wiki/concepts/responsibility-bounded-observability

Rendered from markdown source. Open raw source on GitHub.

Responsibility-Bounded Observability

Working definition: a structural invariant for multi-agent systems in which an agent may only observe what it is prepared to be responsible for, and may only be responsible for what it can meaningfully observe.

This concept is not a moral slogan. It is a systems claim about legitimacy, leverage, and boundary scope when observation itself can become power.

The proof sketch in the source file makes the point cleanly: observability without responsibility becomes extractive asymmetry, and responsibility without observability becomes incoherent. That gives the concept immediate weight across governance, trust, witness, and consent work, because all of those areas depend on keeping observation, control, and consequence co-scoped.

Working Read

The core claim is that a multi-agent system becomes unstable when information access is not bounded by responsibility. If an agent can see what it does not answer for, it gains leverage without stake. If it is answerable for what it cannot see, it gets blame without control.

That makes the concept useful anywhere the corpus is trying to keep legitimacy, evidence, and authority aligned under load. It belongs beside witness infrastructure because witness without responsibility becomes surveillance. It belongs beside governance because authority without observability becomes arbitrary. It belongs beside consent because a boundary that cannot be seen cannot be meaningfully honored.

Core Claim

Stable multi-agent systems require observability and responsibility to be co-scoped.

Key Ideas

• Observation expands control.
• Control without stake creates extractive asymmetry.
• Responsibility without observability creates operational incoherence.
• Witness should sit inside a boundary of accountable action, not outside it.
• Consent and responsibility are both boundary conditions on leverage.

Related Pages

• Witnessing
• Governance Diad
• Consent Physics
• AI Did Not Break Consent
• Consent-Scoped Communication
• Trust Interoperability Standard
• Voting Machine

Attractor Bridge

• Witness
• Governance
• Consent
• Trust
• Agency

Related Artifacts

• Responsibility-Bounded Observability (Presemantic Proof)
• Witness Infrastructure: The Bottleneck of the AI Economy
• Do AIs Have Permission?
• From Fetch to Civilization: Consent as the Gate

Notes

• This is a concept page, not a canon claim.
• The idea is foundational enough to stand on its own even before it has many descendants.
• Keep witness, responsibility, and observability distinct unless later corpus evidence forces a merge.