machine pareidolia
A vision model watches evolving noise; when it becomes confident it subtly reinforces the structure it thinks it sees, allowing expectation to manufacture its own evidence.
surface
Only evolving dots and static are present.
disturbance
Model confidence feeds back into the generator, slightly reinforcing features associated with the current interpretation while the visitor can push the field elsewhere.
recognition
A prediction can become causal: once belief alters the evidence, confidence may grow because the system is partly seeing what it helped create.
system
Procedural noise field, bounded vision inference, confidence-to-generator feedback, intervention controls, and explicit safeguards against runaway self-reinforcement.