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
B.Behaviourism
Pavlovwasstudyingre exesinanimals,andpublishedin1903hisfamousexperimentonhowaconditionedsalivationre excouldbecreatedwithadog.HisworkwascarriedonbyThorndike,whocametotheconclusionthatlearningisimprovedwhenitachievedasatisfactoryresult,withtheextensionthatbeingwrongdoesnotteachtocorrecterrors.Then,Watsonpublishedin1913“Psychologyasthebe-haviouristsviewsit,”which,alongwithPavlov’sconditioningcreatedanewparadigm.Thoseconceptswerere nedbySkinnerin1936whenhecameupwiththeSkinnerBox(torecordanimalorbabybehaviourwhenconfrontedwithlevers,keysordiscsthatprovidedreinforcessuchasfoodandwater)inwhichhisseconddaughterspentmuchofherbabyhood.Skinnerthendevelopedteachingmachinessostudentscouldlearnbitbybit,uncoveringanswersforimmediate“reward.”Thesetechniqueswerethesourceofthe1970/80drillsandpractice(i.e.aseriesofexercisesfollowingthesamepattern)andMultipleChoiceQuestions.Mostcomputer-basedself-instructionsoftwarearestillbasedonSkinner’stechniques,usinggenerallyalinearpath,orbranchesforevolvedsystems.Thefollowingstageisthetutorial,thatappearedinthe70’s.Themainadditionisremediation,i.e.givingnecessaryinformationtocorrecttheerror(whentheansweriswrong)ratherthanabuzzingsoundortheimageofanexplodingbombonthescreen.Thismeansthatthesoftwareshouldtrytoanticipateonpossibleanswers,tobeabletogivethegoodremediationwhentheansweriswrong.Unfortunately,itisimpossibleforadeterministicsoftwaretoanticipateallpossiblewronganswers,meaningthatthereisnorealsolutiontothiscrucialproblemwithtutorials.Anotherproblemwithsuchsoftwareisthatonlyonewayisusedtosolvetheproblemamongmanypossibilities.
Finally,themostadvancedstageinbehaviourismisIntelli-gentTutorialSystems[11].ThistechniquecombinestutorialswithArti cialIntelligencetechniquestoanalysewrongan-swersand ndtheremediationthatwillbethemosthelpfultothestudent.Unfortunately,thismeansthattheAIassociatedwiththetutorialneedstobedevelopedspeci callyforthedomainaddressedbythetutorial.
Currentresearchisfocussedonre ningtheAIalgorithms,aswellascollaborativelearningenvironments,wherestudentscommunicateoverawideorlocalareanetwork[22].
III.PARASCHOOL’SE-LEARNINGSOFTWAREAND
REQUIREMENTSA.CurrentSoftware
Theinternet-basede-learningsoftwareoftheFrenchParaschoolcompanyoffersacomplementtohigh-schoolteachinginthreequitedifferentdomains:Mathematics,PhysicsandFrench.Thesoftwareisbasedonbehaviourismandremediation(notionoftutorial),withshortlessons,exer-cisesandmultiplechoicequestions.Morethan10000stu-dentsareusingthesoftware,eitherindividuallyathome,overtheinternet,oringroupsinhigh-schools,overalocal
areanetwork,withasupervisingteacher.Whenworkingathome,ifinteractionwithahumantutorisneeded,thestudentcan“chat”withateacheronduty.Tougherquestionsaresubmittedandansweredbymailwithinoneday.Theoriginalsoftwareprovidesdifferentsubjectsthatoftenstartwithashortanimatedlesson(Flash le),followedbyaseriesofexercises.Whenenteringanexercise,thesystemmonitorsthestudentactivity,suchasthetimespenttosubmitananswer,soastoprovideaglobalevaluationofthestudentonthegivensubject.Iftheansweriswrong,remediationisimplementedundertheformofcorrectiveexplanationsthataregiventothestudent.Parentsortutorsmayselectsubjectsorexercisesandincludethemintoanagenda.Itemsdonotdisappearoftheagendauntiltheyarevalidated.
Althoughthesystembehavesdifferentlydependingontheanswerprovidedbythestudent,thenavigationremainsdeter-ministicinthesensethatonceasubjectischosen,allstudentswillhavethesamelistofexercisestosolve,presentedinanorderthatisdeterminedbythepedagogicalteam.Individualorcollectivespeci cities(dynamicorstatic)arenottakenintoaccount.B.Requirements
Paraschoolwaslookingforasystemthatwouldenhancesitenavigationbymakingitadaptivetotheuser,accordingtothehistoryofthesystem(recordingofsuccessandfailureforeachstudent,plusotherparameters,suchasthetimespentonanexercise,etc.),sothatbothindividualpro lesandgeneraltimechangingconstraintscouldbetakenintoaccount.
Differentstudentsshouldseeadifferentsoftware,adaptedtotheirlevelandspeci cities,asdetectedbythesoftwareorthetutor.Goodstudents(whohaveahighrateofsuccess)shouldnotbebotheredwithelementarylessonswhich,onthecontrary,mustbesubmittedtostudentswithalowsuccessrate.
C.ProposedSolution
Thenumerousdif culties(multiplecontradictoryobjectives,fuzziness,complexity)pertainingtothisproblemsoonrangthebellsofSoftComputingandEvolutionaryTechniques,amongwhichAntColonyOptimisation(ACO)seemedparticularlywellsuited.
ApplyingthistechniquetotheParaschoolproblemisrela-tivelystraightforwardwhentheParaschoolapplicationisseenasagraphwithvaluatedarcs:nodesarepedagogicalitems,arcsarehypertextlinksandprobabilitiesareassociatedtoarcs.Thenumerousstudentsareimplementedasants,thatroamthroughthegraphwhileleavingevaporatingpheromonesalongthearcstheyfollowtore ecttheirrateofsuccess.
AntmodelsandmoregenerallySwarmIntelligenceheuris-ticsandtheiremergentpropertiesseemtobeespeciallyadaptedtothisproblemwhereonetriestocomeupwithasolutionthatsatis esmultipleandsometimescontradictoryin uences(fromstudents,teachersandevenexercises,whichcanbesaidtohaveanin uencebybeingmoreorlessdif cult)inagradualanddynamicfashion.
热门文档