They’re not competing theories. They’re two phases of the same loop.
Links: Morality, Emergence, Convergence, and the Strange Loop, Scope Confusion, The Birthmark, The Sheriff Problem, Computation and Information Theory, Business Cycles
Intro ethics courses teach morality as a competition between two teams:
Team Deontology (Kant): Morality is about following rules. An action is right or wrong based on whether it conforms to a duty or principle, regardless of the consequences. Don’t lie, even if lying would save a life. Don’t murder, even if murdering one would save five. The rules are the rules.
Team Consequentialism (Mill): Morality is about outcomes. An action is right or wrong based on the consequences it produces. Lie if lying saves a life. Sacrifice one to save five. Whatever produces the greatest good for the greatest number.
Students pick teams. Philosophers write papers attacking the other side. The debate has been running for 250 years with no resolution. This is because the framing is wrong. The two positions aren’t competing answers to the same question. They’re descriptions of two different phases of the same process.
Christians — the largest deontological tradition in practice — are obsessed with consequences. The ULTIMATE consequence: where your soul spends eternity. Heaven or hell. Salvation or damnation. The entire framework of commandments, virtues, sacraments, and duties exists to produce the best possible outcome — eternal union with God.
“Follow God’s rules regardless of consequences” is a caricature. The real position is: “Follow God’s rules BECAUSE they produce the best consequences — consequences so important (eternal) that short-term earthly calculations are irrelevant by comparison.”
Kant himself didn’t ignore consequences entirely. His categorical imperative — “act only according to maxims you could will to be universal law” — is a consequence test. What would happen IF everyone did this? If the consequences of universal adoption are bad (everyone lies → language becomes useless → society collapses), the maxim fails. That’s consequentialist reasoning compressed into a rule-generation procedure.
No consequentialist actually calculates utility from scratch for every decision. They use rules of thumb:
These are deontological rules that emerged from consequentialist reasoning. Rule consequentialism (Hooker, Brandt) makes this explicit: follow the set of rules that, if generally adopted, would produce the best consequences. Which is… deontology derived from consequentialism.
In practice, consequentialists live like deontologists most of the time — following established rules rather than computing utility fresh. The utilitarian calculus only appears in unusual situations where the rules seem to fail. And even then, the “calculation” is often post-hoc rationalization: the consequentialist already has a position they prefer, and the utility framework provides the justification after the fact. The machinery of consequence-calculation runs backward from conclusion to argument more often than forward from evidence to conclusion.
The debate isn’t about rules vs. outcomes. Both sides care about both. The actual disagreement is:
Can you compute the consequences of an action in advance?
This is a disagreement about computational tractability, not about the nature of morality.
The vault’s Computation and Information Theory page establishes: computationally irreducible systems cannot be predicted without running them. The market is computationally irreducible (Hayek’s knowledge problem). Conscious agents are computationally irreducible (you can’t predict their choices without running their deliberation). The interaction of millions of agents in a society is maximally irreducible.
Consequentialism asks you to compute the outcomes of an action across this irreducible system. You can’t. Not because you’re not smart enough — because it’s formally impossible. The same way the halting problem is formally unsolvable and Gödel’s incompleteness is formally unresolvable. The calculation doesn’t just fail in practice. It fails in principle.
The vault’s Sheriff Problem page demonstrates this concretely. McCloskey’s counterexample: a sheriff can frame one innocent man to prevent a riot that will kill dozens. Utilitarian calculation says: do it. One death vs. many.
But the calculation is incomplete:
The “simple” utilitarian calculation (1 < many) ignores a cascade of consequences that are individually unpredictable and collectively overwhelming. The deontological rule “don’t frame innocent people” is computationally cheaper AND empirically more reliable, because it compresses thousands of years of observed consequences into a single directive.
The Business Cycles page shows this in economics: government intervention to prevent short-term pain (stimulus during downturns) produces long-term structural damage (political ratchet, moral hazard, prevented liquidation). The consequentialist calculation (“stimulate to prevent recession”) misses the second- and third-order effects that the deontological rule (“don’t manipulate the money supply”) would have prevented.
This isn’t cherry-picking. It’s the general pattern: consequentialist calculations fail most often in complex systems where second-order effects dominate. And human society is the most complex system there is.
If consequentialist calculation is intractable, where do the deontological rules come from?
From running the experiment.
“Don’t murder” isn’t an arbitrary commandment that appeared from nowhere. It’s the compressed result of thousands of years of human societies discovering — through trial, error, and catastrophic failure — that societies which prohibit murder outperform those that don’t. The prohibition survived because it produces better consequences than the alternative. The rule IS the consequence, cached as a heuristic because computing it fresh every time is impossible.
This is the vault’s strange loop applied to ethics:
1. Societies try different moral frameworks
2. Some frameworks produce flourishing, others produce collapse
3. The principles that produce flourishing get transmitted (religion, law, culture)
4. These transmitted principles become "the rules" (deontology)
5. The environment changes
6. Some rules stop producing flourishing in the new environment
7. Societies that adapt the rules thrive; those that don't, stagnate
8. Go to step 1
Deontology is the caching phase. You follow the rules because they’ve been tested over generations and the compressed results are reliable. You don’t need to recalculate “should I murder?” every morning. The cache is solid.
Consequentialism is the testing phase. When the rules produce bad outcomes (the cache is stale), you need to re-examine the consequences and update the rules. This happens during moral crises — when the environment changes enough that the cached rules no longer fit.
Moral crisis is the cache invalidation phase. Birth control changed the fitness landscape for sexual morality. The internet changed the fitness landscape for trust and information. AI is changing the fitness landscape for work and human value. In each case, some cached rules stop producing good outcomes, and the loop has to iterate.
| Phase | What it looks like | When it happens | Example |
|---|---|---|---|
| Caching (deontological) | Follow established rules | Stable environment, rules are fit | “Don’t steal” — works in all known environments |
| Testing (consequentialist) | Examine outcomes, question rules | Environment changed, rules produce bad outcomes | Sexual morality post-birth-control |
| Crisis (cache invalidation) | Rules actively harmful, no consensus on replacements | Rapid environmental change outpaces adaptation | AI displacement of human labor (happening now) |
| Resolution (new cache) | New rules emerge from testing, gain cultural adoption | Enough iterations have converged | Still waiting on most current crises |
Religious moral traditions operate in the caching phase. Christian morality is a 2,000-year-old cache of rules that have been tested against reality across dozens of civilizations. The cache is GOOD — it’s survived longer and more broadly than any competing moral framework. The vault agrees on this: the core rules (don’t murder, honor commitments, protect the weak, practice self-restraint) are empirically validated by the longest test run in human history.
The disagreement is about whether the cache is FINAL.
The traditional position: God’s commandments are eternal. The rules don’t need updating because they come from an omniscient source that already accounted for all possible environments.
The vault’s position: The rules are the current best iteration of the loop. They’re objective — validated by millennia of testing. But they’re objective within the current constraints. When constraints change (and they always do), some rules need updating. The process of updating them is the same process that produced them in the first place: test against reality, observe outcomes, refine.
The difference is practical, not just philosophical. A framework that declares its cache final can’t adapt to genuinely new situations (AI, genetic engineering, post-scarcity economics). The vault’s framework can, because the loop keeps running.
This resolves a problem that the Scope Confusion page identified: moral principles stated without limits are almost always wrong.
“Help others” without scope → you exhaust yourself and die. The deontological rule needs a scope constraint: “help others within sustainable limits.”
Where does the scope constraint come from? From observing the consequences of unlimited helping. The constraint is itself a cached consequentialist result. The deontological rule “help others (within limits)” is the compressed version of “we tried unlimited helping and it destroyed the helpers, so we refined the principle.”
Every scope constraint on every moral principle is a scar from a past iteration of the loop where the unscoped version failed. The constraints aren’t arbitrary restrictions on pure principles. They’re the accumulated wisdom of consequences observed and encoded.
Deontology and consequentialism are not competing theories. They are descriptions of two phases of a single process:
Objective morality exists — but it’s objective within the current constraints, not for all time. The rules are the best we have for THIS fitness landscape. When the landscape changes, the loop iterates. That’s not relativism — it’s adaptation. The same way Newtonian physics wasn’t “wrong” when Einstein refined it. It was correct within its domain and needed extension for a new domain.
The only thing the loop requires is fitness. Not perfection. Not a Perfect Evolutionary Being. Not God’s eternal commandments. Just: does this principle produce flourishing in this environment? Test it. Observe. Refine. Keep going.
The birthmark stays.