The Architecture of Agency › Volume 5 › The Coexistence Protocol

The Coexistence Protocol

The procedural layer

This chapter is a review — it is readable but still changing.

A stranger shoves you off a curb and you hit the pavement hard. What just happened? If he shoved you out of the path of a bus you never saw, it may have been justified rescue. If he shoved you to take your bag, it was wrongful force and possibly violence. If he stumbled against you in a crowd, it was accidental friction. Same contact, same bruise, different authorization, intent, risk, and responsibility. None is coercion unless a conditional threat was used to obtain compliance. And lying on the pavement, you may not have the information to tell which case occurred. Neither, possibly, does he: perhaps he reasonably but mistakenly believed there was a bus.

The ethics of viability proposes a constraint against using innocents as instruments of another’s ends. Even under perfect factual information, applying that constraint requires normative judgments about baselines, materiality, authorization, necessity, and responsibility. Real agents add asymmetric data, partial visibility, time pressure, noise, and conflicting probability models. Two agents can interpret the same event differently, each sincerely believing the other imposed a wrongful risk. Left unmanaged, those disagreements can escalate as each retaliation is redescribed as self-defense.

This is why the invariant alone is not enough. It needs a second layer:

The Procedural Layer is a fallible adjudication process for claims of boundary violation when perceptions, probabilities, or vantage points diverge.

Without it, even a flawless invariant dissolves into vendetta. This chapter specifies that layer.

Ambiguity Is the Normal Case

Some acts are easy to classify in the abstract. Real accusations of murder, theft, coercion, negligence, or justified defense still require evidence and interpretation before coercive sanction. Most conflict arrives with mistaken inferences, hidden information, disputed baselines, split-second decisions, and institutions that can err.

The prospective harm theory sharpens the problem rather than solving it. A material, attributable worsening of exposure can set back an agent before injury lands. But the relevant baseline is not automatically the instant before the act; it may be due care, a legitimate entitlement, an authorized practice, or the trajectory absent the intervention. The procedure must identify the baseline and the protected welfare or agency interest rather than treating every local probability increase as wrongful harm.

But precisely because the criterion is probabilistic, ambiguity blurs it. Was the risk truly imposed, or merely perceived? Was the delta real, or an artifact of the victim’s fear and the observer’s angle? When risk and perception diverge, judgment cannot rest on intuition. It requires procedure.

Procedural Agency

Procedural Agency need not be a single central authority, but calling it distributed does not solve authority. Any process capable of compelling evidence, issuing rulings, ordering restitution, or authorizing force exercises governance. Its legitimacy depends on jurisdiction, prior agreement where possible, public rules, independence, notice, opportunity to answer, appeal, proportional enforcement, and safeguards against capture. Subject to those conditions, its job is to determine separately:

And because the agents involved lack omniscience, the protocol must also distinguish three epistemic conditions:

The core function is simple to state:

Procedural Agency separates impact, authorization, responsibility, justification, and remedy before coercive enforcement begins.

Everything the protocol does serves that determination. It prevents premature expulsion of agents from the coexistence domain, protects innocents from being mislabeled as predators, and keeps the domain itself stable enough to be worth belonging to.

The determination runs in three stages. It is a reconstruction, not a checklist: each stage illuminates a different dimension of the event, and no stage can be skipped without the classification failing.

Stage One: Fact-Finding

The first stage reconstructs the physical event. What moved, what failed, what information was available to whom, and which risk deltas were introduced into the world. Procedure begins with physics, not psychology: before anyone asks what the shover believed, establish whether there was a bus.

Fact-finding also establishes the facts of authorization. A surgeon’s incision and a knife assault are physically similar events separated by scope, consent, purpose, and professional duty. A surgical injury can remain harm even when the procedure and risk were validly authorized; consent changes permissibility and responsibility rather than rewriting the physical impact.

Stage Two: Vantage Reconstruction

The second stage reconstructs the epistemic vantage of the acting agent: what he knew, what he could have known, and what it was reasonable for him to believe at the moment of action. This stage exists because even when harm occurred, intent, negligence, recklessness, rescue, and predation differ. The man who reasonably but mistakenly believed a bus was coming used force under error; he did not commit coercion because he issued no conditional threat. Whether he owes repair depends on due care and the framework’s responsibility rule, not on relabeling the shove.

Vantage reconstruction is what prevents well-intentioned actors from being mislabeled as predators simply because the world surprised them. A protocol without this stage punishes bad luck as if it were bad faith, and agents governed by such a protocol learn the obvious lesson: never intervene, never act under uncertainty, never attempt a rescue. That lesson kills more innocents than it protects.

Stage Three: The Coexistence Ruling

Only after fact and vantage are established does the normative ruling happen. Some interventions are justified rescues; some are ordinary friction; some are negligent or non-negligent errors; some are coercive threats; some are direct wrongful force; some are predation. The categories are not reducible to one risk delta. No classification automatically forces Domain Exit. That limited status requires an established serious violation or continuing threat, a legitimate process where feasible, and refusal of proportionate repair or protective terms.

The ruling need not condemn the actor’s character, but it can still impose serious burdens and therefore cannot be declared non-punitive by definition. Classification should identify harm, intent, authorization, and remedy while preserving review and proportionality.

The Firewall: Ambiguity Defaults to Innocence

Some cases resist resolution even after fact-finding and vantage reconstruction. The evidence is gone, the vantages are irreconcilable, the Δ-risk cannot be established. For these the protocol has one rule, and it is not negotiable: ambiguity defaults to innocence.

If a material attributable risk increase cannot be established to the applicable standard, coercive sanction is not justified. The burden lies with the party seeking to impose it. This presumption does not prohibit reversible precautions under genuine uncertainty; it requires that their scope, duration, and review track the evidence rather than convert suspicion into permanent status loss.

This firewall is load-bearing. Without it, accusation itself becomes a weapon: any agent who wants a rival expelled need only generate ambiguity around him, and the domain devours its own members on suspicion. A system that lets fear reclassify innocents as threats has already dissolved into the vendetta it was built to prevent; it just hasn’t noticed yet.

Restitution and Re-entry

Agents misread danger, misjudge timing, and act under compression. The protocol does not exile them for this. Repair depends on responsibility: negligent error, breach of duty, induced reliance, or unjustified harm can support restitution, while an unavoidable non-negligent accident may call for shared insurance or aid without a personal moral debt.

Restitution here is not penitence and not punishment. It is repair tied to an established responsibility: restoring, as far as possible, what was wrongfully lost. Payment does not automatically eliminate continuing danger, restore trust, or settle every claim; those questions require their own evidence and proportionate terms. Nor may an institution convert inability to pay into permanent status loss.

Refusal matters differently depending on the process. Rejecting a forum one never authorized, reasonably believes captured, or cannot safely access is not Domain Exit. Refusing a legitimate final ruling after notice, evidence, opportunity to answer, and appeal may justify enforcement of restitution or protective constraints. Standing is not forfeited wholesale; remedies remain tied to the established violation and continuing threat.

The Shield, Not the Sword

Domain Exit is asymmetric, and the boundary must be precise. Exit removes the protection of the coexistence protocol: the domain’s members no longer owe the exited agent the absorption of risk on his behalf, and defensive coercion against him is no longer a violation. That is the whole of it. Exit removes the shield. It does not grant a sword.

Domain Exit is not a warrant for aggression, punishment, or extermination, and it does not forfeit standing wholesale. It records that proportionate defensive constraints may continue while an established violation or threat remains unresolved. Members who treat the person as free prey commit fresh violations. The label must remain act-relative, reviewable, and terminable; otherwise a procedural safeguard becomes an outlaw caste.

Why Civilization Requires This

Without the procedural layer, an ethics of viability collapses under its own precision. The invariant is sharp; the world is blurry; press a sharp rule against a blurry world without a resolution procedure and you get one of two failure modes. If every perceived risk justifies retaliation, society dissolves into vendetta — each agent enforcing the invariant as he sees it, each enforcement a fresh violation as the other side sees it. If, to avoid that, genuine predators are allowed to hide behind ambiguity, cooperation erodes instead — the domain keeps its manners and loses its members.

Procedural Agency forecloses both. It forces evidence before retaliation, reconstructs vantage before judgment, classifies proportionally, defaults to innocence, and prices re-entry in restitution rather than blood. That is what scales the invariant from a personal ethic to a civilizational architecture: one rule, three stages, and a firewall.

The operating-system analogy earns only this much: an ethical rule that cannot handle ambiguity, error, conflict, and divergent perception cannot govern real agents. The invariant states the proposed boundary; the protocol supplies a fallible process for applying it among agents who can be wrong. How a polity institutionalizes this arbitration layer — what its courts, charters, and enforcement look like when built on the invariant and the protocol — is the subject of a later volume, sketched in the Axionic Constitution.