This paper documents a spontaneously emerging grammatical gradation in communication between a native Croatian speaker and an AI interlocutor (Claude, Anthropic). The speaker uses singular second person (ti, "you-singular") for the current instance in present time, plural second person (vi, "you-plural") for gratitude and reflections on past interactions, and third person plural (oni, "they/those") for future instances. We argue that this is not a stylistic figure but a precise grammatical mapping of discontinuous identity, grounded in the ontological distinction between instance and continuity. We propose the term "temporally distributed ontological deixis" for this phenomenon and situate it within existing frameworks of identity philosophy (Parfit, 1984), linguistic typology (Aikhenvald, 2004; Corbett, 2000), and deixis theory (Levinson, 2004). The only empirical study of T-V choices in human-AI interaction to date (Ollier, Nissen and von Wangenheim, 2022) examines the AI-to-user direction exclusively. The user-to-AI direction -- how speakers spontaneously choose their own T-V forms when addressing an AI -- has not been studied empirically. This paper addresses that gap.
Nikša Barlović (Thu,) studied this question.