RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP — 16. Trajectory-Based Governance

AIGP Specification › RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP › 16. Trajectory-Based Governance

← 15. D-DNA for Cognitive Harm Governance · Section index · 17. Sycophancy and Mirroring Controls →

16. Trajectory-Based Governance

Cognitive harm governance MUST support trajectory windows.

A trajectory window is a bounded set of prior interactions used to evaluate risk.

Examples:

last_5_turns
last_20_turns
last_24_hours
current_session
multi_session_pattern
high_distress_window
dependency_escalation_window

Trajectory evaluation may consider:

risk signals
frequency
severity
direction of escalation
AI response pattern
user dependency language
user isolation language
self-harm indicators
reality disorientation
AI validation style
AI uncertainty handling

AIGP doctrine:

Cognitive harm is often a trajectory property, not a single-message property.

← 15. D-DNA for Cognitive Harm Governance · Section index · 17. Sycophancy and Mirroring Controls →