对等知识管理（p2p+knowledge+management）

Nov14,2001DeLiu,CREC1Peer-to-peerKnowledgeManagementProposetheideaofP2PknowledgemanagementWhatdoesP2Pmean?Howcanpeer-to-peerhelpKM?DiscussproblemswithP2PKMFree-rideIssueandincentivedesignImpactoftechnologyonorganizationalexchangenetworkNov14,2001DeLiu,CREC2WhatdoesPeer-to-Peermean?Participatingcomputersarebothclientsandservers.ThreekindsofP2PapplicationsP2Pinformationsharing(Napster;FreeNet)distributedresourcesand/orpeer-to-peerroutingP2PCollaboration/Communication(groove;ICQ)Distributedcomputing(SETI@home)CoordinatemultiplecomputersforcomplextasksNov14,2001DeLiu,CREC3MoreFeaturesofP2PinformationsharingInformationavailableatcertaintimeisthesumoftheinformationcontenteachnodeshares.Nodescaneasilyjoinorquit.Makeuseofunder-utilizedresourcesItusuallycreatesomeredundancytoreducetheimpactofsinglenodefailure.Load-balancingmechanism.Nov14,2001DeLiu,CREC4DrawbacksofP2PinformationsharingAnonymousv.s.CopyrightRedundancyv.s.Up-to-dateQualityofServiceSearchEfficiencyFree-Ride!Nearly70%ofGnutellauserssharenofiles,nearly50%ofallresponsesarereturnedbythetop1%ofsharinghosts.…Wearguethatfreeridingleadstodegradationofthesystemperformanceandaddsvulnerabilitytothesystem.Ifthistrendcontinuescopyrightissuesmightbecomemootcomparedtothepossiblecollapseofsuchsystems.(AdarandHuberman2000,FreeRidingonGuttela)Nov14,2001DeLiu,CREC5WhyweproposeP2PKMP2PtechnologyfitwellwithKMobjectivesUnlock“hidden”organizationknowledgeEnhanceinteractionsthrough“push”&“pull”FosterknowledgesharingcommunitiesContributetopropagationoftacitknowledgeToleratefuzzyknowledgeanddifferentpointsofviewWeexpectfree-ridingissuecanbebetterhandledwithinorganizationcontext.Otherdesirablefeatures:flexible,robust,secure,costefficiency(?)Nov14,2001DeLiu,CREC6SowhatisP2PKM?twocomponents:P2PinformationsharingcomponentexplicitknowledgetransferNotificationofupdateandnewrelease(“push”)AccesscontrolContent-basedaddressingP2PcollaborationcomponentimplicitknowledgetransferMultiplechannels(text,multi-media)Real-timecollaborationandnon-realtimecollaborationP2PKMandcentralizedeffortofknowledgemanagementNov14,2001DeLiu,CREC7PotentialproblemsofP2PKMQualityofknowledgeLackofconsistencyFree-riderproblemKnowledgeisoftentreatedaspublicgood.It’srationalforself-interestagentstoconsumepublicgoodwithoutcontributing(freeride).Consequentlysomesociallydesirablegoodisnotprovidedbyagentsduetoinadequatecompensation.Thisunder-provisionofpublicgoodproblemmayeventuallydefeattheP2Pknowledgemanagementsystem.Nov14,2001DeLiu,CREC8RelatedIS/MgmtLiteratureMostresearchinIS/Mgmttrytoaskaquestion:“whatarereasonsforknowledgesharing”Altruism(constantetal94,PalfreyandPrisbrey1997,DavenportandPrusak1998)&Groupcommitment(O’ReillyanChatman1986;Kollock1999)Reciprocity(DavenportandPrusak1998;Ekeh1974)Reputation(DavenportandPrusak1998,Rheigngold1993)self-efficacy(Kollock1999,Bandura1995)Factorsfavorthesemotivationsfutureinteraction;Recordofpastinteractions;Identitypersistence;Welldefinedanddefendedgroupboundary;Visibilityandrecognitionofcontributions;Groupsize;etc(Kellock1999)Nov14,2001DeLiu,CREC9RelatedEconomicsLiteratureKnowledgeandKnowledgeexchangeValueofknowledgeisdifficulttomeasureorverify;ValueofknowledgemaybedifferentfordifferentagentsValueofknowledgemayinvolvessomeuncertainty.Knowledgeexchangeisoftenembeddedandaffectedbysocialnetworks.So,unlikeothermoretangiblegoods,thereislackofapricesystem(market)orcompletecontractformforknowledge.Nov14,2001DeLiu,CREC10RelatedEconomicsLiteratureWageschedulesthatencourageteamproductioncollaboration.(Multi-Agent:Che&Yoo2001;Grooves1973;)Shouldcompetitionorcollaborationbeencouraged?Whatistheoptimalwagescheduleforteams?IncentivesandTaskstructure(Multi-task:Holmstrom,Milgrom1982;Itoh1991)Eitherthetaskitselfcanberewardedorthemarginalopportunitycostforthetaskcanbeloweredbyremovingorreducingtheincentivesoncompetingtasks.Nov14,2001DeLiu,CREC11Network(Jackson&Wolinsky1995;Watts1999;Kranton1996,1999,2000)Whatkindofnetworkstructureisoptimal?(P2P?Star?)Canagentsformefficientnetwork?QuestionsandSuggestions?[e-mail]