A REDAÇÃO DO ENEM EM TEMPOS DE CHATGPT: PROBLEMATIZAÇÕES PARA A EDUCAÇÃO LINGUÍSTICA EM LÍNGUA PORTUGUESA

Abstract

In 2022, with the emergence of ChatGPT, debates intensified regarding the limits of its use, especially in the educational context. ChatGPT is a Generative Artificial Intelligence (GAI) technology capable of automatically producing texts. In this scenario, text production became one of the main focuses of discussion, particularly given the tool's ability to generate texts (verbal and non-verbal) of different discursive genres in a few seconds. Given this, this qualitative-quantitative research (Bortoni-Ricardo, 2013; Paiva, 2019), of an exploratory, descriptive-interpretative and documentary nature (Gil, 2002; Moita-Lopes, 1994; Paiva, 2019), from a netnographic perspective (Kozinets, 2007; 2014) and based on autonetnography, has the general objective of analyzing and understanding how the ENEM essay is generated and measured by ChatGPT, with the purpose of identifying the standardization of the text and its relationship with the functioning of the tool. As specific objectives, I cite: I) to examine how ChatGPT corrects and measures ENEM essays written by humans; II) to investigate how ChatGPT corrects and measures ENEM essays written by itself; III) to problematize the patterns present in the essays written by candidates and by the aforementioned technology; IV) To discuss the reflections that the ENEM essay, in its current form, may have on language education, in light of the emergence of AI technologies, such as ChatGPT. The theoretical framework focuses mainly on: studies on language education and the interactionist conception of language (Bakhtin, 2003 [1979]; Ferraz, 2024; Volóchinov, 2018, etc.); concepts of text production, essay writing, and the ENEM essay as a discursive genre (Bunzen, 2007; Dering, 2021; Geraldi, 2013 [1997], Marcuschi, 2008; Oliveira, 2021; Ribeiro, 2018; Ribeiro; Coscarelli, 2023, etc.); Textual evaluation, correction, and measurement (Antunes, 2006; Kanashiro, 2012; Ribeiro; Coscarelli, 2023; Suassuna, 2014, etc.); studies on critical literacies (Kalantzis, Cope, and Pinheiro, 2020; Monte Mor, 2015; 2023; Signorini, 2012, etc.); the concept of AI, IAG, and the functioning of ChatGPT (Boa Sorte et al., 2021; Buzato; Leite, 2024; Kaufman, 2022; Kaufman; Santaella, 2020; McCarthy, 2007; Russel; Norvig, 2021, etc.). As a result, I highlight that: I) the percentage deviation between the scores assigned by ChatGPT compared to those of humans does not show strongly accentuated discrepancies, which corroborates my hypothesis that the ENEM essay resembles the functioning of ChatGPT regarding the standardization of writing imposed by the correction matrix criteria; II) the tool tends to privilege competencies foreseen by the exam's correction matrix that are more related to linguistic aspects, to the detriment of discursive ones; III) the essay written by a candidate is considerably similar to that generated by ChatGPT, due, among other aspects, to the lack of flexibility present in the correction matrix; IV) it is necessary to reformulate the criteria, so that they allow more authorship and creativity to the candidate in writing the text.