You are on page 1of 25

Reducingproblemsofsociabilityinonlinecommunities:Integratingonline * communicationwithofflineinteraction forthcomingin:AmericanBehavioralScientist

Abstract Socialnetworkresearchersclaimthatrelationsofflineaffectrelationsonline.However,itis unclearwhich characteristicsofonlineinteractionareaffectedbytheexistenceofoffline relations.Thispaperattemptstoanswerthequestion of whetheramixtureofvirtualand'real life'interaction between atleastsomemembersofknowledgesharingonlinecommunitiesin contrastto'purelyvirtual'interactionreducestheprevalenceoftheproblemsthatthewhole communityfaces.Typicalproblemsofsociabilitythatincreasetheriskoffailurein knowledgesharingare:1)alackoftrustbetweenmembers,2)freeriderbehavior,and3)a lackofstablemembership.Thisanalysisusessurveydatafrom 26onlinecommunitiesthat arepartofavirtualorganizationthathostscommunitiesforteachers.Thefindingsprovide evidencefortheexistenceofmoretrustandlessfreeridingin'mixed'communities,butnot forgreater membershipstability.Resultsleadtotheconclusionthatofflinenetworkshave beneficialeffectsononlineknowledgesharingandcontributetoourknowledgeof how offlinenetworksinfluenceonlinerelations.Moreover,theymodifyearlierclaimsaboutthe integrationofonlinecommunicationwithofflineinteractionbyshowingthatacomplete integrationisunnecessary.

Keywords:sociability,trust,freeriding,embeddedness,socialnetworks,elearning,blended communities,onlinecommunity
UweMatzat Sociology /SchoolofInnovationScience EindhovenUniversityofTechnology P.O.Box513,5600MBEindhoven TheNetherlands u.matzat@tue.nl

The research project was funded by a grant from the Dutch Science Foundation (Nederlandse Organisatie voor

WetenschappelijkOnderzoek,NWO)withintheprogram'Societyandtheelectronichighway'(Projectnumber01443618). IwouldliketothankChrisSnijders,MarianneMulders,FransSchouwenburg,andErikVerhulpfortheirhelpfulinputand support.

Introduction:Barrierstoonlineknowledgeexchangereachingfull potential Onlinecommunitiesarefrequentlyusedforknowledgeexchangeandlearning.Someare linkedto'communitiesofpractice',groupsofprofessionalswithsimilartaskresponsibilities whosharetheirexperiences(Wick2000).Whilecommunitiesofpracticedonotnecessarily containanonlinecomponent, onlinecommunities(OLCs)ofpractice,bydefinition,do (Johnson,2001). Researchintotheeffects ofOLCsforcompaniesindicatesthatsuch communitieshavethecapacitytostimulateexchangeofknowledge(Ardichvili,Page,& Wentling,2003),butahighnumberofcommunitiesdonotfulfilltheseexpectations. Membersareoftennotsufficientlymotivatedtosharetheirknowledgeactively (Johnson, 2001Stoddart,2001Gal2004).SomeOLCsofpracticehavebeen builtfortheprofessional developmentof teachers,andafewteachersclaimtohavebenefitedfromtheiruse.However, ingeneral,researchershaveoncemorecometotheconclusionthattheOLCsforteachersthat havebeen studieddonotyetmeettheirambitiousaspirations(Kling&Courtright,2003 Yang&Liu,2004Barab,MaKinster,&Scheckler,2003).Thispapercontributesto identifyingthetypicalbarriersthataffectsuccessfulsharingofknowledgeinOLCsandhow thesebarrierscanbereduced.Itfocuseson theproblemsofsociability(Preece,2000),which areproblemsrelatedtosocialinteractioningroupsthatareimportantbarrierstoknowledge exchange(Barab,2003). Technologicalcharacteristicsofonlinecommunitieshavebeen showntobeimportantforhighusability(Chen,2007),butotherproblemsreceivemuchless attention(Schlager&Fusco,2003).Thispaperdrawsattentiontoaparticularsocial characteristicofanonlinecommunity,namelyitsembeddednessinsocialnetworksin the offlineworldandhowthataffectssociability. Thisstudyextendstwostreamsofresearch.First,socialnetworkanalystsclaimthatoffline relationsaffectonlineinteraction(Wellman&Gulia,1999Wellman,2001).However,how thatmighthappenisleftopen.Thispapercontributestoansweringthisquestion.Second, communityresearchersarguethatitwouldbeusefultobuildOLCsaroundpreexisting offlinecommunities(Kling&Courtright,2003Barab,MaKinster,&Scheckler,2003), therebysuggestingacompleteintegrationofonlinewithofflineactivities.Iagreewiththese authorsthatintegrationofonlineandofflineactivitiesisbeneficial.However,Iarguethata mixtureofvirtualand'reallife'interactionbetween somemembersisenoughtoreduce problemsofsociability facedby thewholecommunity,makingacompleteintegration unnecessary. Thepaperisstructuredasfollows.First,adistinction ismadebetweenthreeimportant problemsofsociability.Then,afterdescribinghowofflineandonlineinteractionin communitiesarerelatedtoeachother,thepaperproceedsbyspecifyingmechanismsthat showhowembeddednessinofflinenetworksreducesproblemsofsociability,leadingtothree 1

hypotheses.Thedesignofthestudyanditsmeasurementsareintroducednext,followedby testsofthehypothesesusingsurveydatacollectedfrom 26onlinecommunitiesofteachersin theNetherlands.Theimplicationsofthefindingsforthedesignandmanagementofonline communitiesarediscussed,andconclusionsforfurtherresearcharedrawn. Typicalproblemsof sociability Successfulinteractioninonlinecommunitiesnotonlydependson ahigh usability,butalso oneffortsthatsupportthemembers'socialinteractionasagroup.Thisiscalledsolving problemsofsociability(Preece,2000).Threeproblemsofsociabilityaredistinguished. First,a lackoftrustcandevelopduringexchangeofinformationbetweenmembers.Ifone member,ego,providesinformationtoanothermember,alter,thenegocanhopethatthese rolesmay bereversedinthefuture.Altermaybewillingtoprovideinformationtoego becausehewantstorewardegoforhispasthelpandtoensurethattheystayengagedin mutuallyadvantageousexchanges.However,ego'sinitial provisionofinformationisriskyfor him.Altermaynotreciprocate ormaychoosetousethereceivedinformationinwaysthat areharmfultoego. Thismayhappen,forinstance,whenteachershelpothersbysharingtheir teachingmaterial.They takerisksby puttingteachingexamplesonline.Inthetraditional classroomsettingteacherskeeptheirteachingactivitiesprivateafterclosingtheclassroom door.Thisisdifferentinanonlinecommunityofteachers.Membersdonotknowwhere their materialswilltravel iftheyputthemon theInternet(Barab,MaKinster,&Scheckler,2003). Ifproblemsoftrustbetweenmembersaretoosevere, they hesitatetosharetheirinformation andtheOLCfails(Ridings,Gefen,&Arinze,2002Jarvenpaa&Leidner,2000Kling& Courtright,2003). AusefulonlinediscussionisacollectivegoodforthewholeOLC.Everymemberprofits from thediscussion (independentofhisowncontributionintensity)becausethediscussion 'consumption'isnonexcludable(Kollock,1999).Ausefulcontribution toadiscussionor a commondatabaseiscostly.Atthevery least,ittakestimeandeffort.Whilethecontribution costsarerestrictedtoactivemembers,thediscussionbenefitsaredistributedamongallactive andpassivemembers.Withoutselectiveincentivesforindividualstobecomeactive,rational individualswilldecidenottoincurthecosts,butprofitfrom othermembers'contributions nevertheless(Dawes,1980).Freeridingisthetendencyofmemberstowithholdinformation andletothersincurcontribution costs thisisanothersociabilityproblem (McLureWasko,& Faraj,2000Thorn&Connolly,1987Cress,2004).Ifthetendencytofreerideistoohigh, thecollectivegoodisnotproducedandmembersdonotgainenoughbenefits.

Finally,itisoftenassumedthatOLCswill sufferfrom insufficientmembershipstability (Johnson,2001 Butler,2002Dannecker&Lechner,2007Chen,2007). Komito(1998) arguesthatmanyOLCscanbeleftatwill,limitingtheamountofsocialcontrol,andthereby reducingthemembers'motivationforlongerlastinginvestments.Insufficientmembership stability thereforecouldmakethedevelopmentoftrustmoredifficultandmayreducethe motivationtoparticipateactivelyinknowledgesharing.Intheend,amembercould anticipatethatone'scontributiontoanongoingdiscussionmightnotbereturnedifthosewho profitleavethecommunityquicklyaftertheyhavefoundwhattheysought. Socialembeddednessofonlinecommunitiesinofflinenetworks Onlineinteractionisoften embeddedin offlinenetworks. Itcanleadtonewcontactsthatare transferredtotheofflineworldinunplannedways.Userscanmaintainexistingoffline contactsviaonlinecommunication,someonlinecommunitiesarenaturally affiliatedwith offlinecommunities,andsomecommunitymanagersandmembersintentionallycreate opportunitiesforofflinemeetingsamongmembers. AnumberofstudieshaveshownthatInternetusecomeswithneworpreexistingoffline contacts(Hampton&Wellman,2003Hlebec,LozarManfreda,&Vehovar,2006Wellman, Boase,&Chen,2002Franzen,2002).Onlinegroupcommunication,inparticular,canlead tonewofflinecontacts(Zhao,2006Parks,1996Parks&Roberts,1998Utz,2000 Matzat 2004).Forinstance,Heintz(2000),inastudyofSwissintensiveusersofnewsgroupsand chatgroups,foundmanypersonalcontactsbetweenusers.Sixtysevenpercentoftheonline contactsmetfacetoface.Matzat(2009) foundthatmanyacademicemailinglistswere affiliatedwithwellintegratedacademiccommunitiesthatmetregularlyoffline.Inastudyof onlinehealthcarecommunities,DanneckerandLechner(2007)showedthatmanagersand membersactivelycreated opportunitiesforofflineinteraction.IshiiandOgasahara(2007),in astudyofastratifiedrandomsampleofadultinhabitantsofSeoulandTokyo,suggestthat thedegreeofembeddednessinofflinenetworksmaydependonculturalfactors. Online communitiesusedbyKoreansweremore often embeddedinofflinenetworks than thoseused bytheJapanese(48%versus33%).Whileofflineinteractioncanbeplanned,natural circumstancesalsoplay arole.Yuan,Gay,andHembrooke(2006)demonstratethatinonline learningcommunities,studentswhobelongtothesameuniversityhadahigherlikelihoodof facetofacemeetings. Consequencesofembeddednessinofflinenetworks Thereisonlylimitedempiricalevidenceshowinghowembeddednessaffectsonline interaction.Accordingtosocialnetworkanalysts,embeddednessshouldchangeonline 3

relations(Wellman&Gulia,1999).Unfortunately,itisnotclearhow thishappens.The literatureonknowledgesharinggroupsdiscussestherelevanceoffacetofacerelationships betweenmembers(Wenger,1998 Brown & Duguid,1991). However,itdoesnotclarify howcharacteristicsofthewholenetworkmightaffectonlineinteraction.Also,online community researchershavewonderedwhetherfacetofacemeetingsarenecessary(e.g., Johnson,2001). Somearguethatthegreatestpotential foronlinecommunitiesofpracticeis tofunction asasupplementtoalreadyexistinglocalofflinecommunities(Schlager& Fusco, 2003 Kling& Courtright, 2003) Barabetal.(2003)describefacetofaceinteractionasan essentialsupplementtoonlineinteraction.Iagreewiththeseauthorsthatonlinecommunities ofpracticehavethepotential tosupportpreexistinglocalcommunities.However,Iargue thatonlinecommunitiesofpracticehavepotentialbeyondthissupport.Theauthorsneglect thefactthatofflineinteractionof somemembersmaychangethesituationalconstraintsnot onlyforthosememberswhoknoweachother,but alsoforthewholegroup.Understoodin thissense,theperspectivethatispresentedhereimpliesthatfacetoface(oroffline) interactionisnotanessentialpreconditionforamemberchanginghisbehavior.Itisenough forsomeothermembersinteractoffline.Therefore,OLCscanincludeindividualswhoare notmembersofapreexistinglocalcommunityaslongasothermembersinteractfaceto face. Experimentalresearchoncomputermediatedcommunicationin small groupsshowsthat communicationbetweenmembersincreasestrust, andprecedingcommunicationbetween membersincreasesthecontribution rate,reducingfreeriderproblems(Riegelsberger,Sasse, &McCarthy,2003).Muchlessfieldresearch hasbeen conductedontheeffectsof communicationbetweenmembersof largeorlongerlastingonlinegroups.Thesituationin largeonlinegroupsisdifferentfrominteractioninsmalllaboratorygroupsfortworeasons. First,onecannotexpectallmembersinlargegroupstocommunicatewitheachotheroutside thegroup. In largeOLCs,itmightwellbethatmemberswhodonotknoweachotherand whodonotknowany othermemberwishtosharetheirknowledge.Second,inlongerlasting OLCs,theconsequencesofcommunicationevenamongthosememberswhohave communicatedwitheachothermightbeweakerbecausethetimelagbetweenthepreceding communicationandthesubsequentonlineinteractioncan beverylarge.Thelimited empiricalevidencefrom fieldstudiesisbrieflypresentedbelow. Fieldstudiesof smallormediumsizedgroupssuggestthatprecedingfacetofacemeetings amongstudentsinonlinelearningcoursesmightincreasetherateofcontributionto subsequentonlinediscussions(Bluemink& Jrvel,2004KleineStaarman,2003). Haythornthwaite(2000 2001)examinedanonlineclassconsistingof14studentsina15 weekcourse.Afterperiodswithfacetofacemeetings,onlinesocializingactivitiesand emotionalexchangetookplacemuchmorefrequently.Shearguesthatfacetoface 4

interactionhadacatalyticeffect.However,OLCsaremuchlarger,andtheinteractionperiod ismuchlonger.Itisunclearwhetherthesefindingscanbegeneralizedtosuchcommunities. Thecrucialquestionforfieldresearchiswhetherthosewhodonotinteractoutsidethegroup withothermemberswillnevertheless,inthelongrun,beaffectedbythecommonoffline interactionofothermembers.Therefore,I nextfocuson mechanismsthatreveal how embeddednessinofflinenetworkschangesthesituationforthewholegroupandthereby influencessociabilityproblemsinlargeOLCs. Coleman(19881990)arguesthatdensenetworksdiminishproblemsofcooperationin groupsinanumberofways.Cheatingismoredisadvantageousbecauseindividualswhodo sodamagetheirreputations.Informationaboutcheatingbehaviorcanspreadmoreeasilyina higherdensitynetwork (seealsoRaub&Weesie,1990).Moreover,cooperativenormsand expectationsoftrustareeasiertoestablishbecausenormviolatorsfaceahigherlikelihoodof beingsanctionedthroughothermembers'coordinatedaction,whichisfacilitatedbythedense network.TheseandrelatedargumentsbyGranovetter(1985)stimulatedalargebodyof researchexemplifyinghowdensenetworks(or,intheterminologyof Granovetter(1985),the embeddednessofactioninsocialnetworks)facilitatesolutionstoproblemsofcollaboration intheofflineworld. Fieldresearchonlargeonlinegroupsprovidesevidencefortwoeffectsofembeddednessin offlinenetworks.First,Matzat(2009)foundthatinacademicemailinglistsembeddedness providesincentivesforresearcherstogain reputation withintheirscientificcommunityby contributingactivelytotheonlinediscussion.Second,underahighdegreeofembeddedness, group normsthatprescribehelpingothersaremorelikelytoemerge(Matzat2004b). Therefore,embeddednessaffectsthequantityoftheonlinediscussioncontributionsbecause itprovidesselectiveincentives(namely opportunitiestogainorlosereputation), thus reducingfreeriderproblems.Furthermore,Matzat(2009b)arguesthatthereareotherreasons toexpectembeddednesstoaffectsociability.Evenifsomemembersdonothaveanyoffline contacts, theyrecognizeovertimethattheinformalnetworkamongmembersisdensethey realizethatinformationaboutmisbehaviorspreadsquicklyandthatstrongformsof misbehaviormaybesanctionedbycollectiveaction (e.g.,coordinatedpubliccomplaints) facilitatedby denseinformalrelationships.Undersuch conditions,membersaremore dependentoneachothertogainreputationandavoiddisapproval.Ifmembersaremore interdependent,thentheyaremorelikely todevelopaninterestinmaintainingasatisfying relationshipwitheachother. Thepointisthatformembersitismuchhardertoleavean OLCthathasmaintenanceof relationshipsbetweenmembersasagoal than toleavean OLCthatfocusesonlyon informationinterests.Leavingan OLCwith additional relationalinterestsismuchmore 5

disadvantageousforthemember.Ifmembersstepoutofsuchacommunity they clearly indicatealackofrelationalinterests.Thismayoffendsomeothermembers'feelingswhohad differentexpectationsabouthowonehastobehave.Evenifmemberswholeavemaynotcare muchaboutothermembers'feelings,itisnotintheirowninteresttooffendthembecausethat preventsaccesstoinformationthatthey mightneedinthefuture.Comingbacktothe communitylateroncemoreandaskingforinformationisthennolongeravalidoption.In addition,themaintenanceoftherelationshipsislikelytobecomeapleasantexperienceinand ofitself.Inextremecases,memberswill notleaveatallbecausethemaintenanceof relationshipswithspecificindividualsisimportant.So therewillbelessmembership fluctuationinembeddedcommunitiesthaninpurelyvirtualcommunities. Finally,underconditionsof highsocialembeddednessandthereforerelationalinterests betweenmembers,problemsoftrustarediminished.Theabuseofplacedtrustwouldbea clearsignalofalackofrelationalinterest.Justlikeexplainedabove,inOLCswithrelational interestssignalingalackofrelationalinterestsisselfdamagingforamember.Therefore, underahighdegreeofembeddednessandrelationalinterests, trustislesslikelytobeabused andmembersaremorelikelytobehavecooperatively.MembersinOLCswithhigh embeddednessanticipatethehigherlikelihoodofcooperativebehaviorfrom others. Therefore,itismorelikelythattheywill placetheir trustinmembersoftheOLC(Matzat, 2009b).Hence,therewillbefewerproblemsoftrust. Thisleadstothefollowinghypothesesabouteffectsofembeddednessinofflinenetworks. H1:Inonlinecommunitiesofpractice,therewillbefewerinstancesoffreeridingundera highdegreeofembeddednessthanunderalowdegreeofembeddedness. H2:Inonlinecommunitiesofpractice,therewillbefewerproblemsoftrust(orhigherrates ofplacementoftrust)underahighdegreeofembeddednessthanunderalowdegreeof embeddedness. H3:Inonlinecommunitiesofpractice,therewillbegreatermembershipstabilityundera highdegreeofembeddednessthanunderalowdegreeofembeddedness. Studydesignandmeasurements Thehypothesesweretestedwith questionnairedatathatwascollectedbetween November 2005andFebruary2006from 33onlinecommunitiesofpracticein alargeDutch virtual 1 organizationforteachersofsecondaryeducation. Arandomsampleof membersreceivedan emailinvitationforaweb survey (n=1259,37% responserate) TheOLCsworkinthesame 6

technologicalenvironment,includingdiscussionfora,anelectronicnewsletter,and opportunitiesforuploadinganddownloadingteachingmaterial. Dependentvariables Thehypothesesmakepredictionsabouttheeffectsofembeddednessonthreeproblemsof sociability,namelyfreeriding,alackoftrust,andmembershipfluctuation.Thesewere measuredinthreeways. 1.)Theperceivedintensityof allthreeproblemswasmeasuredthrough experienced members'assessmentsofhowseriousthedifferentproblemsareforthefunctioningofthe community.Thisleadstothreedependentvariables. 2.)Theintensityoftheproblemsoffreeridingandtrustweremeasuredbymembers'self reportedwillingnesstoparticipateinanumberofcommonactivities.Theseactivitieswould contributetotheproductionofcollectivegoodsforthecommunity ordemandtheplacement oftrustinothermembers.Inthecaseofthefreeriderproblemtheactivitiesdifferedwith respecttotheamountoftimetheyaskedfor.Inthecaseoftrustproblemstheactivitieswere riskyinthesensethatthey includedthepossibilitythat,tosomeextent,theymightbeself damagingbecausetrustplacementmightbeabused.Theseactivitieswerepresentedas hypothetical,butrealisticscenarios.Thefollowingmeasuresweretakentoincreasethe validityofthemembers'answersinthescenarios.Thevirtualorganization regularly publishesanelectronicjournalthatissenttoall members theintentiontoincreasetheactive participationofteachersintheOLCswasexplicitlyannouncedin thejournal shortlybefore thestartofthedatacollection.Itwasalsoannouncedthatthevirtualorganizationwouldtake specificmeasurestoincreaseparticipation.Thequestionnairereferredtothisinitiative,and thescenarioswerepresentedinsuchawaythatmembersweretoassumethatthe managementmightusetheinformation providedbythemembersinthenearfuturetotake certain actions(asindeeditmight). Inthescenariositwasstressedthatthemember'sanswers wouldbeofcrucialimportanceforthefurtherdevelopmentofhis/herowncommunity. Additionally,thesurveyincludeditemsthatforma'socialdesirabilityscale',sothatitis possibletocontrolforageneraltendencytoanswerinsociallydesirableways.Thescenario answersleadtotwodependentvariablesmeasuringthewillingnesstocontributetocollective goodsandtoplacetrust. 3.)Fortheproblemoffreeriding,themembers'selfreported behaviorwastakenasan indicator.Memberswereaskedquestionsabouttheirpastcontributionstoanumberof collectivegoodsforthewholecommunity.Thisleadstooneadditionaldependentvariable indicatingtheactualparticipationintheproductionofcollectivegoods.

Thethreemeasurementprocedures(perceptionsofproblems,willingnesstoparticipatein collectiveactionandtoplacetrust,andselfreportedbehavior)thusproducesixdependent variablesandaredescribedinmoredetail below. Table1:Measurementof six dependentvariables Perceivedintensity:7pointLikertitems(variables13: onlyanswersfrom respondentswithmorethansixmonthsofmembershipwereusedtoensurevalid answers) 1.Perceivedintensityoffreeridingproblem:
Therearetoofew active membersinthecommunity. Themembersdonotsendenoughteachingmaterialtothecommunity. Ifthereissomethingtobedoneforthewholecommunity,thenthememberswaituntilothermembers taketheinitiativebeforetheycontributethemselves.

Principalaxisfactoringresultinginonefactorscore:=.86,explanationof67%ofitem variance,KMOmeasure=.73. 2.Perceivedintensityofproblemsoftrust:


Membersdonottrusteachother. Themembershesitatetomakeuseoftheinformationthattheycanfindinthecommunity. Themembershesitatetowriteabouttheirproblemsatwork. Themembersareafraidthattheinformationthattheysendtothecommunitymightbeusedinways unfavorabletothemselves.

Principalaxisfactoringresultinginonefactorscore:=.81,explanationof53%ofitem variance,KMOmeasure=.76. 3.Perceivedintensitymembershipfluctuation:


Therearefluctuationsinthenumberofmembersinthecommunity.(singleitem)

Members'willingness(scenarios):Higherscoresindicategreaterwillingness (variables45:answersfromallrespondentsused) 4.Willingnesstocontributetocollectivegoods:5pointLikertitems


Inthenearfuture,thecommunitywillbemakingmoreuseoftheknowledgeandcapabilitiesofits members.Therefore,wewouldliketoknowtowhatextentthemembersarewillingtocommit themselvestothefuturedevelopmentofthecommunity.Youranswerstothesequestionsarethus importantforthefutureofyourowncommunity. Wouldyouorwouldyounotbewillingtoparticipateinthefollowingactivities: a)distributingflyersaboutyouronlinecommunityduringaneveningofinformation ataschoolinyour neighborhood b)givingashortpresentationaboutthecontentandutilityofyouronlinecommunityduringaneveningof informationataschoolinyourneighborhood c)mediatingconflictsbetweenmembersinthecommunity.Youdonotneedspecialtechnicalskills,and yourinvolvementisonlynecessaryincasesofconflict. d)participatinginamembers'advisingcommittee.Thecommitteewouldmeetonceayear.Youdonot needspecialtechnicalskills,andtheexpectedtimeinvestmentwillbeabouteighthoursperyear(four hoursofpreparationandafourhourmeeting).

Principalaxisfactoringresultinginonefactorscore:=.82,explanationof53%ofitem variance,KMOmeasure=.74. 5.Willingnesstoplacetrust:5pointLikertitems


Weareplanningtodevelopanumberofsoftwaretoolsforyourcommunity,butwewouldliketoavoid offeringnewsoftwaretoolsthatnooneneeds.Therefore,wewouldliketoknowwhetheryouwould makeuseofthefollowingtoolsforyourcommunity. Thinkcarefullyaboutyouranswers.Theyaffectthefuturedevelopmentofyourowncommunity. Becausemuchoftheteachingmaterialnowadaysispreparedinelectronicform(PowerPoint,Word, PDF,etc.),itispossibletoprepareadiscussionforuminawaythatmakesiteasyforteachersto providethematerialtocolleaguessothatitcanbeevaluatedbyothers. a)Wouldyouprovideyourownteachingmaterialwithyournameonittocolleagueswithinyour community? b)Wouldyouprovideyourownteachingmaterialtocolleagueswithinyourcommunityifitis possibletodosoanonymously? Assumethateverymonththereisareviewofthelatestteachingmaterialbymembersofacommittee fromyouronlinecommunity.Thisreviewofthematerialandthediscussionsthatledtothereview wouldbepostedandaccessibletoeverymemberofthecommunity.Assumethatyouhaveteaching materialinelectronic form. a)Wouldyoubewillingtoprovideyourownteachingmaterialwithyournameonitforreview? b)Wouldyoubewillingtoprovideyourownteachingmaterialanonymouslyforreview?

Principalaxisfactoringresultinginonefactorscore:=.88,explanationof65%ofitem variance,KMOmeasure=.67. Selfreportedbehavior:Answeroptions:yes/no(answersfromallrespondentsused) 6.Pastcontributiontocollectivegoods:Twoitemsaskedrespondentswhetherthey "hadevercontributedtothenewsletterbywritingsomethingforit" and"hadeversentteaching materialtotheOLC",withascoreof'one'ifrespondentagreedtoatleastoneitem Independentvariables Embeddednessof an OLCwasmeasuredindirectlybyaproxyasproposedbyMatzat(2009). Experiencedcommunitymembershadtoassessthedegreeofembeddednessoftheirown communitybygivinganswerstofouritems.Asacharacteristicofthecommunity,the resultingembeddednessscoreshouldhavesignificantvariationbetweentheonline communities.However,directmeasurementusingfouritemsabouttheamountofoffline communicationbetweenmembersledtoascalewithhighreliability(alpha=.90),butno discriminationbetweenthecommunities(F=1.0,df1=24,df2=106, p=0.47).Measurementthus hadtoproceedinanindirectway.First,theextenttowhich thegroupof teachers constitutean integratedcommunitythatsharesmanyactivitiesandinterestsincommon wasmeasured. Second,it wasshownthatthis scoreisassociatedwithitems that directly measureinteraction betweenmembersoutsideofthe community.

Forthefirststep, embeddednesswasmeasuredindirectly withthehelpofanunfoldingscale analysis(vanSchuur&Kiers1994)ofthreesixpointLikertscaleditems,whichresultedina Mudfoldscale.AMudfoldscaleissimilartothewellknownGuttmanscale,withthe exceptionthatintheMudfoldscalesomedeviationsfromtheperfectanswerpatternofthe Guttmanscaleareallowed.Onecanthentestwhetherthesedeviationscanbeexplainedby chance,allowinganassessmentofthequalityofthescale.In aMudfoldscale,anumberof itemsandpersonscanbeorderedonedimensionallywithregardtoalatenttrait(inthiscase, theperceiveddegreetowhichagroupconstitutes anintegratedcommunitythatsharesmany interestsandactivitiesincommon).Thefollowingthreeitemswerepresentedonlyto experiencedOLCmembers(thosewhosubscribedformorethan twoyears).
Towhatextentdoyouagreewiththefollowingstatementsaboutyourowncommunity? a)Itisasetofgroupsandcliqueswiththeirowninterestsandactivitiesthatdo not have muchincommonasa community. b)Itisasetofgroupsandcliqueswiththeirowninterestsandactivitiesthat also have somecommoninterests asacommunity. c)Itisamoderatelyintegratedcommunitythatsharessomeinterestsandsomeactivities.

Answersrangedfrom"completelydisagree"to"completelyagree".AMudfoldscaleanalysis resultedinastrongMudfoldscaleconsistingof theitemsintheorderabc(H=0.86, assumptionsofanunfoldingscalemet). Inaddition,itdiscriminatedbetweenthe communities(11%ofvariancebetweengroups,F=1.52,df1=25, df2=314, p=0.05). Since embeddednessisacharacteristicofthewholeOLC, thearithmeticmeanofthemembers' answersforeveryOLCistakenastheindicatorofembeddedness. Theassumptionunderlyingthemeasurementprocedureisthatthemorethegroupof teachers constitute anintegratedcommunitythatsharesactivitiesandinterestsincommon,thehigherthe degreeofembeddedness.Embeddedness consistsofthefrequencyofthemembers'interaction outsidetheonline community. TheorderingoftheOLCsaccordingtothescalescoresof embeddednesscanbeexternallyvalidatedinameaningfulway,providingevidenceforthe assumption thatthescalecapturesofflinecommunicationbetweenmembers.InOLCsthat scorehigherontheembeddednessscaletherearemoremembershavingcontactwitheach otheroutsideofthecommunity thaninOLCsthatscorelowerontheMudfoldscale. Respondentswhoagreedthat"in(their)onlinecommunitytherearegroupsofmemberswho regularlymeeteachotherinfacetofacemembershipmeetings"tendtobeinonline communitiesthatscoresignificantlyhigherontheembeddednessscale(t=2.6,df=221.1,p=.01). Table2:Othercontrolvariables Digitalliteracy:SlightlyadaptedversionofHargittais(2005)scale:Howfamiliar

10

areyouwiththefollowinginternetphenomena?,downloads,advancedsearching, preferencesettings,newsgroups,pdfdocuments,refresh/reload,mp3files,blogs, emailinglists,spamfilters,allanswerson5pointLikertscalesresultinginonefactor score(alpha=.91,KMOvalue=.92). Sociallydesirableansweringtendency:ShortenedversionofBIDR6scaling procedure (Paulhus,1991):answerson7pointLikertscalestothefollowingitems: Myfirstimpressionsofpeopleusuallyturnouttoberight.Itwouldbehardforme tobreakany ofmybadhabits.IalwaysknowwhyIlikethings.Isometimestelllies if Ihaveto.WhenIhearpeopletalkingprivately Iavoidlistening.Idon'tgossip aboutotherpeople'sbusiness. Trustingdisposition:ThreeitemsadaptedfromJarvenpaa,Knoll,andLeidners (1998)scale:"Mostpeoplearehonestin describingtheirexperiencesandabilities. Mostpeopleanswerpersonal questionshonestly.Mostpeoplecanbecountedonto dowhattheysaytheywilldo."Answersleadtoonefactorscore(alpha=.83,KMO value=.70). Prosocialorientation:Shortenedversionofthesocalled"decomposedgames method"(Snijders&Weesie,1999) Digitalexperience:yearsofinternetuserangingfrom0to8(8=8ormoreyears) Groupsize:Informationfrommanagersorwebsite Age(loge), gender,andextentofhavingafulltimejobasteacher(loge[percentage]) arealsoincluded. Descriptivefindings Datacleaningresultedintheremovalof7%ofrespondentstheserespondentsdidnotyet haveateachingposition(e.g.,students)orgaveinconsistentanswerstothequestionson pro socialorientationleadingtomissingvalues(6%). Thefinal samplesize,whichisadequate forthetestsofthe threehypotheses,consistsof725respondentsin26OLCsofpractice.The "averagemember"hassubscribedforbetween712months,usesthecommunityoneday per week,andisnotvery activeinthecommunity(mean=2onascalefrom1to7).About6%of respondentshavecontributedtotheelectronicnewsletter, 15%reported havingsentsome teachingmaterialto theOLC,and10%attendedatleastonefacetofacemeetingofOLC membersorganizedbythemanagementofthecommunity.Fiftyfivepercentofrespondents arewomen,andthemedianageis46years.SizeoftheOLCsvariesfrom28to3682 members.Moreinformationcanbefoundintheappendix. TheOLCsorderedaccordingtotheirdegreeofembeddednessdifferwithrespecttotheir members' densityof offlinerelations.Thefollowingdata clarifytowhatextentthedifference

11

betweenhighlyandweaklyembeddedOLCscapturesdifferencesintheamountofoffline relations.Therespondentswereasked toagreeordisagreewith thefollowingthreeitems:"In ourcommunitytherearegroupsofmemberswho:a)areinregularcontactwitheachother outsideofthecommunity,b)meetwitheachotherregularlyatfacetofacemembership meetings,andc)calleachotherregularly."Themedianoftheembeddednessscoreswastaken asacutpointandthetworesultingclustersofcommunities(weaklyversushighlyembedded OLCs)werecomparedwithrespecttotheaverageproportionofmemberswhoagreedwith theabovementioneditems.Forthethreeitemstheproportionsofagreeingmembersgrew from28%to41%,18%to33%,and11%to19%,respectively.ThusinOLCswithhigher embeddednessscores,moremembersperceiveotherstobeinteractingfacetofaceorvia telephone. Resultsofhypothesestesting Forhypothesistesting,multilevelmultiplelogisticandlinearregressionanalysesareused sincethedataareclustered(Bryk&Raudenbush,1992).Unlessotherwisestated,the reportedpvaluesarebasedononesidedtestsbecausethehypothesesaredirectional.Table 35presentthetestsofthefirsthypothesisaboutfreeridingbyanalyzingtheactual contributionbehavior(Table3),thewillingnesstocontributetocollectivegoods(Table4), andtheintensityoftheperceiveddegreeoffreeriding(Table5). Theanalysesstartwiththemultilevellogisticregressionanalysisoftheactualcontribution behavior.Hereitdoesnotmakesensetosplitthevarianceintothetwolevelssothatno "emptymodel"ispresented.Instead, Ipresentamodelthatincludesonlytheeffectof embeddedness.Table3showsthatembeddednesshasapositiveandsignificanteffectonthe likelihoodthatamemberhascontributedtothecommunity,eitherby submittingteaching materialorwritinganarticlefortheelectronicnewsletter.Thereisasignificantbivariate associationbetweenembeddednessandcontributingbehavior.Inonlinecommunitieswitha strongembeddednessmembersaremorelikelytohavecontributedtocollectivegoodslike theelectronicnewsletterortothedatabaseconsistingofteachingmaterial .Moreover,the effectofembeddednessremainssignificantafteranumberof potentiallyinterveningfactors areincludedinthemodel.Additionally,thelikelihoodof contributionincreaseswith respondentsdigitalliteracy,digitalexperience,and theirindividual degreeoftrusting disposition,allofwhichcomesasnosurprise. Table3:Twolevellogisticregressionanalysisofactualcontributionbehavior

Variable

Model1

Model2

12

Estimatedvalue (standarderror) Groupleveleffect: Socialembeddedness Groupsize Individualleveleffects: Digitalliteracy Digitalexperience Trustingdisposition Prosocialorientation Socialdesirability Gender(1=female) Age Extentof fulltimejob asteacher *:p<=.05**:p<=.01(onesided) N=725 n=26
2 t0 =0.513 (0.207)

Estimatedvalue (standarderror) .212* .298 .436** .147* .217* .278 .030 .189 .506 .418 (.101) (.265) (.119) (.067) (.109) (.494) (.113) (.235) (.413) (.326)

.185*(.088)

2 t0 =0.691(0.300)

N:Level1samplesize(numberofindividuals) n: Level2samplesize(numberofgroups)
2 t0 :Level2varianceoftheintercept

Next,I analyzewhetherthewillingnesstocontributetocollectivegoodsinthefutureis affectedbythedegreeofembeddedness.Table4showsthatanumberofvariableshave significanteffects.Thosewhoaremoredigitallyliterate,whoaremoretrusting,andwho haveastrongerprosocialorientationaremorewillingtoparticipateinthecontributionof collectivegoods.Mostimportantly,theeffectofembeddednessissignificantandpositive. HighlyembeddedOLCsaremoresuccessfulthansolelyvirtualonesinthesensethatthey havemorememberswhoreportbeingwillingtocontribute.Furthermore,asModel2inTable 5demonstrates,thevariancebetweentheOLCsreducestozeroaftertheinclusionofthe individualcharacteristicsandthegroupcharacteristics.AmodelthatincludesaLevel2 2 residualvariancedoesnotfitbetterwiththedata( =1.6,df=1,p>.1). Thefindingsshownin Table3andTable4bothsupportHypothesis1. Table4:Twolevellinearregressionanalysisofwillingnesstocontribute

Model1 Variable
(emptymodel) Estimatedvalue (standarderror)

Model2
Estimatedvalue (standarderror)

13

Groupleveleffect: Socialembeddedness Groupsize Individualleveleffects: Digitalliteracy Digitalexperience Trustingdisposition Prosocialorientation Socialdesirability Gender(1=female) Age Extentof fulltimejob asteacher *:p<=.05**:p<=.01(onesided) N=725 n=26 s2=0.809 (0.043)
2 t0 =0.01 (0.01)

.041* .001 .117** .020 .062* .543** .015 .052 .015 .001

(.019) (.047) (.036) (.019) (.035) (.159) (.037) (.071) (.131) (.100)

s2=0.789 (0.041)
2 t0 =0()

N:Level1samplesize(numberofindividuals) n: Level2samplesize(numberofgroups)
2 t0 :Level2varianceoftheintercept, s2: Level 1variance

InTable5,Ishowtheresultsfor theproblemoffreeriding.Theanalysistestswhether membersinembeddedOLCsperceivelessfreeridingthanmembersinvirtual OLCs.The resultsimplythatthosewhohaveaweakerprosocialorientation,scorehigheronthesocial desirabilityscale,andthosewhobelongtolargercommunitiestendtoperceivelessfree riding.Mostimportant, theeffectof embeddednessisnegative,asexpected.However,this 2 doesnotreachstatisticalsignificanceatthe5%level(p=.076). Table5:Twolevellinearregressionanalysisofperceivedfreeriding Model1 (emptymodel)
Estimatedvalue (standarderror)

Model2
Estimatedvalue (standarderror)

Variable

Groupleveleffect: Socialembeddedness Groupsize Individualleveleffects: Digitalliteracy Digitalexperience Trustingdisposition .037 .015 .008 (.045) (.025) (.042) .034 (.023) .196** (.056)

14

Prosocialorientation Socialdesirability Gender(1=female) Age Extentof fulltimejob asteacher *:p<=.05**:p<=.01(onesided) N=513 n=26 s2=0.764 (0.049)
2 t0 =0.053 (0.028)

.412* .087* .125 .032 .088

(.184) (.045) (.084) (.161) (.123)

s2=0.778 (0.049)
2 t0 =0()

N:Level1samplesize(numberofindividuals) n: Level2samplesize(numberofgroups)
2 t0 :Level2varianceoftheintercept, s2: Level 1variance

TheresultsofTables35providetwotimesstrongandonetimeonly weaksupportforthe firsthypothesis,leadingtotheconclusionthatahighdegreeofembeddednessindeedcomes alongwithlessfreeriding. Table6andTable7examinethewillingnesstoplacetrustinothercommunitymembersand theperceptionofproblemsinducedbyalackoftrustbetweenmembers,therebytesting hypothesis2abouttrustandembeddedness.Table6showsthatthosewhohavehigherdigital literacy,moredigitalexperience,aremoretrustful,andhaveahigherprosocialorientation aremorewillingtoplacetrustinothermembersbysendingtheirteachingmaterial.Older memberstendtohavesomewhatlesstrustthanyoungermembers.Themostimportantpoint isthattheeffectofembeddednessispositiveandsignificant.Thisisinaccordancewiththe predictionof Hypothesis2.InOLCswithahigherdegreeofembeddedness,memberstendto bemorewillingtotrustplacingtheirmaterialsonlinethanmembersinsolelyvirtualOLCs. Again,thevariancebetweenthegroupsisreducedtozero. Table6:Twolevellinearregressionanalysisofwillingnesstoplacetrust

Model1 Variable (emptymodel)


Estimatedvalue (standarderror) Groupleveleffect: Socialembeddedness Groupsize Individualleveleffects: Digitalliteracy Digitalexperience Trustingdisposition 15

Model2
Estimatedvalue (standarderror) .043* .051 (.019) (.047)

.069* (.037) .037* (.020) .161** (.035)

Prosocial orientation Socialdesirability Gender(1=female) Age Extentof fulltimejob asteacher *:p<=.05**:p<=.01(onesided) N=725 n=26 s2=0.861 (0.046)
2 t0 =0.007(0.01)

.346* .031 .086 .303* .101

(.162) (.038) (.072) (.133) (.103)

s2=0.817 (0.043)
2 t0 =0()

N:Level1samplesize(numberofindividuals) n: Level2samplesize(numberofgroups)
2 t0 :Level2varianceoftheintercept, s2: Level 1variance

Table7analyzesthemembers'perceptionofalackoftrustinothermembers.Thosewho haveastrongerprosocialorientation,thosewithlowerdigitalliteracy,males,andmembers insmallercommunitiestendtoreporta greaterlackof trustinmembers.Mostimportantly, theeffectofembeddednessisnegativeandstatisticallysignificant,ashypothesized. EmbeddedOLCstendtohavefewerproblemsoftrustaccordingto theirmembersthan virtual OLCs. Table7:Twolevellinearregressionanalysisofperceivedproblemsoftrust Model1 (emptymodel)
Estimatedvalue (standarderror)

Model2
Estimatedvalue (standarderror)

Variable

Groupleveleffect: Socialembeddedness Groupsize Individual leveleffects: Digitalliteracy Digitalexperience Trustingdisposition Prosocialorientation Socialdesirability Gender(1=female) Age Extentof fulltimejob asteacher *:p<=.05**:p<=.01(onesided) N=513 n=26 s2=0.758 (0.048)
2 t0 =0.027 (0.019)

.058** (.023) .159** (.055) .093* .012 .061 .334* .037 .142* .255 .121 (.043) (.025) (.041) (.181) (.044) (.082) (.158) (.121)

s2=0.753 (0.047)
2 t0 =0()

16

N:Level1samplesize(numberofindividuals) n: Level2samplesize(numberofgroups)
2 t0 :Level2varianceoftheintercept, s2: Level 1variance

BothresultspresentedinTable6andTable7supporthypothesis2,leadingtotheconclusion thatembeddedOLCsfacelessproblemsoftrustthanpurelyvirtualones. Table8analyzestheproblemofmembershipfluctuation,testinghypothesis3.Onlygender anddigitalexperiencehavesignificanteffectsontheperceivedproblemofmembership fluctuation.Women andthosewithmoredigitalexperienceregardtheproblemof membershipfluctuationaslessintensive.Mostimportant,embeddednessdoesnotshowan effectontheperceivedintensityoftheproblemofmembershipfluctuation.Ingeneral, differencesintheperceivedproblemofmembershipfluctuationcannotbeexplainedvery wellbythemodelonlytwooutoften variablesshowsignificanteffects. Table8:Twolevellinearregressionanalysisofperceivedmembershipfluctuation

Variable

Model1 (emptymodel)
Estimatedvalue (standarderror)

Model2
Estimatedvalue (standarderror) .039 .118 .018 .055* .040 .170 .015 .278** .244 .036 (.039) (.104) (.055) (.031) (.052) (.228) (.056) (.111) (.202) (.154)

Groupleveleffect: Socialembeddedness Groupsize Individualleveleffects: Digitalliteracy Digitalexperience Trustingdisposition Prosocialorientation Socialdesirability Gender(1=female) Age Extentof fulltimejob asteacher *:p<=.05**:p<=.01(onesided) N=513 n=26 s2=1.192 (0.076)
2 t0 =0.056 (0.035)

s2=1.17 (0.08)
2 t0 =0.078 (0.043)

N:Level1samplesize(numberofindividuals) n: Level2samplesize(numberofgroups)
2 t0 :Level2varianceoftheintercept, s2: Level 1variance

17

TheresultspresentedinTables38showthatOLCsembeddedinofflinenetworksfaceless problemsoftrustandfreeridingthanpurelyvirtualones.Onemightwonderwhetherthisis becauseofthedenserofflinenetworkofthecommunity(amacroorgroup characteristic)or becausesomemembersindividuallyhavemoreofflinerelations(anindividualormicro characteristic).Thetheoryexpectsthateventhosewhohavenoindividualofflinerelations areaffectedbyacommunity'shighdegreeofembeddednessinofflinenetworks.Thus,the macrocharacteristicofthewholecommunityisexpectedtohaveaneffect,independentofa potentialeffecttheindividualcharacteristic.Therefore,inaddition,allmodelspresentedin Tables38werereanalyzedwithtwoadditionalvariables:adummyindicatingthemember's personalparticipationinfacetofacemeetingsandacrosslevelinteractionbetweenthis dummyandtheembeddednessvariable.Theresultsdidnotchangebothvariableswere insignificantinallanalysesandthereforeleftout(tablesavailableonrequest).The insignificanceofthecrossleveleffectsshowsthatevenforthosewhodonotparticipatein facetofacemeetings,embeddednessintheothermembers'offlinerelationshasaneffect. Discussionoftheresults ThreehypothesesaboutdifferencesbetweenembeddedandpurelyvirtualOLCsweretested. WithrespecttoHypothesis1aboutfreeridingbehavior,theamountof reportedcontribution behaviorandwillingnesstoparticipateinfuturecontributingactivitieswashigherin embeddedthaninpurelyvirtualOLCs.Fortheeffectofembeddednesson theperceived intensityoffreeridingonlyveryweakevidencewasfound.Intotal,Iregardthesethree findingsassupportingHypothesis1.WithrespecttoHypothesis2aboutalackoftrust, differencesinthewillingnesstoplacetrustanddifferencesintheperceivedintensityoftrust problemscanbeexplainedbydifferencesinthedegreeofembeddedness.Ialsoregardthe findingsassupportforHypothesis2,which claimsthattherearefewerproblemsoftrustin embeddedOLCsthaninvirtualones.Finally,thetestsshowedthatdifferencesinthe perceivedintensityoftheproblemofmembershipfluctuationcannotbeexplainedbysocial embeddednessnoempiricalevidenceforHypothesis3canbefound. Thereareseveralpossiblereasonsforthefailuretofindsupportforthehypothesison membershipfluctuation.ItcouldbethattheDutch onlinecommunitiesofteachersingeneral aremorestablethanotheronlinecommunities.Many membershavealongtermprofessional interestinthecommunity,andtherearenoseriousalternativeonlinecommunitiesfor teachersintheNetherlands.Sotherecouldbeakindofceilingeffectwithrespecttostability whichmakesthesecommunitiesspecial.Itcouldalsobethattheindicatorused,which consistedofonlyasingleiteminthiscase,isnotadequate.Inaddition,thebeneficialeffects ofsocialembeddednessmaybeofsuchlimitedsizethattheydonotaffectamember's decisionaboutstayinginthecommunityorleavingit.Thelowproportionofmemberswho 18

reportthatmembershipfluctuationisaproblemsupportsthefirstexplanation.Only2%of membersclaimthatthereisfluctuation'toaconsiderabledegree'orhigher,andonly10% claimthatthereisfluctuation'tosomeextent'.Atthegrouplevel,theaveragecommunity scoresa'2'onascalefrom'1'(nofluctuation)to'7'(verylargefluctuation).Thenumbers suggestthatonlinecommunitiesforteachersdonotsufferfromfluctuation.Thesectionof descriptivefindingsshowedthatcontrarytotheproblemoffluctuation,theOLCssufferfrom freeridingandalackofplacementoftrust(seeMatzat,2006formoredetails).Nevertheless, itcanbethatembeddednessreducesfluctuationinothertypesofonlinecommunitiesthat sufferstrongerfromit.Furtherresearchisneededinorder toshedmorelightonthisquestion. Generalsummaryanddiscussion Thispaperteststheclaimthatmixedcommunitiesthatincludeacombinationofonlineand offlineinteractionoffercertainadvantageswithrespecttosociability whencomparedwith purelyvirtualones.Thisisbecausemixedcommunitiesarecharacterizedbyahigherdensity ofthemembers'offlinenetworkwhichprovidesopportunitiesfortheapplicationofsocial control.AccordingtoColeman's(19881990)socialcapitaltheoryindividualsrealizethat underanetworkwithhighdensitymisbehaviorislikelytodamageone'sownreputationand maybesanctionedbycoordinatedaction.Matzat(2009)arguesthatunderahighdensity networkmemberswilldeveloprelationalinterests,makingthemmorereceptiveforwhat othermembersthinkaboutthem.Whenmembershaverelationalinterests,theytrytoavoid uncooperativebehaviorthatwouldsignalrelationaldisinterest.Bothmechanismsshould reducefreeriding,encouragetheplacementoftrust,andstabilizemembership. Thecombinationofonlineandofflineinteraction,calledembeddednessinofflinenetworks, wasindicatedbytheextenttowhichtheonlinecommunityconstitutesanintegrated communitythatsharesmanyactivitiesandinterestsincommon.Itwasshownthatinthose integratedcommunitiesmembersmoreoftenparticipateinfacetofacemeetings,arein regularcontactwitheachotheroutsidetheonlinecommunity,andcalleachother.Moreover, othermembersperceivethesedifferenceswithrespecttoembeddednessinofflinenetworks. Earlierorlater,theyrealizethatsomemembershaveofflinerelationswitheachother. Theseanalysesprovideevidencethatembeddedonlinecommunitiesarelesssusceptiblethan virtual onlinecommunitieswithregardtoproblemsoftrustandfreeriding.Thisispromising forthemanagementofonlinecommunitiesandforfutureresearchoninteractioninonline communities.Thesefindingsdonotprovidegroundsforclaimingthatthepresentedanalyses offerevidenceof acausal relationshipinfact,forsomeoftheindicators,thecausal relationshipmightfunctionineitherdirection.Moreover,acausalanalysisshouldtestin moredetailtheunderlyingtheoriesthatspecifyhowembeddednessaffectstheproblemsof 19

sociability.Thisisbeyondthescopeofthispaperandhastobedonebyfutureresearch. However,otherresearchhasshownthatdensenetworkshavesomeofthehypothesized effectsinother onlineandofflinecontexts(Coleman,1988Matzat,20092004b).In addition,futureresearchshouldstudytheeffectofembeddednessonmembershipstabilityin communitiesthatsufferstrongerfromit. Thesefindingshaveinterestingimplicationsforthesocialdesignandmanagementofonline communitiesingeneral,notonlyforteachers.Thedegreeofsocialembeddednesscanbe influencedinanumberofways.Whenanewonlinecommunityisfounded,onecanthink aboutthetargetgroupof thatcommunity.Thetargetgroupmay beverylarge,which often impliesamoreheterogeneousgroupofmembers.Itistobeexpectedthatinsucha community,somedegreeofofflineinteractionisinterestingonlyforsmallersubgroupsof members.Theconsequenceisthatalthoughaverybroadtargetgroupmayleadtoalarger onlinecommunity,itwillinalllikelihoodalsoleadtoalimitedamountofsocial embeddedness.Duringthisearlyphase,thecommunityfounderscandecidewhetherthey wanttoacceptahigherdegreeoffreeridingandlackof trustinsuchalargecommunity. Theymayprefertohaveacommunityoflimitedsizewithfewerproblemsof sociability.In analready existingonlinecommunity,thedegreeofembeddednesscanbeinfluencedin differentways.Forexample,theadministratorcandecidenottofocusmarketingactivities fornewmembersonisolatedindividuals.Rather,hecanfocusonclustersofindividualswho arealreadyincontactwitheachother.Ifacompleteclusterofrelatedindividualssubscribes tothecommunity,thenthemembers'import'theirembeddednesstotheonlinecommunity. Anotherstrategytoincreaseembeddednessistoorganizeortosupportofflinemeetings.The virtualorganizationofthisstudyfollowedsuchastrategy.Otherresearchononline communitiesalsoindicatesthatsuchinitiativesbythemanagementorbymembersare promising(Dannecker&Lechner,2007).Iwouldliketoaddthattheorganization ofmember meetingscanbebeneficialtothegroupwhenmembersaremadeawarethatofflinemeetings havetaken (orwilltake)place.Themembers'attentioncanbedrawntothefactthatanumber ofofflineactivitiestakeplaceandthatmanymembersknoweachotherandhavepleasant relationshipswith eachother.Throughsuchefforts,boththeactualandperceiveddegreeof embeddednessincrease,whichislikelytohavesomebeneficialeffects. ThesefindingsalsohaveimplicationsforresearchonOLCs.Theysupportthehypothesisthat embeddedonlinecommunitiestendtobemoresuccessfulthanpurelyvirtualones.Theyshed lightonthelinkbetweenofflineandonlinerelationsandtherebysubstantiatetheclaimsof socialnetworkresearchers(Wellman,2001).Thesefindingsprovideevidencethatoffline relationshavebeneficialeffectsforsociability inonlinecommunitiesandpointtothehigh densityastherelevantnetworkcharacteristic.Moreimportantly,theycontributetoclarifying 20

how theoutcomesofonlineinteractionareaffectedbyofflinerelationsandnetworks.Social networksoftheofflineworldfacilitateinteractioninonlinecommunitiesbyreducingtheir problemsofsociability.Inaddition,thefindingsmodify earlierclaimsabouttheintegration ofonlinecommunicationwith offlineinteraction.Whileearlierresearchhasrightlysuggested thatofflinecontactsbetweenmembersarebeneficial(Barab,2003Kling&Courtright, 2003),whethermemberswithoutofflinecontactsareaffectedbytheofflinecontactsofothers withintheircommunitywasleftopen.Thisstudydemonstratesthatnoteveryonehastomeet facetofaceforanonlinecommunitytoactinsolidarityitisenoughifonly somemembers meetfacetofacetoreduceproblemsofsociabilityandtofostertheproductionof collective goodsinthewholecommunity. References
Ardichvili,Alexander,VaughnPage,andTimWentling,"Motivationandbarrierstoparticipationinvirtual knowledgesharingcommunitiesofpractice,"JournalofKnowledgeManagement7(1):6477(2003). Barab,S.A.(2003).Anintroductiontothespecialissue:Designingforvirtualcommunitiesintheserviceof learning.TheInformationSociety,19(3),197201. Barab,SashaA.,JamesG.MaKinster,andRebeccaScheckler,"Designingsystemdualities:Characterizinga websupportedprofessionaldevelopmentcommunity,"TheInformationSociety19(3),237256 (2003). Bluemink,JohanaandSannaJrvel,"FacetofaceencountersascontextualsupportworWebbased discussinsinateachereductioncourse,"TheInternetandHigherEducation7(3),199215(2004). Brown,J.S.&P.Duguid,"Organizationallearningandcommunitiesofpractice:Towardaunifiedviewof wokring,learning,andinnovation,"OrganizationScience2(1):4057(1991). Bryk,A.S.&Raudenbush,S.W.(1992).HierarchicalLinearModels.NewburyPark,London,NewDelhi:Sage. Butler,BrianS.,"Membershipsize,communicationactivity,andsustainability:Aresourcebasedmodelof onlinesocialstructures,"InformationSystemsResearch12(4):346362(2002). Chen,I.Y.L.(2007).Thefactorsinfluencingmembers'continuanceintentionsinprofessionalvirtual comunitiesalongitudinalstudy.JournalofInformationScience,33(4),451467. Coleman,J.S.(1988).'SocialCapitalintheCreationofHumanCapital.'AmericanJournalofSociology94 (supplement),S95S120. Coleman,J.S.,1990.FoundationsofSocialTheory.Cambridge:HarvardUniversityPress. Cress,Ulrike.2004.Strategic,metacognitive,andsocialaspectsinresourceorientedknowledgeexchange.In Proceedingsofthe25thAnnualConferenceoftheCognitiveScienceSociety.EditedbyR.Alterman andD.Kirsch.Mahwah,NJ:LawrenceErlbaum.:282287. Dannecker,A.,&Lechner,U.(2007)."Onlineandofflineintegrationinvirtualcommunitiesofpatientsan empiricalanalysis.in:Steinfield,J.,Pentland,D.,Ackerman,J.,&Contractor,J.(eds).Communities andTechnologies.ProceedingsofthethirdCommunitesandTechnologiesConference.London:151 170. Dawes,R.M.(1980).SocialDilemmas.AnnualReviewofPsychology,31(1),169193. Franzen,A.(2002).SocialCapitalandtheNewCommunicationTechnologies.InJ.Katz(Ed.),Machinesthat BecomeUs(pp.105116).NewBrunswick,NJandLondon:TransactionPublishers. Gal,Yoav,"Therewardeffect:Acasestudyoffailingtomanageknowledge,"JournalofKnowledge Management8(2),7383(2004). Granovetter,Mark.1985.'EconomicActionandSocialStructure:TheProblemofEmbeddedness.'American JournalofSociology91(3),481510. Hampton,KeithN.andBarryWellman,"NeighboringinNetville:HowtheInternetsupportscommunityand socialcapitalinawiredsuburb,"City&Community2(4),277311(2003). Hargittai,E.(2005).Surveymeasuresofweborienteddigitalliteracy.SocialScienceComputerReview,23(3), 371379.

21

Haythornthwaite,Caroline,"Onlinepersonalnetworks,"NewMedia&Society2(2):195226(2000). Haythornthwaite,Caroline,"ExploringMultiplexity:Socialnetworkstructuresinacomputersupporteddistance learningclass,"TheInformationSociety17(3),211236(2001). Heintz,Bettina.2000.GemeinschaftohneNhe?VirtuelleGruppenundrealeNetze.InVirtuelleGruppen. CharakteristikaundProblemdimensionen.EditedbyUdoThiedeke.Opladen:WestdeutscherVerlag.: 188218. Hlebec,V.,Manfreda,K.L.,&Vehovar,V.(2006).Thesocialsupportnetworksofinternetusers.NewMedia& Society,8(1),932. Ishii,K.&Ogasahara,M.(2007).Linksbetweenrealandvirtualnetworks:Acomparativestudyofonline communitiesinJapanandKorea.Cyberpsychology&Behavior,10(2),252257. Jarvenpaa,SirkkaL.Knoll,Kathleen,andDorothyE.Leidner,"IsAnybodyoutthere?AntecedentsofTrustin GlobalVirtualTeams,"JournalofManagementInformationSystems14(4):2964(1998). Johnson,ChristopherM.,"ASurveyofCurrentResearchonOnlineCommunitiesofPractice,"Internetand HigherEducation4(1),4560(2001). KleineStaarman,J.,"Facetofacetalktosupportcomputermediateddiscussioninaprimaryschoolliteracy practice,"Reading2(37):7380(2003). Kling,RobandChristinaCourtright,"Groupbehaviorandlearningineelctronicforums:asociotechnical approach,"TheInformationSociety19(3),221235(2003). Kollock,P.1999.TheEconomiesofOnlineCooperation:GiftsandPublicgoodsinCyberspace.InCommunities inCyberspace.EditedbyM.SmithandP.Kollock.London:Routledge.:220239. Komito,L.,(1998)."TheNetasaforagingsociety:Flexiblecommunities,"in:TheInformationSociety14(2), 97106. Matzat,U.(2009)."TheEmbeddednessofAcademicOnlineGroupsinOfflineSocialNetworks:ReputationGain asaStimulusforOnlineDiscussionParticipation?"in:InternationalSociology,24(1),6392. Matzat,U.(2009b)."ATheoryofRelationalSignalsinOnlineGroups."in:NewMedia&Society,11,3:375 394. Matzat,U.(2006)."KnowledgeManagementinaVirtualOrganization:Areembeddedonlinecommunitiesof practicemoresuccessfulthanexclusivelyvirtualones?"ReportfortheDutchScienceFoundation/NWO withintheResearchProgram'SocietyandtheElectronicHighway'(NWOMES),NWOprojectnumber 01443618. MatzatU.(2004)."AcademicCommunicationandInternetDiscussionGroups:TransferofInformationor CreationofSocialContacts?"in:SocialNetworks,26(3),221255. Matzat,U.(2004b)."TheSocialEmbeddednessofAcademicOnlineGroupsasaNormGeneratingStructure:A TestoftheColemanModelonNormEmergence"in:ComputationalandMathematicalOrganization Theory,10(3),205226. McLureWasko,M.andS.Faraj,""Itiswhatonedoes":whypeopleparticipateandhelpothersinelectronic communitiesofpractice,"JournalofStrategicInformationSystems,9(2/3),155173(2000). Paulhus,D.L.(1991).Measurementandcontrolofresponsebias.InJ.P.Robinson&L.S.Wrightsman(Eds.), Measuresofpersonalityandsocialpsychologicalattitudes(pp.1759).NewYork:AcademicPress. Parks,M.R.,"MakingFriendsinCyberspace,"JournalofCommunication46(1),8097(1996). Parks,M.R.andL.D.Roberts,"'MakingMOOsic':Thedevelopmentofpersonalrelationshipsonlineanda comparisontotheirofflinecounterparts,"JournalofSocialandPersonalRelationships15(4),517537 (1998). Preece,Jennifer.2000.OnlineCommunities:DesigningUsability,SupportingSociability.Chichester,NewYork, Weinheim,Brisbane,Singapore,Toronto:JohnWiley&Sons,LTD. Raub,W.andJ.Weesie,"Reputationandefficiencyinsocialinteractions:Anexampleofnetworkeffects," AmericanJournalofSociology,96(3),626654(1990). Ridings,C.M.,D.Gefen,andB.Arinze,"Someantecedentsandeffectsoftrustinvirtualcommunities,"Journal ofStrategicInformationSystems11(34),271295(2002). Riegelsberger,Jens,AngelaM.Sasse,andJohnMcCarthy,"TheResearcher'sDilemma:EvaluatingTrustin ComputerMediatedCommunications,"InternationalJournalofHumanComputerStudies58(6),759 781(2003). Schlager,MarkS.andJudithFusco,"Teachingprofessionaldevelopment,technology,andcommunitiesof practice:Areweputtingthecartbeforethehorse?,"TheInformationSociety19(3),203220(2003).

22

Snijders,C.&Weesie,J.(1999).Socialeorintaties,tijdspreferentiesendestabiliteitvanrelaties.In M.Kalmijn,W.Bernasco,&J.Weesie(Eds.),HuwelijksensamenwoonrelatiesinNederland:De organisatievanafhankelijkheid(pp.209226).Assen:vanGorcum. Stoddart,Linda,"Managingintranetstoencourageknowledgesharing:opportunitiesandconstraints,"Online InformationReview25(1),1928(2001). Thorn,BrianK.andTerryConnolly,"DiscretionaryDataBases:ATheoryandSomeExperimentalFindings," CommunicationResearch14(5),512528(1987). Utz,Sonja,"SocialInformationProcessinginMUDs:TheDevelopmentofFriendshipsinVirtualWorlds", retrieved20112000fromtheWWW:http://www.behavior.net/JOB/v1n1/utz.html(2000). VanSchuur,W.&Kiers,H.A.L.(1994).Whyfactoranalysisoftenistheincorrectmodelforanalyzingbipolar concepts,andwhatmodeltouseinstead.AppliedPsychologicalMeasurement,18(2),97110. Wellman,B.(2001).Computernetworksassocialnetworks.in:Science293(14),20312034. Wellman,Barryetal.,"DoestheInternetIncrease,Decrease,orSupplementSocialCapital?SocialNetworks, Participation,andCommunityCommitment,"AmericanBehavioralScientist45(3),436455(2001). Wellman,B.,Boase,J.,&Chen,W.(2002).TheNetworkedNatureofCommunityOnlineandOffline.IT& Society,1(1),151165. Wellman,BarryandMilenaGulia.1999.Netsurfersdon'tridealone:virtualcommunitiesascommunities.In NetworksintheGlobalVillage.EditedbyBarryWellman.Boulder,CO:WestviewPress.:331367. Wenger,EtienneC.1998.Communitiesofpractice:Learning,meaning,andidentity.Cambridge:Cambridge UniversityPress. Wick,C.,"Knowledgemanagementandleadershipopportunitiesfortechnicalcommunicators,"Technical Communication47(4),515529(2000). Yang,ShuChingandShuFangLiu,"Casestudyofonlineworkshopfortheprofessionaldevelopmentof teachers,"ComputersinHumanBehavior20(6),733761(2004). Yuan,Y.C.,Gay,G.,&Hembrooke,H.Focusedactivitiesandthedevelopmentofsocialcapitalinadistributed learning"community".InformationSociety,22(1),2539(2006). Zhao,S.(2006).DoInternetUsersHaveMoreSocialTies?ACallforDifferentiatedAnalysesofInternetUse. JournalofComputerMediatedCommunication,11,article8.

NOTES
1. Forthemultivariatetestsofthehypothesesonlydataof26communitiesisusedbecauseofmissingvalues. 2. Matzat(2006)presentsadditionalanalysesofthedatashowingthattheuseofasomewhatdifferentembeddednessscalewithalower scalability(H=.3)leadstoasignificantnegativeeffectofembeddednessontheperceivedintensityoffreeriding.Thedifferentfinding betweenthetwoanalysescanbeexplainedbythefact thathigherscalabilitywas obtained throughtheinclusionof onlythoserespondents withamembershipofatleasttwoyears(insteadofatleastoneyearasdoneinMatzat[2006]).Thisledtotheexclusionof two communities.

23

Appendix Variables
1)Perceivedfreeriding 2)Perceivedproblemoftrust 3)Perceivedfluctuation 4)Willingnesstocontribute 5)Willingnessto trust 6)Actualcontributionbehavior(yes=1) 7)Digitalliteracy 8)Digital experience 9)Trustingdisposition 10)Prosocialorientation 11)Socialdesirability 12)Gender(female=1) 13)Age(loge) 14) Extent of fulltimejob asteacher(loge[%]) Groupvariables(N=26) Groupsize (rawdata) Embeddedness
740 26.4 797 4.5

Mean SD
.04 .02 2.1 0 0 .21 .08 6.41 .03 .87 .58 .56 3.75 4.33 .93 .90 1.2 .90 .94 .40 1.03 1.88 .95 .21 .89 .50 .27 .34

1)
1 .53*** .35*** .17*** .13** .09* .03 .07 .01 .11* .09* .07 .02 .03 Group size 1 .46*

2)

3)

4)

5)

6)

7)

8)

9)

10)

11)

12)

13) 14)

1 .46*** 1 .15*** .05 .05 .04 .09 .06 .07 .07 .06 .06 .03 .05 .00 .09* .02 .10* .02 .02 .03 .10* .02 .03 1 .32*** 1 .21*** .24*** .15*** .13** .05 .07 .12** .02 .03 .03 .01 .11* .17*** .07 .03 .04 .12** .01 1 .16*** 1 .13** .06 .06 .02 .00 .02 .02 .05 .04 .02 .002 .07 1 .09 .06 .12* 1 .01 1 .00 .04 1 .02 .05 .03 1 .15** 1

.21*** .00 .09* .01 .04

.23*** .14** .02 .05

.06 .03

.28*** .12* 1 *

Embeddedness

n=725 (n=516forvariables13),N=26,*p<.05**p<.01***p<.001 (twosided)

24

You might also like