Improving MOOCs Through the Development of Automated Tools for Learner Modelling

Simple item page
dc.contributor.advisorFalkner, Nickolas
dc.contributor.advisorAtapattu, Thushari
dc.contributor.advisorThilakaratne, Menasha
dc.contributor.authorSridhara, Abhilash
dc.contributor.schoolSchool of Computer and Mathematical Sciences
dc.date.issued2025
dc.description.abstractMassive Online Open Courses (MOOCs) have seen a massive rise in popularity in recent years. The COVID-19 pandemic resulted in many educational providers adopting more MOOCs to augment and replace conventional brick-and-mortar classrooms. The number of fully online degrees from universities and fully online micro certifications have also increased in popularity. In these online learning contexts, there is a lack of personalised interactions with both peers and instructors for the enrolled learners. To benefit more learners, there is a need for automated systems that can tailor the experience of learners in MOOCs to ‘supplement’ and ‘replace’ the loss of interaction(s) observed in online contexts. This thesis focuses on the task of automation in MOOC contexts through Learner Modelling (LM). LM comprises of techniques to utilise learner-item interaction in educational systems to create an estimate of the learner which can be used for a variety of uses including predicting the next response correctness probability, optimisation of instructional policies, personalising the learning path of learners and others. This thesis is divided into five main research objectives that address the following topics: Utilising partial credits in learner modelling, development of lightweight learner modelling techniques for performance prediction, development of a deep learning model for performance predictions with a high level of transferability, dveloping a model for grading short text responses for a wide variety of ordinal ranges and generating knowledge graphs to visualise the different skills and their structure as well as the paths learners take as they progress through the course. Utilising partial credits in learner modelling involves awarding partial marks when learners answer questions after requesting a hint or after multiple attempts. This can improve learner hint-seeking behavior by rewarding learners with partial marks when they use scaffolding questions to answer questions they initially got wrong. The scalable performance prediction model presented in the thesis addresses the need for accurate performance prediction techniques that require limited computing resources. This would enable widely adapted LMS systems such as Moodle to predict learner proficiency. The transferable deep-learning model is aimed at utilising the ability of deep learning model to learn relationships within existing question items. This could automate the process of creation and labelling of assessment items with Knowledge Components (KC) or skills which is a necessity for adaptive scheduling systems The short text grading system focuses on improving the accuracy of existing text grading models while retaining better transparency. The system is more informative on why a particular mark was awarded compared to existing models. The generation of knowledge graphs is a powerful visualisation technique for domain modelling analysis where the different skills and their relationship with each other are expressed as a graph. This also enables identifying the most important skills and any hierarchical skill connections if they exist. Visualising the paths learners take through the course when they struggle with a skill and navigating to other skills before revisiting and attempting the skill they struggled with can be very informative for optimising educational content. In summary, the findings presented here could alter how Learning Management Soft wares (LMS) are integrated into MOOC suites. The development of scalable automated tools for learner modelling that can be utilised in a wide context and can be transferred across subjects with minimal effort could aid in wide adoption of such systems in existing LMSs. This could result in the development, publication and maintenance of MOOCs with learner engagement on par with ITSs specifically designed for different contexts.
dc.description.dissertationThesis (Ph.D.) -- University of Adelaide, School of Computer and Mathematical Sciences, 2025en
dc.identifier.urihttps://hdl.handle.net/2440/144882
dc.language.isoen
dc.provenanceThis electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at: http://www.adelaide.edu.au/legalsen
dc.subjectMOOCs
dc.subjectIntelligent Tutoring Systems
dc.subjectLearner modelling
dc.titleImproving MOOCs Through the Development of Automated Tools for Learner Modelling
dc.typeThesisen

Files

Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
Sridhara2025_PhD.pdf
Size:
29.61 MB
Format:
Adobe Portable Document Format
Download

Collections

Research Theses