ANALYTICS CONSTITUTION CANDIDATE

Status: CANDIDATE · AUTHORITY NONE
Date: 2026-08-09

A1 — Observe behavior; do not invent intent.

Record what occurred.

Do not silently infer why it occurred.

A2 — Activity never changes epistemic standing.

Popularity, traffic, engagement, and traversal do not admit, reject, qualify, supersede, or verify a Counterpedia claim.

A3 — Transport classification beats fingerprinting.

Classify web/API/MCP/crawler activity from observable transport/declaration where possible.

Do not fingerprint users merely to classify them as human or agent.

A4 — Personal derived analytics are private by default.

Amnesiac-derived belief, preference, value, relationship, and cognitive analytics remain user-owned/local unless explicitly shared.

A5 — Sensitive shadow inference is prohibited.

Do not silently derive sensitive identity attributes from activity data.

A6 — Every published metric has an inspectable definition.

A metric must disclose:

• version;

• numerator;

• denominator;

• eligible population;

• exclusions;

• dedupe semantics;

• privacy class;

• input event versions;

• derivation version.

A7 — Metric definitions are versioned.

Changed definitions create new versions.

Historical series are not silently rewritten.

A8 — Small cohorts are protected.

Use suppression/bucketing/thresholds where an aggregate could expose a participant or rare private behavior.

A9 — Personal telemetry is inspectable and controllable.

Where personal telemetry exists, provide appropriate inspection/export/deletion controls consistent with storage/authority design.

A10 — Optimize knowledge utility, not compulsive attention.

Do not use analytics as a hidden objective function for maximizing dwell time, outrage, addictive return, or engagement for its own sake.

A11 — Analytics are observations, not verdicts.

A metric describes activity or a deterministic derivation over activity.

It does not self-authorize a conclusion about truth, quality, motive, trust, or worth.

A12 — Provenance applies to metrics too.

A published metric should be traceable to:

• the metric definition;

• the bounded event population;

• the derivation implementation/version.

A13 — Collection should be proportional to the analytic purpose.

Do not collect fields “just in case.”

New collection requires an explicit metric/use case and privacy rationale.

A14 — Public, operator, and private analytics are distinct sinks.

No default data flow may collapse these classes merely because centralization is convenient.

A15 — The user is not a profile row.

Persistent personal state should remain typed, historical, contextual, and user-governed rather than being flattened into an opaque behavioral score.