Evaluation of recommender technology using multi agent simulation

Evaluation of Recommender Technology Using Multi-Agent SimulationZinaPetrushyna, Ralf KlammaMarch 22nd, 2011CELSTEC, Open University, Heerlen

AgendaMotivationTeLLNetGame TheoryNetwork Formation GamesMulti-Agent SimulationsFuture Work

TeLLNet = TeachersLifelong Learning NetworkWhy do someteacherscollaboratewithothersandsome not?163.330 registered teachersonly 29.119 teacherscollaborate in 19.128 projectsHowtocreatebettersupportforteachers?TeLLNet

Game Theory BasicsEvery situationas a game [Borel38, NeMo44]A player – makesdecisions in a gamePlayers choosebeststrategiesbased on payofffunctionsPayoffsmotivationsofplayersA strategydefines a setofmovesoractions a player will follow in a givengame (mixedstrategy, pure strategy)

Game TheoryA gameis a tuple , whereNis a nonempty, finite setofplayersEachplayerhasa setofactions (strategyspace) payofffunctionspayoffmatrix

Socialnetworksareformedby individual decisionsCost: write an e-mailUtility: cooperatewithothersSocialnetworksbetweenpupilsCost:make a jokeUtility:getappreciationfromothersLifelonglearnernetworksCost:take a learningcourseUtility: find learnerswithsimilarwayofreasoningNetwork Formation Games

Set ofagentswhichareactorsof a network. andaretypicalmembersof a setA strategyof an agentis a vectorwhereforeachActorandareconnectedifNetwork Formation

Nash Network : Win-Win SituationEvery agentchangesitsstrategyuntil all agentsaresatisfiedwiththeirstrategiesand will not benefitiftheychangestrategies (thenetworkisstable)  Nash equilibriumA networkis a Nash networkifeachagentis in Nash equilibriumChosen strategiesdefeatothersforthegoodof all players [Nash51, FuTi91]

Network Formation StrategiesHomophily – loveofthe same [LaMe54, MSK01]similarsocio-economicalstatusthinking in a similarwayContagiositybeinginfluencedbyothersHowtorepresentstrategiesfor alifelonglearner?

Epistemic Network Analysis: Assesmentof LearningLearning in action [Gee2003]Assessmentofisolatedskillsis not effectiveFocus on performance in context (actions)Evidence of learning: linking models of understandingobservable actionsevaluation [SHS*09]

Multi-Agent Simulation SystemA multi-agentsystemis a collectionofheterogeneousand diverse intelligent agentsthatinteractwitheachotherandtheirenvironment [SiAi08]Simulation of a real-worlddomain [LMS*05]Approximation ofthe real worldSimulation model consistsof a setofrulesthatdefineshowthesystemchangesover timePurposesofsimulationsystem:Betterunderstandingof a systemPredictions

Examples / State ofthe ArtRecommendationsYenta [Foner97] – lookingforuserswithsimilarinterestsbased on datafrom Web mediaMarket-bindingmechanismsLookingforthebest item (a rewardagent, setofitemsandusersagents) [WMJe05]Team formationFormingteamsforperforming a task in dynamicenvironment [GaJa05]

Multi-Agent Simulation QuestionsWhich kind of behavior can be expected under arbitrarily given parameter combinations andinitialconditions?Which kind of behavior will a given target system display in the future?Which state will the target system reach in the future? [Troitzsch2000]200920102008

Agent Based SimulationHeterogeneous, autonomous and pro-active actors, such as human-centered systemsAgents are capable to act without human interventionAgents possess goal-directed behaviorEach agent has its own incentives and motivesSuited for modeling organizations: most work is based on cooperation and communication [Gazendam, 1993]

Inputs forsimulation modelAgent =TeacherTeacherproperties: LanguagesSubjectsCountryInstitution roleAny Awards? (European Quality Label orPrize)Project properties:LanguagesToolsSubjectsNumberofpupils in a projectAge ofpupils in a projectAny Award? (Quality Label)

RecommendationTechniquesCollaborative filtering [Breese et al.1998]Memory-based: user-based, item-basedModel-based: Bayesian, pLSA, Clustering, etc.Content-based Recommendation [Sarwar et al.2001]Items featuresUsers‘ profilebased on featuresofrateditemsHybrid Techniques [Burke2002]Partner?

Simulation of Network Formation using Data Mining Compareteacherprofiles:subjects ,institutionalroles, experiences in projectsFind teachersthatsuittoeachotherCosinesimilarityBelief NetworksDecisiontrees The relationshipconcernsonly 2 teachersandomitsteachers in a network!

Network Formation Game Simulation Payoffdefinition: payoffmatrixiscalculateddynamicallybased on Epistemic Frame vector:teachers‘ subjects, subjectsofprojects (experiences)teachers‘ languages, languagesofprojects (experiences)toolsused in projects (experiences)countries pastcollaboratorsarecomingfrom (beliefs)...Strategydefinition: homophilyorcontagiosityLookingfor a suitablenetworkfor a teacherand not for a suitablepartner!

Nash Equilibrium forNetwork FormationFinding a Nash Equilibrium (NE) is NP-hardComputer scientists deal withfindingappropriatetechniquesforcalculating NE with a lotofagentsWeare not interested in thebestsolution but in a bettersolution

Future workRunningsimulation model withmanyagents (>100)Evaluation ofsimulationsresultscomparingnetworksEvaluation ofteacherssatisfactionofproposednetworksTools/techniquesforcomputing Nash equilibrium

ReferencesLuck, M., McBurney, P., Shehory, O., & Willmott, S. (2005). Agent technology: computing as interaction (a roadmap for agent based computing). Liverpool, UK: AgentLink.Troitzsch, K.G. Approaching agent-based simulation: FIRMA meeting 2000, Available via http://www.uni-koblenz.de/~moeh/publik/ABM.pdfGazendam, H.W.M. (1993). Theories about architectures and performance of multi-agent systems. In: III European Congress of Psychology. Tampere, Finnland.Burke, R. Hybrid recommender systems: Survey and experiments, User Modeling and User-Adapted Interaction12 (2002), pp. 331–370Helou, S. El, Salzmann C.,Sire S., Gillet, D. The 3A Contextual Ranking System: Simultaneously Recommending Actors, Assets, and Group Activities, in: Proc. of the ACM Conference On Recommender Systems, ACM, New York, 2009, 373–376.Herlocker J.L., Konstan J.A., Terveen L.G., Riedl J.T. (2004). Evaluating Collaborative FilteringRecommender Systems, ACM Transactions on Information Systems, Vol. 22, No. 1, January 2004, pp. 5–53.Manouselis, N. , Drachsler, H., Vuorikari, R., Hummel, H., Koper, R. (2010) Recommender Systems in Technology Enhanced Learning, in Kantor P., Ricci F., Rokach L., Shapira, B. (Eds.), Recommender Systems Handbook: A Complete Guide for Research Scientists & Practitioners.Brusilovsky P., Nejdl W., (2004) “Adaptive Hypermedia and Adaptive Web”, Practical Handbook of Internet Computing, CRC Press LLCWalker, A., Recker, M., Lawless, K., Wiley, D., “Collaborative information filtering: A review and an educational application”, International Journal of Artificial Intelligence and Education,14, 1-26, 2004.Nadolski, R., Van den Berg, B., Berlanga, A., Drachsler, H., Hummel, H., Koper, R.,& Sloep, P. (2009). Simulating light-weight Personalised Recommender Systems in learning networks: A case for Pedagogy-Oriented and Rating based Hybrid Recommendation Strategies. Journal of Artificial Societies and Social Simulation (JASSS), vol. 12, no 14, http://jasss.soc.surrey.ac.uk/12/1/4.html, Accessed 17 November, 2009.Drachsler, H., Pecceu, D., Arts, T., Hutten, E., Rutledge, L., Van Rosmalen, P., Hummel, H.G.K., Koper, R.: ReMashed - Recommendations for Mash-Up Personal Learning Environments. In: Cress, U., Dimitrova, V., Specht, M. (eds.): Learning in the Synergy of Multiple Disciplines, EC-TEL 2009, LNCS 5794, Berlin; Heidelberg; New York: Springer, pp 788-793, 2009aFoner, L. 1999. Political artifacts and personal privacy: The Yenta multi-agent distributed matchmaking system. Ph.D. thesis, Massachusetts Institute of Technology.Gaston, M.E. and des Jardins, M. Agent-organized networks for dynamic network formation. In ACM AAMAS’05, pp. 230-237, New York, USA, 2005Anderson, C. The Long Tail: why the future of business is selling less of more. New York: Hyperion, 2006Siebers, P.-O. and Aickelin, U. Introduction to multi-agent simulation. Computing research repository, 2008von Neumann, J. and Morgenstern, O. (1944), Theory of games and economic behavior, Princeton University PressBorel E. (1938) Applications aux Jeux de HasardMcPherson, M., L. Smith-Lovin, and J. Cook. (2001). Birds of a Feather: Homophily in Social Networks. Annual Review of Sociology. 27:415-44. Lazarsfeld, P., and R. K. Merton. (1954). Friendship as a Social Process: A Substantive and Methodological Analysis. In Freedom and Control in Modern Society, Morroe Berger, Theodore Abel, and Charles H. Page, eds. New York: Van Nostrand, 18-66.Gee, J.P. 2003 What video games have to teach about learning and literacy. New York: Palgrave Macmillian

Recommender Systems in TELTEL User Tasks supportedbyRecommender System [HKTR04, MDV*10] : Find peers!Adaptive systems (educationalhypermedia) [BrNe04] – contentselection, navigationsupport, presentationAltered Vista System [WRL*04]3A Contextual Ranking System [ESS*09]Recommenderalgorithmssimulations [NBB*09]ReMashed - tags andratingsof Web media [DPA*09]

What Do We Query in the Dataset?How do teachers(agents) maketheirdecisions?Whatpropertiesshouldthecollaboratorpossess?Whatpreferencesdoes a teacherhasaccordinghisfuture/currentpartners?How do teachers form theirfuturebehaviours?Whatpreferenciesmaybechanged in thefuture in definingtheircollaborationpartnersandwhy?How do theyremember he past? How do theylearnandreflect in theirbehaviour?

Evaluation of recommender technology using multi agent simulation

More Related Content

Similar to Evaluation of recommender technology using multi agent simulation (20)

More from Zina Petrushyna (8)

Recently uploaded (20)

Evaluation of recommender technology using multi agent simulation