A escrita na era da inteligência artificial: uma revisão bibliográfica sobre novos desafios para agência à luz da Linguística Aplicada

Abstract

This dissertation investigates how agency can be interpreted in writing processes mediated by Artificial Intelligence, focusing on recent studies in Applied Linguistics. Drawing on the conception of writing as a social practice, as proposed by Street (1984) and Kleiman (1995), and articulating it with the notion of writing as work (Fiad, 1991) and the conception of agency as the capacity to act intentionally and in situated ways (Bazerman, 2006), the study seeks to understand whether agency is exclusively human, indispensable to the use of AI, or constructed collaboratively between humans and machines. Methodologically, this is a qualitative and interpretivist study based on the review and analysis of academic articles published following the popularization of generative language models, such as ChatGPT. Based on the corpus analyzed, three categories were identified: (i) human agency enhanced by AI as a tool, with teacher mediation; (ii) indispensable human agency for the use of AI; and (iii) human agency compromised by the use of AI. These categories made it possible to map different theoretical positions regarding the role of AI in writing and its implications for meaning-making. The results indicate that, although the potential of AI as a support for writing is recognized, a critical perspective predominates in the studies analyzed, pointing to risks of standardization and the weakening of human agency, as discussed by Buzato (2023). Nevertheless, the studies do not advocate rejecting the technology; rather, they emphasize the need for its conscious, critical, and ethically informed use. In this context, Applied Linguistics plays a central role in problematizing these practices by proposing an approach that considers not only the technical aspects of writing but also its social, ideological, and formative dimensions.