Moderation Framework

Two-stage complaint moderation — Stage A operations and Stage B trust eligibility before any reputation impact.

Trust must never react to intake noise. CVT separates operational complaint handling (Stage A) from trust eligibility (Stage B). Raw accusations, rumours, and uninvestigated reports do not affect the Trust Profile.

Why two stages?

Informal economies run on word of mouth. A single anonymous complaint can destroy a livelihood if it instantly hits a public score. CVT’s moderation framework ensures:

Complaints are investigated before trust impact
Anonymous complaints are excluded by default
Single upheld complaints have light v1 impact — repeat patterns matter more
Rehabilitation is built in — sealed records after clean recovery

Stage A — Operational workflow

Stage A is existing CVT complaint ops — unchanged in spirit:

State	Meaning
`new`	Intake recorded
`under_review`	Investigation in progress
Ops notes, evidence, plate linkage	Internal only

Stage A does not affect trust. Investigators work the case without changing Trust Profile, trust level, or passport standing.

Stage B — Trust eligibility

A complaint affects trust only after trust moderation completes:

New
  ↓
Investigating
  ↓
Verified          → may affect trust (standard upheld)
  ↓
Severe            → may affect trust (stronger factor)

Trust moderation state	Affects Trust Profile?
New	No
Investigating	No
Verified	Yes — negative factor (light in v1)
Severe	Yes — stronger negative factor
Dismissed / duplicate / insufficient info	No

When Stage B completes as Verified or Severe, CVT emits conduct trust_events:

COMPLAINT_TRUST_VERIFIED
COMPLAINT_TRUST_SEVERE

These events carry conduct_category for Hard Search (e.g. dangerous_driving) but are never on the public timeline.

What never affects trust

Raw accusations not yet through Stage B
Rumours or unverified third-party reports
Anonymous complaints (unless identity verified to CVT standard — default: exclude)
Single complaints without repeat pattern (v1 caps meaningful penalty)
Complaints linked only to unverified plates with no verified subject
Dismissed, duplicate, or insufficient-info outcomes

Visibility after Stage B

Surface	Conduct visible?
Public passport	Standing indicator only — Good / Review Required / Not in Good Standing
Soft Search	Aggregates — count, severity band, months since last event
Enhanced Search	Counts + recovery period — still no categories
Hard Search	Category + severity + month + status — insurers with contract only
Owner view (Tier 0)	Full conduct history including categories and outcomes

Narratives, witness statements, complainant identity, and moderator notes never leave CVT — at any tier.

v1 penalty philosophy

v1 is overwhelmingly positive (~85–90% of trust level from positive signals). Negatives are light and late:

Signal	v1 impact
`COMPLAINT_UPHELD_TRUST` (Stage B Verified)	Light; single event capped
`COMPLAINT_SEVERE_TRUST` (Stage B Severe)	Moderate
`COMPLAINT_REPEAT_UPHELD` (2+ in 24m)	Moderate
`SUSPENSION_ACTIVE`	Strong — blocks Gold+
`FRAUD_CONFIRMED`	Strong

A long clean record with one old upheld complaint should not destroy Platinum tenure — policy recomputes under versioned rules without rewriting history.

Audit requirements

Every trust-affecting factor must reference:

complaint_id or evidence document ID
Moderator staff ID
Timestamp
Stage B outcome

Conduct events are append-only in trust_events. Corrections use EVENT_REVERSED compensating entries — never silent edits.

Rehabilitation and sealed records

After policy thresholds (direction: 5 years + clean recovery + moderate events only):

Event marked is_sealed = true
RECORD_SEALED emitted
Excluded from Soft, Enhanced, and Hard Search
Registrant sees indicator: “sealed after rehabilitation”
CVT Trust Team and lawful disclosure only

Severe events may follow stricter seal rules — policy TBD.

Permanent punishment destroys trust systems. Complaints should not permanently destroy reputations.

Moderator workflow checklist

Stage A (ops)

Complaint intake with verifiable subject link where possible
Investigation notes documented
Do not change trust moderation until investigation complete

Stage B (trust team)

Confirm identity standard met (not anonymous unless verified)
Select outcome: Verified, Severe, or Dismissed
Assign conduct_category if Verified/Severe (for Hard Search tier)
Emit trust event — triggers snapshot recompute
Registrant may dispute → Trust Team review → possible EVENT_REVERSED

Connect and public layers

Passengers and riders on Connect see trust level and Good Standing badge — not conduct detail. Partner tiers are not exposed in consumer apps.

Required positioning

CVT maintains an accountability record — not a rumour ledger
Higher trust levels mean verified positive signals over time
Not “safe driver”, “trustworthy person”, or legal finding of guilt