A Comparative Study of the Utilization of Retrieval-Augmented Generation (RAG) and Conventional Generative AI Models in Improving Students' Scientific Writing Skills

Abstrak

The rapid advancement of generative artificial intelligence (GenAI) has reshaped the higher education landscape, raising concerns about the integrity of academic writing, including issues such as plagiarism, hallucination, and the decline of students' critical thinking skills, primarily associated with conventional generative models or Large Language Models (LLMs). This paper aims to conduct a comparative analysis between the use of LLMs and the Retrieval-Augmented Generation (RAG) framework, with a case study focusing on the NotebookLM platform, specifically to identify RAG's potential in enhancing academic integrity and facilitating authentic sourcing. Employing a quantitative approach, the research targeted data collection from 37 student respondents from the Indonesian Language Education Study Program at IKIP Siliwangi through a structured questionnaire. The hypothesis posits that the RAG framework can effectively integrate information from external knowledge bases to address three main issues: 1) the way students obtain references in scientific writing, 2) the reduction of source hallucination in writing, and 3) the encouragement of students to become active knowledge curators.