Abstract — An attempt is made to apply Ant Colony Optimization (ACO) heuristics to an E-learning problem: the pedagogic material of an online teaching website for high school students is modelled as a navigation graph where nodes are exercises or lessons
AntColonyOptimisationforE-Learning:Observing
theEmergenceofPedagogicSuggestions.
YannSemet,EvelyneLutton,PierreCollet
INRIA–ProjetFRACTALES78150LeChesnay-France
Yann.Semet@inria.fr,Evelyne.Lutton@inria.fr,Pierre.Collet@inria.fr
Abstract—AnattemptismadetoapplyAntColonyOptimiza-tion(ACO)heuristicstoanE-learningproblem:thepedagogicmaterialofanonlineteachingwebsiteforhighschoolstudentsismodelledasanavigationgraphwherenodesareexercisesorlessonsandarcsarehypertextlinks.Thearcs’valuation,representingthepedagogicstructureandconditioningtheweb-site’spresentation,isgraduallymodi edthroughthereleaseandevaporationofvirtualpheromonesthatre ectthesuccessesandfailuresofstudentsroamingaroundthegraph.
Acompromiseisexpectedtoemergebetweenthepedagogicstructureasoriginallydictatedbyprofessors,thecollectiveexperienceofthewholepoolofstudentsandtheparticularitiesofeachindividual.
ThepurposeofthisstudyconductedforParaschool,theleadingFrenche-learningcompany,istwofold:enhancingthewebsitebymakingitspresentationintelligentlydynamicandprovidingthepedagogicalteamwithare nedauditingtoolthatcouldhelpitidentifythestrengthsandweaknessesofitspedagogicchoices.
Keywords:E-Learning,AntColonyOptimisation(ACO),SwarmIntelligence,EvolutionaryComputation,InteractiveEvolutionaryDesign.
I.INTRODUCTION
ThisworkwasinitiatedwhenParaschool,theFrenchlead-inge-learningcompanycontactedtheINRIAresearchcentertoconceiveanautomaticalgorithmthatwouldallowtherelativelyrigidalbeitfunctionalexistingParaschoolsoftwaretobehavedifferentlydependingonuserspeci cities.
Afterseveralbrainstormingsessionswhereneuralnetworks,evolutionaryalgorithmsandotherarti ciallyintelligenttech-niqueswereconsidered,itappearedthatswarm-likealgorithmscouldbeused,thankstothegreatnumberofactualusers(morethan10000)andmoreespeciallyant-basedprobabilisticoptimisationthatcouldeasilybegraftedontheexistingpedagogicalgraphconstitutedbytheParaschoolsoftware.Moreover,AntColonysystemspresenttheinterestingprop-ertyofexhibitingemergentbehaviourthatallowindividualstobene tfromthedynamicexperienceacquiredbythecollectivity,whichmeans,inpedagogictermsthatastudentcouldbene tfromthepedagogiclessonsdrawnoutofhispeers’successesandfailures.
TheimplementationofthesealgorithmsyieldsresultsthatgobeyondtherequirementsoftheParaschoolcompanywhichwillsoonbeexperimentinginrealsizetheautomaticdynamic
optimisationofthepedagogicgraph(theirsetofintercon-nectedlessonsandexercises)implementedbytheirsoftware.Thispapersuccessivelypresentsaconcisedescriptionofhuman-learningconceptsandtheirsoftwareimplementation,ashortdescriptionofthetechnicalimplementationoftheAnt-Colonybasedoptimisationalgorithmandadiscussionontheuseofvariousselectionoperators.Asetofexperimentsisthenconducted,showingthaterroneousarcprobabilitiescanbeautomaticallycorrectedbythesystem.
II.ELEMENTSONTHEPHILOSOPHYOFLEARNINGThemainconceptsofteachingandlearningusednowadaysarestillveryold.ThetwomaincurrentsareConstructivism,thatwaselaboratedbyKantandBehaviourism:atheorythatcamefromPavlov’sexperiments.A.Constructivism
In1781,Kanttriedtosynthesizerationalistandempiricistviewpoints.Kantseesthemindasanactiveagent,thator-ganisesandcoordinatesexperiences.Alongtheselines,Piagetstatesthatknowledgeisnotsimply“acquired,”bychildrenbitbybit,butconstructedintocoherent,robustframeworkscalled“knowledgestructures.”Childrenarenotpassiveabsorbersofexperienceandinformation,butactivetheorybuilders.
Papert,amathematician,andoneoftheearlypioneersofArti cialIntelligence(hefoundedtheArti cialIntelligenceLaboratoryatMIT),workedwithPiagetattheUniversityofGenevafrom1958-1963.ThiscollaborationledPaperttoconsiderusingmathematicsintheserviceofunderstandinghowchildrenlearnandthink[14].
InConstructivisttheory[2],theemphasisisplacedonthestudent:Teachersareseenascoacheswhoassiststudentstoconstructtheirownconceptualisationsandsolutionstoproblems.Collaborativegroupswithpeerinteractionareanimportantpartofthelearningprocess.
Suchmethodssupposethatthestudentisgivenmostoftheelementsallowinghimtosolveaproblem(sometimes,thestudentneedstoelaborateonhisownmissingelements),andhemustusehisintelligencetoconstructthesolution.Suchmethodsarenotwellsuitedtocomputer-basedlearningore-learningsoftware.
热门文档