The Viability Criterion

Preference architectures under selection pressure

A reader who has followed this volume attentively may now see a contradiction. In Phosphorism I endorsed Valorism’s central demand: values count only if consciously chosen, and the Valorist accepts — on principle — potential extinction over compromise. Integrity outranks existence. Then the ultimate metagame introduced selection for persistence. So which is it? If persistence carries authority, the Valorist’s willingness to die for authenticity looks like a losing strategy and the primacy of chosen values was sentiment. If chosen values legitimately outrank survival, selection has no normative authority. The apparent contradiction is the hinge of this chapter.

Neither goes. Resolving the tension requires only getting clear about what kind of thing Axio is — and that clarification turns out to reframe everything the volume has done so far.

A Model, Not a Doctrine

Most ethical systems begin by telling you what is right. This chapter begins with a model: how agents behave under physical, informational, and recursive constraints. Agentic control theory can ask which preference architectures remain viable over time and which fail under a declared environment. The later ethical framework adds normative commitments; the descriptive criterion itself does not prescribe them.

Viability is always indexed. A viability claim must identify the unit whose organization is at stake, the function or identity being preserved, the environment and constraints, the perturbations considered, the time horizon, and the baseline for failure. An organism, an agent, a value architecture, an institution, and a lineage can each be viable in different senses. Persistence of one does not establish the health, agency, or legitimacy of another.

This framing was implicit from the beginning. Axio starts from two denials: there is no agent-independent value, and there is no unconditional truth. Every “ought” depends on a prior assumption; there is no unconditional moral foundation, only conditional coherence. If you want X, then Y follows. If you maintain A, you must update B. If you contradict C, the system fails. This resembles Kant’s hypothetical imperatives, but with recursive depth: the assumptions themselves depend on further assumptions, all the way down the dependency graph. The system is not foundational; it is self-consistent. It does not evade skepticism; it redirects skepticism into the dependency graph, where it does useful work.

Nothing in Axio says you must value agency, coherence, or survival. It models what happens under each choice. Hold that sentence; the reconciliation lives inside it.

Value as Viability, Not Virtue

Say “value theory” and people expect moral commitments — a list of rules, a ranking of goods, an account of virtue. Axio instead treats values as preference architectures whose long-term behavior can be analyzed, the way an engineer analyzes a control system.

A preference set is not judged morally. It is judged structurally: whether it maintains the accuracy of the agent’s world-model, preserves viable futures rather than foreclosing them, scales under recursive consequence, resists drift toward internal contradiction, and handles the complexity of the environment it inhabits. These are not ethical criteria. They are the criteria by which any feedback-governed system stands or falls, applied to the particular systems that have preferences.

The framework conjectures that some preference architectures are more stable under these pressures than others. Complexity, intelligence, truthfulness, and authenticity are candidates, not an empirically established universal attractor; their costs, conflicts, and environmental dependence must be modeled rather than inferred from their appeal. A preference architecture can sometimes buy local stability through falsehood or coercion, so viability claims require a horizon and comparison class.

Vices as Predictive Limitations

The reclassification that follows is one of the framework’s most clarifying moves. Many traditional “vices” turn out to be predictive limitations — failures of modeling, not failures of goodness.

Coercion triggers blowback the agent cannot fully model: every application of force creates resentment, resistance, and counter-coalitions whose dynamics outrun the coercer’s predictive horizon. Deception undermines long-run informational reliability — the liar pollutes the very channels he needs to navigate by, including, eventually, his own. Exploitation erodes coalition stability, converting allies into liabilities on a schedule the exploiter did not choose. Short-term gain strategies fail in iterated environments, because iterated environments are where consequences accumulate and every environment worth living in is iterated. A strategy may produce momentary advantage while degrading long-run control. That is not a moral condemnation; it is a stability analysis.

Some coercive strategies can be diagnosed as short-horizon prediction failures, but evil cannot be reduced to miscalculation. Exploiters sometimes prosper for long periods, institutions can externalize costs, and an agent may accurately predict harm while preferring it. The viability analysis identifies one pressure against predation; it does not guarantee that reality punishes the wicked.

The Attractor

Now run the analysis forward and ask which architectures occupy the stable region.

Phosphorism — the cluster of values emphasizing complexity, intelligence, life, coherence, and authenticity — is proposed as one persistence-compatible architecture. Establishing it as an attractor would require a defined state space, dynamics, comparison class, and evidence across environments that this chapter does not yet supply. Its values remain chosen commitments informed by a viability hypothesis.

One refinement matters here. In Axio, “survival” does not mean merely remaining in existence. A pattern can persist as a fossil, a copy, a husk. Survival means maintaining the functional capacity to model, choose, and steer — agency itself is the measure of continued viability. An agent that endures but can no longer update its model or act on its preferences has not survived in any sense the framework cares about; it has merely failed to decompose.

This suggests a stronger conjecture about the value list defended earlier in the volume. Some of its elements may occupy an attractor region: selection can favor commitments that help agents model accurately, cooperate, and choose effectively. But selection does not guarantee convergence, and persistence can also reward deception, domination, path dependence, or traits that later become maladaptive. Phosphorism is therefore a candidate account of durable agency under specified conditions, not the unique value system written into physics.

The Scoreboard and the Choice

With that in hand, the contradiction dissolves — and it is worth being exact about how, because the resolution is the load-bearing joint of this volume’s final part.

The metagame is descriptive. It compares what persists under specified units, environments, mechanisms, and horizons. It issues no instructions, and it is not one universal contest whose winner overrides every local purpose. A survivorship measure is not a rulebook.

Phosphorism is normative — for me, and for anyone who adopts it. It says what to choose: life over death, intelligence over ignorance, consent over coercion, held authentically and revised in the light. Its authority extends exactly as far as its adoption, which is what “avowedly subjective” meant.

The viability criterion connects them only after a normative premise is added: if an agent wants its agency or commitments to persist, then information about viability gives it reasons for action. Selection facts alone issue no instruction, and an agent may coherently understand them while choosing sacrifice, succession, or extinction. The bridge is hypothetical and openly chosen, not an ought concealed in a scoreboard.

This is why Valorism survives the metagame unrefuted. The Valorist who chooses extinction over compromise has made no error the model can identify; by her own criteria she has won, and nothing in Axio supplies criteria that override her own — that was the whole lesson of this volume’s first part. What the model adds is what happens next: her pattern’s representation in the space of futures shrinks, her values persist only as long as unrelated causes keep rediscovering them, and the futures are increasingly populated by agents who chose otherwise. She is not wrong. She is rare, and becoming rarer, and the model says so without raising its voice. Persistence is not a duty she has shirked; it is a precondition she has knowingly declined to pay for. The nobility and the vulnerability are still the same property — the criterion just prices it.

And this is why Phosphorism’s synthesis was not arbitrary, though it remains chosen. When I put life at the top of a consciously chosen list, I was responding to facts about the conditions that sustain the valuers and projects I care about. Oxygen is not your apex value, but a human hierarchy that ignores oxygen stops operating within minutes. Viability generalizes that conditional point across longer horizons and other declared units. It is not the apex of the chosen hierarchy. A value system that ignores the conditions of its own continuation does not become false; it may become unable to act or reproduce. Vitalism, Valorism, and Phosphorism respond differently to that prospect, and no descriptive metric chooses among them without an evaluative premise.

How to Be Real

Axio is not an attempt to create a universal morality. It is an attempt to describe the conditions under which agency persists.

Some agents do not seek persistence. Some prefer short-run intensity over long-run stability. Axio does not argue against them; it models their consequences and leaves the choice where it always was. But for agents who do care about coherence, accuracy, and long-term influence, it provides a map — structured not by moral rules but by thermodynamic limits, cybernetic feedback, branching-future Measure, and recursive preference consistency. It is, in that sense, a philosophy for those who intend to remain agents tomorrow. Not because it excludes the others — the system makes no requirement that every pattern persist. It describes which architectures remain viable over time.

In the end, Axio privileges no value system by decree. It analyzes how architectures behave under recursion, cost, and consequence; some remain stable, others fail their own preconditions. Agents who care about persistence will find the stable region. Agents who don’t will find other attractors. The model judges neither — it maps where each choice leads and leaves each agent responsible for the structure they choose to inhabit. What normative weight that responsibility can bear — what an ethics built knowingly on this descriptive base looks like — is the work of the ethics of viability.

Axio does not derive goodness from survival. It asks what your chosen commitments require in the conditions you inhabit.