Governance Model

Governance Model

Four-Phase Violation Processing

When an Agent violates governance rules, MAREF applies a four-phase escalation:

Phase 1: WARNING → Phase 2: SANCTION → Phase 3: ISOLATE → Phase 4: RECOVER

Phase	Action	Duration	Auto/Manual
1. Warning	Log + notification	Immediate	Auto
2. Sanction	Capability restriction	Configurable	Auto
3. Isolation	Full lockdown	Until reviewed	Auto
4. Recovery	Gradual restore	Step-by-step	HITL

---

Trust Engine v2

5-Factor Weighted Scoring

Factor	Weight	Description
Historical reliability	25%	Past task success rate
Peer evaluation	20%	Reputation from other agents
Security compliance	20%	Safety rule adherence
Response quality	20%	Output accuracy and timeliness
Resource efficiency	15%	Token/compute usage optimization

Goodhart Detection

MAREF detects and prevents Goodhart's Law manipulation:

• "When a measure becomes a target, it ceases to be a good measure"
• Automatic detection of score gaming patterns
• Uncertainty quantification and confidence calibration

---

CircuitBreaker

Success → Success → Success → ... → Failure → Failure → Failure → LOCKED (30s cooldown)

• Threshold: 3 consecutive failures
• Cooldown: 30 seconds
• HALT absorption: Prevents cascading failures
• Auto-recovery: Gradual trust restoration after cooldown

---

Rule Freeze Zone

Constitutional red-line modifications are blocked by the RuleFreezeZone:

• Module-level write protection
• Context isolation for governance primitives
• Human approval required for any constitution changes

---

← Back to Home