Human languaging and large language models

Abstract This paper explores the interplay between human languaging and AI-powered language models from the view of the distributed language perspective (DLP). According to this perspective, human linguistic activity can be described as ongoing, heterogenous activity, thus making proponents of DLP prefer the term ‘languaging’ to that of ‘language-use’. Following Maturana’s (Ir J Psychol 9(1):25–82, 1988; Cybern Hum Knowing 9(3–4):5–34, 2002) theorizings on languaging, the notion rests on a foundational distinction: human languaging occurs within the realm of autopoietic living systems, while AI-powered language models in being non-living, man-made machines operate within the bounds of allopoiesis, characterized as fundamentally open systems. On this view, languaging in humans is enabled by a bio-logic, and it plays out in structurally coupled organisms that co-exist in an ontogenetic drift; something which is impossible for LLM-based systems given their lack of biological constitution. Considering that languaging in this specific sense is inapt for thematizing human–LLM interaction, the paper turns to the broader notion of ‘enlanguaged cognition’ to theorize on this particular phenomenon in a DLP context. Specifically, the paper explores how LLM-based systems differ fundamentally from traditional tools, not only in terms of the dialogicality they afford but also in their practical openness and transsituational functionality. Rather than operating as predictable instruments, these systems actively shape cognitive processes, enabling new forms of coordinative dynamics that transcend specific situational contexts.