Skip to main content

Scoring Constitution

Status: Supreme Domain Authority

1. Purpose

Scoring is a constitutional capability of the ZAYAZ platform.

This Constitution governs how scores are produced, governed, calibrated, interpreted, and evolved.

It applies to every scoring system implemented within ZAYAZ.

2. Constitutional Philosophy

Scores are governed evidence.

They are not opinions.

Scores summarize governed evidence into explainable decision support.

Every score shall remain:

  • explainable;
  • reproducible;
  • governed;
  • calibrated;
  • replayable;
  • auditable.

3. Scope

This Constitution governs:

  • trust scores;
  • confidence scores;
  • quality scores;
  • risk scores;
  • materiality scores;
  • knowledge confidence;
  • Bayesian confidence;
  • constitutional recommendations;
  • future scoring systems.

4. Constitutional Principles

CS-001 Evidence First​

Scores shall originate from governed evidence.

Never from undocumented assumptions.

CS-002 Governance First​

Weights, thresholds and modifiers are governance.

Runtime engines consume governance.

They never create it.

CS-003 Explainability​

Every score shall explain:

  • evidence;
  • computation;
  • governing policy;
  • applicable replay profile;
  • calibration bundle.

CS-004 Replayability​

Every score shall declare replay semantics.

Replay behaviour shall conform to the Replay Constitution.

CS-005 Calibration​

Calibration is a governed constitutional process.

Calibration shall never occur implicitly during runtime.

CS-006 Confidence​

Confidence expresses the constitutional certainty associated with a score.

Confidence shall not be confused with probability.

5. Constitutional Components

Every constitutional scoring system shall define:

  • evidence;
  • dimensions;
  • aggregation;
  • governance;
  • replay;
  • explanation.

Missing components constitute constitutional defects.

6. Weight Governance

Weights shall originate exclusively from governed policy bundles.

Implementations shall not define scoring weights in:

  • source code;
  • configuration files;
  • databases outside constitutional governance;
  • AI prompts.

7. Threshold Governance

Thresholds are governed constitutional artifacts.

Thresholds may be:

  • public;
  • tenant-private;
  • regulator-defined.

Visibility policies shall themselves be governed.

8. Anti-Gaming

Scoring systems shall actively detect adversarial optimisation.

Examples include:

  • threshold bunching;
  • modifier abuse;
  • behavioural drift;
  • constitutional inconsistencies;
  • statistical anomalies.

Detection itself constitutes constitutional evidence.

9. Constitutional Invariants

CS-100​

Every score possesses evidence.

CS-101​

Every score possesses governance.

CS-102​

Every score possesses replay semantics.

CS-103​

Every score possesses explainability.

CS-104​

Every score possesses calibration identity.

CS-105​

Governance precedes scoring.

CS-106​

Calibration precedes deployment.

CS-107​

Gaming detection is mandatory.

10. Conformance

Every scoring specification shall demonstrate constitutional conformance.

Platform-specific implementations—including TrustGate Score Governance (TG-SG)—shall document how constitutional scoring principles are realized.

11. Future Evolution

Future scoring domains may introduce additional scoring methodologies, dimensions, calibration techniques, and confidence models.

Future extensions shall preserve the constitutional principles established by this Constitution while strengthening explainability, governance, and replayability.



GitHub RepoRequest for Change (RFC)