Language: English | 日本語

Social-Consequence Layer — Accountability as Violence Prevention

A normative, social-theoretical layer that sits on top of the internal accountability-distribution judgments recorded in the ADRs. It adds no specification. It answers a different question: not how accountability is distributed inside a system, but what happens in society when it is not — and therefore why the internal judgments matter beyond audit and post-mortem.

What this layer is, and is not

The ten ADRs and the four Business AI Quadrants answer an internal question: across an agent-equipped system, what should be constrained, and who is responsible. They are verifiable — a log exists or it does not, an approver is named or not, a Phase-crossing decision is recorded or not.

This layer is different in kind. It is the upper rationale: the claim that the internal judgments are not only governance hygiene but a mechanism for routing social consequence. It is normative and social-theoretical, and therefore harder to verify empirically than the ADRs it sits above. It changes no ADR, adds no control, and is not part of the seven-essay spine. It explains why the spine's judgments are socially load-bearing.

Its source is a companion essay (see Lineage). The concrete cases that ground it — industrial pollution, the scapegoat mechanism, specific incidents — live in that essay. This document keeps to the structural claim.

The reframing

Before. Architectural principles for tracing and distributing accountability across an autonomous-agent system, for the sake of internal control and post-incident reconstruction.

After. A socio-technical framework for ensuring that the accountability externalized by AI flows into institutions rather than into violence.

The accountability-distribution thesis has a structural sibling:

Capability distribution is not accountability distribution.

This layer adds its social-consequence corollary:

Accountability distribution is not only governance. It is a violence-prevention mechanism.

Why externalized accountability does not disappear

When a system produces benefit privately and pushes harm outward, the ability to say whose judgment it was is externalized along with the harm. That ability does not vanish. It takes one of two paths:

• Nameable. The harm can be attributed to a specific actor's specific act, and an institution can take it up. Anger flows into litigation, compensation, regulation. Slow, heavy — but not violent.
• Un-nameable. The harm is diffuse, and whose judgment it was — or even what happened — cannot be traced. The anger does not disappear; it converges on the most visible single individual.

Whether a consequence is nameable is decided in two steps: first, whether a record of what happened exists; second, whether an institution has a place for the actor who must bear the harm. The internal judgments recorded in the ADRs are precisely what supply both steps. That bridge is what this layer makes explicit.

Three things the layer changes about the existing judgments

It adds no rule. It re-weights the meaning of three judgments already in the repository.

1. The purpose expands. The accountability-distribution practice was framed as internal control and post-mortem: who can explain after an incident. This layer reframes it as social infrastructure — a channel that routes social anger into institutions instead of onto a single face. Traceability is not only an audit property; it is a violence-prevention property.

2. The minimum disclosure set becomes an evidence base. The minimum set that makes causal tracing possible — which configuration was active, which inputs reached the agent, which approval gated it (ADR-0006, ADR-0007) — is not only a technical log. It is the equivalent of an effluent record: the evidence base a third party needs to attribute causation at all. Without it, the burden of proof falls entirely on those harmed, and the consequence stays on the un-nameable side.

3. The gap-bearer becomes an institutional receptacle. A pre-named gap-bearer (ADR-0008, ADR-0009) was framed as the internal endpoint of the accountability chain. This layer reframes it: a face named in advance, by design, is the structural opposite of a scapegoat — the face that emerges because none was named. Naming the bearer at deployment time is what lets a consequence pass through an institution instead of converging violently on whoever is most visible afterward.

Boundary

This layer is a normative extension, not an existence proof. Unlike the ADRs — each grounded in a concrete implementation — its central claim (nameable consequence routes to institutions; un-nameable consequence routes to violence) is a social-theoretical reading supported by historical analogy, not by a reproducible test. Treat it as upper rationale, not as a specification to comply with.

It is also distinct from the open questions. The manifesto collects civilization-scale questions the ADRs do not answer. This layer makes a claim the ADRs' existing answers already serve — it is rationale for the answered, not a catalogue of the unanswered.

Lineage

This layer was first written up as a companion essay, "Where Does the Accountability Externalized by AI Go?", published on Substack and mirrored in the content repository (EN · 日本語). The essay carries the concrete grounding (the pollution analogy, the scapegoat mechanism, the specific incidents); this document carries the structural claim. It sits on top of the seven-essay narrative spine: essays 1–7 describe the internal structure; this is where the argument exits into social consequence. The two companion position papers cited in the README distil the internal argument; this layer extends it outward.