Human Decision Authority (Value)
- Entity ID:
ent-20260423-b002c1000001 - Type:
decision - Scope:
private - Status:
active
Description
First of the paper's five motivating values. Humans retain ultimate decision authority, organized through a principal hierarchy (Anthropic > operators > users) that formalizes who holds authority over what. The response to the finding that users approve 93% of permission prompts was not more warnings but to restructure the problem via defined boundaries (sandboxing, auto-mode classifiers). Motivates deny-first defaults, graduated trust, append-only state, externalized policy, and values-over-rules.
Key claims
- Anthropic's response to 93% prompt-approval fatigue was restructured boundaries, not more warnings
Relations
- Human Decision Authority (Value) --[motivates]--> Deny-First with Human Escalation (Principle)
- Human Decision Authority (Value) --[motivates]--> Graduated Trust Spectrum (Principle)
- Human Decision Authority (Value) --[motivates]--> Append-Only Durable State (Principle)
- Human Decision Authority (Value) --[motivates]--> Externalized Programmable Policy (Principle)
- Human Decision Authority (Value) --[motivates]--> Values Over Rules (Principle)
- Principal Hierarchy (Anthropic > Operators > Users) --[operationalizes]--> Human Decision Authority (Value)