Artificial Communication and Semiosis in Large Language Models (LLMs) From the Perspectives of Esposito’s Communication Theory and Peircean Semiotics

Byongchang, Kang (2026) “Artificial Communication and Semiosis in Large Language Models (LLMs) From the Perspectives of Esposito’s Communication Theory and Peircean Semiotics“

---

This paper redefines large language models (LLMs) not as “artificial intelligence” but as systems of “artificial communication,” and analyzes their operative mechanisms and inherent limitations through the concept of semiosis drawn from the semiotics of Charles Sanders Peirce. Since the advent of ChatGPT in 2022, generative AI has blurred the boundaries of language, thought, and creativity, prompting urgent humanistic reflection. Yet the prevailing AI discourse remains trapped in a category error that, through the anthropomorphizing label “artificial intelligence,” positions humans and machines as competitors on the same plane. To move beyond this replacement-versus-defense dichotomy, this study draws on Elena Esposito’s notion of “artificial communication” and Niklas Luhmann’s communication theory to clarify the medial nature of LLMs. In Luhmann’s framework, communication is not the transmission of a sender’s intention but a social operation completed through the observer’s interpretation. Although LLMs do not understand the meaning of the texts they generate, they participate in communication by producing outputs that human users can interpret. In doing so, LLMs parasitically exploit the intentionality embedded in vast human data to generate a “virtual double contingency,” functioning as semiotic agents beyond mere pattern repetition. Within the Peircean semiotic framework, LLMs are positioned as “quasi-minds” capable of performing semiosis without consciousness.