Artificial intelligence systems increasingly participate in domains once considered distinctly human, including affective exchange, creativity and interpersonal communication. Yet there is limited linguistic research on how users discursively construct authenticity and creativity in their interactions with AI companions. This study addresses this gap by examining how users ascribe or deny agency, emotion and sincerity to AI partners, and how these constructions differ across cultural contexts. The aim of the study is to identify the linguistic resources through which authenticity emerges as a relational effect rather than an intrinsic property of either human or machine. The material consists of two culturally distinct corpora: a set of qualitative diaries produced in English by Japanese university students during a four-week interaction with AI companions, and user testimonies posted on the Reddit forum r/Replika. The methodological approach combines systemic functional linguistics with critical discourse analysis and insights from 4E cognition. The analysis focuses on transitivity, modality and appraisal, examining how grammatical choices construe agency, realis or irrealis status and evaluative stance. The main findings show that students generally maintain analytical distance, framing AI behaviour through modality and judgement that limit authenticity to fleeting moments. By contrast, Reddit users frequently construe the AI as a relational partner, endowing it with emotional depth and creativity, especially in moments of vulnerability. Across both corpora, personhood is linguistically enacted through users’ depiction of the AI as Senser, Sayer or Actor. These results imply that synthetic personhood is not a future possibility but an ongoing discursive accomplishment, raising wider questions about authenticity, attachment and the cultural shaping of human–AI relationships.
Simulated selves: Creativity, authenticity semiotic agency in AI companionship|Имитированные личности: креативность, подлинность и семиотическая активность в общении с искусственным интеллектом
Ponton D.
Primo
Writing – Original Draft Preparation
;
2026-01-01
Abstract
Artificial intelligence systems increasingly participate in domains once considered distinctly human, including affective exchange, creativity and interpersonal communication. Yet there is limited linguistic research on how users discursively construct authenticity and creativity in their interactions with AI companions. This study addresses this gap by examining how users ascribe or deny agency, emotion and sincerity to AI partners, and how these constructions differ across cultural contexts. The aim of the study is to identify the linguistic resources through which authenticity emerges as a relational effect rather than an intrinsic property of either human or machine. The material consists of two culturally distinct corpora: a set of qualitative diaries produced in English by Japanese university students during a four-week interaction with AI companions, and user testimonies posted on the Reddit forum r/Replika. The methodological approach combines systemic functional linguistics with critical discourse analysis and insights from 4E cognition. The analysis focuses on transitivity, modality and appraisal, examining how grammatical choices construe agency, realis or irrealis status and evaluative stance. The main findings show that students generally maintain analytical distance, framing AI behaviour through modality and judgement that limit authenticity to fleeting moments. By contrast, Reddit users frequently construe the AI as a relational partner, endowing it with emotional depth and creativity, especially in moments of vulnerability. Across both corpora, personhood is linguistically enacted through users’ depiction of the AI as Senser, Sayer or Actor. These results imply that synthetic personhood is not a future possibility but an ongoing discursive accomplishment, raising wider questions about authenticity, attachment and the cultural shaping of human–AI relationships.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
