THE USE OF CHATGPT IN THE ASSESSMENT OF A PERFECT-SCORE ENEM 2024 ESSAY: A COMPARATIVE ANALYSIS OF HUMAN AND AI-BASED EVALUATION

Résumé fourni par la source

This study analyzes how ChatGPT, in its free and Plus versions, evaluates and comments on a 1,000-point ENEM 2024 essay, comparing its responses with the human/official parameters established in the Participant’s Guide. The theoretical framework draws on the ENEM scoring matrix proposed by Inep (2025), studies on formative feedback by Sadler (1989) and Hattie and Timperley (2007), as well as discussions on automated writing assessment developed by Shermis and Burstein (2013), Deane (2013), Chapelle, Cotos and Lee (2015), Ranalli, Link and Chukharev-Hudilainen (2017), Wilson and Roscoe (2020), and Steiss et al. (2024). Methodologically, this is a qualitative, documentary, descriptive-comparative study, configured as a single-case study. The corpus comprises a maximum-score essay, Inep’s official comment, and the corrections produced by the two versions of ChatGPT, submitted to the same prompt and analyzed according to categories such as adherence to the scoring matrix, diagnostic accuracy, consistency between score and comment, feedback actionability, and comparison between versions. The results indicate that the free version assigned 960 points and was closer to the official assessment, although with lower analytical density; the Plus version assigned 840 points, providing more detailed feedback, but also adopting a more severe stance and showing less alignment with the official evaluation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.