DISCOURSE-BASED COMPARISON OF NATURAL AND SYNTHETIC TEXTS OF THE THESIS GENRE

Abstract

The irruption of generative artificial intelligence poses challenges and opportunities for established academic writing practices. In this context, this paper proposes a descriptive study of the differences, in discourse variables, between natural and synthetic texts in undergraduate and doctoral theses in disciplines such as aquaculture, law and linguistics. A corpus linguistics methodology was used to measure paragraph and sentence length, lexical richness, discourse markers, deixis and modality. The main findings indicate that natural texts exhibit characteristic length patterns, greater lexical richness, as well as more frequency and diversity in the use of discourse resources compared to synthetic texts. In particular, natural texts present greater concentration of subjectivity markers, while the synthetic ones show a frequent use of certain structuring markers, less variability of mechanisms, and a more uniform structure.