimulationmodel. indevelopinganalternativetheory.Ifad- BeyondtheirtheoreticalobjectionstoPopperian ditionalresources,creditandotherrewards falsificationism,WoodwardandGoodsteinclaimthat willflowtothewinner,perhapswecanrea- thisapproachhasseriouspracticaldisadvantages: sonablyexpectthattherivalwillactasase- verePopperiancriticofthetheory,andvice versa.Aslongasothersinthecommunity Supposeanoveltheorypredictssomepre- willperformthisfunction,failuretobehave viouslyunobservedeffect,andanexperi- likeagoodPopperianneednotberegarded mentisundertakentodetectit.Theex- asaviolationofsomecanonofmethod. perimentrequirestheconstructionofnew instruments,perhapsoperatingatthevery edgeofwhatistechnicallypossible,andthe Turningnexttothepsychologyofindividualsci- useofanovelexperimentaldesign,which entists,WoodwardandGoodsteinexplorethediffi- willbeinfectedwithvariousunsuspected cultyofsustainingthenecessarylong-termcommit- anddifficult-to-detectsourcesoferror.As mentoftimeandresourcestoahypothesiswithout historicalstudieshaveshown,inthiskind mentallyexaggeratingthesupportingevidenceand ofsituationtherewillbeastrongtendency downplayingthecontraryevidence—especiallyinthe onthepartofmanyexperimentaliststocon- earlystagesofaprojectwhenbeliefinthehypothesis cludethattheseproblemshavebeenover- maybeextremelyfragile: comeifandwhentheexperimentproduces resultsthatthetheorypredicted. Such Allthingsconsidered,itisextremelyhard behaviorcertainlyexhibitsanti-Popperian formostpeopletoadoptaconsistentlyPop- dogmatismandtheoretical“bias,”butit perianattitudetowardtheirownideas. maybethebestwaytodiscoveradifficult- to-detectsignal.Hereagain,itwouldbe Giventheserealisticobservationsaboutthe unwisetohavecodesofscientificconductor psychologyofscientists,animplicitcodeof systemsofincentivesthatdiscouragesuch conductthatencouragesscientiststobea behavior. bitdogmaticandpermitsacertainmea- sureofrhetoricalexaggerationregardingthe ThescenarioofWoodwardandGoodsteinisare- meritsoftheirwork,andthatdoesnotre- quireanexhaustivediscussionofitsdeficien- markablyaccuratedescriptionoftheexperimental cies,maybeperfectlysensible....Infact settinginwhichoccurredallofthecasesofpathologi- partoftheintellectualresponsibilityofasci- calsciencedetailedbyLangmuirandHall(1989)and entististoprovidethebestpossiblecasefor BroadandWade(1982).Moreover,thisscenariode- importantideas,leavingittootherstopub- scribesthenotoriouscoldfusionexperimentsofMar- licizetheirdefectsandlimitations. tinFleischmannandB.StanleyPonsasdocumented inthebookColdFusion:TheScientificFiascoof Incontrasttothispointofview,PeterMedawar,the theCenturybyJohnR.Huizenga(1993).Itseems winnerofthe1960NobelPrizeinmedicineforhis clearthatinsuchascenario,thescientist’sforemost workontissuetransplantation,madethefollowing concernshouldbetoavoidlapsingintoself-deception statementinhisbookAdvicetoaYoungScientist andpathologicalscience. (Medawar1979,p.39): 1410 Wilson

Icannotgiveanyscientistofanyagebetter Scientistsmustreportwhattheyhavedone advicethanthis:theintensityoftheconvic- sofullythatanyotherscientistcanrepro- tionthatahypothesisistruehasnobearing ducetheexperimentorcalculation. onwhetheritistrueornot.Theimportance ofthestrengthofourconvictionisonlyto Theyclaimthatsciencehasalarge“skill”or“craft” provideaproportionatelystrongincentive component,andthat tofindoutifthehypothesiswillstandupto criticalevaluation. Conductinganexperimentinawaythat producesreliableresultsisnotamatterof (TheemphasisinthequotedstatementisMeda- followingalgorithmicrulesthatspecifyex- actlywhatistobedoneateachstep. war’s.)LikeLangmuirandHall(1989),Medawar’s AdvicetoaYoungScientistshouldberequiredread- Thismaybetrueofsomeareasinthebiologicalsci- ingforindividualsatallstagesintheirscientificca- encesandotherexperimentalsciencesinwhichthe reers. behavioroflivingorganismsorthefunctioningof Overthepasttwentyyears,Ihaveaccumulated complicatedinstrumentationmaynotbewellunder- considerableexperienceinmediatingextremelyacri- stood,butthisdoesnotapplytocomputersimulation moniousdisputesbetweenresearchersactingas“se- experiments.Wecanandmustinsistonexactrepro- verePopperiancritics”ofeachother’swork.Muchof ducibilityofsimulationexperiments;andthisshould, thishard-wonexperiencewasgainedduringthenine infact,beamatteroffollowingpreciselystated,fully yearsthatIservedasadepartmentaleditorandfor- documentedalgorithms. merdepartmentaleditorofthejournalManagement Thereisofcoursealarge“craft”componentin Science.Toavoidreopeningwoundswhichhavenot buildingandusingsimulationmodels.Differentindi- hadmuchtimetoheal,Iwillnotgointotheparticu- vidualspresentedwiththesamesystemtobemod- larsofanyofthesecases;butIfeelcompelledtodraw eledwillneitherbuildidenticalsimulationsnorapply somegeneralconclusionsbasedonthesecases. thosemodelsinpreciselythesameway,justasdiffer- IneveryoneofthedisputesthatImediated,the entresearchersinanyotherscientificdisciplinewill troublestartedwithextensiveclaimsaboutthegen- neitherbuildthesameexperimentalapparatusnor eralapplicabilityofsomesimulation-basedmethod- carryoutexactlythesameexperimentalprotocolto ology;andthenfailingtovalidatetheseclaimsinde- studyagiveneffect.Neverthelessinthesesituations pendently,reviewersandotherresearchersproceeded differentsimulationistsshouldbeabletoreproduce towriteupanddisseminatetheirconclusions.This eachother’sresultsinordertojudgethesignificance inturngeneratedaheatedcounterreaction,usually andlimitationsoftheconclusionsbasedontheexper- involvingclaimsoftechnicalincompetenceortheft imentsinquestion.Moregenerally,thereisalarge ofideasorboth.EarlyinmycareerIservedasthe “craft”componentindoingsimulationresearchjust “specialprosecutor”inseveralofthesecases.Later asthereisalarge“craft”componentindoingother onImoveduptobecomethe“judge,”andintheend typesofscientificresearch—butthisstateofaffairs Iwasoftenforcedtoplaytheroleofthe“jury”as doesnotmitigatetheneedforreproducibilityofthe well.Ineveryoneofthesecases,ultimatelythetruth mainexperimentsassociatedwithsuchresearch. emerged(asitmust,ofcourse)—buttheprocessof sortingthingsoutinvolvedtheexpenditureofmas- 6.PEERSANDPUBLICATION siveamountsoftimeandenergyonthepartofmany dedicatedindividualsinthesimulationcommunity, 6.1IstheScientificPaperaFraud? nottomentionthenumerousprofessionalandper- sonalrelationshipsthatwereseverelydamagedalong WoodwardandGoodsteincitePeterMedawar’s theway.Insummary,Iclaimthatwhenindivid- (1991)paperentitled“IstheScientificPapera ualresearchersviolateFeynman’spreceptsof“utter Fraud?”toarguethatbecausemostarchivalpapers honesty”and“leaningoverbackwards,”thecostto inthescientificliteraturedonotaccuratelyportray thescientificenterpriseofpolicingtheseindividuals thewayscientificresearchisactuallydone,thesepa- rapidlybecomesexorbitant. persfailtomeasureuptoFeynman’sidealof“leaningoverbackwards.”Itiscertainlytruethatprimary journalarticlesinthescientificliteraturedonotdocu- 5.SCIENCEASCRAFT mentallofthemistakes,deadends,andbacktracking thatareaninevitablepartofvirtuallyeverysuccess- WoodwardandGoodsteinquestionthegeneralvalid- fulscientificinvestigation.Medawar(1982,p.92) ityofthefollowingprinciple: himselfadmittedthat Conduct, Misconduct, and Cargo Cult Science 1411

Ireckonthatforalltheuseithasbeentosci- ingaremuchworseinthesefieldsthaninthesimula- enceaboutfour-fifthsofmytimehasbeen tioncommunity.Perhapsthemostegregiousfailure wasted,andIbelievethistobethecom- oftherefereeingsysteminrecentyearswasthepub- monlotofpeoplewhoarenotmerelyplay- licationoftheinitialpaperoncoldfusionbyFleis- ingfollow-my-leaderinresearch. chmannandPons(1989a).Thispaperwaspublished Inmyview,thefundamentalissuehereisthat intheJournalofElectroanalyticalChemistryinjust theresimplyisnotenoughspaceinallthescientific fourweeks;andalonglistoferratasoonfollowed journalstodocumentthewaythatscienceisactu- (FleischmannandPons1989b)—includingthename allydone;moreovernoonehasthetimetoabsorball ofM.Hawkins,acoauthorwhowassomehowomit- thefinalresultseveninarelativelynarrowareaof tedfromtheoriginalpaper.Adetailedaccountof specialization,muchlesstoreadtheassociatedback- thisinfamousepisodecanbefoundonpp.218–220of groundmaterial.Nowadaysmanyhighschoolstu- Huizenga(1993). dentsaresufficientlysophisticatedtorealizethatpri- maryjournalarticlesarevehiclesforefficientlycom- 6.3RefereeingRemedies municatingsignificantdiscoveriesratherthanfordoc- umentingtheprocessesbywhichthosediscoveries Thetwomainreasonsforbreakdownsintheoper- weremade.Moreover,thisissueisrapidlybecoming ationoftherefereeingsystemare(a)misconceptions mootbecauseofcurrenttrendstowardcomplement- byrefereesaboutthejobtheyaresupposedtodo,and ingtheprintedversionofaprimaryjournalarticle (b)lackofincentivesfordoingagoodjobofreferee- withcomprehensivesupportingdocumentation(such ing.AsGleser(1986)pointsout,manyrefereesthink asappendicescontaininglengthyproofsordetailed thatamanuscriptmustbecheckedlinebylinefor descriptionsofexperimentalprotocols)archivedona errors;andseeingthatthiswillbeextremelytime- WorldWideWebserverthatismaintainedbythe consuming,theycontinuallyputoffthetask.On journal’ssponsoringorganization. thecontrary,thereferee’smainresponsibilityisto servetheeditorasan“expertwitness”inanswer- ingcertainkeyquestionsaboutthemanuscript—and 6.2ProblemswiththePeerReviewSystem mostofthesequestionscanbeansweredundertheas- FinallyWoodwardandGoodsteinexaminethepeer sumptionthatthemanuscriptiserror-free.Thesekey reviewsystemforevaluationofresearchproposals questionsaregiveninTable2andareelaboratedin andprimaryjournalarticles,concludingthatthecon- Forscher(1965),Gleser(1986),andMacrina(1995, flictofinterestinherentinaskingcompetitorstoeval- pp.84–89)alongwithgeneralguidelinesforreferee- uateeachother’sworkhasinflictedgenuinedistress ingthatshouldberequiredreadingforeveryresearch onthesystem.Inmyownexperience,byfarthe workerinthefieldofcomputersimulation. mostcommonformofmisconductbypeerreviewers Ifapaperpassestheinitialscreeningthatcon- hasnothingtodowithconflictsofinterest;instead sistsofansweringquestions1–8inTable2,thenit theproblemissimplederelictionofdutybyreviewers isnecessarytoundertaketheverificationoftechnical whocannotbebotheredtoreadandevaluatecare- correctnessrequiredtoanswerquestions9and10. fullytheworkofotherresearchers.Althoughthis Ifcompetentrefereeshadscrutinizedtheinitialpa- remarkappliestoevaluationofresearchproposalsas peroncoldfusionbyFleischmannandPons(1989a) wellasrefereeingofprimaryjournalarticles,Iam withtheobjectiveofansweringquestions9and10 mostconcernedwithproblemsinrefereeing.Inmy inTable2,thenthefatalflawsinthisworkwould judgment,theproblemofnonperformancebyrefer- havebeenuncoveredimmediately.Inmyviewitis eeshasreachedepidemicproportions,andIbelieveit imperativethatweprotectthesimulationliterature isurgentlynecessaryforthescientificcommunityto againstthelong-lastingstigmathatresultsfromper- addressthisscandalousstateofaffairs. mittingthepublicationoftechnicallyincorrectwork. InpreparingtheseremarksIsolicitedcomments Ifeveryoneinthesimulationcommunityfollowedthe fromnumerouscolleaguesnotonlyinthesimulation guidelinesinTable2forpreparingreferee’sreports, communitybutalsointhe“hard”scientificdisci- thenIbelieveourproblemswithpeerreviewwould plines,andIhavebeenstartledbythevehemence largelydisappear. oftheiragreementwithmyevaluationofthecurrent Additionaltipsoneffectiverefereeingaregiven stateoftherefereeingsystem.Basedonnumerous byWaser,Price,andGrosberg(1992).Asetofques- conversationswithcolleaguesinbiology,electricalen- tionssimilartothosegiveninTable2canbefound gineering,industrialengineering,mathematics,and onthehomepageoftheACMTransactionsonMod- statistics,Ihaveasensethatproblemswithreferee- elingandComputerSimulationbyusingtheURL 1412 Wilson

Table2:KeyQuestionstobeAnsweredinaReferee’sReport

1.Aretheproblemsdiscussedinthepaperofsubstantialinterest?Wouldsolutionsoftheseproblems materiallyadvanceknowledgeoftheory,methods,orapplications? 2.Doestheauthoreithersolvetheseproblemsorelsemakeacontributiontowardasolutionthatimproves substantiallyuponpreviouswork? 3.Arethemethodsofsolutionnew?Cantheproposedsolutionmethodsbeusedtosolveotherproblems ofinterest? 4.Doestheexpositionofthepaperhelptoclarifyourunderstandingofthisareaofresearchorapplication? Doesthepaperholdourinterestandmakeuswanttogivethepaperthecarefulreadingthatwegive toimportantpapersinourareaofspecialization? 5.Arethetopicandnatureofthispaperappropriateforthisjournal?Aretheabstractandintroduction accessibletoageneralreaderofthisjournal?Istherestofthepaperaccessibletoareadilyidentified groupofreadersofthisjournal? 6.Aretheclarityandreadabilityofthemanuscriptacceptable?Isthewritinggrammaticallycorrect? 7.Doesthemanuscriptcontainanadequatesetofreferences?Isadequatecreditgiventopriorworkin thefielduponwhichthepresentpaperisbuilt? 8.Isthematerialappropriatelyorganizedintoaneffectivemixoftext,figuresandtables?Aredatagiven intablesbetterpresentedinfiguresorinthetext? 9.Istheworktechnicallycorrect?Arethemainconclusionsjustifiedbytheexperimentaldataandby logicallyvalidarguments?Arethetheoremsstatedandprovedcorrectlygiventheassumptions?In practicalapplicationsofthetheoreticalresults,dotheauthorscheckthevalidityoftheunderlying assumptions? 10.Aretheregapsinthediscussionoftheexperimentalmethodsorresults?Iftherearesuchgaps,canthe closingofthesegapsbeconsidered(i)essential,(ii)desirable,or(iii)interesting?Aretheexperimental methodsdescribedinsufficientdetailsothatotherinvestigatorscanreproducetheexperiments?

http://www.acm.org/pubs/tomacs/review/review centralthesisissimplythis:asscientistsweshould .html. allstrivetoliveuptothestandardsofprofessional Thereremainsthequestionofadequateincentives conductsomemorablyarticulatedbyFeynman.So- forgoodrefereeing. Inreviewingpreliminaryver- phisticated(ormerelysophistic)rationalizationsof sionsoftheseremarks,severalindividualscomplained anythingshortofthisstandardservenoconstruc- aboutgenerallackofeditorialfeedbackon(a)the tivepurposeandshouldbeavoided.Inatimewhen strengthsandweaknessesoftheirreviews,and(b) publicesteemforsciencehasbeendamagedbyhigh- theissuesidentifiedinotherreferees’reportsonthe profilecasesofscientificmisconduct,weinthesimu- samepaper.Asaroutineprofessionalcourtesy,ed- lationcommunityhaveauniqueopportunitytolead itorsshouldincludesuchfeedbackwiththeirletters thewayinachievingFeynman’sidealsnotonlyinthe ofappreciationtoreferees.Moreover,editorsshould designandexecutionofourexperimentalprocedures strivetoensurethatindividualswhoprovideprompt butalsoinourcollectiveresponsetothechallenges andthoroughrefereeingwillreceivecomparableser- ofresponsible,professionalpeerreview. vicewhenthoseindividualssubmittheirownpapers forreview.Ultimatelyrefereeingisoneofthepro- ACKNOWLEDGMENTS fessionalresponsibilitiesthateachofusmustfulfill toensurethevitalityofourchosenfield,butdoing Althoughtheymaynothavefoundtheseremarks thisjobwellshouldbeasourceofprideandsatisfac- tobecompletelycongenial,IthankDavidGood- tioncommensuratewiththatofourotherprofessional steinandJamesWoodwardfortheircommentson contributionstothefield. thisarticle. Ialsothankthefollowingindividu- alsforinsightfulsuggestionsconcerningthisarticle: 7.CONCLUSION R.H.Bernhard,L.F.Dickey,S.E.Elmaghraby, andS.D.Roberts(NorthCarolinaStateUniv.); Toclosetheseremarks,Icomebacktotheopen- F.B.ArmstrongandB.J.Hurley(ABBPower ingquotationbyRichardFeynman.Inessencemy T&DCo.);C.Badgett(U.S.NavyJointWarfare Conduct, Misconduct, and Cargo Cult Science 1413

AnalysisCenter);K.W.Bauer(AirForceInsti- Huizenga,JohnR.1993.Coldfusion:Thescientific tuteofTechnology);R.C.H.Cheng(Univ.ofKent fiascoofthecentury.NewYork:OxfordUniver- atCanterbury);M.M.Dessouky(Univ.ofSouth- sityPress. ernCalifornia);P.L’Ecuyer(Univ.deMontréal); Knepell,PeterL.,andDeborahC.Arangno.1993. D.Goldsman(GeorgiaInstituteofTechnology); Simulationvalidation:Aconfidenceassessment P.Heidelberger(IBMT.J.WatsonResearchCen- methodology. LosAlamitos,California:IEEE ter);M.Irizarry(Univ.ofPuertoRico);R.W.Klein ComputerSocietyPress. (RegenstriefInstituteforHealthCare);R.E.Nance Langmuir,Irving,andRobertN.Hall.1989.Patho- (VirginiaPolytechnicInstituteandStateUniv.); logicalscience.PhysicsToday42(10):36–48. B.L.Nelson(NorthwesternUniv.);A.A.B.Pritsker Macrina,FrancisL.1995.Scientificintegrity:An (PritskerCorp.andPurdueUniv.);R.G.Sargent introductorytextwithcases.Washington,D.C.: (SyracuseUniv.);B.W.Schmeiser(PurdueUniv.); ASMPress. T.J.Schriber(Univ.ofMichigan);R.W.Seifert Medawar,PeterB.1979.Advicetoayoungscientist. (StanfordUniv.);A.F.Seila(Univ.ofGeorgia); NewYork:BasicBooks. P.M.Stanfield(ABCOAutomation,Inc.andNorth Medawar,PeterB.1982.Pluto’srepublic.Oxford: CarolinaAgriculturalandTechnicalStateUniv.); OxfordUniversityPress. J.J.Swain(Univ.ofAlabama–Huntsville);and Medawar,PeterB.1991.Isthescientificpapera M.A.F.Wagner(BoeingInformationServices).The fraud?InThethreatandtheglory:Reflections quotationbyRichardFeynmanappearingatthebe- onscienceandscientists,ed.DavidPyke,228– ginningofthisarticleisreproducedwithpermission 233.Oxford:OxfordUniversityPress. fromW.W.Norton&Company. Nye,MaryJo.1980.N-rays:Anepisodeinthehis- toryandpsychologyofscience.HistoricalStudies inthePhysicalSciences11(1):127–156. REFERENCES Onbeingascientist:Responsibleconductinre- search.1995.2ded.Washington,D.C.:National Bacon,Francis. [1620]1994. Thenovumorga- AcademyPress. num;withotherpartsof“Thegreatinstauration.” Popper,KarlR.1972.Thelogicofscientificdiscov- Chicago:OpenCourt. ery.3ded.London:Hutchinson. Broad,William,andNicholasWade.1982.Betrayers Sargent,RobertG.1996.Verifyingandvalidating ofthetruth.NewYork:SimonandSchuster. simulationmodels. InProceedingsofthe1996 Elliott,Deni,andJudyE.Stern,eds.1997.Research WinterSimulationConference,ed.J.M.Charnes, ethics:Areader.Hanover,NewHampshire:Uni- D.J.Morrice,D.T.Brunner,andJ.J.Swain,55– versityPressofNewEngland,fortheInstitutefor 64.Piscataway,NewJersey:InstituteofElectrical theStudyofAppliedandProfessionalEthicsat andElectronicsEngineers. DartmouthCollege. Waser,NickolasM.,MaryV.Price,andRichardK. Feynman,RichardP.1985.“Surelyyou’rejoking, Grosberg.1992.Writinganeffectivemanuscript Mr.Feynman!”:Adventuresofacuriouscharac- review.BioScience42(8):621–623. ter.NewYork:W.W.Norton&Co. Wood,RobertW.1904.Then-rays.Nature70 Fleischmann,Martin,andStanleyPons.1989a.Elec- (1822):530–531. trochemicallyinducednuclearfusionofdeuterium. Woodward,James,andDavidGoodstein.1996.Con- JournalofElectroanalyticalChemistry261(2A): duct,misconductandthestructureofscience. 301–308. AmericanScientist84(5):479–490. Fleischmann,Martin,andStanleyPons.1989b.Er- rata.JournalofElectroanalyticalChemistry263: AUTHORBIOGRAPHY 187–188. Forscher,BernardK.1965.Rulesforreferees.Sci- JAMESR.WILSONisProfessorandDirectorof ence150:319–321. GraduateProgramsintheDepartmentofIndustrial Gardner,Martin.1957.Fadsandfallaciesinthe EngineeringatNorthCarolinaStateUniversity.He nameofscience.NewYork:DoverPublications. wasProceedingsEditorforWSC’86,AssociatePro- Gleser,LeonJ.1986.Somenotesonrefereeing.The gramChairforWSC’91,andProgramChairfor AmericanStatistician40(4):310–312. WSC’92.Currentlyheservesasacorepresentative Honorinscience.1986.2ded.NewHaven,Con- oftheINFORMSCollegeonSimulationtotheWSC necticut:SigmaXi,TheScientificResearchSoci- BoardofDirectors.HeisamemberofASA,ACM, ety. IIE,andINFORMS.