Explainable scoring
Trust methodology
Scoring formula
Base change = 10.
Confidence multiplier = confidence / 50.
Correct = +base_change * multiplier.
Incorrect = -base_change * multiplier.
Clamp score between 0 and 100.
Trust rules
Meridian does not score opinions. It scores resolved records. Every record has a timestamp, confidence level, outcome, evidence, and public audit trail.
- Every user starts at Meridian Score 50.
- Correct resolved records increase score.
- Incorrect resolved records decrease score.
- Higher confidence increases both upside and downside.
- Every reputation change creates a reputation_history row.
- Scores are clamped between 0 and 100.
- No paid score boosting or hidden score changes.
- Claims without objective resolution criteria can be recorded but should not affect score until they are resolvable.
- Records that cannot be fairly resolved should be marked inconclusive instead of forced into a false binary.
- Every resolution must cite evidence and remain challengeable.
Identity
A reputation record belongs to an authenticated user with a public profile.
Record
A claim is only scoreable when it has a timestamp, event, confidence, and unambiguous resolution criteria.
Resolution
Outcomes are written separately from predictions so the audit trail can explain what changed and why.
Reputation
Score movement is a consequence of resolved outcomes, never payment, status, or manual promotion.
Record standard
A proof record needs claim text, timestamp, confidence, deadline, and resolution criteria before it is trusted.
Resolution parameter standard
Scoreable records should define the source, condition, deadline, and evidence standard that will decide the outcome before the record locks.
Evidence standard
Primary sources beat secondary sources. A source URL and resolver identity should be attached to every resolution.
Evidence score standard
Records receive an evidence score from 0 to 100 based on source quality, snapshot state, hash availability, identity level, criteria precision, and challenge state.
Challenge standard
If a resolution is disputed, the dispute must be public, evidence-linked, and resolved without altering the original record.
Calibration standard
Meridian should evolve toward proper scoring and calibration views so overconfident wrong claims are visibly punished.
Domain standard
A record is scored inside its domain. World Cup judgment does not imply finance, policy, health, hiring, or AI safety credibility.
Inconclusive standard
If the selected source fails, the claim was vague, or the evidence cannot fairly resolve the outcome, the record can be marked inconclusive without creating reputation authority.
Consent standard
Meridian records claims and actions a user or organization chooses to put on record. It does not audit a person's whole life.
Methodology changelog
Meridian cannot be trusted if scoring rules change quietly. Material methodology updates must be dated, versioned, and readable by the public.
| Date | Version | Change | Impact |
|---|---|---|---|
| 2026-06-03 | v0.1 | Record formula published | Initial public score model: base change 10, confidence multiplier, full downside weight, and 0-100 clamp. |
| 2026-06-03 | v0.1 | Trust Profile standard published | Identity type, verification level, risk status, evidence quality, and challenge window become visible on records. |
| 2026-06-03 | v0.1 | Paid trust prohibition published | Payments may unlock workflow, API volume, verification review, or audit export. Payments cannot move reputation. |
| 2026-06-04 | v0.2 | Inconclusive state added | Unresolvable records are not forced into correct or incorrect outcomes. |