Modelos de Linguagem e Educação a Distância: Revisão Sistemática da Literatura Sobre o Uso Ético e Pedagógico no Ensino Superior

Abstract

The adoption of large-scale language models (LLMs) in distance education (DE) has grown rapidly in higher education; however, their pedagogical and ethical impacts are still poorly explored in an integrated way in the literature. This scoping review aimed to map the available evidence on the use of LLMs in DE in higher education, identifying ethical aspects, benefits, limitations, and the pedagogical role of these tools. Eligible articles were empirical and theoretical articles that addressed the use of LLMs in the context of DE in higher education, with analysis of ethical, pedagogical, or learning impact dimensions, published in English or Portuguese. The search was conducted in the Scopus and Web of Science databases, following the PRISMA-ScR protocol, in which 229 articles were identified in Scopus and 136 in Web of Science, totaling 365; after removing 92 duplicates and applying the eligibility criteria, 55 articles were included in the final analysis. The results show that the use of LLMs can strengthen learning in distance education by offering personalization, immediate feedback, and support for study organization, but also reveals important limitations, such as inaccuracies, biases, excessive dependence, and risks to academic integrity. Studies converge in indicating that positive effects occur mainly when the use of LLMs is mediated by structured pedagogical practices aligned with institutional guidelines. It is concluded that LLMs have significant potential to improve distance education in higher education, provided they are integrated into ethical and structured pedagogical models, with clear institutional policies and digital inclusion strategies.

Keywords: Ethics. Artificial intelligence. Distance education. Learning process