Sara Ghaoui2025-09-212025-09-212025-07http://dspace.univ-khenchela.dz:4000/handle/123456789/9377In anincreasinglydigitaleducationalcontext,Computer-BasedHumanLearningEnviron- ments(CBHLEs)playacrucialroleinsupportingbothindividualandcollaborativelearn- ing. Collaborativelearningsupportedbytheseenvironmentsoffersamajoropportunityto strengthen learnerengagementandimprovethequalityofpeerinteractions.However,to fully leveragethispotential,itisessentialtounderstandandassesslearningdynamicsusing reliable indicatorsderivedfrominteractiontraceanalysis. This thesisfocusesonthedesignandcomputationofcollaborationindicatorsinComputer- Based HumanLearningEnvironments(CBHLEs).Itproposesamodel-drivenarchitecture (MDA)-basedapproachtomaketheindicatorcomputationplatform-independent. Three maincontributionsarepresented: • First contribution:amodeltransformation-basedapproachisproposedtocompute collaborationindicatorsindependentlyofthelearningplatform. • Second contribution:aformalmodel(DECINModel)isintroducedtoassistteachers in designingvalidcollaborationindicatorstailoredtotheirobservationneeds. • Third contribution:thecomputationprocessisautomatedthroughthegenerationof Sequences oftransformations,generatedbyaformalsystem(DECIN-AGSTSystem). Tosupportthisapproach,aplatform-independenttracemodel(T-PIM)isproposed, along withformalconstraintsexpressedinOCLtoensurethevalidityandconsistencyofthe collected traces.Similarly,agenericmodelforcollaborativeindicators(CI-PIM)isdefined, supportedbyconstraintsthatensurethequalityofthegeneratedindicators. Finally,theautomationoftransformationsequencegenerationisachievedthroughthe implementationoftheDECINmodelinaformalsystem(DECIN-AGST),whosevalidityhas beenverifiedusingthemodelcheckingmethodviatheUPPAALtool,thusensuringsafety, accessibility,andabsenceofdeadlocks.enThesis