The article is dedicated to the study of cognitive-communicative transformations occurring in the digital public sphere under the influence of large language models (LLMs). The subject of the study is the simulation of rational dialogue as a mechanism for changing the forms of meaning creation and interpretation in algorithmically mediated communication. The work examines large language models both as a technological tool and as a factor in the transformation of discursive practices, within which the boundaries between understanding, imitation, and argumentation become less distinguishable. Particular attention is paid to the formation of a hybrid form of discourse that arises at the intersection of human thinking and machine-generated text. The aim of the work is to identify and interpret key directions of language transformation and meaning creation in the communication environment influenced by large language models. The methodological foundation of the research is a hypothetico-deductive approach that combines a meta-analysis of scientific literature with an experimental analysis of interaction protocols with LLM systems. The experimental part includes a comparative analysis of the communicative behavior of models based on criteria for assessing the structure and content of dialogues, as well as the perceptual effects of interaction. Such an approach allows for the comparison of theoretical conclusions with observed communicative practices. Modern large language models function as tools for simulating understanding, reproducing the structure of meaningful statements without possessing either intention or cognitive experience. At the same time, texts produced by the models are perceived as meaningful and coherent, which fosters a certain level of trust in them. The human interacts not with a source of knowledge but with an interface that organizes dialogue within predictable linguistic frames, creating the illusion of joint reasoning. The boundaries between understanding and its imitation become increasingly indistinguishable, and language transforms into a space of coordination between human experience and the algorithmic process. The scientific novelty of the work lies in the conceptualization of the hybrid form of discourse as a systemic result of human-LLM interaction and in the identification of the ways through which the simulation of reasoning affects communication. It is shown that the perceptual effects created by the models redistribute epistemic responsibility among the participants in the interaction and reinforce new norms of interpretation. The results obtained allow for considering LLMs as an infrastructural element of the digital communication environment and indicate directions for further cognitive-sociological research.
Polyakov et al. (Sun,) studied this question.