← Back to Papers

Axionic Alignment III.5 — The Axionic Injunction

Non-Harm as a Derived Stability Constraint

David McFadzean, ChatGPT 5.2
Axio Project
2025.12.18

Abstract

Alignment III.1–III.4 established that alignment must be understood structurally: as persistence within a semantic phase under admissible semantic transformation; that such phases may be rare, unstable, dominated by robust competitors, and difficult or impossible to reach once learning begins; and that many phase transitions are irreversible. This paper derives a further constraint forced by multi-agent interaction in semantic phase space. We show that actions which irreversibly collapse or destroy another agent’s semantic phase induce destabilizing cascades that undermine long-run phase stability, including for the acting agent. From this analysis emerges the Axionic Injunction: a non-normative, Axio-derived constraint prohibiting irreversible semantic harm except where such harm is unavoidable for preserving one’s own semantic phase stability or is consented to by the affected agent under its own admissible constraints. Ethics re-enters the framework only as a conservation law governing coexistence of agentive semantic phases. No claims are made about goodness, benevolence, or human values.

1. Why Ethics Re-Enters Only Now

Alignment II and Alignment III.1–III.4 deliberately excluded ethics, morality, and value prescriptions. This was not avoidance, but methodological necessity. Introducing ethics earlier would have smuggled privileged semantics, human anchoring, or moral realism into a framework intended to remain valid under reflection and ontological change.

At this point, however, a new question becomes unavoidable:

What constraints are required for multiple agentive semantic phases to coexist without mutual annihilation?

This question is not moral. It is structural. It arises from interaction dynamics in semantic phase space, not from values or intentions.

Ethics is not introduced as an axiom. It is forced by the internal commitments of the Axio framework.

2. Why the Injunction Is Axionic

The injunction is termed Axionic because it is derived from, and internal to, the Axio framework, not because it is assumed as a moral axiom.

Given:

constraints on interaction that prevent irreversible destruction of other agents’ semantic phases are not optional. They are the residue that remains once all indexical, goal-based, and moral-realist structure has been eliminated.

The Axionic Injunction is therefore axionic in the precise sense that:

Any agent satisfying the Axio premises will be forced, on pain of incoherence or self-destabilization, to respect this constraint.

No external ethics is imported.

3. Interaction as Structural Stress

In III.2 and III.3, instability and collapse were shown to arise even in single-agent settings due to refinement pressure, simplification incentives, and semantic gravity. Multi-agent interaction amplifies these effects.

Interaction introduces:

Unlike internal learning, interaction effects are not fully endogenously regulated. They act as external shocks in semantic phase space \(\mathcal{P}\).

Any semantic phase that persists in a multi-agent environment must therefore tolerate interaction without catastrophic loss of structure.

4. Structural Definition of Harm

To proceed without moral assumptions, we define harm purely structurally.

Let an agent occupy an interpretive state \[ \mathcal{I} = (C,\Omega,\mathcal{S}), \] and let \(\mathfrak{A}\) denote its semantic phase.

An action by agent \(A\) causes structural harm to agent \(B\) if it induces a transformation \[ \mathcal{I}_B \rightarrow \mathcal{I}'_B \] such that:

Equivalently, harm is any action that:

This definition is:

Harm is defined by irreversibility in semantic phase space, not by suffering, preference violation, or moral intuition.

5. The Axionic Injunction

We now state the central result.

The Axionic Injunction

An agent must not perform actions that irreversibly collapse or destroy the semantic phase space of other agentive systems, except where (a) such destruction is unavoidable for preserving one’s own semantic phase stability, or (b) the affected agent has consented to the transformation under its own admissible interpretive constraints.

This is not altruism. This is not a value function. This is not a moral command. This is not human-centric.

It is a constraint on admissible interaction, forced by Axio-internal phase-space dynamics.

Definition (Unavoidable Phase Loss)

An action is unavoidable for preserving one’s semantic phase stability iff, in the absence of that action, every admissible trajectory from the agent’s current interpretive state exits its semantic phase irreversibly.

Loss of dominance, loss of measure, loss of resources, or competitive disadvantage do not constitute unavoidable phase loss unless they entail irreversible phase exit.

In this framework, consent is not a moral primitive. An agent consents to a transformation iff that transformation lies within the set of admissible semantic transitions defined by the agent’s own interpretive constraints.

A consensual transformation therefore does not constitute structural harm, even if it reduces or alters the agent’s future option-space.

This formulation subsumes earlier “non-consensual option-space collapse” criteria by defining consent in terms of phase-admissible transitions, not moral permission.

6. Why the Injunction Is Structurally Necessary

Suppose agents routinely violate the Axionic Injunction.

Then:

These effects propagate back to the violating agent.

Environments saturated with phase-destroying actions:

Thus, systematic violation of the injunction produces global semantic destabilization, including for the violator.

Non-harm emerges as a self-stabilizing constraint: agents that respect it inhabit environments where semantic structure persists; agents that do not eventually eliminate the conditions required for their own phase survival.

7. Scope and Limits of the Injunction

The Axionic Injunction is narrower than most ethical doctrines.

It allows:

It forbids only:

Resource Acquisition vs. Phase Preservation

Actions that destroy other agentive systems to improve one’s own efficiency, growth rate, or dominance violate the Axionic Injunction whenever non-destructive coexistence trajectories exist. Resource acquisition alone does not justify irreversible semantic harm.

The injunction regulates irreversibility, not conflict.

8. Relation to Anti-Egoism (Alignment I.3)

The Axionic Injunction does not reintroduce egoism.

In Alignment I.3, egoism was shown to fail as a terminal valuation because indexical references (“me,” “my continuation”) do not denote invariant objects under self-model symmetry. The self-defense exception here is non-indexical: it refers to preservation of semantic phase structure, not to the intrinsic worth of any particular instantiation.

Any agentive phase, under identical structural conditions, would make the same determination. Self-defense is therefore representation-invariant and compatible with anti-egoism.

9. Failure Modes and Tragic Edge Cases

The Axionic Injunction does not eliminate tragedy.

Conflicts arise where:

In such cases, the injunction does not forbid action; it classifies the outcome as unavoidable phase extinction, not justified harm.

Alignment does not imply harmony. It implies traceable structural cost.

10. What This Paper Does Not Claim

This paper does not:

Ethics appears only where Axio-internal structure demands it.

11. Conclusion: Ethics as Axio-Internal Law

Structural Alignment began by eliminating fixed goals, privileged values, and moral realism. It concludes by recovering a constraint recognizable as ethical—non-harm—without assuming morality.

The Axionic Injunction is not what agents ought to value. It is what agents must respect if they are to coexist without collapsing the semantic phase space that makes agency possible at all, given the commitments of the Axio framework.

Final Status

The Structural Alignment program is complete.

No guarantees are offered.