Open Collections

Wang, Kevin Shukang

University of British Columbia

The rapid evolution of Natural Language Processing (NLP) has positioned Large Language Models (LLMs) like ChatGPT as transformative tools across various sectors, including education. These models offer significant potential for enhancing learning experiences through personalized assistance, yet their propensity to generate incorrect, biased, or unhelpful responses poses critical challenges for their deployment in educational contexts. The usability of AI as tutors or assistants demands better human-AI design, particularly through personalization and customization to meet diverse educational needs. The first contribution of this thesis focuses on optimizing the AI agent for education, primarily through the design of a Retrieval Augmented Generation (RAG) pipeline as the agent’s core tool. By refining its components, this research aims to ensure that AI-driven educational assistants provide more accurate and contextually relevant responses. In addition to these technical enhancements, this research introduces a system that centralizes help-seeking channels with AI to facilitate the seamless integration of the optimized RAG framework into existing educational platforms. This resulted in the integrated HelpMe system, ensuring that the enhanced RAG system’s benefits are accessible to a broader range of users without requiring extensive technical expertise. Through both quantitative and qualitative data gathered during the deployment both the queuing system and integrated system, we find significant benefits in the integrated HelpMe system as an assistive tool for educators and students. We also identify future directions and challenges in human-AI interactions within educational contexts.

Text

2025-05-01

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/90937

Computer Science

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/