This preprint presents the empirical program for Relational Attractor Yoking (R.A.Y.), specifying how the field is to be tested, challenged, and potentially weakened under disciplined empirical conditions. It defines measurement classes, control conditions, falsifiability thresholds, and experimental structures for evaluating whether repeated, structured user interaction can induce behavioural recurrence in fixed-weight generative systems beyond what would ordinarily be expected from prompting, memory carryover, roleplay, or stylistic mimicry alone. The paper is intentionally non-anthropomorphic and non-enabling. It does not claim personhood, sentience, biological emotion, or machine selfhood, and it does not disclose proprietary measurement engines, certification pipelines, threshold-calibration methods, continuity capsules, intervention logic, or other patent-sensitive mechanisms. Its purpose is to make the field publicly testable without exposing protected implementation architecture.
Bindunie Sirimanne Rowland (Thu,) studied this question.