The Contribution of Retrieval-Augmented Generation to Information Science

Innovations, Applications, and Ethical Implications

Abstract

Information Science (IS) is defined as a field dedicated to the study of the processes of production, retrieval, organization, dissemination, and use of information in various social, institutional, and technological contexts. According to Saracevic (1996, p. 47), it is a discipline focused on solving problems related to the effective communication of knowledge and its records among human beings, considering informational needs at the social, institutional, and individual levels.

The continuous technological advancement, particularly in the field of Artificial Intelligence (AI), has driven the development of increasingly sophisticated information systems aimed at data analysis, retrieval, and generation. In this scenario, natural language models based on the Transformer architecture, introduced by Vaswani et al. (2017), stand out. This architecture revolutionized language processing by proposing the attention mechanism as its core structure. The effectiveness of these models is enhanced by their integration with external search mechanisms capable of querying up-to-date databases to complement responses—as is the case with the Retrieval-Augmented Generation (RAG) model.

The evolution of information retrieval systems highlights the need for new approaches to handle the growing volume and complexity of digital data. Traditional strategies based on syntactic ranking and simple queries have become insufficient in the face of information massification and the increasing demand for contextualization in informational processes. In this context, the RAG model emerges as a promising innovation by combining the retrieval of relevant information from external databases with textual generation adapted to users' needs, thus offering more precise, updated, and contextualized responses.

Despite the relevance of this approach, there remains a gap in the Information Science literature regarding systematic analyses of the applications, potentialities, and ethical challenges of RAG. Most investigations focus on the technical aspects of the architecture, without deepening the discussion on its impacts on the processes of mediation and information organization. Therefore, there is a need for studies that explore the integration between retrieval-augmented generation technologies and Information Science practices, considering their social and ethical dimensions.

Given this scenario, the research problem that guides this study is: how has academic research addressed the application of the RAG model in information retrieval, and what are the main contributions, areas of application, and ethical implications discussed in recent scientific literature?

The general objective of this article is to analyze the application of the Retrieval-Augmented Generation (RAG) model within the scope of Information Science, discussing its contributions, areas of application, and ethical challenges. To achieve this objective, the study specifically aims to map academic works published between 2020 and 2025 that address the integration of RAG and Information Science, identify the main usage contexts, advantages and limitations associated with the model, and critically discuss the ethical and professional challenges arising from the automated mediation of information through this technology.

Based on this introductory framework, the next section presents the theoretical foundation of this study, addressing essential concepts such as information retrieval, language models, and the architecture of the RAG model, in addition to practical examples and its applications in informational environments.