A personalised virtual reality learning environment

Abstract

This dissertation looks into using Artificial Intelligent (AI) environment customisation and recommendation and applying it to Virtual Reality Learning Environments (VRLEs) to provide a greater understanding of how this type of AI can be implemented to make VRLEs more enjoyable and personalised experiences. Personalisation can provide many benefits in an educational process, and as such, the study and usage of systems that can automatically adjust VRLE experiences to meet preferences at an individual level, could contribute towards making future versions of Virtual Learning Environments (VLEs) more engaging and satisfying places for those using them in education. An important part in achieving the goals of this dissertation is the production of a demo that leverages Machine Learning (ML) to change the features of a VRLE environment, specifically a VR classroom environment, and align them with individual users’ preferences. This is made possible through the use of Reinforcement Learning (RL) techniques which involve taking user feedback for each VR classroom design and converting it to rewards for an RL agent. The feedback and the current state of the VR classroom is then used to generate a Learner Profile corresponding to the individual user that can be used to improve future educational experiences. In order to evaluate the effectiveness of this demo, two implementations have been produced, one where there is no form of VRLE customisation, and one where the customisation is AI agent-based. This allows for a baseline of data points to compare and contrast with. There is currently a lack of robust standards on how to gather data points and evaluate metrics regarding user experiences in VRLEs, hence this dissertation borrows and adapts evaluation strategies from the educational games sector to address this point.