Bitutor – Bayesian Intelligent Tutoring System

Abstract

Computershavebeenusedineducationsince1960.Thepurposeofusingcomputersinassistinginstructionistohelpstudentslearnmoreefficiently.Untilrecently,thecomputerizedtutoringresearchhasbeenfocusedonstudentmodeling.Decisionmakingundertheenvironmentofuncertaintyhasalsobeenanareaoffocus oftheseresearchers.Inliterature,thestudentmodelhasbeenclassifiedbasedonvariousfactors.Literaturesurveyrevealsthattheconstructionofstudentmodelcanbebroadlyclassifiedintostereotypes,overlay,andextendedoverlay.Diagnosisofstudent'smisconceptionsisgenerallythroughmodeltracingorconstraint-basedmodeling.AgeneralmethodologyforbuildinganIntelligentTutoringSystemisproposedanddemonstrated.ThemethodologyadvocatesbuildinganadaptiveandgeneralizedBayesiannetworkstudentmodel.TherandomvariablesofthisBayesiannetworkdenotethestochasticinformationassociatedwiththemasterystatesfortheknowledgenodeofthedomainknowledge.Inaneducationalenvironment,agoodstudentmodelmustincludeallthefeaturesofthestudent’sknowledgeandpreferencesrelatedtohislearningandperformance.Thisinformationisusedtoadaptthesystemtothestudent.Assessmentisanecessarypartoftutoring.Anumberofdecisionsneedtobetakenaboutstudents’knowledgemasterywhiletutoring.Eachdecisiontakenbythetutorforremediationisbasedonthecurrentinformationinthestudentmodel.Inthisresearch,decisiontheoreticpedagogicalactionselectionstrategyisusedformakingdecisionsunderuncertainty.Forassessingtheoverallprogressofthelearners’domaincompetence,thefocusofassessmentshouldbeontheacquisitionofskillsintheapplicationoffactsindifferentcontextualandnon-contextualscenariosandemphasisshouldbemoreonthecognitiveskills.Thediagnosticcapabilityofthestudentmodelmustbeextendedtoincludex theselectionofitemstomatchthestudent'smasterystate.Thismakesthesystemadaptiveandthetestitemschallengingtothestudents.ProvidingadaptivefeedbacktothestudentswhenthestudentneedsfeedbackorhelpfromthetutorisalsoaprominentprobleminIntelligentTutoringSystem.Thetutoringstrategycommonlyusedbytheteachersneedstobeformulated.Tutoringstrategyisasequenceofoptimalactionstobetakenbythetutor dependingonthestudent'smastery.Thisstrategyispersonalizedtothestudentandthissequenceofactionsisabletodirectthestudent'slearninginthemosteffectiveandefficient manner.ForasuccessfulIntelligentTutoringSystem,therearethreekeyfactorsidentified:(i)thegenerationofacognitivemodeltofacilitateitscommunicationtostudent;(ii)studentsmustbeprotectedfromthepotentiallyhighcostoferrorsbyminimizinglearningtime;and(iii)theabilitytomonitorstudent'sunderstandingoftheconcepts.OnesuchtutoringsystemdevelopedtoaddressthesekeyfactorsisBiTutor(BayesianIntelligentTutoringSystem).BiTutorattemptstobuildamodelofthestudent,andusethatmodelinconjunctionwithknowledgeaboutthedomainofinstructionandinstructionalstrategiestomodifytheorderofpresentationofmaterial,selectionofitemandfeedback,andstyleorfinteractionwiththestudent.BiTutorcanbeusedtoteachanycomputersciencecourses.TheevaluationshowsthatBiTutorisabletogeneratethetutoringstrategyinpolynomialtime.Finally,thisthesissuggestssomeareasofpossiblefuturework,whichmayleadtoimprovementsinIntelligentTutoringSystem.xi