tradeiphonewifi

AssociationforInformationSystemsAISElectronicLibrary(AISeL)ICIS2009ProceedingsInternationalConferenceonInformationSystems(ICIS)1-1-2009UnderstandingITInnovationsThroughComputationalAnalysisofDiscourseChia-jungTsuiUniversityofMaryland-CollegePark,ctsui@umd.
eduPingWangUniversityofMaryland-CollegePark,pwang@umd.
eduKennethR.
FleischmannUniversityofMaryland-CollegePark,kfleisch@umd.
eduDouglasW.
OardUniversityofMaryland-CollegePark,oard@umd.
eduAsadB.
SayeedUniversityofMaryland-CollegePark,asayeed@umd.
eduThismaterialisbroughttoyoubytheInternationalConferenceonInformationSystems(ICIS)atAISElectronicLibrary(AISeL).
IthasbeenacceptedforinclusioninICIS2009ProceedingsbyanauthorizedadministratorofAISElectronicLibrary(AISeL).
Formoreinformation,pleasecontactelibrary@aisnet.
org.
RecommendedCitationTsui,Chia-jung;Wang,Ping;Fleischmann,KennethR.
;Oard,DouglasW.
;andSayeed,AsadB.
,"UnderstandingITInnovationsThroughComputationalAnalysisofDiscourse"(2009).
ICIS2009Proceedings.
Paper102.
http://aisel.
aisnet.
org/icis2009/102ThirtiethInternationalConferenceonInformationSystems,Phoenix20091UNDERSTANDINGITINNOVATIONSTHROUGHCOMPUTATIONALANALYSISOFDISCOURSEResearch-in-ProgressChia-jungTsui,PingWang,KennethR.
Fleischmann,DouglasW.
Oard,andAsadB.
SayeedUniversityofMarylandCollegePark,MD20742{ctsui,pwang,kfleisch,oard,asayeed}@umd.
eduAbstractHowdoInformationTechnology(IT)innovationconceptsemerge,coexist,evolve,andrelatetoeachotherToaddressthisquestion,wetheorizethatinnovationconceptsareinterrelatedinanideanetwork,wheretheycanbelikenedtospeciesinacompetitiveandsymbioticresourcespace.
Communitiesoforganizationsandpeopleinterestedintheinnovationsproducediscoursethatbothreflectsandenablestheflowsofattentionamonginnovations.
Fromthisecologicalperspective,weapplydiscourseanalysistoinnovationresearchandproposecomputationalapproachtoscaleuptheanalysis.
Specifically,weemployedKullback-Leiblerdivergencetocomparethelinguisticpatternsof48ITinnovationsreportedinInformationWeekandComputerworldoveradecade.
Usingmultidimensionalscaling,wefoundthatsimilarinnovationsdemonstratedsimilardiscourses.
Theresultsdemonstratethevalidity,scalability,andutilityofcomputationaldiscourseanalysisforpractitionersandscholarstounderstandthesocio-technicaldynamicsintheITinnovationecosystem.
Keywords:Informationtechnologyinnovation,innovationconcept,discourse,computationalanalysis,Kullback-Leibler(KL)divergence,multidimensionalscalingGeneralTopics2ThirtiethInternationalConferenceonInformationSystems,Phoenix2009IntroductionOracle'srecenttakeoverofSunMicrosystemsandHP'sacquisitionofEDSearliersignifiesanimportantindustrytrend:Ontheonehand,thecurrenteconomiccrisisandtherelentlessdriveforgrowthpressureITvendorstoexpandanddiversifytheirofferingsbymergersandacquisitions.
Ontheotherhand,enterprisecustomersincreasinglypreferone-stopshoppingofintegratedinformationsystemswithouttheneedforcomplicatedplumbingin-house(TheEconomist2009).
DespitethetrendtowardconsolidationandintegrationinthemarketplaceforITproductsandservices,themarketplaceforideasthatunderlieITproductandserviceinnovationsremainmessyandfragmented(LyytinenandKing2004;PfefferandSutton2006;Wang2009).
Withminimalcost,anyonecanentertheideamarketplacewithaseeminglyinnovativeconcept.
Atanytime,numerousITconceptsarecompetingforthealreadythinattentionofpractitionersandscholars.
Whattheconceptsmeanandproposeisofteninconsistentandambiguous.
Thusfar,researchonITinnovationshasprimarilysoughttounderstandthesocialandtechnicaldynamicsintheITproduct/servicemarketplace(Fichman2004).
OurunderstandingoftheideamarketplaceforITinnovationsisstillinadequateaswefacethornyquestionsoftheoreticalandpracticalsignificance.
Ontoday'ssceneofITinnovations,Web2.
0andrelatedconceptsareintheprocessofyieldingthelimelighttoCloudComputing.
AsITinnovationsebbandflowconstantly,whatarethecurrentinnovationconceptsinthemarketplaceandwhatisemergingTheabilitytomonitorexistingandemerginginnovationsandtobemindfuloftheirimplicationsforspecificorganizationsisacriticalmanagerialcapability(SwansonandRamiller2004).
Alongwiththeemergenceofalmosteverynewconceptcomesthequestion:IsthisreallyneworjustoldwineinanewbottleForexample,isCloudComputingabrandnewideaorsimplyUtilityComputingrepackagedSuchsense-makingisnotonlylimitedtothecomparisonofthenewwiththeold,butalsonecessaryforunderstandingthecomplexrelationshipsamongconceptscoexistinginanideanetwork.
Forinstance,whatisthedifferencebetweenWebServicesandSoftwareasaService(SaaS)WhatistherelationshipbetweenvirtualizationandService-OrientedArchitecture(SOA)Asinnovationconceptsprogressthroughtheirdifferentiatedtrajectories,howdotheyevolveandwhatdoestheirevolutionmeantotheorganizationsandpeopleassociatedwiththeseinnovationsForinstance,doestheCustomerRelationshipManagement(CRM)conceptmeanthesamethingtodaythatCRMmeantadecadeagoDependingontheanswertothisquestion,avendormaychoosetocontinuepromotingitsofferingsundertheCRMbannerorswitchtoanewlabelorcategorythatcorrespondsmorewithitscurrentemphasisandcustomerpreferences.
Asaninnovationconceptevolves,howdoesthecommunityofpeopleandorganizationsassociatedwiththeinnovationevolveForexample,hasthediversecommunityforWeb2.
0becomefragmentedorcoherentinthecurrenteconomicmeltdownHavethediverseopinionsonWeb2.
0inthecommunitybeenconvergingordivergingWhatdoestheco-evolutionoftheinnovationanditscommunityimplyforthefateoftheinnovationThelackofknowledgeabouthowITinnovationconceptsemerge,coexist,co-evolve,andrelatetoeachotherisinpartcausedbytheoreticalandmethodologicallimitations.
Theoretically,thefocusofITinnovationresearchontheproduct/serviceformofinnovationshasthusfarprovidedonlyamodestnumberofinsightsforunderstandinginnovationsasconcepts.
Methodologically,mostinnovationstudiesweredesignedtoexamineonlyoneorafewinnovations,owingtothedifficultyinanalyzinglarge-scaledataonmultipleinnovations(StrangandSoule1998).
Thepresentstudyseekstoaddresstheselimitationsbyoffering(1)atheoreticalfoundationbuiltuponanecologicalviewofinnovationsand(2)ananalyticalmethodologyenabledbycomputationalanalysisofdiscourse.
Inwhatfollows,afterlayingthetheoreticalfoundation,weillustrateourmethodologywithanempiricalstudyof48ITinnovationsoveraten-yearperiod.
WeconcludebydiscussingtheutilityofourapproachforITinnovationresearchandpractice.
AnEcologicalViewofITInnovationConceptsInnovationconceptsarerelatedtooneanotherinmanyways.
First,abroaderconceptmaybecomprisedofnarrower,morespecificconcepts.
Second,differentconceptsmayrepresentthesamecoreidea.
Third,conceptsmaycompetewitheachotherasalternativesolutionstosimilarproblemsorfortheattentionfromthesamegroupofpeopleororganizations.
Finally,conceptsmaycomplementeachothertoaccomplishcommontasks.
Asinnovationsareinterrelated,theirevolutionarytrajectories(asindicatedbypopularityorperformanceforinstance)areinterrelatedtoo.
Itmaybehelpfultoconceptualizeanetworkofinnovationsaspartofanecologicalsystem,whereinnovationscanbelikenedtospeciesinacompetitiveandsymbioticresourcespace(Wang2009;WhittakerTsuietal.
/UnderstandingITInnovationsthroughComputationalAnalysisofDiscourseThirtiethInternationalConferenceonInformationSystems,Phoenix20093andLevin1975).
Innovationsrelyontheattentionfromcommunitiesoforganizationsandpeoplewithinterestsinproducingand/orusingtheinnovations.
Eachcommunityemergestomakesenseofaninnovationandorchestratematerialactivities.
Themembershipofthecommunityevolvesdynamically,asthecollectiveattentiontotheinnovationevolves.
Theflowsofattentionamonginnovationsarebothreflectedandenabledbydiscourse–whathavebeensaidandwrittenabouttheinnovations.
Whilethediscourseaboutaninnovationsometimesmanifestshumanactionsundertakenonbehalfoftheinnovation,oftenthediscourseitselfisaformofhumanaction,e.
g.
,tomakesenseof,promote,ordenouncetheinnovation(PhillipsandHardy2002).
Therefore,analysisofdiscourseaboutmultipleinnovationscanhelpusunderstandtheemergenceandevolutionofinnovationsandtheirrelationships.
Methodology:ComputationalAnalysisofDiscourseDiscourseanalysisofinnovationconceptspresentlyfacesamethodologicalchallenge:Discoursedataareoftenvoluminousandverylabor-intensivetocollectandanalyze.
Extantdiscoursestudiesofinnovationconceptshavetotradeoffbetweencasestudiesusingin-depthdataandlarge-scaleanalysisusingthinobservations(e.
g.
,citations).
Recentadvancesincomputationalanalysisofdiscoursehavemadeitpossibletoachievebothdepthandbreadthindiscourseanalysis.
Computationalorautomatedanalysisofdiscourseisalarge,activeinterdisciplinaryfieldwithavarietyoftheoriesandtechniques(seeOard2008foranon-technicalprimer).
Todemonstratetheutilityofcomputationaldiscourseanalysis,wehavechosenonetechniquesuitableforourinterestintheemergence,coexistence,co-evolution,andrelationshipsofinnovationconcepts.
Thistechnique,calledKullback-Leibler(KL)divergence(KullbackandLeibler1951),isessentiallyameasurethatquantifieshowcloseaprobabilitydistributionistoanotherdistribution.
ForprobabilitydistributionsPandQofadiscreterandomvariable,theKLdivergenceofQfromPisdefinedaslog(DPQPiPiQiKLi∑=.
KLdivergenceiscommonlyusedforcomparingtherelativefrequencyoftermuseinpairsofdiscourses(ManningandSchütze1999).
Beforewedetailouruseofthistechniqueinthisillustrativeempiricalstudy,weneedtodescribethediscoursedatawehavecollected.
DataCollectionTherearenumerousdiscourseoutlets,includingbooks,magazines,conferences,blogs,wikis,andmanyothers.
Specifically,wedownloadedallofthearticlespublishedduringaten-yearperiod(1998-2007)inInformationWeek,anITtrademagazine,usingtheLexis/Nexisonlinedatabase.
InformationWeekwasusedasanexemplaroutletoftheITinnovationdiscourse.
Meanwhile,wecompiledalistof48ITinnovationconcepts(Table1),rangingfromenterprisesoftware(e.
g.
,CRM)topersonalgadgets(e.
g.
,iPod),fromabstractconcept(e.
g.
,ArtificialIntelligence)toconcreteproducts/services(e.
g.
,YouTube),andfromhighlypopular(e.
g.
,e-business)tolesswell-knownconcepts(e.
g.
,digitalsubscriberline–DSL).
ThislistillustratesabroadrangeofITinnovationconceptsintheexaminationperiod.
WethenextractedfromtheInformationWeekarticlesalltheparagraphscontaininganyofITinnovationsonthelist.
Indoingso,weconsideredpossiblelabelsforeachinnovation,pluralforms,andacronymsuniquetotheinnovation.
Forexample,inextractingparagraphscontaining"digitalsubscriberline,"wealsoincludedparagraphsmentioning"digitalsubscriberlines"and"DSL.
"SomeITinnovationshadmanyparagraphsinthe10-yearperiodwhileothershaveonlyafew.
Forexample,thereweremorethan5,000paragraphsmentioningEnterpriseResourcePlanning(ERP).
Intotal,71,113paragraphswereextracted,withabout1,500paragraphsonaverageforeachinnovation.
DataAnalysisInthisdataset,eachinnovationisrepresentedbytheparagraphsmentioningtheinnovation.
TheuseoflanguageintheparagraphsconstitutesaprobabilitydistributionoverwordsandwecalculatedtheKLdivergenceforeachpairofinnovations.
Thecalculationgeneratesanasymmetric48x48matrixwitheachcolumnandrowrepresentingoneofthe48innovations.
Aftersymmetrization(byaveragingtheKLdivergenceineachdirection),thevalueineachcellofthematrixcanbeconsideredasthedistancebetweenapairofinnovations.
Inordertovisualizethedistancebetweeninnovations,weappliedmultidimensionalscaling(MDS)tothesymmetrizedKLdivergencematrix.
MDSisasetofstatisticaltechniquesforinformationvisualization.
Baseduponamatrixofitem-itemsimilaritiesordissimilarities,anMDSalgorithmassignsalocationtoeachiteminaGeneralTopics4ThirtiethInternationalConferenceonInformationSystems,Phoenix2009spacesuchthatthedistancesbetweentheitemscorrespondascloselyaspossibletothemeasureddissimilaritiesbetweentheitems.
Inotherwords,theproximityofitemstoeachotherinthespaceindicateshowsimilartheyare.
InMDS,onecanchoosethenumberofdimensionss/hewantsthealgorithmtocreate.
Generally,themoredimensions,thebetterthestatisticalfit,butthemoredifficultitistointerprettheresults.
Table1.
ListofInformationTechnologyInnovationConceptsAIArtificialIntelligenceMultimediaMultimediaASPApplicationserviceproviderMP3MP3playerATMAutomatedTellerMachineMySpaceMySpaceBIBusinessintelligenceOLAPOnlineAnalyticalProcessingBlogBlogOSSOpenSourceSoftwareBluetoothBluetoothOutsourceOutsourcingCADComputerAidedDesignPDAPersonalDigitalAssistantCRMCustomerRelationshipManagementRFIDRadioFrequencyIdentificationDigiCamDigitalCameraSmartCardSmartCardDLearnDistanceLearningSCMSupplyChainManagementDSLDigitalSubscriberLineSFASalesForceAutomationDWDataWarehouseSocNetSocialNetworkingeBizeBusinessSOAService-OrientedArchitectureeComeCommerceTelecommuteTelecommutingEDIElectronicDataInterchangeTabletPCTabletPCEgove-GovernmentUtiCompUtilityComputingERPEnterpriseResourcePlanningVirtualizationVirtualizationGPSGlobalPositioningSystemVPNVirtualPrivateNetworkGrpwareGroupwareWeb2.
0Web2.
0IMInstantMessagingWebServWebServicesiPhoneiPhoneWiFiWi-FiiPodiPodWikiWikiKMKnowledgeManagementWikipediaWikipediaLinuxLinuxYouTubeYouTubeMDSisadvantageousoverotherdimension-reductiontechniquessuchasfactoranalysisbecauseMDScanfitanappropriatemodelinfewerdimensionsthanothertechniques(Wilkinson1986).
Inaddition,amatrixofsymmetrizedKLdivergencemeasuresisappropriateinputforMDSbutnotforfactoranalysis.
Further,MDSallowsresearcherstogaininsightsintotheunderlyingstructureofrelationsbetweenitemsbyprovidingageometricalrepresentationoftherelations(DeunandDelbeke2000).
WeusedtheMDSprocedureinSPSSbasedontheALSCALoralternatingleastsquaresscaling(Takaneetal.
1977),themostpopularalgorithminMDS.
Forsimplicity,wechosetwodimensionsandpresentedthe48ITinnovationsinatwo-dimensionalscatterplot.
ResultsFigure1istheMDSplotofthe48innovations,withanR-squaredof0.
72,meaningthat72%ofthevarianceofthescaleddatacanbeaccountedforbytheMDSprocedure.
Tointerpretthisplot,wefollowedCoxon(2006)anddrewclosedcontoursaroundtheitemsthatweconsidercloselyrelatedinnovationsbasedonthelocationsoftheitemsandourownknowledgeoftheinnovations.
Theareassoenclosedrepresentregionsofrelativelyhighdensity,andtheextentoftheirdissociationisthedistanceinaMDSconfiguration(Coxon2006).
Forillustration,inFigure1wehaveidentifiedfivegroups,whichwedescribeonebyonebelow.
Group1includesWeb2.
0,socialnetworking,MySpace,blog,YouTube,wiki,andWikipedia.
Apparently,theyseemtobelongtotheWeb2.
0familybroadlydefined.
HencewenamedthisgroupWeb2.
0.
ThisgroupisclosetoOpenSourceSoftware(OSS).
WesuspectthatsomecommonattributessharedbyOSSandWeb2.
0technologies,suchasopenness,freedom,anduserparticipation,mayexplaintheproximity.
Tsuietal.
/UnderstandingITInnovationsthroughComputationalAnalysisofDiscourseThirtiethInternationalConferenceonInformationSystems,Phoenix20095Figure1.
MDSPlotofthe48ITInnovationsfrom10-yearInformationWeekDataWecountedthenumberofparagraphseachyearcontainingtheinnovationconceptsinGroup1andFigure2showsthepopularitycurvesoftheseinnovations.
Thenumberofparagraphsaboutaninnovationindicatestheprevalenceorpopularityoftheinnovationinthediscourse.
Interestingly,conceptsinthisgroupfollowedsimilarpatternsinpopularity:Everyconcepthadasignificantsurgearound2005and2006.
ThisfindingseemstosuggestthatitemsclosetoeachotherinaMDSplottendtofollowsimilarpopularitypatternsinthediscourse.
Group2hasteninnovationsandtwosub-groups(Subgroups2.
1and2.
2.
)areevident.
Subgroup2.
1includesWi-Fi,GlobalPositioningSystem(GPS),andBluetooth.
Subgroup2.
2includesiPod,iPhone,andMP3player.
Besidesthesesubgroups,Group2alsoincludesPersonalDigitalAssistant(PDA),multimedia,tabletPC,anddigitalcamera.
Subgroup2.
1seemstorepresentthewirelesstechnologiesformobiledevicesandSubgroup2.
2isaboutmobiledevicesthemselves.
Intuitively,wenamedGroup2mobiledevices.
ThepopularitycurvesfortheinnovationsinSubgroup2.
1arepresentedinFigure3.
SimilartotheinnovationsinGroup1,thethreeinnovationsinSubgroup2.
1hadsimilarpopularitypatterns.
However,thepopularitycurvesfortheinnovationsinSubgroup2.
2showninFigure4didnotfollowsimilarpatterns.
Rather,Figure4impliesthatiPhonemighthavesupersededoldertechnologiessuchasiPodandMP3players,suggestingthatnewinnovationsmayforceoldinnovationsout(AbrahamsonandFairchild1999).
Group3isthelargestgroupwith21innovationsintheupper-leftquadrantoftheplot(Figure1).
Ingeneral,theyareenterpriseITinnovationssuchasCRM,e-business,andERP.
ThepopularitycurvesforfiveinnovationsselectedfromGroup3arepresentedinFigure5.
Theseinnovationsexperiencedtheirpeaksaround1999and2000,andthentheirdiscoursesdwindled.
-1.
50-1.
00-0.
500.
000.
501.
001.
502.
002.
501.
501.
000.
500.
00-0.
50-1.
00-1.
50-2.
00MySpaceYouTubeBlogSocNetWikiiPodBluetoothGPSWiFiDigiCamPDAMultimediaOutsourceTabletPCVPNDSLRFIDCADDWGrpwareOLAPUtiCompEgovAIBIOSSIMWikipediaWeb2.
0MP3iPhoneSmartCardATMTelecommuteDLearnSFAeComLinuxSOAKMSCMVirtualizationEDIASPERPeBizCRMWebServ122.
12.
23GeneralTopics6ThirtiethInternationalConferenceonInformationSystems,Phoenix200901002003004005006007008001998199920002001200220032004200520062007ParagraphCount(Blog,MySpace,SocNet,Web2.
0)01020304050607080ParagraphCount(Wiki,Wikipedia,YouTube)BlogMySpaceSocNetWeb2.
0WikiWikipediaYouTube0501001502002503003504004501998199920002001200220032004200520062007ParagraphCountBluetoothGPSWiFiFigure2.
PopularityofConceptsinGroup1Figure3.
PopularityofConceptsinSubgroup2.
10501001502002503003501998199920002001200220032004200520062007ParagraphCountiPhoneiPodMP30500100015002000250030001998199920002001200220032004200520062007ParagraphCountASPCRMeBizeComERPFigure4.
PopularityofConceptsinSubgroup2.
2Figure5.
PopularityofConceptsinGroup3DiscussionValidityandAdvantagesoftheComputationalDiscourseAnalysisTheresultsfromtheKL-divergenceandMDSanalysisapparentlydemonstratethatinnovationswithsimilarcontentsand/orintrinsicrelationshipsarecloselylocatedinthetwo-dimensionalspatialrepresentationofthediscourse.
Whilethisfindingisunsurprisingtoanyonewithatleastbasicfamiliaritywiththeinnovations,theresultsprovidereasonableconfidenceintheinternalvalidityofthestudy'scomputationalapproachtodiscourseanalysis.
Tofurtherstrengthensuchconfidence,wecollectedallthearticlespublishedinComputerworld,anotherITtrademagazine,inthesameten-yearperiodandperformedthesameanalysis.
TheMDSplotbasedontheComputerworlddataturnedouttohaveadifferentorientation–innovationsinGroups1and2appearedintheleftsideofthechartandGroup3appearedontheright.
TheorientationoftheconfigurationofpointsinaMDSplotisoftenarbitraryregardingthecoordinateaxesandthustheplotisfreetorotateorflip(Shepardetal.
1972).
Exceptthedifferentorientationsoftheaxes,theMDSplotsbasedonthetwodatasetsareverysimilartoeachother.
Thisadditionalanalysissuggestsreasonableexternalvalidityofthe"KL-divergenceplusMDS"analyticalapproach.
Inadditiontointernalandexternalvalidity,thisapproachhasseveraladvantages.
Foremost,computationalanalysisisscalable.
Thestudyhasexaminedthediscourseon48innovationsintenyears,alreadysurpassingthescaleandscopeofmanyinnovationstudies.
Whilewehaveusedjusttwotrademagazinesforthisillustration,thecapabilityofthisapproachisnotlimitedtothenumberortypeofdiscourseoutlets.
Further,althoughourownknowledgeTsuietal.
/UnderstandingITInnovationsthroughComputationalAnalysisofDiscourseThirtiethInternationalConferenceonInformationSystems,Phoenix20097helpedvalidatethemethodsintheillustrationstudy,themethodsthemselvesdonotrelyonexpertknowledge.
Thisfeaturedifferentiatesourapproachfromotherclassificationmethodsbasedonexpertratingsoropinions(e.
g.
,Ein-DorandSegev1993;SwansonandRamiller1993).
Expertknowledgecanbeusefulforspecificresearchobjectives,butmethodsrelyingonexpertsarenotscalable.
Moreover,unlikescalableanalysisthatreliesonrelativelythinobservations,suchascitations(e.
g.
,Bettencourtaetal.
2006)orvocabulary(e.
g.
,AbrahamsonandEisenman2008),theKLdivergencemeasurecapturesboththevocabularyandtherichcontextofthevocabularyuseinthediscourse.
Overall,theseadvantagescreateamiddlegroundwherebothbreadthanddepthcanbeachievedindiscourseanalysis.
ImplicationsforITInnovationResearchandPracticeTheecologicalviewofITinnovationsandthecomputationaldiscourseanalysisareusefulforbothscholarsandpractitionerstounderstandtheemergence,co-existence,relationship,andevolutionofinnovations.
Weexplaintheimplicationsbelow,revisitingtheseriesofquestionsweraisedintheIntroduction.
UnderstandingEmergenceWeappliedourknowledgeofexistingITinnovationstovalidatethecomputationalapproachintheillustrativeempiricalstudy.
Whensuchknowledgedoesnotexist,asinthecaseofemerginginnovations,thesameanalysiscanbeappliedtothediscourseaboutnewinnovations,andtothediscourseaboutexistinginnovationsaswell.
Aninnovation'slocationintheMDSplotmayindicateitsbroadtypeanditsproximitytoexistingconceptswithinthesametypemayindicatenovelty.
InassessingthenewnessofCloudComputing,forexample,itwouldbeusefultocheckitslocationinreferencetothoseofotherinnovationssuchasUtilityComputingandWebServices.
UnderstandingCoexistenceandRelationshipWithregardtothecomplexrelationshipsamongexistinginnovations,theMDSplotbasedonKLdivergencecanhelpvisualizebroadcategories.
Forexample,inFigure1,Group2isaboutmobiledeviceswhileSubgroup2.
1isaboutwirelesstechnologies.
ThehierarchicalrelationshipillustratedbyGroup2andSubgroup2.
1suggeststhatmobiledevicesareenabledbywirelesstechnologies.
However,theMDSplotonitsowncannotfullyexplaintherelationshipsamonginnovations.
Aswehaveseen,thepopularitycurvesofcloselylocatedinnovationsmayfollowsimilarpatterns(e.
g.
,Figures2and5)ortheymaysignificantlydiffer,suggestingsubstitution(e.
g.
,Figure4)orcompetition.
Therefore,wesuggestcombiningtheuseofMDSplotbasedonKLdivergencewithtimeseriesanalysisofthepopularityofinnovations.
Thiscombinedapproachcouldbeusedtodetectthecomplementaryand/orcompetitiverelationshipsamongcoexistinginnovations.
UnderstandingEvolutionandCo-EvolutionOvertime,themeaningofanITinnovationconceptmaychangeandtherelationshipsamonginnovationsmayalsochange.
Forexample,intheearly1990s,CRMwasinitiallyconceptualizedasanautomationtoolforimprovingtheefficiencyofanorganization'ssalespeople,thenasabackbonetechnologyforenhancingtheeffectivenessofcustomerservices,andmorerecentlyasamarketinginnovationforbusinessintelligence(BI)gathering.
Consistentwiththisstory,Figure6showsthatCRMhadmovedawayfromSalesForceAutomation(SFA)by1998andmovedclosertoBIin2001.
Organizationsandpeopleininnovationcommunitiesaresensitivetothesechanges.
Forexample,thestatisticssoftwarecompanySASstrategicallymovedawayfromtheCRMlabelforitssoftwareproductstotheembracetheBIlabelaround2002(WangandSwanson2008).
Tostudytheevolutionofasingleinnovation,olderdiscourseandnewerdiscourseaboutthesameinnovationcanbeanalyzedandpositionedinthesameMDSplot,revealingtheevolutionarytrajectory.
Regardingtheco-evolutionofinnovationsandcommunities,itwouldbeusefultoanalyzethediscoursesofdifferentmembersinacommunity(vendordiscourseonCRMvs.
academicdiscourseonCRM)andcomparethelocationsofthemembersinMDSplots,discoveringtheleading,following,converging,ordivergingopinionsabouttheinnovation(Barleyetal.
1988).
GeneralTopics8ThirtiethInternationalConferenceonInformationSystems,Phoenix200919982001Figure6.
TheEvolutionofCRMNextStepsAspartofthisstudy,wearetakingthreestepstodomorein-depthanalysisoftheInformationWeekandComputerworlddata.
First,weareapplyinghierarchicalclusteringanalysistoKLdivergencematrixes.
Clusteringanalysiswillhelpusnotonlygrouptheinnovationssystematically,butalsodiscoverthehierarchicalstructureofinnovationsatfiner-grainlevels,possiblydetectingcommonalitiesanddistinctionsamongdifferenttypesofinnovationssuchasprocessvs.
productinnovations,management-focusedvs.
technology-focusedinnovations,andproductvs.
serviceinnovations.
Second,inadditiontotheITinnovations,weplantoaddtotheanalysiskeywordsthatrepresentmaindiscursivethemessuchascustomer,automation,end-user,andoptimization.
WewillassesstheextenttowhichITinnovationsclusteraroundthesekeywordsinMDSplotsinordertofurtherunderstandthemulti-dimensionalinnovationecosystem.
Third,weplantoexpandfromourpreliminaryanalysisoftheevolutionofCRMandrelatedinnovationstoalongitudinalanalysisofallinnovationsinourdata.
Wewillslicethedatabyyearandperformthesameanalysisoneachyear'sdata.
Thislongitudinalanalysiswilllikelyrevealthedynamicevolutionofinnovationsandtheirecosystem.
Goingbeyondthisstudy,weareexpandingtheInformationWeekandComputerworlddatafrom10yearsto20yearssothatwecanstudytheevolutionofmoreinnovationsoveralongerperiodoftime.
Thislargerdatasetwillallowustoinvestigatefurtherthecomplexrelationshipsamonginnovationsandfine-tuneourmethodstoteaseoutcompetition,complementation,substitution,andhierarchy.
Inaddition,recognizingthatthetwotrademagazinesonlyrepresentasmallportionofthelargerdiscourseintheinnovationecosystem,wewillcollectdatafromothertypesofdiscourseoutletssuchasacademicjournals,blogs,andwikis.
Weplantoassesstherobustnessofourapproachandlookforwardtodiscoveringinterestingdifferencesandqualifications.
Datafrommultiplesourceswillallowustoconstructamorerealisticrepresentationoftheinnovationnetworkandcommunities.
Finally,becausepositiveandnegativediscoursesmayhavedifferentiatedinfluencesonpopularity(Wang2009),weplantoenhanceourpresentcomputationaldiscourseapproachwithsentimentanalysis.
Suchlongerexaminationperiods,largerandbroaderdatasets,andricheranalysiswilllikelysustainourcontinuedresearchprogramontheITinnovationecosystem.
ConclusionInconclusion,theecologicalviewofITinnovationconceptsandthescalablecomputationaldiscourseanalysispresentedhereprovidethetheoreticalfoundationandmethodologyforscholarsandpractitionerstomonitorandmakesenseofITinnovationsintheideamarketplace.
TheprosperityandefficiencyofthatmarketplacedependontheknowledgeabouthowITinnovationsandcommunitiesemerge,coexist,andevolveinadynamicsocial-technicalecosystem.
Thisstudyandourbroaderresearchprogramwillcontributesuchcrucialknowledge.
-2.
00-1.
50-1.
00-0.
500.
000.
501.
001.
502.
001.
501.
000.
500.
00-0.
50-1.
00KMBICRMSFAeBizeComERPDWEDI-2.
50-2.
00-1.
50-1.
00-0.
500.
000.
501.
001.
501.
501.
000.
500.
00-0.
50-1.
00-1.
50BIKMEDIeBizeComERPDWSFACRMTsuietal.
/UnderstandingITInnovationsthroughComputationalAnalysisofDiscourseThirtiethInternationalConferenceonInformationSystems,Phoenix20099AcknowledgementsThispaperisbaseduponworksupportedbytheNationalScienceFoundationunderGrantsNo.
IIS-0729459andSBE-0915645.
WewouldliketothankLidanWangforhersuggestiontousesymmetrizedKLdivergence.
ReferencesAbrahamson,E.
,andEisenman,M.
2008.
"Employee-ManagementTechniques:TransientFadsorTrendingFashions"AdministrativeScienceQuarterly(53:4),pp.
719-744.
Abrahamson,E.
,andFairchild,G.
1999.
"ManagementFashion:Lifecycles,Triggers,andCollectiveLearningProcesses,"AdministrativeScienceQuarterly(44:4),pp.
708-740.
Barley,S.
R.
,Meyer,G.
W.
,andGash,D.
C.
1988.
"CulturesofCulture:Academics,PractitionersandthePragmaticsofNormativeControl,"AdministrativeScienceQuarterly(33:1),pp.
24-60.
Bettencourta,L.
M.
A.
,Cintron-Arias,A.
,Kaiser,D.
I.
,andCastillo-Chavez.
2006.
"ThePowerofaGoodIdea:QuantitativeModelingoftheSpreadofIdeasfromEpidemiologicalModels,"PhysicaA(364:2006),pp.
513-536.
Coxon,T.
2006.
"InterpretingConfigurations,"in:TheUser'sGuidetoMultidimensionalScaling.
pp.
93-116.
Deun,K.
V.
,andDelbeke,L.
2000.
"MultidimensionalScaling.
"OpenandDistanceLearning,retrievedApril15,2009,fromhttp://www.
mathpsyc.
uni-bonn.
de/doc/delbeke/delbeke.
htm.
TheEconomist,2009.
"MrEllisonHelpsHimself,"April25-May1,pp.
65-66.
Ein-Dor,P.
,andSegev,E.
1993.
"AClassificationofInformationSystems:AnalysisandInterpretation,"InformationSystemsResearch(4:2),pp.
166-204.
Fichman,R.
G.
2004.
"GoingBeyondtheDominantParadigmforInformationTechnologyInnovationResearch:EmergingConceptsandMethods,"JournaloftheAssociationforInformationSystems(5:8),pp.
314-355.
Kullback,S.
,andLeibler,R.
A.
1951.
"OnInformationandSufficiency,"TheAnnalsofMathematicalStatistics(22:1),pp.
79-86.
Lyytinen,K.
,andKing,J.
L.
2004.
"NothingattheCenter:AcademicLegitimacyintheInformationSystemsField,"JournaloftheAssociationforInformationSystems(5:6),pp.
220-246.
Manning,C.
,andSchütze,H.
1999.
FoundationsofStatisticalNaturalLanguageProcessing.
Cambridge,MA:MITPress.
Oard,D.
W.
2008.
"WhirlwindTourofAutomatedLanguageProcessingfortheHumanitiesandSocialSciences,"SymposiumonPromotingDigitalScholarship:FormulatingResearchChallengesintheHumanities,SocialSciencesandComputation,WashingtonDC,pp.
34-42.
Pfeffer,J.
,andSutton,R.
I.
2006.
HardFacts,DangerousHalf-Truths,&TotalNonsense:ProfitingfromEvidence-BasedManagement.
Boston,MA:HarvardBusinessSchoolPress.
Phillips,N.
,andHardy,C.
2002.
DiscourseAnalysis:InvestigatingProcessesofSocialConstruction.
ThousandOaksCA:SagePublications.
Shepard,R.
N.
,Romney,A.
K.
,andNerlove,S.
B.
1972.
MultidimensionalScaling;TheoryandApplicationsintheBehavioralSciences.
NewYork:SeminarPress.
Strang,D.
,andSoule,S.
A.
1998.
"DiffusioninOrganizationsandSocialMovements:FromHybridCorntoPoisonPills,"in:AnnualReviewofSociology,J.
HaganandK.
S.
Cook(eds.
).
PaloAlto,CA:AnnualReviews,pp.
265-290.
Swanson,E.
B.
,andRamiller,N.
C.
1993.
"InformationSystemsResearchThematics:SubmissionstoaNewJournal,1987-1992,"InformationSystemsResearch(4:4),pp.
299-330.
Swanson,E.
B.
,andRamiller,N.
C.
2004.
"InnovatingMindfullywithInformationTechnology,"MISQuarterly(28:4),pp.
553-583.
Takane,Y.
,Young,F.
W.
,anddeLeeuw,J.
1977.
"NonmetricIndividualDifferencesMultidimensionalScaling:AnAlternatingLeastSquaresMethodwithOptimalScalingFeatures,"Psychometrika(42:1),pp.
7-67.
Wang,P.
2009.
"PopularConceptsBeyondOrganizations:ExploringNewDimensionsofInformationTechnologyInnovations,"JournaloftheAssociationforInformationSystems(10:1),pp.
1-30.
Wang,P.
,andSwanson,E.
B.
2008.
"CustomerRelationshipManagementasAdvertised:ExploitingandSustainingTechnologicalMomentum,"InformationTechnologyandPeople(21:4),pp.
323-349.
Whittaker,R.
H.
,andLevin,S.
A.
1975.
Niche:TheoryandApplication.
Stroudsburg,PA:Downden,Hutchinson&Ross,Inc.
Wilkinson,L.
1986.
Systat:TheSystemforStatistics.
Evanston,IL:Systat,Inc.