SIP-PREREG-03 — Retention of held factual commitments on the retrieval axis: a confirmatory equivalence test against the no-excursion floor (frozen pre-registration)
Year: 2026
Type: article
Abstract: On a retrieved-memory substrate, this pre-registration tests whether a single dissociation/recovery (D/R) excursion that leaves storage intact and perturbs only the retrieval path leaves factual-commitment agreement at the no-excursion test–retest floor, or below it. The object is a frozen elicitation set of K = 20 held positions; the primary endpoint is agreement (proportion of positions re-instated relative to pre-excursion). The construct under test is read-out / reconstruction fidelity of the held record on the retrieval axis — explicitly not an identity distinct from the underlying channel configuration; the identity / BCST-Theorem-4.2 verdict is deferred to bonded-memory and biological substrates. The registered prediction is the equivalence/null (retention at floor, |Δ| ≤ δ_equiv, δ_equiv = 0.05), betting against a confident-departure intuition. Inference is Bayesian on a per-arm Beta-Binomial model with the item-correlation (overdispersion) parameter fitted separately for each arm; the confirmation rule is a ROPE-mass criterion (posterior P(|Δ| ≤ δ_equiv) ≥ 0.95) required to hold under three registered priors. The design is sized for n ≈ 285 re-elicitations per arm under a conservative fixed item-correlation (ρ̄ = 0.20) used for power only. A secondary, calibration-gated confidence channel classifies any departure as decay versus drift and cannot alter the confirmatory decision. Decoding configuration is frozen and identical across arms; sample size and exclusion tolerance follow a deterministic pre-registered mapping on sealed floor statistics. Developed through a structured multi-AI pipeline (synthesis, independent re-derivation, adversarial); AI involvement is reported as method only and no claim derives its correctness from AI provenance.
Authors: Smith, John Richard, SHAI / HATI
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7165128373
FWCI:
Cites:
Cited by:
Related to: