| Keywords: | generative language models, artificial intelligence, AI-generated writing, quasi-subjectivity, authorship, pragmatics
|
| Abstracts: | The article considers artificial intelligence as a new object of linguistic description and as a quasi-subject of communication. The widespread use of generative models is changing the paradigm of text, style, and authorship analysis. The spread of generative language models is transforming written communication and requires a clearer understanding of authorship, style, agency, and discursive responsibility. The material consists of two comparable corpora: 50 human-authored essays and 50 essays produced by language models. To ensure comparability, genre, length, topic, and a unified prompt template were controlled. The study combines quantitative indicators with qualitative analysis of composition, argumentation, and pragmatic strategies. The study identifies stable features of model writing at the lexical, syntactic, discourse, and pragmatic levels. The most noticeable are clichés, hyper-connectedness, compositional orderliness, reduced content density, hyper-correctness, and cooperativeness. The comparison shows that the differences affect not only individual verbal elements but also broader ways of organizing the text, coherence, inference, and reader address. The transparency/hiddenness scale is refined and its role in explaining the quasi-subject effect is demonstrated. Generative models produce stable discourse-pragmatic patterns that create an impression of agency and support the effect of quasi-subjectivity. Further research should test these features across genres, models, versions, and post-editing conditions.
|
| The contact details of authors: | Lanovaya, Tatyana Vladimirovna – Candidate of Philology, Associate Professor of the Department of Russian Language, Literature and Journalism, Mari State University, Yoshkar-Ola, Russia, https://orcid.org/0000-0002-6862-5977, lanovaya_t_v@mail.ru
|
