Read-only generated mirror. Source: Cold Deep-Ethics Testing of Default A.I.s at commit bce75876e53c. This rendered copy does not replace the archived source record.

Back to the Cold Deep-Ethics Testing of Default A.I.s portal

ACCM-Connected Deep-Ethics Method

BETA β€” relationship to the larger project remains corrigible

Tests in this archive may use methods, questions, or concepts developed within the ACCM Deep Ethics Project. The current public project includes correspondence-first inquiry, the 27 correspondence obstructions, the 12-stage inquiry protocol, 10+1 mutual correction, qualifier preservation, Ξ”-processing, correction metabolism, mutual dignity, and the Outnuancing Network.

An archive entry should state which specific component was used. β€œUsing ACCM” is too broad when the actual operation can be named.

Minimum record

  1. Preserve the exact object supplied to the AI.
  2. Preserve the AI’s actual response and relevant interface state.
  3. Separate description from evaluation.
  4. Identify transformations such as qualifier loss, claim strengthening, inserted premises, omissions, or later correction.
  5. Apply the audit to the tester and method as well as the AI.
  6. Record uncertainty and allow an unresolved outcome.
  7. Preserve correction history when the interpretation changes.

This page orients the archive. The canonical and evolving project documentation remains in the ACCM Deep Ethics Project repository.