Prompting and RAG vs. student engagement and comprehension in educational technology

Rana, Soumya Prakash, Dey, Maitreyee, Patel, Preeti, Requena, Jesus and Fu, Colin (2025) Prompting and RAG vs. student engagement and comprehension in educational technology. In: IEEE EDUCON 2025 - 16th Global Engineering Education Conference, 22nd April - 25th April 2025, Queen Mary University of London. (Unpublished)

Abstract

Generative AI (GenAI) has emerged as a valuable tool in education technology, offering potential to enhance learning and teaching processes. While concerns like fraudulent practices, algorithmic bias, privacy issues, and overreliance on technology persist, GenAI’s benefits are significant when used strategically. It is essential, however, to view GenAI as a supplementary aid for students and educators rather than a replacement for human-led teaching. Effective use of GenAI requires thoughtful implementation, including techniques like prompting and retrieval-augmented generation (RAG). Prompting involves formulating questions or tasks for the AI, while RAG enhances the AI’s ability to retrieve relevant information based on its training. This study focuses on the relationship between GenAI and students, excluding educators’ roles. A mixed-method survey evaluated students’ interactions with GenAI-generated answers in two scenarios: one where they had prior topic knowledge and another where they did not. Five chatbots—ChatGPT, Gemini, Copilot, Perplexity AI, and Sana AI—were tested with varied prompts. Results showed that students benefit most when they are engaged and have foundational topic knowledge. These findings underscore the role of educators in fostering student engagement and guiding effective GenAI use. By prioritizing understanding, educators ensure GenAI enhances learning, reinforcing that AI should support, not replace, education.

Documents
10105:51606
[thumbnail of EDUCON 2025  Prompting RAG.pdf]
Preview
EDUCON 2025 Prompting RAG.pdf - Accepted Version
Available under License Creative Commons Attribution 4.0.

Download (308kB) | Preview
Details
Record
View Item View Item