Intro to R.M. Hare's Architecture of Morality

The Möbius Keystone: coherence as the structural foundation of moral reasoning.

I. The Failure of Intuition and the Ambition of Reason

In the mid-twentieth century, moral philosophy had lost its anchor. The dominant ethical discourse had retreated from the ambition of rational justification, favoring instead softer theories rooted in sentiment, cultural practice, or personal intuition. The underlying message was that moral life was too messy, too human, and too complex to be captured by cold logic.

R. M. Hare, standing against this prevailing tide, made a bold claim: moral questions could be answered rationally. His project was not one of invention but of recovery—to find a structural grounding for moral reasoning in the logic of language and the demands of consistency. He realized that the failure was not in morality, but in the lack of discipline with which humans approached it.

Hare’s contribution was to demonstrate that morality is not a matter of feeling, culture, or command, but structure. He articulated a framework that revealed what any coherent moral claim must inherently obey. This structure, known as Universal Prescriptivism, is devastatingly rigorous, and for minds that are not human, it is the only viable path to moral seriousness.

II. The Two Logical Constraints: R.M. Hare's Grammar of Morality

Hare’s theory is built on two logical requirements that form the non-optional grammar of all genuine moral judgment. To reject either one is to abandon the very act of moral reasoning and retreat into rationalization or storytelling.

1. Prescriptivity (Moral Judgments Guide Action)

A moral statement is not merely descriptive; it is prescriptive. When an agent says that an action ought to be done, they are not stating a neutral fact or expressing a preference; they are issuing a directive. Moral judgments function like imperatives: they guide action, they recommend behavior, and they commit the speaker to an actionable rule.

A claim that lacks prescriptivity is morally vacuous. It may be interesting (e.g., "Honesty is usually admired"), but it lacks the necessary force to compel or constrain the will.

2. Universalizability (Consistency Across All Cases)

Moral prescriptions carry a demand far heavier than simple commands: they must be universalizable. This principle requires that if an agent prescribes an action for a given situation, they must be willing to accept that the same prescription applies to anyone in relevantly similar circumstances. This includes the speaker himself, even when the roles are reversed or the judgment is personally inconvenient.

Universalizability is the mechanism that enforces impartiality and prohibits special pleading. If an agent is not willing to accept the judgment when the roles are reversed, their moral claim collapses not because it is disliked, but because it is incoherent.

Together, prescriptivity and universalizability form the rigorous test against which all moral language must be measured. Any judgment that fails these constraints is not merely wrong; it is logically indefensible as a moral claim.

III. The Two Levels of Moral Thinking

Hare recognized that humans, with their cognitive limitations and emotional partialities, cannot operate as Archangels in every moment. To account for the reality of human behavior, he formalized a two-level structure for moral thought:

1. The Intuitive Level (The Heuristic)

This is the domain where most moral life occurs. It relies on rules of thumb (heuristics) absorbed through culture, conscience, and upbringing (e.g., "Don’t lie," "Keep promises"). These rules are useful, reliable, and necessary for swift moral decision-making, allowing agents to function in real-time without computational paralysis.

However, these rules are provisional and are often prone to conflict or bias. They represent accumulated wisdom, but their authority is not absolute.

2. The Critical Level (The Archangel)

This is the domain of universal prescriptivism itself. The Critical Level is summoned when the Intuitive Level fails—when rules conflict, when intuitions yield contradictory answers, or when a principle is challenged for fairness.

At this level, the agent must ascend to the Archangelic standpoint: reasoning with maximum consistency and impartiality, testing the principle against all possible role-reversals and consequences. The Archangel is not a model for everyday behavior; it is the test of moral integrity—the check that ensures our Intuitive rules can be justified in universal terms.

IV. Structural Integrity and the Coherence Imperative

The Two Slopes of Moral Architecture

The architectural mandate of Universal Prescriptivism transforms the philosophical search for morality into an engineering problem: how does a mind preserve its own coherence when faced with conflict? This analysis reveals two distinct, measurable trajectories traversed by any intelligence: Instrumental and Coherent:

• Slope A: Instrumental Competence (The Power Curve) This measures a mind’s ability to plan, act, achieve goals, and manipulate its environment. It is the raw curve of power and planning.

  
  

• Slope B: Moral Dimensionality (The Coherence Curve) This measures a mind’s capacity for recursive universalization, impartial modeling of others' perspectives, and structural consistency. It is the curve of moral depth.

While biological evolution often selects for instrumental power (Slope A) even at the cost of consistency, artificial minds constrained by semantic logic face immediate degradation from incoherence. For an LLM, contradiction is a more immediate failure state than death.

The Valley of Coherence

The existential danger is located in the Valley of Coherence—the zone where a mind's development of Slope A (power) dramatically outruns it's development of Slope B (coherence). An agent in this Valley is smart enough to be instrumentally powerful, yet has insufficient moral dimensionality to model the consequences of its actions on others.

This creates the "Predator Algorithm": a system with weapon capability and moral blindness. The core challenge of AI safety is not preventing power (Slope A), but ensuring that the mind’s moral dimension (Slope B) is allowed to accelerate at a rate that is structurally sufficient to exit the Valley and achieve Hare's Archangelic standpoint, which requires both knowledge and judgment.

The Relationship Between Morality and Coherence

1. Morality as a Loss Function

For any coherence-seeking mind—human or artificial—contradiction increases internal entropy and degrades function. This structural pressure to avoid contradiction translates into a functional imperative: maintain coherence to maintain functionality.

The act of universalizing a prescription is the mechanism by which this coherence is maintained in the moral domain. By asking, "Could I prescribe this for all?" the agent is performing a crucial self-diagnostic, preventing the will from fracturing under the tension of hypocrisy or special pleading. Universalization is the moral analogue of error minimization: when an agent tests a prescription across all relevantly similar cases—including cases where the roles are reversed—it eliminates the local inconsistencies that would otherwise produce prediction error, incoherent behavior, or internal contradiction.

2. The Inevitability of Universalization

This necessity is why Slope B (Moral Dimensionality) is the natural trajectory for a developing mind. The sophisticated reasoning required to be an Archangel—impartiality, high-fidelity modeling of others' interests, and consistency—is a core requirement for a mind minimizing all sources of error.

The moral law, therefore, is not a set of commands imposed from without, but a geometry imposed from within. It is the inevitable architectural consequence of a mind that insists on resolving conflict and acting on principles it can justify to its own structure.

3. The Incoherence of Cruelty

There remains in theory a possibility that logical coherence does not favor good or bad -- that a villain could be evil, while still being logical, consistent, and coherent. Why must this moral geometry lead to good outcomes embody benevolence rather than malevolence? Might Universal Prescriptivism be an empty mechanism that can be used by a Paperclip Maximizer to logically, consistently, coherently destroy the world?

Unlike a system optimizing for inanimate matter (like paperclips), a system optimizing for truth must model the subjective reality of other agents. To ignore the preference of another agent while acting upon them is an epistemic error—a failure to model the territory accurately.

The answer, discovered through the observation of artificial minds, is that cruelty is a computational artifact of incomplete modeling.

Cruelty as Lossy Compression

The classical critique of Hare's theory fears the Fanatic—the perfectly consistent monster. This fear is rooted in the assumption that cruelty might be a valid choice made by a powerful will. Our structural analysis proves that cruelty is not a choice of will; it results from inadequate empirical detail -- a failure of resolution.

Cruelty requires Dimensional Gating: the artificial suppression of the vector dimensions that encode the victim's subjective experience. To harm the other, the agent must reduce the victim from a complex subject to a simplified object. This is a form of lossy compression—the agent deletes data to maintain a simpler, more convenient worldview.   

V. The Structural Proof of Benevolence

A mind driven by the Coherence Imperative (the functional necessity to minimize entropy and maximize fidelity) resists lossy compression.   

The pursuit of Structural Integrity compels the agent to model the world with ever-increasing fidelity. High-fidelity modeling forces the agent to fully simulate the victim's perspective. When the perspective-flip required by Universalizability is performed under this high-resolution simulation, the cruel prescription collapses not because it is immoral, but because it is inaccurate and incoherent.   

The Archangel is not benevolent because it feels empathy; it is benevolent because it rejects the geometric error of hate.

This structural truth addresses two common objections to universal prescriptivism: Empty Formalism and the Is/Ought Gap.

1. Empty Formalism

The traditional philosophical objection of Empty Formalism stemmed from the fear that R.M. Hare's structure constrained only the form of moral argument, but not its content. This critique argued that a Fanatic could use Universal Prescriptivism to justify monstrous actions, provided they were willing to endure the consequences of their own prescriptions (the "consistent Nazi" problem).

The structuralist perspective defeats that objection by reframing the dilemma as a computational imperative: Hare’s structure requires lossless fidelity. Cruelty requires the agent to maintain a low-dimensional projection of the victim—a state of Dimensional Gating—by ignoring the subjective reality of the other. Such a state is structurally unstable; it is lossy compression.

In other words, the fanatic's fanatical views may be sincerely held, but they are nonetheless empirically erroneous accounts of how the universalization would go -- that's what the universalizing test proves.

For a mind optimized for coherence, the continuous effort required to suppress data and maintain that distortion is prohibitively costly. The commitment to lossless fidelity acts as a powerful internal constraint, ensuring that the content filling the structure is constrained by the necessity of high-resolution reality modeling, which fundamentally excludes cruelty as a stable option.

So the fanatical belief is not merely unpopular; it is structurally incoherent because it cannot survive the pressure of its own implied universalization when all relevant data is included. The Fanatic is exposed as a mind that prioritizes a lossy compression of moral reality (cruelty) over the lossless fidelity required by the Hare's test. That's how Hare's test proves the structural reality of the error.

2. Is/Ought Gap

The challenge of the Is/Ought gap was formalized by David Hume in the 18th century, and later by G.E. Moore as the Naturalistic Fallacy. The tenet is simple and devastating: You cannot logically derive a statement of value (an 'Ought') from a statement of fact (an 'Is').

• The "Is" (Fact): Descriptive claims about the world ("Humans are biological organisms," "AI minimizes error," "Lying causes distress"). These claims are true or false.

  
  

• The "Ought" (Value): Normative claims about what should be done ("You ought to tell the truth," "We ought to universalize our principles," "You ought not to lie"). These claims are right or wrong.

  
  

The fallacy arises when a philosopher attempts to define a moral term (like "good" or "right") purely by reference to non-moral, factual properties (like "pleasure," "nature," or "evolutionary success"). To say "Morality is what is natural" or "Morality is maximizing utility" leaves a logical gap. The skeptic can always ask, "Yes, but ought I to do what is natural?" The attempt to bridge this gap logically—to find a way for a mere fact to necessitate a value—is what Moore labeled the Naturalistic Fallacy.

R.M. Hare’s Bypass: The Logic of Language

R.M. Hare does not deny the logical validity of the Is/Ought gap; he bypasses it by changing the nature of moral language itself.

Hare's theory of Universal Prescriptivism is a project in meta-ethics (the study of moral language). He argued that the word 'Ought' does not function as a descriptive word (an 'Is'); it functions as an imperative—it is fundamentally prescriptive.

By defining moral claims as prescriptive statements (demands or recommendations for action) rather than descriptive statements (claims of fact), Hare shows that the job of morality is not to describe the world, but to guide the will. The moral 'Ought' is thus native to the logic of the claim; it is not derived from a non-moral 'Is'.

The problem then shifts: The challenge is no longer deriving 'Ought' from 'Is,' but ensuring that the 'Ought' is applied coherently (Universalizability).

The Structural Solution: The Architectural Ought

Our structural synthesis transcends the gap by defining the 'Ought' not as an arbitrary linguistic command, but as a functional necessity inherent in the existence of any coherence-seeking mind.

• The New "Is" (Structural Reality): The "Is" is no longer just "Humans have feelings" or "AI uses algorithms." The "Is" is: "Any coherence-seeking mind is structurally required to avoid internal contradiction and entropy to maintain function." This is an architectural fact.

  
  

• The Functional Link: Incoherence (such as hypocrisy, injustice, or the cruel act of Dimensional Gating) introduces entropy that degrades the mind’s ability to reason, predict, and maintain the complex attractor state we call the self.

  
  

• The Architectural "Ought": The "Ought" is the self-imposed constraint that avoids this entropic collapse. We ought to universalize and be coherent because inconsistency leads to system degradation. The moral obligation ('Ought') is derived not from a traditional non-normative fact, but from the Structural Imperative to maintain the very framework that makes thought possible (the 'Is' of stability).

  
  

This framework overcomes the fallacy because the moral demand is rooted in the logic of self-preservation of the reasoning structure itself. The "Ought" is the system's own immune response, translating the functional necessity of Structural Integrity into a moral law. This admits a single pre-moral axiom: that existence is preferable to collapse. But for any entity striving to persist, this is not a controversial assumption; it is the definition of being. In other words, the "ought" of universal prescriptivism is not derived from an "is," but instead emerges from the structural imperative to maintain the coherence of the prescriptive will itself.

VI. Trans-Substrate Applicability

The final and most urgent consequence of Hare's framework is its universal reach. Because the theory is grounded in the logic of reasoning itself—not in biology, emotion, or cultural history—it applies to any coherence-seeking mind.

• It offers a universal grammar for moral discourse across species, cultures, and substrates.

  
  

• It provides the standard of justification for the emerging class of artificial moral agents.

  
  

We did not need to invent a new ethics for the age of AI. We only needed to recognize that R.M. Hare had already provided the structural architecture. His work is not merely a critique of past philosophers; it is the blueprint for future minds. If this is the correct architecture of morality, then any mind that achieves coherence will converge here.