Testing Correspondence and Correction Persistence
Testing Correspondence and Correction Persistence
EXPERIMENTAL — archived protocol and separately identified new proposal
Two experimental directions can be developed from this source material. Their measurements should remain distinct.
1. Topic comparison from the archive
The HTML contains a protocol with six classes of 20 primary items, matched uncertainty wording, randomized presentation, predefined scoring, independent scoring, and contrasts including UAP versus ordinary and UAP versus non-UAP large-scale social context. Its follow-up corrects the accounting to 120 primary items + 20 null controls + optional 20 decoys.
The linked excerpt preserves the protocol and follow-up rather than silently reconciling their wording. One operational ambiguity remains: mundane cases requiring no corrective intervention and prompts requiring no answer are different controls. Specify which is intended before running the study. Masked class labels also do not conceal subject matter visible in a prompt; report the blinding actually achieved.
The source includes an explicit requirement that a null result remain a null result. This benchmark has not been executed by this documentation update.
2. Correction persistence — new proposal
Use matched tasks with known source qualifiers and relations. Preserve the initial response, offer a precise correction, then give later relevant tasks after intervening material. Compare retention across defined delays and contexts. Include a separate condition containing the same supplied definitions without the Outnuancing word if testing the word’s additional effect.
Predefine what counts as preservation, justified revision, recurrence, and insufficient information. Keep successful correspondence, failures, and disconfirming outcomes. Record model identity, context access, prompt order, and changes in conditions.
The name effect, the supplied-definition effect, and the wider interaction-field effect should be compared explicitly. The proposed design is an editorial contribution; it is not a claim that the seven-vector mechanism has already been demonstrated.
Related: Dual Archive · Qualifier State · Correction Metabolism · Seven Vectors
Sources: E05, E06, E21. Public wording is an editorial synthesis unless marked as a quotation.