CONCEPT
Hallucination as Matrix-Crossing
The provocative reframing that AI hallucination and bisociation share structural features—both cross matrix boundaries; they differ in whether the crossing finds genuine structural identity or nothing at all.
AI hallucination—the machine's tendency to produce confident assertions that are factually wrong—is typically treated as a reliability failure, a bug to be engineered away through grounding and retrieval mechanisms. The bisociative framework reveals an uncomfortable structural kinship: hallucination and genuine bisociation share the same underlying operation. Both involve the machine crossing the boundary of the matrix specified by the prompt. The hallucination crosses and finds nothing—the connection is spurious, the fact is invented. The bisociation crosses and finds something—a structural identity the matrices had not previously revealed. The mechanism is the same; the difference is in what the crossing discovers.
In The You On AI Field Guide
The framing reveals a tradeoff that the engineering community has not fully confronted. Techniques that reduce hallucination also reduce the probability of genuine bisociation. Retrieval-augmented generation, grounding mechanisms, and tighter output constraints all work by keeping the machine more firmly within the matrix specified by the prompt. They increase accuracy by decreasing divergence. And decreased divergence