ChatGPT’s Performance and the Use of Argumentative Structures in Romanian Educational Contexts
Roxana Rogobete, West University of Timisoara (Romania)
Ana-Maria Pop, West University of Timisoara (Romania)
Abstract
This study examines the performance of ChatGPT in generating argumentative essays for various levels of Romanian education, assessing its adherence to required argumentative structures and evaluation criteria. The research involves three sets of real-life writing tasks: the 8th-grade national assessment, the 12th-grade baccalaureate, and the Titularizare and Definitivat exams for teachers (Tenure Exam for Teachers, Final Teaching Qualification Exam). These tasks are processed by ChatGPT, and the resulting essays are analyzed in collaboration with MA students specializing in curriculum development for Romanian language and literature. By applying corpus linguistics methods, this study seeks to analyze the frequency, variation, and function of lexical bundles in ChatGPT-generated essays for Romanian exams. The aim is to determine whether AI writing displays the same level of syntactic and semantic variation as human writing or whether it follows predictable, repetitive structures, reinforcing concerns about its formulaic nature [1]. Additionally, the research focuses on the coherence and cohesion of AI-generated arguments, and the degree of conformity to the structural and stylistic expectations outlined in national evaluation standards. The findings of this study will provide insights into both the pedagogical implications of AI-assisted writing [2,3] and potential challenges in assessing AI-generated texts within standardized educational frameworks. By comparing AI outputs with human-graded essays, the research will contribute to the ongoing debate on the role of artificial intelligence in language education, assessment, and the development of argumentative competence in Romanian academic contexts.
Keywords: AI-generated writing; AI-assisted writing; argumentative structures; writing evaluation; Romanian national evaluation standards
REFERENCES
[1] Jiang, F., & Hyland, K. (2024). Does ChatGPT argue like students? Bundles in argumentative essays. Applied Linguistics, amae052. https://doi.org/10.1093/applin/amae052
[2] Bašić, Ž., Banovac, A., Kružić, I., & Jerković, I. (2023). ChatGPT-3.5 as writing assistance in students’ essays. Humanities and social sciences communications, 10(1), 1-5. https://www.nature.com/articles/s41599-023-02269-7.pdf
[3] Su, Y., Lin, Y., & Lai, C. (2023). Collaborating with ChatGPT in argumentative writing classrooms. Assessing Writing, 57, 100752. https://doi.org/10.1016/j.asw.2023.100752