Title: | An intelligent VR-based classroom simulator for student teachers |
Subject: | Computer science |
Level: | Basic, Advanced |
Description: |
Many disciplines require, apart the theoretical training, practical exercises to let learners experience scenarios they will face in real life. This need has been addressed extensively for phisically demanding activities (e.g. flying an aircraft or doing surgery) where there exist very advanced simulation environments reproducing reality with high fidelity. Activities requiring the interation with persons are intrinsically more difficult to simulate and often training activities are limited to workshops discussing potential simulations "on-paper". Recently, the advent of AI/ML and LLMs technologies disclosed new training opportunities: notably, artificial agents can be trained to play specific roles in conversations and other situations. In this respect, a trainee practicing teaching in elementary school could be virtually interact with a 8-years old student, maybe with poor language skills, timid, etc.
This thesis work aims to investigate and propose appropriate technical foundations to realize an intelligent VR-based classroom simulator for student teachers. In particular, a virtual classroom would be populated with learners, played by independent AI-agents, where student-teachers can practise various aspects of teaching their subject. An exemplary scenario could be the following: • Select the grade they wish to practise their teaching in (eg. 7th grade English);
• Select the topic they wish to practise with (eg. “summer holidays” with the focus on creating an advertisement in English for an eco-holiday spot);
• Select the lesson phase they wish to practice (eg. giving instructions to complex and multi-phase tasks, or giving feedback to learner presentations).
The fully completed system would consist of a virtual classroom populated by AI agents in avatar form. The agents will also reflect the heterogeneity typical of many classrooms such as mixed ability (gifted learners, weaker learners, average learners), learners displaying disruptive behaviour, learners with recent migrant background without knowledge of English or Swedish, and learners requiring general or specific learning support. For the scope of the thesis, the goal is to propose a training procedure targeting a specific agent behaviour. This would allow the future introduction of new behaviours/characters through new training procedures. |
Start date: | 2025-01-01 |
End date: | 2025-06-08 |
Prerequisites: |
- knowledge of AI/ML techniques is required; - knowledge of VR simulation environments is required; - knowledge of advanced programming in Python or Java is required; - knowledge of LLMs is considered a plus. |
IDT supervisors: | Antonio Cicchetti |
Examiner: | |
Comments: |
This thesis is suitable for two students |
Company contact: |
This thesis is developed in collaboration with UKK and Joanna Baumgart joanna.baumgart@mdu.se |