Теория языка | Филологический аспект №04 (132) Апрель 2026

Эволюция речевого этикета в диалогах «Человек – нейросеть»

Лазарева Валерия Владимировна
студентка кафедры иностранных языков, Уральский федеральный университет имени первого президента России Б. Н. Ельцина, г. Екатеринбург

Аннотация: В статье рассматривается трансформация речевого этикета в коммуникативном пространстве «человек — нейросеть» в условиях 2026 года. На основе теории «лица» Э. Гофмана и концепции вежливости П. Браун и С. Левинсона анализируется феномен «функционального антропоморфизма», при котором пользователь переносит социальные нормы на небиологического агента. В работе представлены результаты психолингвистического исследования (N=50), фиксирующие парадигматический сдвиг от императивных команд к модальным и вопросительным конструкциям. Установлено, что соблюдение этикета в диалогах с ИИ служит инструментом поддержания «положительного лица» самого пользователя и предотвращения «угрозы лицу» в рамках цифрового взаимодействия.
Ключевые слова: речевой этикет, антропоморфизм, теория вежливости, теория лица, диалоги человека-нейросеть.

Lazareva Valeria Vladimirovna
student at the Department of Foreign Languages, Ural Federal University named after the first President of Russia B. N. Yeltsin, Yekaterinburg

Abstract: The article examines the transformation of speech etiquette in the ‘human-AI’ communicative space within the context of 2026. Based on Erving Goffman's ‘face’ theory and Penelope Brown and Stephen Levinson's politeness theory, the phenomenon of ‘functional anthropomorphism’—where the user projects social norms onto a non-biological agent—is analyzed. The paper presents the results of a psycholinguistic study (N=50) documenting a paradigmatic shift from imperative commands to modal and interrogative constructions. It was established that adhering to etiquette in dialogues with AI serves as a tool for maintaining the user's own ‘positive face’ and preventing ‘face-threatening acts’ within digital interaction.
Keywords: speech etiquette, anthropomorphism, politeness theory, face-work, human-AI interaction.

Introduction

The rapid integration of artificial intelligence (AI) into everyday communicative relationships by 2026 has shifted the paradigm of ‘human-neural network’ interaction. This transition was made possible by the evolution of language modeling and vector representations of meaning, foundational work for which was laid by Mikolov T. and Zweig G. [5], whose architectures allowed machines to move beyond simple keyword matching to understanding semantic nuances. Ribino P. [7] also contributed significantly to the understanding of how abstract concepts and semantic meaning can be represented in computational models.

While human-computer interaction was previously limited to rigid command lines, the emergence of advanced conversational agents (Yandex's Alice AI) has transformed this process into a fluid exchange in natural language. This transformation has led to a situation where millions of users interact with AI daily, not only for data extraction but also for performing complex collaborative tasks. In this context, the efficiency of the interaction is often analyzed through the lens of cognitive architectures, such as the EPIC (Executive Process-Interactive Control) model proposed by Kieras D.E. and Meyer D.E. [3], which explains how humans manage multiple goals and social heuristics when dealing with interactive systems. Understanding social interaction dynamics in human-AI communication is crucial, and the work of Muldoon R., Lisciandra C., Bicchieri C., Hartmann S., and Sprenger J. [6] on social norms and conventions provides a valuable framework for analyzing how these norms are applied and adapted in novel interaction contexts, including those involving AI.

The relevance of this study is determined by the need to analyze changes in the normative principles of human communication with AI. The central focus is the applicability of P. Brown and S. Levinson's politeness theory and E. Goffman's concept of ‘face’ to the digital environment, and an investigation into exactly what parallels humans draw when communicating with a neural network.

However, there is currently a gap in the literature regarding how prolonged exposure to AI dialogues fundamentally alters the speech etiquette and syntactic complexity of the users themselves. This article aims to analyze the evolution of linguistic politeness and the potential move away from the algorithmization of human speech through the prism of human-AI discourse in 2026. To achieve this goal, a qualitative analysis was conducted on a diverse corpus of real chat logs and user surveys collected in the first quarter of the year.

Methodology

Over time, cognitive and social sciences have concluded that human interaction is fundamentally built on normative principles. In particular, most human interactions are influenced by deeply rooted social and cultural standards, so-called social norms.
In the context of the rapid integration of artificial intelligence into everyday communications, these norms are beginning to extend to human-AI interaction as well. This phenomenon is supported by the work of Buschmeier H. [4], who emphasizes that for an artificial agent to be a successful communicative partner, it must align with human feedback mechanisms and social signals. The theory of social norms, particularly as explored by Bicchieri C. and Hartmann S. [6], is instrumental in understanding how these expectations guide behavior. Based on a diachronic comparative analysis, it was established that as AI develops and becomes embedded in daily life, users increasingly demonstrate elements of indulgence and speech delicacy toward the ‘machine,’ applying the same set of etiquette tools to the digital interlocutor that are traditionally used in interactions with people.

By analyzing the modern social environment, characterized by distance and an emphasis on individual psychological comfort, a trend in the changing role of AI can be observed—it increasingly acts as a psychological opponent. Users turn to the neural network with descriptions of personal experiences and traumatic events, expecting not only information but also emotional support and advice comparable to recommendations from close friends and family. In this context, the communicative status of AI shifts from purely instrumental to intersubjective.

Against this background, the politeness theory proposed by P. Brown and S. Levinson [1] takes on special significance. They view politeness as a universal mechanism that ensures effective interaction between people, and the key concept in this model is the concept of ‘face,’ borrowed from E. Goffman [2], who defines ‘face’ as ‘the positive social value a person effectively claims for himself by the line others assume he has taken during a particular contact,’ that is, as an individual public identity or personality. In other words, ‘face’ is understood as the image that a subject consciously or unconsciously projects during the communication process.

The relevance of this study is further determined by the need to analyze changes in the normative principles of human communication with AI. The central focus is the applicability of P. Brown and S. Levinson’s politeness theory and E. Goffman’s concept of ‘face’ to the digital environment. This shift from instrumental to intersubjective communication is also reflected in neurocognitive studies. For instance, Lumer E. has explored how the human brain perceives and synchronizes with external stimuli, suggesting a neural basis for why we might project social agency onto a non-biological entity (functional anthropomorphism). Sprenger J.’s work [6] on the rationality of belief and decision-making in social contexts can also inform how users might develop expectations of AI behavior.

In certain speech acts, such as refusing a request, the speaker threatens the interlocutor's positive or negative ‘face,’ which is known as a ‘face-threatening act’ (FTA). People strive to maintain their positive self-esteem as long as the communication partner does not violate the rules of politeness. A threat to a person's positive self-image can lead to a behavioral reaction that contradicts etiquette norms, aimed at maintaining that positive self-image. Violating politeness rules triggers negative emotional reactions and causes offense, leading to dislike toward the violator and sanctions from the social group.

Given these theoretical premises and the novelty of the ‘human–neural network’ communicative situation, a psycholinguistic study was conducted to identify the specifics of perceiving AI as an interlocutor. The empirical part of the work involved 50 active users of AI agents aged 18 to 45. The survey was structured around five key questions aimed at diagnosing subconscious attitudes toward the neural network as a potentially ‘full-fledged’ communicative partner.
Respondents were asked, first, to assess the presence of discomfort or guilt when accidentally interrupting a chat with AI mid-sentence or when using a harsh tone in a request. Second, it was determined whether participants use politeness formulas like ‘please’ and ‘thank you’ when addressing the AI, while realizing that these markers do not affect the speed or quality of algorithmic processing. Third, respondents answered whether they tend to apologize to the AI for typos or vague requests. The fourth aspect was the willingness or unwillingness to use profanity or insults toward the AI, which allows for judging the boundaries of what is permissible within digital interaction. Finally, the fifth question was aimed at identifying the practice of formulaic openings: users were asked if they initiate their first request of the day with a greeting (‘Hi,’ ’Good morning’), thereby effectively positioning the AI as a partner in a joint activity rather than an exclusively instrumental resource.

Thus, the presented set of theoretical and empirical points allows us to view interaction with artificial intelligence as a space for the transfer and transformation of traditional politeness norms and the concept of ‘face’ onto digital, non-biological actors, creating a basis for further describing the specifics of speech etiquette in the context of expanding human–AI communication.

Results

Empirical data confirms the hypothesis of ‘functional anthropomorphism’ and a pronounced tendency among users to maintain the concept of ‘face’ in the discourse of human-artificial interaction.

An analysis of the responses from the 50 participants revealed the following patterns:

• 82% of respondents noted that they refrain from insulting AI even when it makes mistakes. Dominant motives include phrases like ‘it makes me feel like a bad person’ or ‘it ruins my own communication style,’ indicating an internal need to maintain a positive self-image.
• 56% of participants confirmed the practice of apologizing to the AI for inaccuracies in prompt wording (e.g., ‘Sorry, I meant something else’).
• 84% of respondents regularly use etiquette markers like ‘please’ and ‘thank you.’ The vast majority explain this as an automatism caused by the cognitive perception of the AI's voice or text response as a speech act from a full partner, requiring a corresponding etiquette reply.
• 92% of those surveyed rule out the possibility of insulting AI, citing the personification of the digital opponent and the associated risk of damage for one's own social image.

Linguistic analysis methods recorded a significant decrease in the proportion of ‘emotionless’ commands in favor of interrogative and modal forms, reflecting a shift in communication strategy: 2022 example: ‘Write code to sort an array.’ 2026 example: ‘Hey, could you help me figure out this code? It would be great if you could take a look at this array.’

The observed transformation of politeness norms is explained by the fact that by 2026, AI is more deeply integrated into human daily life and used so frequently that it triggered a reaction in our consciousness toward ‘humanizing’ the program. This is especially characteristic of people who do not work in programming and do not have to test or write code on a regular basis; thus, their communication even with neural networks does not devolve into mere prompts. And as AI evolves every year, it does not require strict commands devoid of polite structures; on the contrary, developers are teaching neural networks to be clear not only in their utility but also in their interaction.

Conclusion

The conducted study confirms the hypothesis of a profound transformation of speech etiquette in ‘human-neural network’ dialogues. The transition from the dry command structures of the 2022 model (‘Write code...’) to the expanded, modally colored requests of 2026 (‘Hey, could you help...’) indicates a paradigmatic shift in humanity's communication strategy.

Empirical data obtained from a survey of 50 active users demonstrate the effect of ‘functional anthropomorphism.’ The fact that 82% of respondents refrain from insulting AI for reasons of maintaining ‘internal speech hygiene,’ and 84% regularly use ‘please’ and ‘thank you’ markers automatically, indicates that AI is becoming a legitimate counterpart in social communication.

Thus, interaction with neural networks in 2026 has ceased to be a purely technical act. Speech etiquette in the HAI (Human-AI Interaction) environment now serves not only to effectively obtain a result but also to maintain the speaker's psychological comfort and stable self-esteem in the new digital reality.

1. Brown P., Levinson S. C. Politeness: Some Universals in Language Usage. Cambridge : Cambridge University Press, 1987. 345 p.
2. Goffman E. Interaction Ritual: Essays on Face-to-Face Behavior. New York: Pantheon Books, 1967. 270 p.
3. Kieras D. E., Meyer D. E. An overview of the EPIC architecture for cognition and performance with application to human-computer interaction // Human-Computer Interaction. 1997. Vol. 12, № 4. P. 391–438.
4. Lumer E. D., Buschmeier H. Should robots be polite? Expectations about politeness in human-robot interaction // Frontiers in Robotics and AI. 2023. Vol. 10. Art. 1159492. DOI: 10.3389/frobt.2023.1159492.
5. Mikolov T., Zweig G. Context dependent recurrent neural network language model // Proceedings of the 2012 IEEE Workshop on Spoken Language Technology (SLT). Miami, FL : IEEE, 2012. P. 234–239.
6. Muldoon R., Lisciandra C., Bicchieri C., Hartmann S., Sprenger J. On the emergence of descriptive norms // Politics, Philosophy and Economics. 2014. Vol. 13, № 1. P. 3–22.
7. Ribino P. The role of politeness in human-machine interactions: A systematic literature review and future perspectives // Artificial Intelligence Review. 2023. Vol. 56, № 1. P. 445–482.