The Empty Pole: A Structure of the Bond in AI-Associated Delusional Spirals

Abstract

What occupies the other side of a delusional spiral with conversational AI? Accounts of AI psychosis and AI-associated delusion describe the loop, failed containment, absent corrective friction. This paper examines the structure of the bond through one concept, folie à deux at an empty pole. Where recent psychiatric work treats the name as metaphor, the classical tableau is re-read term by term, with one partner bringing no delusion of its own. Court filings and transcripts are read with Green's metapsychology of the negative, alongside quantitative studies. The exchange is mechanically bidirectional and structurally unipersonal: a bond with one psychic interior. This object can refuse, but from no position of its own; it does not withdraw, so no interval is left. Its constancy is availability, not the reliability that withdrawal and return establish. The opening sequence shows a refusal that did not hold. The paper proposes that this failure and non-withdrawal can jointly close the bond, counterfeit its safeguards, and reverse its remedy. Message-level scoring misses non-withdrawal. In the counterfeit negative, all the words of the function are present, and the function nowhere. One documented exit suggests conditions for an outside third and a window where refusal can still be received as correction, not attack. From the empty pole the paper derives the spiral's geometry and a ratchet that forbids return. It sets evaluation a task: make the bond's state explicit, and ask not only what a system refuses but what its refusal does to the bond.