LLMs and the Collapse of Representational Interiority: Preface — On Method and Constraint

This series is not an argument about artificial intelligence.

Large language models function here as a point of exposure, not an object of analysis in their own right. What is under examination is not whether machines think, understand, or become conscious, but the set of assumptions that make those questions appear self-evident in the first place.

In particular, the series targets a deeply embedded but rarely articulated commitment:

that meaning, understanding, and intelligence are fundamentally internal properties of bounded entities, and that language is the external expression of those internal states.

This commitment will be referred to, in various forms, as representational interiority.

It is not treated as a formal theory. It is treated as an implicit civilisational habit: a background metaphysics embedded in ordinary language, institutional practice, and reflective self-understanding.

The aim of this series is not to replace this framework with another system of representation.

It is to trace the points at which it becomes unstable.

Accordingly, several methodological constraints are in place:

First, “meaning” will not be treated as a substance contained inside words, minds, or systems. It will be treated as something that becomes determinate through relational construal under structured conditions.

Second, “understanding” will not be treated as the possession of internal semantic objects. It will be treated as a mode of participation within systems of symbolic coordination.

Third, “intelligence” will not be treated as a private property of individuals or machines. It will be treated as a distributed capacity for adaptive participation within relational fields of constraint.

These are not definitions offered for acceptance. They are constraints on how the analysis proceeds.

The purpose is not to persuade the reader to adopt a new doctrine, but to prevent familiar doctrines from silently reasserting themselves under new terminology.

This matters because the intuitions targeted by this series are not merely intellectual. They are structural.

They shape how it seems natural to talk about:

• minds and machines
• thought and expression
• knowledge and representation
• agency and attribution
• selfhood and interiority

In ordinary usage, these distinctions feel stable. They feel like descriptions of obvious features of reality.

One of the central claims of this series is that this stability is itself a historical achievement of a particular metaphysical orientation, not a neutral reflection of how things must be.

Large language models are introduced here not as a turning point in the history of technology, but as a disturbance within this orientation.

They produce coherent symbolic behaviour without cleanly fitting the assumptions that normally anchor interpretations of meaning and understanding. This creates pressure not on the machines, but on the interpretive framework itself.

The series therefore proceeds in a deliberately sequential manner:

Each part isolates one component of representational interiority, allows it to reach a point of conceptual strain, and then follows the implications of that strain without prematurely restoring coherence.

The intention is not demolition for its own sake. It is clarification through controlled destabilisation.

What emerges, if the sequence holds, is not a denial of mind, meaning, or intelligence, but a shift in where those phenomena are located — or more precisely, what it means to say they are “located” at all.

By the end of the series, nothing essential will have been removed.

But something familiar may no longer be available in the same form.

And that absence is not a conclusion.

It is the beginning of the question the series is really asking.