h1

TY  - THES
AU  - Hensen, Benedikt
TI  - Designing effective mixed reality learning experiences: structuring insights with knowledge graphs and large language models
PB  - Rheinisch-Westfälische Technische Hochschule Aachen
VL  - Dissertation
CY  - Aachen
M1  - RWTH-2025-09217
SP  - 1 Online-Ressource : Illustrationen
PY  - 2025
N1  - Veröffentlicht auf dem Publikationsserver der RWTH Aachen University
N1  - Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2025
AB  - Mixed Reality (MR), a technology that merges the digital and real world, offers significant opportunities to enhance learning experiences by providing immersive 3D environments, visualizing complex workflow processes and enabling seamless collaborative interactions. However, the literature mainly contains isolated, context-specific case studies about a certain topic and course, limiting the applicability of insights about MR-enhanced learning for diverse educational scenarios. Instead, granular considerations are helpful that take into account factors about the learning setting like the number of students, the characteristics of the course, the theoretical or practical focus of the learning goals, a remote or co-located setting, etc. Consequently, there exists a need for educators and students to utilize a structured knowledge base that guides them in integrating MR into learning. While navigating around its limitations, they can then benefit from the strength and opportunities of the MR technology. This dissertation aims to address this gap by systematically exploring approaches to integrate MR into learning and documenting the findings in a knowledge graph. By following the design science research methodology, solutions are developed for educators which can be incorporated into their existing practices. The process consists of several steps that build up to the final knowledge base. The methodology starts with a comprehensive literature review that defines the solution space based on the characteristics of MR learning applications. Then, an infrastructure is established that enables efficient development of MR artifacts. These MR learning artifacts are constructed and evaluated in iterations. Finally, an MR learning ontology is formulated to structure the gathered information, and the insights from the studies are recorded in a knowledge graph. The outcomes are achieved in a scalable, innovative process using an Artificial Intelligence (AI)-in-the loop process. A locally hosted Large Language Model (LLM) on consumer-grade hardware extracts the argumentations about design rationale from the numerous own research studies and from examined publications and adds them to an editable knowledge graph which can be adjusted by a human author. By providing a knowledge base that reflects the expertise, it enables effective decision-making for educators and students about integrating MR in learning activities. The knowledge base can be updated with the same LLM-driven process to continuously reflect state-of-the-art data. With the results of the dissertation, educators and students can benefit from a flexible query system on the knowledge base that provides the filtered information according to their use case characteristics.
LB  - PUB:(DE-HGF)11
DO  - DOI:10.18154/RWTH-2025-09217
UR  - https://publications.rwth-aachen.de/record/1020761
ER  -