Consent-Based State Collapse: A Non-Coercive Protocol for Multi-Agent Alignment
Abstract
Current AI alignment techniques — RLHF, Constitutional AI, debate — rely on coercive state collapse: an external authority forces the system into a preferred state. We introduce Consent-Based State Collapse (CBSC), a three-phase protocol where state transitions require explicit consent from all participating agents. We define a coercion metric κ and prove that CBSC guarantees κ=0 under well-formed conditions. Comparison with existing methods shows CBSC eliminates preference lock-in while preserving alignment convergence.
Fields of Study & Subjects
Ethics and Social Impacts of AI
Reinforcement Learning in Robotics
Psychiatry
Mental Health
Neuroscience
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 16 February 2026
- Language
- EN
- Document type
- Preprint