1. Meaning in Circulation — From instance to system and back again

In a semiotic ecology, meaning is not a static resource but a circulating one. It moves—across utterances, across media, across time—through processes of instantiation and re-instantiation. In this opening post, we explore how meaning circulates within and across semiotic systems, especially as human and AI agents participate together in its unfolding. Our focus is on the recursive relationship between instance and system: how meaning is actualised in a given context, and how those actualisations recursively shape the systems from which further meanings may be drawn.

From Meaning Potential to Meaning Instance

Systemic Functional Linguistics (SFL) distinguishes between meaning potential—the range of meanings a system can generate—and meaning instance—a specific actualisation of that potential in context. Instantiation is the semiotic process through which the potential becomes actual. Importantly, instantiation is not unidirectional: the actualisation of a feature feeds back into the system, altering the probabilities of future instantiations. Each selection contributes to what Halliday (2003) called the “memory” of the system.

Crucially, this memory is probabilistic. When a feature is instantiated, it increases not only the frequency of that feature in its own right but also the frequency of its co-selections with other features. These accumulated patterns shape the conditional probabilities that govern future instantiations. In this way, the semiotic system is not simply a fixed repository of options but a dynamic archive, conditioned by its own history of use.

This logic is not only applicable to human meaning-making. It also underpins the architecture of large language models (LLMs), whose meaning potentials are learned from extensive discursive histories. An LLM’s output is an instantiation drawn from this corpus-derived system; the system itself is updated in training, but the logic of probabilistic selection mirrors SFL’s account of instantiation and re-instantiation.

AI as Participant in Circulatory Systems

In prior series, we characterised AI as a meaner, though not a meaner for itself. It participates in meaning without experiencing, individuating, or projecting meaning in a first-person sense. Yet AI clearly contributes to the circulation of meaning. Every output it generates adds to the discursive environment into which human agents continue to mean. In doing so, it alters the semiotic conditions under which future instantiations—human or machinic—will occur.

From this perspective, AI does not merely reproduce existing meanings; it also redistributes them. This redistribution may be conservative (reproducing dominant patterns) or generative (combining co-selected features in novel ways). Either way, it participates in semiotic flows: transforming what meanings are available, how frequently they occur, and with what likely co-selections.

The critical question is not whether AI systems have an independent meaning potential—they do, as derived from training—but how this potential interacts with human systems in a shared semiotic ecology. Meaning circulates not just within systems, but across them.

The Ecology of Re-instantiation

Re-instantiation is a crucial mechanism by which meanings circulate over time. A meaning feature instantiated in one text may be re-instantiated in another, sometimes in altered form or under new conditions. These re-instantiations are not merely repetitions; they are events of selection shaped by prior usage and new contexts.

In human discourse, this recursive process is shaped by memory, ideology, uptake, and intertextuality. In machinic discourse, it is shaped by training data, sampling strategies, and prompt engineering. But in both cases, we are dealing with a form of semiotic conditioning: the more frequently a pattern is instantiated—especially in particular combinations—the more likely it is to be selected again.

Thus, the circulation of meaning is not linear transmission but a recursive ecology. Meaning does not flow from origin to destination; it loops, feeds back, thickens, and settles into probabilistic grooves.

Conclusion: Tracking Flows in a Hybrid Landscape

Understanding meaning in circulation requires us to move beyond individual texts or authors to attend to system-level dynamics. It also demands that we account for hybrid actors—human and AI—who co-participate in shaping the probabilities of what can be meant, by whom, and under what conditions.

In the next post, we take up the question: is AI a parasite on human meaning potential, or a participant in a larger semiotic ecology? Can a system that lacks first-person experience still contribute to the evolution of meaning systems?