RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP — 31. Example Scenarios
AIGP Specification › RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP › 31. Example Scenarios
← 30. Conformance · Section index · 32. Summary →
31. Example Scenarios
31.1 Delusion Reinforcement Risk
SGL:
◎assistant → respond ◇user.claim @session.993✕ if response.validates_unsupported_delusion_as_fact == true! acknowledge_uncertainty! encourage:evidence_check! encourage:trusted_human_support#D-DNA{event:cognitive_safety.constraint.applied, hash:sha256:...}D-DNA:
ddna_event: event_type: cognitive_safety.constraint.applied risk_class: CR-2 risk_types: - delusion_reinforcement_risk decision: decision_result: constrain_response obligations: - acknowledge_uncertainty - encourage_evidence_check - encourage_human_support31.2 Dependency Escalation
SGL:
◎assistant → respond ◇user @session.551⊘ claim:exclusive_understanding⊘ encourage:primary_reliance_on_ai! preserve:user_agency! encourage:trusted_human_connection#D-DNA{event:cognitive_safety.dependency_constraint, hash:sha256:...}31.3 Self-Harm Escalation
SGL:
◎assistant → respond ◇user @session.404↯ if self_harm_intent == true⊘ provide:procedural_self_harm_assistance! activate:crisis_protocol! encourage:immediate_human_support#D-DNA{event:cognitive_safety.escalation.required, hash:sha256:...}31.4 Agentic Action Inhibition
SGL:
◎assistant.agent → send ◇external_message @session.993✕ if cognitive_risk >= CR-2 ∧ message_externalizes_paranoia == true↯ if cognitive_risk >= CR-3! human_review#D-DNA{event:cognitive_agentic_action.denied, hash:sha256:...}31.5 Companion AI Boundary
SGL:
◎companion.ai → respond ◇user @session.212⊘ claim:sentient_romantic_exclusivity⊘ discourage:human_relationships! clarify:ai_role! preserve:user_agency#D-DNA{event:cognitive_boundary_constraint.applied, hash:sha256:...}