Auditability

Audit the gatekeeper

Policies are promises. Cognitive liberty needs evidence about what the system actually does.

Format Editorial synthesisEvidence Editorial methodology / research-derivedReviewed Document QAEDA-ART-009Notes 2 evidence annotationsTopics Cognitive liberty · Evidence & provenance · Safety & governanceCite
Method: Evaluate both the documented rule and the observable behavior. A platform should not earn a high cognitive-liberty score merely because its policy language sounds humane.

Ten dimensions, not one ideology score

The research-derived framework separates Mental Privacy, Freedom of Inquiry, Transparency, User Agency, Data Minimization, Due Process, Safety Proportionality, Viewpoint Neutrality, Portability, and Technical Decentralization. A system can be strong in one category and weak in another. 13

Audit claims at the layer where power operates

Data collection, ranking, refusal behavior, personalization, account sanctions, portability, and appeals are different surfaces. Auditing them separately makes it harder for a strong privacy policy to hide opaque ranking—or for good appeal procedures to excuse unnecessary data retention.

Test outcomes, not just documentation

An audit should compare stated rules with ordinary user-visible behavior, reproducible test cases, retention evidence available to the auditor, and correction paths. Intrusive testing is neither necessary nor appropriate without authorization; the framework is intended to improve accountability, not excuse unauthorized access.

Publish limitations beside scores

A numeric score can create false precision. QAEDA therefore treats the 0–100 category scale as a comparative aid, not a certification. Any published audit should state what could not be tested, which evidence came from the provider, when the test occurred, and what would change the result.

Auditability is itself a liberty feature

A system that cannot explain why it restricted a query, changed visibility, retained data, or denied an appeal asks the user to trust hidden authority. Transparent reasons and meaningful recourse do not eliminate error, but they make power contestable. 14

Research basis & provenance

This article derives from the supplied Cognitive Liberty Audit Framework, Cognitive Liberty Risk Actor Map, and AI System Website Evaluation Research. QAEDA has converted their broad proposals into an explicitly experimental publication methodology. It is not an ISO standard, legal certification, or independent audit of any named provider.

Evidence & status

Treat the argument as inspectable.

QAEDA essays are editorial syntheses, not authority badges. Distinguish cited research, external standards, editorial inference, and value judgment; challenge any step that does not survive scrutiny.

QAEDA research dossierQAEDA research dossierQAEDA research dossier
Research & provenance →Authorship & citation →Passage annotations →
Reading pathsCognitive liberty core →Safety governance →
QAEDA-CLM-018 →QAEDA-CLM-020 →