InternationalJournalofInformationManagementxxx (2012) xxx–xxx
ContentslistsavailableatSciVerseScienceDirect
International Journal of Information Management
jo u rn a l h om epa g e :w w w . e l s e v i e r . c o m / l o c a t e / i j i n f o m g t
An empirical investigation of factors influencing the adoption of data mining tools
Tony Cheng-Kui Huang
a,∗, Chuang-Chun Liu
b, Dong-Cheng Chang
aaDepartmentofBusinessAdministration,NationalChungChengUniversity,Taiwan,ROC
bDepartmentofInformationManagement,Shu-ZenCollegeofMedicineandManagement,Taiwan,ROC
a r t i c l e i n f o
Articlehistory:
Available online xxx
Keywords:
Technologyacceptancemodel Behavioralintention Dataminingtools Informationsystems
a b s t r a c t
Previousstudiesexploredtheadoptionofvariousinformationtechnologies.However,thereislittle empiricalresearchonfactorsinfluencingtheadoptionofdataminingtools(DMTs),particularlyatan individuallevel.ThisstudyinvestigateshowusersperceiveandadoptDMTstobroadenpracticalknowl- edgeforthebusinessintelligencecommunity.First,thisstudydevelopsatheoreticalmodelbasedon theTechnologyAcceptanceModel3,andthenexaminesitsperceivedusefulness,perceivedeaseofuse, anditsabilitytoexplainusers’intentionstouseDMTs.Themodel’sdeterminantsinclude4categories:
thetask-orienteddimension(jobrelevance,outputquality,resultdemonstrability,responsetime,and format),controlbeliefs(computerself-efficacyandperceptionsofexternalcontrol),emotion(computer anxiety),andintrinsicmotivation(computerplayfulness).Thisstudyalsosurveysthemoderatingeffect ofexperienceandoutputqualityonthedeterminantsofDMTadoptionanduse.Anempiricalstudy involving206DMTuserswasconductedtoevaluatethemodelusingstructuralequationmodeling.
Resultsdemonstratethattheproposedmodelexplains58%ofthevariance.Thefindingsofthisstudy haveinterestingimplicationswithrespecttoDMTadoption,bothforresearchersandpractitioners.
© 2011 Elsevier Ltd. All rights reserved.
1. Introduction
With the issue of globalization becoming increasingly widespread, global competition among enterprises to profit isfiercernowthanithasbeeninthepast.Tofacethechallenges arisingglobally,moremanagersareusinginformationtechnology (IT)and informationsystems (IS)inbusiness. Thisallowsthem tobemoreefficientandaccuratewhenacquiringinformationor makingdecisions.AccordingtoaGartnerreport(Gartner,2011a), worldwideenterpriseITspendingisprojectedtototal$2.7trillion dollars in 2012,a 3.9% increase from 2011. Data warehousing technologyisone oftheparamountinvestmentsinestablishing ITinfrastructure(Gartner,2011b),enablingenterprisestocollect andstorevastamountsofdata.Thesedatacanbeextractedand analyzedbydataanalytics,helpingmanagersfindbetterwaysto generatevalueandcompeteinthemarketplace(Goeke&Faley, 2007;LaValle,Lesser,Shockley,Hopkins,&Kruschwitz,2011).A reportbyGartner(2011c)listsdata(next-generation)analyticsas oneofthetop10strategictechnologies.Dataanalyticshasbeen successfully usedin many fields,suchas insurance(Hopkins &
∗Correspondingauthorat:DepartmentofBusinessAdministration,National ChungChengUniversity,168,UniversityRd.,Min-Hsiung,Chia-Yi,Taiwan,ROC.
Tel.:+88652720411x34319;fax:+88652720564.
E-mailaddress:[email protected](T.C.-K.Huang).
Brokaw,2011),e-commerce(Kohavi,Rothleder,&Simoudis,2002), healthcarefraud(LaValle,Hopkins,Lesser,Shockley,&Kruschwitz, 2010),ande-retailing(LaValleetal.,2010).Dataanalyticshashad significanteffectsonbothoperationalandstrategicdimensionsin enterprises(Davenport&Harris,2007).
Forenterprises,datamining(DM)isadataanalysistechnology thatiswidelyapplicabletoavarietyofbusinesses,includingsales, marketing, and customer relations (Davenport & Harris, 2007;
Jackson,2002;Kohavietal.,2002).Softwarecompanieshaveinte- gratedDMfunctionsand launcheddataminingtools(DMTs)in marketstoassistusers(consumersofDMToutputs)performdata analysis.DMT can beusedto predictfuture trendsand behav- iorsfromhistoricaldata,allowingbusinessestomakeproactive, knowledge-drivendecisions(Sharma,Goyal,&Mittal,2008).Man- agerscanuseDMTstospotsalestrends,developsmartermarketing campaigns, and accurately predict customer loyalty (Rygielski, Wang, & Yen, 2002; Sumathi & Sivanandam, 2006). Therefore, DMTachievesthegoalsofimprovingcustomerservice,building long-termcustomerrelationships,reducingmarketingcosts,and increasingsales(Hui&Jha,2000;Nemati&Barko,2002;Rygielski etal.,2002;Shaw,Subramaniam,Tan,&Welge,2001).
In previousstudiesonDMT, computerscienceand informa- tionengineering scholarshavedeveloped variousalgorithms to improvetheefficiencyofDMT(Han&Kamber,2006).Davenport andHarris(2007)andLaValleetal.(2010)arguedthatthecrit- icaldeterminantofsuccessfullyusingdataanalyticsistoreduce 0268-4012/$–seefrontmatter© 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijinfomgt.2011.11.006
2 T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx thegapbetween humanbeings and technologies, and notjust
toenhance thelatter functions.Hence, users liketo useDMTs becauseitoffersavarietyoffunctionsandgoodalgorithmicper- formances,improvestaskperformances,anddecreasesmanagerial costs,i.e.,increaseseffectiveness.Thesameideaappearsinresearch onthetask-technologyfit(Goodhue&Thompson,1995).Because enterpriseshaveinvestedasubstantialamountoffundsintoDMT applications,examiningthefactorsthataffectDMTusersisabene- ficialresearchtopic.Dahlan,Ramayah,andMei(2002)andDahlan, Ramayah,andKoay(2002)addressedthereadiness oftelecom- municationemployeesandthebankingindustryinadoptingDM technologies.Chang,Chang,Lin,andKao(2003)studiedtheadop- tionofDMtechniquesinthefinancialserviceindustryusingfive characteristics:organizationalsize,organizationalculture,attitude ofdataresource,styleofdecision-making,andcompetitivenessof theoutsideenvironment.HuangandChou(2004)proposedanana- lyticalmodeltoexploretherelationshipsinfluencingthestageof webminingadoption.Althoughpriorstudiesinvestigatetheadop- tionofDMT,theydosofromtheperspectiveofthefirm(Changetal., 2003;Dahlan,Ramayah,&Mei,2002;Dahlan,Ramayah,&Koay, 2002;Huang&Chou,2004).Nostudiesinvestigatethisissueatthe individuallevelcompletely,asstatedbyGoodhueandThompson (1995).
ResearchersintheIT/ISdomainhavediscussedtheproblemsof individualadoptionandacceptanceusingdifferenttheoreticalfor- mulationsandconstructs.Thegoalofthesestudiesistounderstand andexplaintheimportantfactorsaffectingacceptance behavior andsubsequentIT/ISusage.FishbeinandAjzen(1975)developed awell-supportedbehavioraltheory,calledthetheoryofreasoned action(TRA),which describesthepsychologicaldeterminantsof behavior.In1989,Davisandhiscolleaguesproposedanextension oftheTRA(Davis,1989;Davis,Bagozzi,&Warshaw,1989),called thetechnologyacceptancemodel(TAM).Thismodelexaminesthe mediatingroleofperceivedusefulnessandperceivedeaseofuse intherelationshipsbetweensystemorindividualcharacteristics (externalvariables) andprobability ofsystemuse(anindicator ofsystemsuccess).TomaketheTAMmorecomplete,Venkatesh and Bala (2008) integrated themodels proposedby Venkatesh (2000)and Venkatesh and Davis (2000)and developed a com- prehensivenomologicalnetwork of ITadoption and use, called TAM3.TAM3investigates thedeterminantsofperceiveduseful- nessandperceivedeaseofuse,anddiscussesvariousinterventions thatcaninfluencetheknowndeterminantsofITadoptionanduse.
ThefindingsandresearchagendaofVenkateshandBala(2008) provideimportantimplicationsforITimplementation.Response timeandformatremainthemajorissuesintheDMresearchcom- munity(Chen,Han,&Yu,1996;Chung&Gray,1999), andhave beenstudiedbycomputerscienceand information engineering scholars. According to previous studies on DMTs, two signif- icant determinants primarily influence behavioral intention to useDMTs.
Asmentionedpreviously,previousresearchfailstoaddressthe problemofbehavioralintentiontouseDMTs.BecauseDMTisatype ofdecisiontoolforusers,researchonindividual-levelITadoptionis aparticularlyimportantpathtounderstandingDMTusage.There- fore,thisstudyproposesacomprehensivetheoreticalmodelbased onTAM3toaddressthisissue.Thefindingsofthisstudyhavesig- nificantimplicationsforDMTimplementationforresearchersand practitioners.
Theremainderof thispaper isorganizedas follows.Section 2 reviewstheliterature onDM and TAM3.Section 3 discusses theresearchmodelandhypotheses,andSection4describesthe researchmethodology.Section5presentsthefindingofanalyzing theempiricaldata.Section6discussestheresultsandconcludes thepaperwithcontributionsandimplications,discussionsofthe limitations,anddirectionsforfutureresearch.
2. Literaturereview
2.1. Datamining(DM)anddataminingtools(DMTs)
TheITcapabilitiesofbothgeneratingandcollectingdatahave beenincreasingrapidly.Thewidespreaduseofbarcodesinmany commercialproductsandthecomputerizationofmanybusinesses andgovernmenttransactionshaveprovideduswithhugeamounts ofdata.Thisinformationisstoredinthedatawarehousesofenter- prises,where itcanthenbeanalyzedtotransforminsightsinto action(LaValleetal.,2011).Datamining(DM)isadataanalysis technologythathasseenwidespreaduse(Jackson,2002;Kohavi etal.,2002).DMextractsimplicit,previouslyunknown,andpoten- tially usefulinformation from databases(Chen &Huang, 2008;
Chenetal.,1996).Itusespatternrecognitionlogictoidentifytrends inasamplingdatasetandextrapolatethatinformationagainsta largerdatapool.BecauseDMfirstbecameavailableforbusiness analytics,itsdevelopmenthasbecomeanimportantresearchfield inacademics(Chenetal.,1996;Chung&Gray,1999).
AlthoughthedevelopmentofDMisavailabletobusinesses,DM algorithmssuchasApriori,FP-growth,GSP,PrefixSpan,k-means, andC5(Han&Kamber,2006)mustbeimplementedonworkable programsbyJavaorC++,andarestillnotpracticalforendusers.
TheseprogramsareoftenexecutedbyDMexpertsorresearchers andnotendusersbecause(1)enduserscannotunderstandpro- gramsettingsand(2)theprograminterfaceisnotuserfriendly.
Therefore,manycompaniesandresearchinstituteshavedeveloped DMsoftware,i.e.,dataminingtools(DMTs),suchasSPSSClemen- tine,XLMiner,andWEKA,andsubsequentlyintroducedthemto themarket.ThreestatementsonDMTsaregivenasfollows.
(1)BecauseDMTsperformdataanalysisanduncoverimportant datapatterns,theycontributegreatlytobusinessstrategies, knowledgebases,andscientificandmedicalresearch(Han&
Kamber,2006).
(2)DMTs includesoftware componentsand theoriesthatauto- maticallysearchforsignificantpatternsorcorrelationsindata (Greenberg,2004).
(3)DMTsprovideindividualsand companieswiththeabilityto gatherlargeamountsofdataanduseittomakedecisionsona particularuserorgroupsofusers(Wisegeek,2011).
Basedonthestatementsabove,thisstudydefinesDMTfromthe Input-Processing-Output(IPO)perspective:softwarethatallows userstoinputtheirrequireddata-miningmodelsandarguments andthenperformthecomputationofthemodels.Thissoftware outputsimplicit,previouslyunknown,andpotentiallyusefulinfor- mationfromdatabasesbyvisualization.Theseoutputscanassist individualsandcompaniesinmakingdecisions.
Very few studies discuss the adoption of DMTs. Dahlan, Ramayah, and Mei (2002) studied the readiness of telecom- municationemployeestoadopt dataminingtechnologies.They investigatedthecontextualfactorsintelecommunicationorgani- zations,andhowthesefactorsinfluenceemployees’datamining readinessindex.Theirtheoreticalframeworkwasadaptedfroma modelforbuildinganalyticalcapability(Davenport,Harris,Russell, DeLong,&Jacobson,2001),statingthatthemoreanalyticallycapa- bletheindividual,thehigherthereadiness.Dahlan,Ramayah,and Koay(2002)appliedthesamemodeltothebankingindustry.Chang etal.(2003)exploredtheeffectsoforganizational attributeson theadoptionofDMTsinthefinancialservicesindustry.Inaddi- tion,HuangandChou(2004)focusedonthewebminingadoption ofB2Cfirmsintermsoforganizationalinnovation.Theirfindings reveal that a firm’s Internet strategy, business complexity, and internationalizedstrategy,alongwithcompetitivepressure,have aneffectonthestageofwebminingadoption.Thoughthesestudies
T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx 3 suggestthatorganizationalfactorsinfluencetheadoptionofDMTs,
noempiricalstudiesdiscusstheindividualfactorsaffectingDMT adoption.Therefore,thisstudyproposesatheoreticalmodeltopre- dictandexplainintentionofusingDMTsfromauser’sperspective.
2.2. ThetheoreticalbackgroundofTAM3
SinceitsintroductionbyDavis(1989),theTAMhasbeenused forpredictingtheacceptance, adoption,anduseofIT.TheTAM playsanimportantroleinthefieldofITadoptionandhasbeen appliedextensively insubsequentresearch.However,moreand moreresearchersarefindingflawsinthismodel(e.g.,theexclu- sionofsignificantvariables).Inresponsetothissituation,Taylor andTodd(1995)addedsomesignificantdeterminantstotheTAM.
VenkateshandDavis(2000)revisedtheoriginalmodelandpro- posed anextended model of thetechnology acceptance model (referred toas TAM2). TAM2 can be separated intotwo parts:
socialinfluenceandcognitiveinstrumentalprocesses.Venkatesh andDavissuggestedthatsocialinfluenceandcognitiveinstrumen- talprocessesarebothsignificantfactorsaffectinguseracceptance.
Theformerincludessubjectivenorms,voluntariness,andimages.
Thelatterincludesjobrelevance,outputquality,resultdemonstra- bility,andperceivedeaseofuse.VenkateshandDavisconducted longitudinalresearchwiththreepointsofmeasurement(i.e.,after initialtraining, onemonthlater, andthree monthlater)infour differentorganizations.TAM2,whichwasdevelopedthroughtwo processestoexplainindividualintentiontowardusingIT,indicates thattheseimportantdeterminantsaffectperceivedusefulness.
Venkatesh (2000) subsequently proposed the determinants of perceived ease of use to foster user acceptance and usage.
Venkatesh’sresearchmodelusesthreedifferentconstructs(con- trol, intrinsic motivation,and emotion)asgeneralanchors that change into adjustments (perceived enjoyment and objective usability)withincreasingexperience.Toensurethecompleteness ofTAMandenhancetherobustnessofthemodel,TAM3(Venkatesh
& Bala, 2008) integrated the studies by Venkatesh (2000) and VenkateshandDavis(2000)andexplaineduseracceptancemore deeply.
TheTAMhasbeenappliedandinvestigated inmanystudies (Chang,2010;Lin&Lu,2000;Liao,Palvia,&Lin,2006;Park,Roman, Lee,&Chung,2009;Pontiggia&Virili,2010;Teo,Lim,&Lai,1999;
Teo,2001; Youngberg,Olsen,&Hauser,2009)andcontinuesto beexploredbyscholars.TAM3synthesizesmanypreviousstud- iesontheTAMandproposesatheoreticalframeworktoexplain useracceptancemoredeeply.Inthismodel,therearefourtypes ofconstructs(individualdifferences,systemcharacteristics,social influence,andfacilitatingconditions)thatinfluenceperceiveduse- fulnessandperceivedeaseofuse.Longitudinalstudieshavebeen conductedonTAM3,includingsurveysduringthreedifferentperi- ods(after initialtraining, onemonthafterimplementation, and three monthsafterimplementation).Because fewstudiesmen- tiontheroleofinterventionbyorganizationsinassistingmanagers makedecisionsonITusage,TAM3proposesaresearchagendafor interventionandindicateshowitcanaffectindividuals’behavioral intentionsandacceptance. Throughdeliberateinvestigationand surveying,VenkateshandBala(2008)developedTAM3toiden- tifythefactorsandinterventionsthatplayrolesintheTAMfield.
Järveläinenand Vahtera(2010)combinedtheUnifiedTheoryof Acceptance and Use of Technology (UTAUT) withTAM3 tothe acceptanceofmobilesystems.
Traditionalinformationsystemsaregenerallyusedtomakepro- grammed decisions for organizational problemsolving through formal(routine) information processing(Lyytinen,1987;Varga, 2005).However,DMTsprovidenonprogrammeddecisions(Varga, 2005)and arecapableof solving nonroutineproblems through knowledgeofthepast (whathashappened), thepresent(what
is happening), and the future (what might happen)(Nemati &
Barko,2002).Thisapproachbridgesmanytechnicalareas,includ- ingdatabases,statistics,machinelearning,andhuman-computer interaction(Pechenizkiy,Puuronen,&Tsymbal,1998),makingit moredifficulttousethanotherIT(e.g.,onlineshoppingoronline games)ortypesofapplicationsoftware(e.g.,spreadsheetsorword processors).Becauseofitscomplexity,itis increasinglycompli- catedforuserstomakedecisionsontheadoptionandacceptance ofDMTs.Instead,usersfocusonutilitarianbenefitssuchasper- formanceandeffectiveness(i.e.,usefulness)(Rygielskietal.,2002;
Shawetal.,2001).ProfessionalstatisticalsoftwareisatypeofDMT.
DesigningDMTstobeeasytouseiscrucialtothesuccessofDM technologies(Smith,Willis,&Brooks,2000).Basedonthediscus- sionabove,this studyuses TAM3toinvestigatetheDMT usage because(1)TAM3isanintegratedmodelofthedeterminantsof perceivedusefulnessandperceivedeaseofuseand(2)TAM3isa morecomprehensivenomologicalnetworkthantheTAM.Theissue ofefficiencyisalsoaconcerninthefieldofDMT(Blockeel&Sebag, 2003).Therefore,thisstudyconsidersthefactorsofresponsetime andformat.Thefollowingsubsectionsexplainwhyresponsetime andformatarecontainedinthismodel.
2.3. Responsetime
ResponsetimeisoneoftheachievementindicatorsforDMTs.
ThefasteraDMT responds,themoreefficientitis.Theissueof responsetimehasalsobeenaddressedinothertypesofsystems.
BaileyandPearson(1983)analyzedcomputerusers’satisfaction withcomputer tools.They interviewed 32 middle managersin eightdifferentorganizationsandidentified39factorsofsatisfac- tion.The32middlemanagersrankedthese39importantfactors accordingtotheirlevelofimportance,andincludedresponsetime inthetop15factors.Conklin,Gotterer,andRickman(1982)con- ductedasetofexperimentstodeterminetheroleofresponsetime.
Theyexploredtheresponsetimeofanonlinesystembyrunning variousjobsinthebackground.Computermanagerscanusethese resultsasareferenceformakingdecisionswhenconductingsuch tasks.ResearchersintheDMfieldhavedevelopedmanyapproaches toimprovetheruntimeofDMalgorithms,suchasFP-growthfor miningassociationrulesandPrefixSpanforminingsequentialpat- terns(Han&Kamber,2006).Becauseoftheseefforts,DMTscanbe usedtotacklehighertime-complexityproblems.Hence,response timeisasignificantfactorinfluencinguserbeliefsandattitudesin usingDMTs(Chenetal.,1996;Chung&Gray,1999;Nelson,Todd,
&Wixom,2005).
2.4. Format
Hendrickson,Massey,andCronan(1993)showedthataneffec- tiveDMprocessrequirestheuserinadataexplorationprocess tocombinehisor herflexibility,creativity, andgeneralknowl- edgewiththeenormousstoragecapacityandthecomputational logicof computers.DMT formatdesignseekstointegrateusers intothedataexplorationprocess.Thebasicideaofthisapproach istopresenttheresultsinavisualformthatallowsuserstogain insightandgenerateconclusions.Formatdesignreducestheusers’
cognitivework,allowingthemtoperformmorespecifictasksand gainmoreunderstanding.Thisconceptisespeciallyusefulwhen aprimitiveideaisknownbutthegoalofexplorationisambigu- ous.Becauseusersaredirectlyinvolvedinthisprocess,thegoal ofexplorationmustbemodifiedaccordingtotheirrequirements.
Forexample,managerswouldliketoknowwhichcustomersbring thegreatest,average,orlowestprofitstocompanies.Therefore, theycanadjustDMTargumentsbyvisualizationtocomprehend theoutcomesfromthedifferentprofitlevels.Somestudiesinthe
4 T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx DMfieldusevisualtechniquestopresent miningresults(Keim,
2001;Kreuseler&Schumann,2002).
Nelsonetal.(2005)proposedameasureofusersatisfactionin theDMsetting.Theirstudyinvestigates465usersfromsevendif- ferentorganizations,indicatingthatusersatisfactionisdetermined byformat.Hence,thisstudytreatsformatasasignificantfactor affectinguserimpressionsofDMTthroughoutletdesigns;thatis, ifusersthinkthatthedisplayispresentedinausefulformat,they havemorepositivebeliefsonDMT.
3. Researchmodelandhypotheses
PriorTAMstudiesshowthatuserperceptionsofusefulnessand easeofusearethekeydeterminantsofindividualtechnologyadop- tion(Legris,Ingham,&Collerette,2003).TheTAMoffersseveral advantages,including itsbasis in social psychology,parsimony, validity,andinstrumentreliability.Responsetimeinthedimen- sionofsystemqualityandformatinthedimensionofinformation qualityaretwodeterminantsofDMTadoptionanduse.Thisstudy developsa theoreticalmodelbasedontheconceptsofTAM3to predictandexplaintheuseofDMTsbyindividuals.Thesedetermi- nantsaregroupedintofourcategoriesbasedonpreviousstudies (Venkatesh, 2000;Venkatesh &Bala, 2008; Venkatesh& Davis, 2000).Thesecategoriesincludethetask-orienteddimension(job relevance,outputquality,resultdemonstrability,response time, andformat), controlbeliefs (computerself-efficacyand percep- tionsofexternalcontrol),emotion(computeranxiety),andintrinsic motivation(computerplayfulness).Twointerventions,experience andoutputquality,areposited toinfluencethedeterminantsof DMTadoptionanduse.Fig.1illustratestheresearchmodel.
Section2reviewspreviousstudiesrelatedtoDMTsandTAM3.
Basedonthefindingsofthestudiesdiscussedabove,thisstudy presentsaresearchmodelandproposesthefollowinghypotheses.
3.1. Thetask-orienteddimension
AccordingtoTAM2(Venkatesh&Davis,2000),therearetwo major processes affecting theperceived usefulness of informa- tionsystems:socialinfluence(subjectivenorm,voluntariness,and image)and cognitiveinstrumental (job relevance, outputqual- ity,resultdemonstrability,andperceivedeaseofuse)processes, respectively.Thisstudyoverlookstheformerbecauseofthechar- acteristicsofDMTs.Accordingtothedecision-typeclassification proposedbyVarga(2005),adataanalysistechniquecanbeused tomakenon-programmeddecisions.Becauseexecutivesandman- agersorknowledgeworkerstypicallymakesuchdecisions,they generallyhavegreaterfreedomtodecidewhetherDMTshouldbe adoptedintheirwork.Anothersystem,datawarehousing(DW),is similartoDMT.Thistypeofsystemfunctionsasaspeciallyprepared repositoryofdatacreatedtosupportdecision-making(Hartono, Santhanam,&Holsapple,2007).BecauseDMTandDWhaveasim- ilarcharacteristic,decision-support,this studyreviewsprevious researchrelated to the adoption of DW(Foshay, Mukherjee,&
Taylor,2007;Goeke&Faley,2007;Gorla,2003).Thoughthesethe- oreticalmodelsarebasedontheTAM,theydonotincludesocial influence.Thesearetworeasonswhytheproposedmodeldoesnot includetheformer.Nevertheless,thelatterismoreimportantthan theformer.Therefore,thisstudyconsidersthelatterasasignificant dimensionfortheuseofDMT.
Thisstudyalsoproposestwo essentialdeterminantsofDMT use:responsetime andformat.Themainreasonwhymanagers use DMTs is to administrate affairs or plan important strate- gies for their company. Time is so precious that it cannot be wastedarbitrarily.Therefore,responsetimeisanimportantfactor here,especiallywhenusingDMTs.Managerscanfindsomeuseful
informationtoenhancetheperformanceoftheirwork.Theycan alsoretrieveusefulinformationifitcanbewellpresentedbyDMTs.
Anunderstandableandinterpretableoutcomecanhelpmanagers completetheirtasksmoreeasily.Consequently,formatistheother significantfactorinDMTusage.
Job relevance. Previous studies show that job relevance has aneffectonusers’perceived usefulnessofinformation systems (Venkatesh&Davis,2000).Theyindicatethatjobrelevanceisan individual’sperceptionofthedegreetowhichaninformationsys- temisapplicabletousers’jobs.Peopleuseinformationsystemsto completeorattainsomegoalsundernormalsituations.Goodhue andThompson(1995)proposedthetask-technologyfit(TTF)the- ory, which is similar to job relevance in its discussionof user acceptance.ThistheoryindicatesthatITismorelikelytohavean effectonauser’sperformanceandbeusedifitscapabilitiesfitthe tasksthattheusermustimplement.DMTsareadoptedinappli- cationsof information management, query processing, decision making,andprocesscontrol,todiscovertask-specificknowledge fromlargedatabases(Chenetal.,1996).Themathematicalmodels inDMTs,suchasassociationrules,sequentialpatterns,classifica- tion,andclustering(Han&Kamber,2006),canfacilitatemanagerial tasks. DMT has been successfully applied in business, includ- ingsalesmarketing,finance,banking,andinsurance(Sumathi&
Sivanandam,2006).Therefore,usersmustconsiderwhetherthese capabilitiesarehelpfultoaccomplishingtheirtasksbyadopting DMTs.Therefore,thisstudyhypothesizesthefollowing:
Hypothesis1. (H1)Jobrelevanceispositivelyassociatedwiththe perceivedusefulnessofaDMT.
Outputquality.Davis,Bagozzi,andWarshaw(1992)discussed theeffectsofoutputqualityonperceivedusefulness.Theysug- gestedthatoutputqualityisanessentialfactorintheuseofIS.
Outputqualityis“thedegreetowhichanindividualbelievesthat thesystemperformshis/herjobtaskswell”(Venkatesh&Davis, 2000).VenkateshandDavis(2000)foundthatjobrelevanceand outputqualityhaveaninteractiveeffectonperceivedusefulness.
Theyomittedthemaineffectsofjobrelevanceandoutputquality becausetheycouldnotbeinterpretedafterincludingtheinterac- tionterm.However,thisstudyretainsthedirecteffectsofoutput qualityandperceivedusefulness.DMTsareappliedtoanalyzethe logicofdataandrevealknowledgeandtrends.Outputqualityisan importantrequirementforDMTs.IfDMTsproducehigheroutput quality,usersmayfeelthatDMTsarehelpfulincompletingtheir tasks.Hence,thisstudyproposesthefollowinghypotheses:
Hypothesis2a. (H2a)Outputqualityispositivelyassociatedwith theperceivedusefulnessofaDMT.
Hypothesis2b. (H2b)Outputqualitypositivelymoderatesthe relationshipbetweenthejobrelevanceandperceivedusefulness ofaDMT.
Resultdemonstrability.UsersnoticeiftheresultsofIS canbe discriminatedeasilywhentheyuseit.Ifuserscannotgraspthe system’scharacteristicseasily,theyaredefinitelyunabletounder- standit.Inotherwords,theydonotrealizehowusefulthesystem reallyis.Evenagreatsystemcannotbeacceptedbyusersifthey havedifficultyattributingtheirjob performancetotheiruseof thesystem(Venkatesh&Davis,2000).Resultdemonstrabilityis
“the tangibility of theresults of using the innovation” (Moore
&Benbasat,1991).Userscangetmeaningful rulesorfindsome previouslyundiscoveredpatternsbyusingaDMT,thusgradually enhancingtheperception ofitsusefulness.However,theabove conditionworksonthepremisethattheyunderstandtheadvan- tages of using the DMT. Mining results can be interpreted by managerswho formulatestrategies in business. Therefore, it is
T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx 5
H9 H8 H7
H6 H5b
H5a
H4b H4a
H3 H2a
H2b H1
Intrinsic Motivation Emotion Control Beliefs
H12d H12c
H12b H12a
H11b H11a
H10 Job
Relevance
Behavioral Intention to Use
DMT Output
Quality
Result Demonstrability
Response Time
Format
Computer Self-efficacy
Perceptions of External Control
Computer Anxiety
Perceived Ease of Use
Perceived Usefulness
Experience Computer
Playfulness Task-Oriented
Dimension
Fig.1.Theproposedresearchmodelandhypotheses.
importanttounderstandtheadvantageofa toolthoroughly.As discussedabove,thisstudyhypothesizesthefollowing:
Hypothesis3. (H3)Resultdemonstrabilityispositivelyassociated withtheperceivedusefulnessofaDMT.
Responsetime.Nelsonetal.(2005) definedresponse timeas
“thedegreetowhicha systemoffersquickortimelyresponses torequestsfor informationoraction.”Thisisthemeasurement oftheperformanceofDMalgorithmsindiscoveringknowledge rules.Whentheamountof rawdata ishuge, it isnecessaryto devisemoreefficientalgorithmstoimprovetheruntimesofthe miningprocess.Therefore,computerscienceandinformationengi- neeringresearchershavedevelopedmanyapproachestoaddress thisproblem(Han&Kamber,2006).AsmentionedbyKohavietal.
(2002),thetimerequiredtoanalyzedatamustbereducedtonar- rowthegapbetweentherelevantanalyticswithDMTsandusers’
strategicbusinessneeds.Asforintuitiveness,thequickertherun- timesofDMTare,themorelikelyitisthatusersperceivethatthe DMTiseasytouseanduseful.Therefore,thisstudyhypothesizes thefollowing:
Hypothesis4a. (H4a)Responsetimeispositivelyassociatedwith theperceivedusefulnessofaDMT.
Hypothesis4b. (H4b)Responsetimeispositivelyassociatedwith theperceivedeaseofuseofaDMT.
Format. Format is “the degree to which information is pre- sentedin amannerthat isunderstandable andinterpretable to auser,andthusaidsinthecompletionofatask”(Nelsonetal., 2005).Thisisespeciallycriticalwhenassessingadecision-making andinformation-processingsystemsuchasaDMT.Information,in termsofpatternsorrules,oftenprovidesmanyinterestingresults tousers.However,itisdifficulttoidentifyvaluableinformation
6 T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx directly.ThisisbecauseaDMTisoftendesignedforquantitative
analysis,andlacksoutputthatistranslatedintolanguageandvisu- alizations(Kohavietal.,2002).Awell-formatteddesigntopresent theoutcomesofaDMTcancreateavirtualenvironmentthatallows userstounderstandtheresultsandfeelthattheyareuseful.Ifusers caneasilyoperatetheformatfunctionsofaDMTtoshowdifferent presentationsfortheresults,theyfeelthatitiseasytouseandhelps themcompletetheirtasks.Thisstudyhypothesizesthefollowing:
Hypothesis5a. (H5a)Formatis positivelyassociatedwiththe perceivedusefulnessofaDMT.
Hypothesis5b. (H5b)Formatispositivelyassociated withthe perceivedeaseofuseofaDMT.
3.2. Controlbeliefs,emotion,andintrinsicmotivation
Venkatesh(2000)showedhowtodecidethedeterminantsof perceivedeaseofuse.Thatstudypresentedandtestedananchoring (control,emotion,andintrinsicmotivation)andadjustment-based (perceivedenjoymentandobjectiveusability)modelofthedeter- minantsofsystem-specificperceivedeaseofuse. Incontrastto thecurrentstudy,Venkateshconductedalongitudinalstudy.Three differentperiods(post-training,onemonthpost-implementation, andthreemonthspost-implementation)wereconductedtomea- surethedifferenteffects ofthedeterminants onperceivedease ofuse. In thecurrent study,themainstress fallsonthecross- sectional aspect. Therefore, the proposedmodel only tests one period,lookingatuserswithdifferentexperiencesofusingDMTs.
Theanchorsinclude internal control,external control,emotion, andintrinsicmotivation.Internalcontrolisconceptualizedascom- puterself-efficacy,externalcontrolreferstofacilitatingconditions (perceptionsof external control),emotion representscomputer anxiety,andintrinsic motivationsignifiescomputerplayfulness.
Thisstudydiscussesthesedeterminantsandascertainshowthey influenceperceivedeaseofuseforDMTs.
Computer self-efficacy. Control is regarded as an individual’s perception of the availability of related knowledge, resources, support,oropportunitiesrequiredtocarryoutataskoraccom- plishaspecificpurpose.Ajzen(1991)andMathieson(1991)noted thatcontrolaffectsindividualintentionsandbehavior.Venkatesh (2000)developedthisideafurtherandreportedthatcontrolisone oftheimportantdeterminantsofperceivedeaseofuse,andcan beseparatedintointernal (computerself-efficacy) and external (perceptionsofexternalcontrol)controls.Computerself-efficacy istherefore“thedegreetowhichanindividualbelievesthathe/she hastheabilitytoperformaspecifictaskorjobusingthecomputer”
(Compeau&Higgins,1995).Usersprefertousetools(e.g.,DMTs) thattheyarecapableofusing.Becausethesuccessfuloperationof aDMTdependsoncomputer-basedskills,usersdislikeDMTsfor whichtheylackrelatedknowledge.Theconfidenceofanindivid- ual’scomputer-basedknowledgeandabilitiesserveasthebasisfor hisorherjudgmentonthelevelofdifficultyforDMTusage.This reasonablyleadstothefollowinghypothesis:
Hypothesis6. (H6)Computerself-efficacyispositivelyassociated withtheperceivedeaseofuseofaDMT.
Perceptionsofexternalcontrol.Perceptionsofexternalcontrol (i.e.,facilitatingconditions)are therelatedresourcesor techni- calinfrastructures in anorganization that help peopleperform theirjobssmoothlyusing thetoolstheybelieve in (Venkatesh, Morris,Davis,&Davis,2003).In thiscase, externalcontrolsare moreimportantfactors,astheyaffectuserbeliefsthattoolshave highlyeffectiveskillsorcomplicatedcharacteristics(e.g.,DMTs).
For instance, thesupport of a consultant is a kind of resource (Harrison,Mykytyn,&Riemenschneider,1997).Thiscanhelpusers solvetechnicalproblemsrelatedtotools,increasingthelikelihood
of theiruse. A DMT isa type of high-use-skill tool. Assuch, it isnoteasytousewithouttheappropriatesupport,training,and resources.Technicalinfrastructureisanotherkeypointrelatedto DMTs.Anorganizationmustbuildtheinfrastructureofadataware- housebeforeexecutingDMT.Therefore,thisstudyproposesthe followinghypothesis:
Hypothesis7. (H7)Perceptionsofexternalcontrolarepositively associatedwiththeperceivedeaseofuseofaDMT.
Computeranxiety.Computeranxietyisanemotionexperienced byITusers.Itisdefinedas“thedegreeofanindividual’sapprehen- sion,orevenfear,whenhe/sheisfacedwiththepossibilityofusing computers”(Venkatesh,2000).BecauseDMTsarehigh-use-skill tools,usersmayencountersometroubles.Theymayfeelanxiety, fearfulness,ormisgivingsbecauseoftheirunfamiliaritywithDMT usage.Becausetheydonotknowhowtousea DMTinitially,it maytakealongtimeforthemtolearnandoperateit.Forman- agers,anewDMTmaybreaktheusualroutine,andtheymusttry toadapttonewworkconditions.Ifusersbecomefamiliarwiththe functionalityofaDMT,theymayreducetheirinitialfeelingsofanx- iety.Somestudiesconfirmthatcomputeranxietyhasaneffecton behaviorandperformance(Anderson,1996).Therefore,thisstudy suggeststhatcomputeranxietyhasanegativeeffectonbeliefson usingaDMT.Inotherwords,ascomputeranxietyincreases,indi- vidualperceptionontheeaseofuseofaDMTdecreases.Therefore, thisstudyhypothesizesthefollowing:
Hypothesis8. (H8)Computeranxietyis negatively associated withtheperceivedeaseofuseofaDMT.
Computerplayfulness.Motivationisabasicaspectforusers,and isnecessarytominimizetheirphysicalpainandmaximizeplea- sure.Previousresearchindicatesthatmotivationhasasignificant effectonbehavior,andcanbecategorizedasintrinsicandextrin- sicmotivation(Vallerand,1997).Extrinsicmotivationcomesfrom outsideusers.Examplesincluderewards,money,orpromotions.
Incontrast,intrinsic motivationcomesfromtheuser’sinherent perceptionsofenjoymentorloveofaspecificbehaviororactiv- ity.Computerplayfulnessisatypeofintrinsicmotivationdefined as“thedegreeofcognitivespontaneityinmicrocomputerinterac- tions”(Webster&Martocchio,1992).Userswithaplayfulattitude experiencemorejoywhenusingaDMTthanthosewithoutplay- fulness.Theyusuallyenjoytheprocessofusageinsteadofsimply usingtheDMTtoaccomplishtasks.Forthisreason,theygenerally concentratetheirattentiononusingDMTandfeelthatDMTiseas- iertouse.Thus,computerplayfulnesshasapositiverelationship withtheeaseofusingaDMT.Therefore,thisstudyhypothesizes thefollowing:
Hypothesis9. (H9)Computerplayfulnessispositivelyassociated withtheperceivedeaseofuseofaDMT.
3.3. ConstructsoftheoriginalTAM
WhydopeopleuseIT?Alotofresearchinthisfieldexamines themotivesforusingIT.TheoriginalTAM(Davis,1989)hasledto manyfollow-upinvestigationsonthisissue.AlthoughtheTAMis notwithoutitsflawsandhasbeencriticizedbymanyscholars,it isthefoundationoftheuseracceptancemodelofIT.Anumberof studiespresentmodifiedorextendedTAMmodelsthataddressthe defectsoftheoriginalTAM.Areviewofthesestudiesrevealsthat bothperceivedusefulnessandperceivedeaseofuseareimpor- tantbeliefs associated withanindividual’sbehavioral intention touse,orattitudetowardusingIT(Davisetal.,1989;Venkatesh
&Davis, 2000).Thus, a DMT is essentiallya complex problem- solvinginformationsystem(Tejaswi,Prakash,Manaswi,Swarup Kumar,&Srinivas,2010).Assuch,theintentiontouseaDMTcan
T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx 7 beexplainedbytheTAM;thatis,bothperceivedusefulnessand
perceivedeaseofuseareindispensablefactorsaffectingbehav- ioralintention.TheeasieritistouseaDMT,themoreusefulan individualperceivesthatDMTtobe.Thatis,aneasy-to-usetoolpro- videshelpfulguidanceindoingjobs,andpeoplefeelthatitisuseful touse.Basedonthestatementabove,thisstudyhypothesizesthe following:
Hypothesis10. (H10)Perceivedusefulnessispositivelyassoci- atedwithusers’behavioralintentionstouseaDMT.
Hypothesis11a. (H11a)Perceivedeaseofuseispositivelyasso- ciatedwiththeperceivedusefulnessofaDMT.
Hypothesis11b. (H11b)Perceivedeaseofuseispositivelyasso- ciatedwithusers’behavioralintentionstouseaDMT.
3.4. Themoderatingvariable:experience
Todeterminewhethertheexperiencemoderatoraffectsindi- vidualbehavioralintentiontouseIT,somestudiesuselongitudinal researchtocollectdatafromdifferentperiods(Venkatesh&Davis, 2000;Venkateshetal.,2003).However,thisstudyfocusesonhow manyyearsanindividualusesDMT,andthereforeadoptsacross- sectionalapproach.Thisstudytreatsthenumberofyearsspent usingDMTasindividualusageexperiences.TAM3(Venkatesh&
Bala,2008)showshowexperienceplaysanimportantroleinthe moderatingeffect.Thisstudyfollowsasimilarconceptofexperi- encebutadoptsamoresuitablemethodofdatacollection.People usuallybecomeanxiouswhenusingDMTwhentheylackrelated experience,andfeelthatitisdifficulttouseaDMT.Increasedexpe- riencedecreasestheeffectofcomputeranxietyregardingtheease ofuseofaDMT.
Thenextchangerelatedtoexperienceiscomputerplayfulness.
Peoplewithcomputerplayfulnessoftenenjoytheprocessofusing aDMT,andmaydecidetouseitwithoutanyspecificpurposes.
Thisintrinsicmotivationurgesthemtousecomputersandaffects theirperceptionoftheeaseofuseofaDMT.Peoplewithabundant experienceusingcomputersoftenexperiencelessplayfulnessthan thoseintheearlystagesofusage.Theyfeelthatitisveryinteresting andcreativeinthebeginning.Astimegoesby,itmaybecomeboring anddrearybecauseoftheroutinenatureoftheprocess.Hence,the effectofcomputerplayfulnessonperceivedeaseofuseofDMTcan bemoderatedbyexperience.
PeopleperceiveITasusefulin theirworkiftheyfeelthatit iseasytouse(Davis, 1989).Theadditionalfactorofexperience moderatestherelationshipbetweenperceivedeaseofuseandthe perceivedusefulnessofaDMT.Userswithenoughexperiencegen- erallyhavemoreknowledgeofhowtouseaDMT,andoperateit moreaccuratelywithoutdifficulty.Therefore,thisstudyinfersthat experiencestrengthenstheeffectofperceivedeaseofuseonthe perceivedusefulnessofaDMT.
Davis(1989)proposedthatintentiontouseITisaffectedbythe perceivedeaseofuse.IfaDMTrequireslesseffort,individualsare inclinedtouseitandfeelthatitiseasiertooperatethantoatypeof DMTthatiscomplicatedtouse.Experiencealsoplaysasignificant rolehere,andhasareverseeffectontherelationshipbetweenper- ceivedeaseofuseandbehavioralintentionsregardingaDMT.An experiencedindividualfamiliarwithDMThaslessdifficultyusing it.Therefore,experienceattenuatestheeffectsofperceivedeaseof useonthebehavioralintentionofaDMT.Basedonthediscussion above,thisstudyhypothesizesthefollowing:
Hypothesis12a. (H12a) Experience negatively moderatesthe relationshipbetweencomputeranxietyandtheperceivedeaseof useofaDMT.
Hypothesis12b. (H12b) Experience negatively moderates the relationshipbetweencomputerplayfulnessandtheperceivedease ofuseofaDMT.
Hypothesis 12c. (H12c) Experience positively moderates the relationshipbetweenperceivedeaseofuseandtheperceiveduse- fulnessofaDMT.
Hypothesis12d. (H12d) Experience negatively moderates the relationshipbetweenperceivedeaseofuseandusers’behavioral intentiontouseaDMT.
4. Researchmethodology
Themeasurementitemsusedinthisstudywerederivedfrom prior studies and adapted slightly to suit thepurposes of this study.TheAppendixpresentsthequestionnaireusedinthisstudy.
Researchvariablesweremeasuredonaseven-pointLikertscale, rangingfrom“stronglydisagree”(1)to“stronglyagree”(7).
4.1. Pretestandpilottest
Toavoidproblemsofvaguewordingandmeaning,apretestwas performedtovalidatetheinstrument.Thepretest involvedtwo expertsintheDMfieldandtwoexperiencedDMTusers.Thesefour participantsreviewedtheentireinstrumentandprovided some commentsregardingthecontentoftheitemsinthesurvey.The contentvalidityoftheinstrumentwasconfirmedbasedontheir feedbackandinstructions.
Toverifyconsistencybetweentheconstructsandscaleitems,a pilottestwasconductedafterthepretest.Thepilottestinvolved 30subjects,allofwhomhadpreviouslyusedarelatedDMT,and marked the appropriate answers based on their experience of usage.Weappliedastatisticaltool(SPSS)totesttheinternalconsis- tencyandcalculatedtheindexofCronbach’s˛,whosevalueshould beabove0.7(Nunnally,1978).AlloftheCronbach’s˛coefficientsof therespectiveconstructsareabove0.7,rangingfrom0.723to0.962, confirmingthattheinstrumenthasnoproblemswithreliability.
4.2. Datacollection
Aweb-basedsurveymethodwasusedtocollectthedata.Sub- jects wererecruited from800peoplerandomlyselectedfroma listof2500informationmanagementandbusinessadministration alumniofauniversitywhoworkinenterprisesinTaiwan.Adirect mailcontainingahyperlinktoonlinesurveywebpageswassent toeachone,invitingthemtoparticipateinthissurvey.Therelated information(includingthedefinitionofDMTinSection2.1)was attachedtothequestionnairetoconfirmthattheserespondents werequalified(i.e.,theyhadpreviouslyusedaDMT).Themajority ofthesubjectswereofficeworkersfromsmallandmedium-sized enterprisesinTaiwan.Thissurveyattracted212respondentsto answerthequestionnaire,but6hadtobedroppedbecausethey wereoutliers(onlystronglydisagreeorstronglyagreetoallques- tionnaireitems)orrepeatedthesamevalues.Thus,thefinalsample consistedof206participants.
5. Dataanalysisandresults
SPSS(v. 15.0 for Windows) and Amos 6.0were adoptedfor thedataanalysisinthis study.Thestatisticalmethods included descriptivestatistics,reliabilityanalysis,confirmatoryfactoranal- ysis(CFA),andstructuralequationmodeling(SEM).Atwo-stage analysismethodof SEM,includingmeasurement andstructural models,wasusedfordataanalysis.
8 T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx
Table1
Descriptivestatisticsofrespondents.
Measure Items Frequency Percentage
Gender Male 134 65.05
Female 72 34.95
Age(years) Under20 1 0.49
21–30 105 50.97
31–40 65 31.55
41–50 33 16.02
Over50 2 0.97
Education College(2years) 1 0.49
Bachelor’sdegree(4years) 67 32.52
Master’sdegree 124 60.19
Ph.D. 14 6.80
Industry Manufacturing 38 18.45
Wholesaleandretailtrade 12 5.83 Informationand
communication
45 21.84
Financialandinsurance 30 14.56
Technology 26 12.62
Education 19 9.22
Others 36 17.48
Seniority Under1year 68 33.01
1–5years 60 29.13
5–10years 41 19.90
10–15years 20 9.71
Over15years 17 8.25
Yearsofusing DMTexperience
Under1year 96 46.60
1–2years 41 19.90
2–3years 25 12.14
3–5years 25 12.14
5–7years 12 5.83
7–9years 2 0.97
Over9years 5 2.43
Thesoftwareof DMTusage
MicrosoftSQLServer 59 28.64
WEKA 15 7.28
IBMIntelligentMiner 15 7.28
SPSSClementine 76 36.89
SPSSStatistics 95 46.12
Minitab 27 13.11
SAS 45 21.84
XLMiner 12 5.83
STATISTICA 32 15.53
Others 22 10.68
Themodelsof DMTusage
Associationrulesmining 72 34.95 Sequentialpatternmining 27 13.11
Dataclassification 123 59.71
Dataclustering 85 41.26
Regression 125 60.68
Timeseries 56 27.18
Others 6 2.91
5.1. Descriptivestatistics
Table1showsthedemographicdistributions,with65.05%male subjectsand34.95%femalesubjects.Mostsubjectsrangedinage from21to50yearsold.Thelargestagegroupwas21–30yearsold (50.97%).Mostsubjectsheldbachelorandmaster’sdegrees.Sub- jectswithmaster’sdegreesmadeupthemajority(60.19%).The informationandcommunicationindustryaccountedforthelargest percentageofoccupation(21.84%),withmanufacturingcomingin second(18.45%).Thelargestgroupofsubjectshadjobseniorityless than10years(82.04%),andmorethanhalfhadlessthan5yearsof seniority(62.14%).
Mostof thesubjects had relatively littleexperience using a DMT(46.60%ofthesampleshadusedaDMT lessthan1 year).
Thesesubjectsmay helpreveal thereasonsor factorsaffecting whypeopleintendtouseaDMT.Subjectsweredividedintotwo groupstodeterminewhetherthefactorofexperienceaffectsper- ceptionsordecisions:thosewholackexperience,andthosewith
Table2
TheresultsofKMOandBartletttests.
Dimensions KMOvalue Bartlett’stestofsphericity Chi-square df Significance
Jobrelevance 0.75 627.98 3 0.000
Outputquality 0.75 444.77 3 0.000
Resultdemonstrability 0.77 465.19 6 0.000
Responsetime 0.64 300.58 3 0.000
Format 0.75 432.34 3 0.000
Computerself-efficacy 0.61 264.86 3 0.000
Perceptionsofexternalcontrol 0.67 313.87 6 0.000
Computeranxiety 0.69 290.90 6 0.000
Computerplayfulness 0.69 202.10 6 0.000
Perceivedusefulness 0.87 953.26 6 0.000
Perceivedeaseofuse 0.67 313.87 6 0.000
Behavioralintentiontouse 0.63 221.31 3 0.000
significantexperience.Theformerincludesthosewhohadnotused DMTformorethanoneyear,andthelatterincludesthosewhohad usedDMTformorethanoneyear.Thisapproachissimilartothat adoptedbyBellman,Johnson,Kobrin,andLohse(2004)andXuand Gupta(2009).
SPSSStatisticswasthetoolmostcommonlyusedbythelargest numberofsubjects(46.12%oftheentiresample).SPSSClementine hadalsobeenusedbyalargenumber(36.89%oftheentiresample).
ManydifferentDMmodelscanbeappliedorselectedtomineuseful informationfromenormousamountsofdata.Regressionanalysis isthemostwidelyappliedmodel.Upto60%ofthesubjectshad previouslyusedthismodeltosolvespecificproblems.Dataclassi- ficationwasanothermodelcommonlyusedbythesubjects(59.71%
oftheentiresample).
5.2. Measurementmodel
Confirmingdatanormalityisthefirststepofmultivariateanal- ysis.Data normalitycanbetested bya Ztest of skewnessand kurtosis.Thisstudyachievedmultivariatenormalitybecausethe absolutevalueofskewnesswasunder3andtheabsolutevalueof kurtosiswasunder8(Kline,1998).
Factor analysis was performed using principal compo- nent extraction with varimax and Kaiser normalization. The appropriateness of factor analysis was determined by the Kaiser–Meyer–Olkin(KMO)testandBartlett’stest.TheKMOval- uesof the sampling adequacy fell withinthe acceptable range of 0.61–0.87, and theBartlett test of sphericity was significant (Table2),providingsupportforthevalidityoftheinstrument.
Someindices must becomputed and tested todemonstrate a reasonable fit for the model. These indices include the Chi- square/degrees of freedom (2/df), Goodness-of-fit Index (GFI), ComparativeFitIndex(CFI),NormedFitIndex(NFI),Incremental FitIndex(IFI),andRootMeanSquareofApproximation(RMSEA).
TheoverallmodelfitindicesforCFAwereasfollows:2/df=1.62, GFI=0.80,CFI=0.94,NFI=0.87,IFI=0.94,andRMSEA=0.05.Allthe indices inourmodel metthesuggestionsof othersresearchers (Bagozzi & Yi, 1988; Etezadi-Amoli & Farhoomand, 1996; Hair, Anderson,Tatham,&Black,1998).
Surveysofteninvolvemeasurementerrors.Toaccuratelyreflect themeasurementerrorinherentineachtest,thisstudyexamines thereliabilityandvalidityoftheproposedmodel.Severalreferral indicescanbeusedtotestreliabilityand validity.Table3indi- catestheresultsofreliabilityandvaliditywithCFA.Theinternal consistencyofeach constructwastested bycomposite reliabil- ity(CR).TheCR foreach constructinthisstudywasabove0.7, achievingthehighestpossiblereliabilitylevelsuggestedbyHair etal.(1998).Theaveragevarianceextractedshowswhatpercent- ageofthevarianceoftheconstructisexplainedbyanobserved
T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx 9
Table3
TheresultsofthereliabilityandvaliditywithCFA.
Construct Variable Factorloading t-Value Compositereliability Averageextractedvariance
Jobrelevance(JR)
JR1 0.90 20.98
0.95 0.85
JR2 0.95 24.55
JR3 0.92 –
Outputquality(OQ)
OQ1 0.90 18.72
0.92 0.80
OQ2 0.87 16.54
OQ3 0.91 –
Resultdemonstrability(RD)
RD1 0.90 16.62
0.92 0.75
RD2 0.90 16.58
RD3 0.87 16.75
RD4 0.79 –
Responsetime(RT)
RT1 0.82 17.79
0.91 0.77
RT2 0.85 17.80
RT3 0.95 –
Format(FOR)
FOR1 0.86 16.57
0.91 0.77
FOR2 0.89 17.46
FOR3 0.89 –
Computerself-efficacy(CSE)
CSE1 0.76 13.92
0.88 0.71
CSE2 0.91 17.37
CSE3 0.85 –
Perceptionsofexternalcontrol (POEC)
POEC1 0.92 19.68
0.90 0.68
POEC2 0.87 –
POEC3 0.79 14.82
POEC4 0.71 12.71
Computeranxiety(CA)
CA1 0.81 14.10
0.93 0.78
CA2 0.86 16.58
CA3 0.95 20.65
CA4 0.90 –
Computerplayfulness(CP)
CP1 0.86 11.65
0.88 0.65
CP2 0.82 11.62
CP3 0.80 10.86
CP4 0.74 –
Perceivedusefulness(PU)
PU1 0.93 –
0.96 0.85
PU2 0.94 25.89
PU3 0.92 25.62
PU4 0.90 20.61
Perceivedeaseofuse(PEOU)
PEOU1 0.83 –
0.93 0.77
PEOU2 0.89 17.52
PEOU3 0.89 17.67
PEOU4 0.90 15.05
Behavioralintentiontouse(BIU)
BIU1 0.89 –
0.90 0.76
BIU2 0.94 25.53
BIU3 0.78 16.67
variable.Iftheconstructspossessanaveragevarianceextracted thatexceedsthebenchmarkof0.5(Fornell&Larcker,1981),the average variance demonstrates satisfactory reliability and con- vergent validity. All the average variances extracted from the constructsinthisstudyrangedfrom0.65to0.85,exceedingthe acceptablevalueof0.5.
Discriminant validity is the lack of a relationship among measuresthattheoreticallyshouldnotberelated.Thus,therela- tionshipsbetweenconstructsmustbelowtoverifythesignificant difference of each construct. Table 4 lists the results of the
discriminantvalidity test.The diagonalsinTable4 aretheval- uesoftheaveragevariance extracted(AVE),andtheothersare thesquaredvaluesofthecorrelationswithotherlatentvariables.
ThesesquaredvaluesmustbelowerthantheAVEofeachconstruct toachievediscriminantvalidity(Fornell&Larcker,1981).These results show that this study achieves satisfactory discriminant validity.
Anexaminationoftheimportantindicesshownaboveensured that there were noproblems related tothereliability, validity, or model fit in this study. All the values achieved the levels
Table4
Squaredinter-correlationamongconstructs.
JR OQ RD RT FOR CSE POEC CA CP PU PEOU BIU
JR 0.85
OQ 0.16 0.80
RD 0.23 0.19 0.75
RT 0.14 0.44 0.18 0.77
FOR 0.08 0.43 0.18 0.48 0.77
CSE 0.06 0.07 0.16 0.09 0.10 0.71
POEC 0.29 0.16 0.32 0.18 0.19 0.16 0.68
CA 0.00 0.01 0.05 0.01 0.03 0.12 0.02 0.78
CP 0.07 0.06 0.21 0.11 0.08 0.18 0.12 0.17 0.65
PU 0.30 0.49 0.23 0.50 0.38 0.12 0.25 0.01 0.10 0.85
PEOU 0.11 0.32 0.17 0.38 0.41 0.15 0.23 0.02 0.10 0.50 0.77
BIU 0.12 0.25 0.12 0.27 0.25 0.08 0.15 0.01 0.06 0.45 0.45 0.76
10 T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx
−0.01 0.02
0.20*
0.15*
0.30***
0.02
0.24**
0.17*
0.26***
0.08*
0.18*** 0.18***
Intrinsic Motivation Emotion Control Beliefs Task-Oriented
Dimension
−0.23**
0.07
−0.02 0.33
0.46***
0.27***
0.36***
Job Relevance
Behavioral Intention to Use
DMT Output
Quality
Result Demonstrability
Response Time
Format
Computer Self-efficacy
Perceptions of External Control
Computer Anxiety
Perceived Ease of Use
Perceived Usefulness
Computer Playfulness
Experience
R2=0.58
R2=0.54 R2=0.74
Fig.2.Theresultsofthestructuralmodelinganalysis.Note:*p<0.05,**p<0.01,***p<0.001;dottedlinerepresentsnosignificance.
recommendedbypreviousresearchers.Theseinternalconsistency andvalidityresultsenabledustosubsequentlyestimatethestruc- turalmodel.
5.3. Structuralmodel
Thestructuralmodelinthisstudywasanalyzedtoidentifythe links betweenvariables. The constructs and theirhypothesized relationships were tested simultaneously. Thirteen hypotheses weresupported,whereastheothersixhypothesesdidnotreach thelevelofsignificance.Onlyonehypothesisofthemoderatorof experiencewassupported,andtherestwereunsupported.Fig.2 showsalltheresultsofthestructuralmodel.Thisresearchmodel explains 58% of the variance of behavioral intentions to use a DMT.
6. Conclusionandimplications 6.1. Discussionoffindings
The proposed model is based on TAM3 (Venkatesh & Bala, 2008),andexplainshowvariousdeterminantsinfluenceindividu- als’intentionstoadoptaDMT.Theempiricalfindingsofthisstudy presentsomeinterestingresults.
Perceivedusefulnessandperceivedeaseofusearetwoimpor- tantfactorsaffectingintentionstouseIT.Thesefactorshavebeen confirmedbymanyresearchersinthepast(Davis,1989;Venkatesh
&Davis,2000).Notsurprisingly,thesefactorsalsohaveasignificant effectonanindividual’sintentionstouseaDMT.Perceiveduseful- nessandperceivedeaseofusedirectlyinfluenceusers’behavioral intentions. Perceived ease of use also has a direct effect on
T.C.-K.Huangetal./InternationalJournalofInformationManagementxxx (2012) xxx–xxx 11 perceivedusefulness.HumanbeingsattempttouseDMTswhen
theyperceiveittorequirelesseffortorcreatebenefitsintheirwork.
IftheyexpendlessefforttouseaDMT,theirperceptionofuse- fulnessincreases.However,thewillingnesstouseDMTsincreases spontaneously.Theseresultsconfirmthatthetwobeliefsdiscussed above,theperceivedusefulnessandperceivedeaseofuseofaDMT, arethemostimportantfactorsaffectinguserintentions.Therefore, thisstudyconductsfurtherresearchtoidentifythekeyantecedents affectingthetwobeliefs.
Theeffectsofjobrelevanceonperceivedusefulnessaresignifi- cantinthismodel.ThegreatertherelationshipbetweenaDMTand theuser’sjob,thegreateritsperceivedusefulness.Theseresults also support result demonstrability as a significant factor that directlyinfluencesperceivedusefulness.Makingtheminingout- comesofaDMTmoreunderstandableforusersisanimportanttask.
Theeasieritisforuserstointerpretfindings,themoretheyperceive theDMTtobeuseful.Thisisconsistentwithpreviousfindingsfrom VenkateshandDavis(2000).TheresultsofTAM2revealaninter- activeeffectofjobrelevanceandoutputqualityindetermining perceivedusefulness.BecauseofthisspecialDMTcharacteristic, outputqualityisacriticalfactorinthisstudy.Withoutthesup- portofexcellentoutputquality,usersdonotacceptatool,evenif astrongrelationshipexistsbetweenjobrelevanceandperceived usefulness.Theproposedresearchmodelnotonlyteststhisinter- activeeffect,butalsoexaminesthedirecteffectsofjobrelevance andoutputquality.Itcomesasnosurprisethatboththeinterac- tiveanddirecteffectsaresignificant.BecauseaDMTisdesigned torevealnewknowledgeandassistmanagersinmakingdecisions, pooroutputqualitymaycausemanagerstomakewrongdecisions, resultingingreatlosses.Thus,inappropriateDMToutcomescan causeseriousproblems.Usersbasetheperceivedusefulnessofa DMTonthequalityofitsoutcomes.Outputqualityalsopositively moderatestheeffectsofjobrelevanceontheperceivedusefulness ofaDMT.Theseresultsshowthatthedegreeoftheeffectofjob relevanceonperceivedusefulnessisreinforcedwhentheoutput qualityisexcellent.
The other two essential factors that should be noted are responsetimeandformat.ThesourceprocessedbyaDMT usu- allycomesfromvolumesofenterprisedata.Howtocopewitha highvolumeofdataefficientlyandquicklyisatrialformanagers.
Thus,theresponsetimeofaDMTisaprimaryconcern.Response timeisasignificantantecedentaffectingperceivedusefulnessand perceivedease ofuse. Timeis animportant resource andlimi- tationinbusiness,especiallyfordecision-makingmanagers.The fasteraDMTruns,themoreusefulitsusersperceiveittobe.Users perceivethatitiseasiertouseaDMTifitrequiresnopatience toexecutetheminingprocess.Formatisanothervitalpredictor ofintentionstousea DMT.However,thehypothesesrelatedto formatareonlypartiallysupported.Formatdoesnothaveasig- nificantlypositiveeffectonperceivedusefulness,butdoeshave apositiveeffectonperceivedeaseofuse.Thisindicatesthatfor- matdoesnot influence thebelief ofperceived usefulness even withwell-demonstratedoutput.Thisresultcouldbeexplainedby consideringuserobjectiveswhenadoptingDMT.Themaingoalis toextractusefulknowledgefromvolumesofdataandthenpro- ceedwithfurtheranalysis.Therefore,awell-formattedoutputis insufficienttosupporttheusefulnessofadoptingthistool. For- matisasignificantantecedentaffectingaDMT’sperceivedeaseof use.
Thisstudyrevealsthatalmostalltask-orientedantecedentsare significantlyrelatedtoperceivedusefulness(exceptfortherela- tionshipbetweenformatandperceivedusefulness).Theseresults revealtheinstrument-orientedcharacteristicofDMTs.Moreover,a DMTcanbeusedtohelpuserssolvesomespecificproblems.Con- sequently,thesefactorsmustbeconsideredwhenanenterprise adoptsaDMT.
DMTismoredifficulttousethanotherIT(e.g.,onlineshop- pingoronlinegames)orvarioustypesofapplicationsoftware(e.g., spreadsheetsorwordprocessors). Becauseofthecomplexityof DMT,it isincreasinglydifficultfor employeestomakeeffective decisionsontheadoptionand acceptancewithDMT.Computer self-efficacycanplaya rolein increasingperceived easeofuse.
Theresultsofthisstudyshowthatcomputerself-efficacyincreases aDMT’sperceivedeaseofuse.OneofaDMT’scharacteristicsis thebarriertoentry,which makesitdifficulttolearn andoper- ate.If users can easily operatea computeron theirown, they areabletousea DMTmore easily.Dahlan, Ramayah,andKoay (2002)showedthatuserskillsandexperience(i.e.,computerself- efficacy)ispositivelyassociatedwithDMusage.Manyfirmshave recently madean efforttodevelop and introduce IT. However, thecommonproblemtheyencounteristhatnumerousemployees mayresistorganizationalchangebecausetheylosetheirauthority or positionin their organization.Controlling employee percep- tionsofexternalcontrolcouldbeameanstosolvethisproblem.
Theempiricalresultsofthis studyindicatethattheappropriate resourcesorsupportsinanorganizationcanhelpusersperform theirjobssmoothlyusingDMTs.Providingresources,opportuni- ties,andknowledgedecreasesemployeeresistancetousingDMTs and increasestheirperceived easeof use. Companies that pro- vide training programs beforeintroducing a DMT candecrease employee resistance to the DMT. In other words, the training coursesenhanceemployees’understandingandskillsrelated to using theDMT. Both computerself-efficacy and perceptions of externalcontrolhavedirecteffectsontheperceivedeaseofuseof aDMT.TheseresultsagreewiththefindingsofVenkatesh(2000) andChangetal.(2003).
Emotionandintrinsicmotivationfactorsincludecomputeranx- ietyandcomputerplayfulness.However,theresultsofthesetwo predictorsarenotconsistentwithpreviousfindings.Accordingto Venkatesh(2000),thesetwoconstructssignificantlyinfluenceper- ceivedease ofuse. Thenegative associationbetweencomputer anxietyandperceivedeaseofusesuggeststhatindividualswith increasedanxietytowardcomputersperceivethataDMTisharder touse.Asaresult,thishypothesisisnotsuitablefortheproposed modelbecauseofthespecialfeatureofthetool.ADMTisaspe- cific tool thatdiffersfromother IT.Ingeneral,peoplewho use aDMTareManagementInformationSystems(MIS)managersor personnelwhohaveabundantexperiencewithcomputers.There- fore, anxietyon computerusage is unlikely to bea significant factor.
DMTscan assistmanagers inmakingimportant decisionsin business. Because of the instrument-oriented characteristic of DMTs, managers are likely to use them to meet certain goals or solve particular problems. Though DMTs are not suited to entertainment-orientedITusage(e.g.,onlinegames),theyarecon- sideredtoimprovetheproductivityoforganizations.Therefore,the antecedentsofbothcomputeranxietyandcomputerplayfulness havearelativelyunimportanteffectonauser’sperceptionofease ofuse.
Withtheexceptionoftheeffectoftherelationshipbetweenper- ceivedeaseofuseandbehavioralintentiononadoptingaDMT,the moderatingeffectofexperiencedoesnotplayanimportantrole intheDMTacceptancemodel.Theseresultsdifferfromthoseof previousresearchonTAM3(Venkatesh&Bala,2008).Accordingto previousdiscussions,computeranxietydoesnothaveadirecteffect onperceivedeaseofuse.Thisisbecauseemotiondoesnothavea directeffectonDMTusage.Becauseusersusetoolstocomplete theirtasksorachievecertainobjectives,computeranxietyisnota criticalfactorinperceivedeaseofuse.Thisinstrument-oriented characteristic ofusers makesDMT less playful.Computer play- fulness,therefore,doesnothaveasignificanteffectonperceived easeofuse.Regardlessofwhatinfluenceexperiencehas,computer