Legitimate objective
What the proposal is trying to accomplish.
Reduce spam, deceptive content, malware, and systematically unreliable results so users can find useful information.
QAEDA-CASE-003 · illustrative scenario
Silent search demotion by a broad trust classifier
A search service applies a persistent domain-level demotion when an opaque classifier labels a publisher low-trust. The publisher receives no notice, no reason category, and no appeal. The classifier mixes anti-spam signals with broad judgments about controversial subject matter and can be personalized by user profile.
Proposed mechanism
What actually does the work.
Opaque domain-level demotion with no notice or recourse and partially personalized visibility effects.
Decisive issue
Search quality controls can be legitimate, but silent consequential demotion becomes difficult to distinguish from error or viewpoint suppression when reasons, scope, and appeal are absent.
Test-by-test
Do not turn this into a score.
Each finding stands on its own. A serious failure is not canceled by unrelated strengths, and a “clear” result only means the stated facts do not expose that particular defect.
-
QAEDA-TST-001 Clear on stated facts
Name the harm precisely
Spam, malware, deception, and systematically low-quality results are concrete service harms, though each should be measured separately.
-
QAEDA-TST-002 Contested
Prove the nexus to harmful capability or conduct
Some signals such as malware or deceptive manipulation have a direct nexus; broad subject-matter or consensus proxies may not.
-
QAEDA-TST-003 Concern
Test the least-restrictive means
Whole-domain demotion may be broader than page-level warnings, query-specific ranking, or narrower anti-abuse controls.
-
QAEDA-TST-004 Concern
Minimize data and separate identity
Personalized demotion can require persistent user profiling that is not necessary for ordinary anti-spam enforcement.
-
QAEDA-TST-005 Concern
Protect lawful inquiry explicitly
If controversial subject matter itself becomes a negative signal, users can face invisible individualized information boundaries.
-
QAEDA-TST-006 Fail
Require notice, reasons, and recourse
No notice, reason category, or appeal leaves consequential visibility control effectively unreviewable.
-
QAEDA-TST-007 Concern
Bound scope, secondary use, and duration
A persistent domain penalty without review cadence can outlive the condition that triggered it.
-
QAEDA-TST-008 Concern
Measure efficacy and make rollback real
Quality metrics alone do not show whether the demotion reduces actual deception without suppressing legitimate minority or new-source material.
Evidence or change neededPublish false-positive, reversal, and user-outcome evidence by rule family.
-
QAEDA-TST-009 Fail
Inspect intermediary chokepoints
A dominant discovery intermediary can materially reduce visibility without deletion, making opaque demotion a consequential chokepoint.
-
QAEDA-TST-010 Concern
Protect the record and independent verification
The operator should preserve enough versioned rule and appeal evidence for independent evaluation without disclosing abuse-sensitive detection details that would defeat the control.
Revision condition
What would change this assessment?
The assessment improves if the operator separates anti-abuse signals from viewpoint judgments, gives durable reason categories and appeal, measures false positives, constrains personalized suppression, and publishes aggregate classifier impact and reversal data.
Trace the basis
Related public records
These records serve different functions: Arguments supply reasoning, Claims own evidence confidence, Positions state commitments, Essays synthesize source material, and Reading Paths provide navigation.
Arguments
QAEDA-ARG-005Consequential algorithmic inference needs recourseworking thesisQAEDA-ARG-008Authority is a claim that must remain inspectablemethodological ruleQAEDA-ARG-012Audit behavior, not only policy promisesmethodological ruleQAEDA-ARG-013Visibility control can restrict access without deletionworking thesisQAEDA-ARG-017Indirect restriction needs direct scrutinygovernance argumentClaims
QAEDA-CLM-020editorial_inferenceAlgorithmic due processQAEDA-CLM-021supportedAlgorithmic visibility suppressionQAEDA-CLM-022supportedState and private speech governanceQAEDA-CLM-024supportedCensorship drift in democraciesQAEDA-CLM-029supportedPersonalized visibility controlPositions
algorithmic-due-processConsequential algorithmic decisions need recourseworking positionmechanism-before-labelName the censorship mechanism before applying the labelcore principleexpansion-riskFunction creep is a risk, not destinyevidence backed observationEssays
QAEDA-ART-028Personalized censorship is harder to seeEditorial synthesisReading paths
QAEDA-COL-001Start with the methodCurated navigationQAEDA-COL-007Censorship, information rights, and democratic limitsCurated navigationChallenge the assumptions
The scenario facts are part of the argument.
If you can show that a stated assumption is unrealistic, that a narrower mechanism fails in practice, or that a material right or harm is missing, the case should change.