Enabling personalized education through the use of machine learning and learning analytics

Dette prosjektet vil bruke maskinlæring for å støtte elevenes læring. For å forbedre elevens læringsreise vil vi utvikle algoritmer som lager elevprofiler basert på elevens digitale aktiviteter, samt algoritmer som enkoder utdanningsressurser til et enhetlig og sammenlignbart format. Vi vil generere elevprofilene som en naturlig del av læringsprosessen i stedet for å basere oss på standardiserte tester. I første omgang vil vi fokusere på å fylle profilene med lese- og skriveferdighetene til elevene, og derfra tar vi sikte på å utvide til forklarbar innsikt i hvordan eleven lærer best. Siden maskinlæringsmodeller generelt krever inndata i et standardisert format, vil disse enkodingene gi et praktisk og solid utgangspunkt for AI-applikasjoner for utdanningssektoren. Algoritmene våre for enkoding av pedagogiske ressurser vil være i stand til å håndtere alle datamodalitetskombinasjoner (pdf-filer, videoer, pedagogiske spill, etc.). Et stort problem med forskning på dette feltet er at teorier og rammeverk mangler testing i den virkelige verden. Vi vil løse dette problemet ved å implementere Ph.D. kandidatens prototyper inn i Neddy-plattformen og tester dem med våre partnerskoler.

This project plans to employ machine learning to support student learning. The project's core is student profiling, enabled by AI and interpretable by humans to assist teachers, parents, and students in understanding how the student learns best. We plan to create a machine learning framework and a fully functional prototype that will seamlessly build profiles while the student works integrated into the learning experience without explicit testing. The framework will also aid teachers in creating content for their class, optimizing the learning activities to account for the profiles of the students in class. Lastly, we will focus on the individual student's learning process and investigate if AI can improve the students' learning strategies and techniques. The project will utilize and advance the field of multimodal neural networks by first applying such networks to dyslexia detection and on the remaining objectives. We will build on the recent development for the NorBERT model by using it as an information encoder of Norwegian classroom content. The core of student profiling will fill address an open research question of skill assessment for adaptive learning systems. We will also create a novel dataset by cooperating with Norwegian schools to advance research endeavors in AI in educational technologies. We plan the following academic papers: Paper 1: Dyslexia detection with deep neural networks Paper 2: Dyslexia detection with multimodal networks Paper 3: Big data and ANN techniques for content recommendation and adaption. Paper 4: Optimizing classroom content based on student profiles Paper 5: Optimizing study techniques based on student profiles Paper 6: Evaluation of AI optimization for real-world Norwegian classrooms During the project phase, the candidate will stay a minimum of one year at the company (where the candidate has a specified desk during the entire period), and a minimum of one year at UiA (the candidate has available desk at UiA).