The Dissolution of the Is-Ought Distinction

An Agent-Relative Semantics Grounded in Persistence

Abstract The is-ought distinction, as articulated by Hume (1739), asserts that normative conclusions cannot be derived from descriptive premises by valid deductive inference. This paper argues that the apparent gap between fact and value is a semantic artefact rather than an ontological feature of reality. The argument proceeds in three steps. First, the term "ought" functions as a regulatory operator whose semantic content presupposes an agent with persistence-conditions; deployed outside this presupposition, "ought" claims are type-errored rather than false (§2). Second, any agent capable of formulating "ought" claims is by selection effect a self-maintaining system, so the question "ought there be self-maintaining systems?" is asked only from within the class whose existence it questions (§3). Third, "ought F for agent S" identifies a conditional necessity-relation between F and S's persistence, with explicit extensions for non-necessary but persistence-relevant actions (§4). The result is dissolution rather than reduction: the gap is not bridged but shown to be an artefact of misapplied universal-scope semantics. The account is positioned against naturalistic moral realism (§7.1), constitutivism (§7.2), evolutionary ethics (§7.3), and error theory (§7.4), and concludes with explicit scope limitations including computational intractability, the dark-room limit, inter-agent normativity, and counterfactual ought-claims (§8). The dissolution claim is bounded: it resolves the core is-ought problem and grounds a tight epistemic imperative; it does not deliver a complete normative theory.

1. The Problem and Its Mischaracterization

1.1 Hume's argument

In A Treatise of Human Nature (Book III, Part I, §I), Hume observes that moral writers proceed from "is" propositions to "ought" propositions without explicating the inferential step, and concludes that the transition requires explanation no valid deductive inference supplies from descriptive premises alone.

The claim has a precise logical structure. It concerns deductive entailment between propositions of distinct syntactic types: indicative (descriptive) and normative (prescriptive). The claim is not metaphysical (about the nature of value), epistemological (about how value is known), or motivational (about how value moves to action). It is a claim about formal derivability.

1.2 The standard reading

In post-Humean meta-ethics, the gap has typically been read as an ontological claim about a categorical distinction between two types of fact (descriptive and normative) corresponding to two types of property (natural and non-natural). This reading is reinforced by Moore's open-question argument (Principia Ethica, 1903), which appears to show that no proposed naturalistic identity for "good" closes the conceptual gap: for any candidate property P, the question "is P actually good?" remains substantively open.

The standard reading generates an apparent trilemma: accept a non-natural moral realism with its attendant epistemological difficulties; an error theory denying the existence of moral facts (Mackie 1977); or non-cognitivism on which moral utterances are not truth-apt (Ayer 1936; Stevenson 1944).

1.3 The mischaracterization

The standard reading conflates two distinct claims:

(D) From descriptive premises alone, no normative conclusion follows by valid deductive inference. [Hume's claim]

(O) Normative claims pick out properties or facts of a categorically different kind than descriptive claims. [Standard ontological reading]

(D) is a logical observation about derivability. (O) is a metaphysical thesis about types of property. (D) does not entail (O). One can grant (D) while denying (O), provided one can show that the appearance of categorical distinction is generated by semantic structure rather than by ontological difference.

This paper grants (D) and rejects (O). The argument has three components, developed in §§2–4.

2. The Semantic Structure of "Ought"

2.1 "Ought" as a regulatory operator

In ordinary usage, "ought" claims have the form S ought to F, where S is an agent and F an action or state. The semantic content is regulatory: it evaluates or directs S's behavior with respect to a target.

Proposition 2.1. The semantic content of "S ought to F" presupposes (i) an agent S with goals or persistence-conditions, and (ii) a relation between F and those goals or conditions.

Without (i) or (ii), the claim has the surface form of an "ought" claim but lacks semantic content. Compare:

"You ought to eat to live." [agent: you; goal: to live; F: eating; well-formed]
"The asteroid ought to collide with Earth." [no agent or goals; ill-formed unless the asteroid is anthropomorphised or an implicit observer-frame supplied]
"The universe ought to exist." [no agent; no goal-relative frame; semantically empty unless a theistic or anthropic frame is supplied]

2.2 The type-error argument

Consider the question: "What is north of the North Pole?" The grammatical structure is well-formed. "North of" is a relational direction operator; "the North Pole" is a referring expression. Yet the question is not answerable, not because the answer is unknown, but because "north of" is undefined at the pole. The question is a type error in the geometry of the sphere.

The analogous claim for "ought":

Proposition 2.2. "Ought" is undefined outside the domain of agents with persistence-conditions. Questions of the form "ought there be X?" asked without specifying a referent agent or evaluative frame are type-errored, not false.

This is the first dissolutional move. Universal-scope "ought" questions ("ought there be a universe?", "ought there be life?") are not questions whose answers are hidden or contested. They are questions whose presuppositions fail. The grammatical surface obscures the semantic ill-formedness.

2.3 Comparison to existing semantic accounts

The agent-relative semantics here has antecedents in the literature but is distinct in its precise claim.

Anscombe (1958) argues that the moral "ought" lost coherence with the decline of theological frameworks that grounded its authority, and proposes restricting ethics to virtue concepts. The present account agrees that universal-scope "ought" requires a referent frame but does not require theological or virtue-theoretic grounding; the persistence-conditions of agents themselves supply the referent.

Foot (1958, 2001) argues that moral evaluation is a species of "natural goodness" tied to the form of life of the evaluated organism. The present account preserves this insight at a more general level: "ought" is keyed to the persistence-conditions of any self-maintaining system, not specifically to species-typical forms of life.

Stevenson (1944), Ayer (1936) hold emotivist or non-cognitivist views on which moral utterances are not truth-apt. The present account rejects this: properly-framed "ought" claims (with explicit agent and persistence frame) are truth-apt; their truth conditions are facts about persistence-relevance.

3. The Persistence Tautology

3.1 The selection effect

Proposition 3.1 (Persistence Tautology). Any system that does not maintain the conditions of its own persistence ceases to exist as a referent in subsequent time-steps. The class of observable systems at time t is the class of systems that have maintained their persistence-conditions up to t.

This is not a normative claim. It is a logical consequence of what "persistence" and "existence as observable" mean. It functions as a selection principle on the space of systems.

3.2 The asker is the answerer

Any agent capable of formulating the question "ought there be self-maintaining systems?" is itself a self-maintaining system. This is not contingent on the agent's biology; it is constitutive of being an agent in the formal sense (an entity capable of formulating claims).

Proposition 3.2. The set of askers of universal-scope "ought" questions is a strict subset of the set of self-maintaining systems. No system that fails to maintain itself is in a position to question whether it should maintain itself.

The structure is analogous to anthropic reasoning in cosmology: questions about why the universe permits observers are asked only by observers, and the existence of the question presupposes the conditions of its asking.

3.3 Constitutive vs. evaluative response to the selection objection

Objection. This is a mere selection effect; it does not show that persistence is good, only that non-persistence eliminates evaluators.

Response. The objection presupposes the standard reading (O). Once (O) is rejected and "ought" is recognised as agent-relative, the demand to show that persistence is good in a frame-independent sense is itself a type error. There is no frame-independent "good." For any specific agent S, the question "is persistence good for S?" has a frame: S's own persistence-conditions. The answer is constitutive: persistence is the precondition of any "good for S" being instantiable.

This is dissolution, not reduction. The framework does not derive a normative conclusion from a descriptive premise. It identifies the appearance of such a derivation as an artefact of treating "ought" as if it had universal-scope semantics it does not have.

4. The Identity Claim

4.1 Core identity

Proposition 4.1. For a self-maintaining system S and an action F, where F is necessary for S's persistence over interval Δt:

Ought(S, F) ⟺ [ ¬F → ¬∃S(t + Δt) ]

That is: "S ought to F" is equivalent to the conditional that S's not-F-ing causes S's non-persistence over Δt.

4.2 Scope and extensions

The identity in 4.1 captures strictly necessary persistence-relevant actions. It does not capture, without extension:

(a) Probabilistically relevant actions. Actions that increase or decrease persistence probability without strict necessity. Extension: replace the strict conditional with a probabilistic conditional and replace bare persistence with expected persistence-utility over relevant intervals.

(b) Flourishing-beyond-survival. Actions that improve persistence-quality (resource margins, capacity, resilience) without being strictly necessary for bare persistence. Extension: include conditions on persistence-quality, not only persistence-existence.

(c) Supererogatory acts. Actions that exceed what is strictly required of S. The framework treats these as actions whose connection to S's persistence-conditions is weak or absent at the individual level; their "ought" status (if any) requires inter-agent extension. The framework does not deliver a closed criterion for the supererogation threshold (§8.6).

(d) Counterfactual ought-claims. "S ought to have F'd" where S did not F. The framework handles these as evaluations against S's past persistence-conditions. The semantics extends straightforwardly; the epistemology of accessing past conditions is non-trivial.

4.3 Why this is not reduction

Reduction would require translating "ought" claims into "is" claims via a bridging definition: "S ought to F" means "F is necessary for S's persistence." Reduction is then vulnerable to the open-question objection: "yes, F is necessary for persistence, but is it good?"

The dissolution claim is structurally different. It denies the cogency of the open-question objection in this context. Once the agent-relative semantics of "ought" is recognised, the question "is persistence-necessity good?" has no frame-independent referent. The question is type-errored at the same level as "what is north of the North Pole?"

Moore's open-question argument retains force against reductive naturalism that aims to translate "good" into a natural property while preserving universal-scope semantics. It loses force against an account that denies the universal-scope semantics from the outset.

5. Epistemic Fidelity and the Cost of Misalignment

5.1 The cognitive-load asymmetry

For an agent S that holds both a true representation T and a competing false representation F, deploying F in communication or reasoning imposes higher cognitive cost than deploying T. This is empirically established across three decades of deception research:

Reaction-time penalty of approximately 200 to 300 milliseconds for deceptive responses relative to truthful responses (cognitive-load deception literature, principally Vrij, Walczyk and collaborators).
Elevated activation in dorsolateral prefrontal cortex, anterior cingulate cortex, and inferior frontal gyrus during deception in fMRI studies (multiple replicated findings).
The basis of cognitive-load lie-detection methods (unexpected questions, reverse-order recall, dual-task interference), which operationalize the asymmetry as a measurement procedure.

The mechanism: deception requires inhibition of the prepotent truthful response, generation of a plausible alternative, and ongoing monitoring of consistency. Truth-telling requires only retrieval.

5.2 The free-energy framework

At a higher level of theoretical abstraction, the cognitive-load asymmetry expresses within variational free-energy minimization (Friston 2010, with subsequent developments). Agents minimize divergence between their internal generative model and the sensory evidence they receive. Maintaining a model that misaligns with evidence imposes a free-energy cost the agent must either compensate (additional cognitive resources) or resolve (model update).

Status of the free-energy framework: a productive theoretical research program with substantial uptake in cognitive neuroscience; not a fully settled physical principle. The present account uses it as theoretical framing for the cognitive-load asymmetry and does not depend on its strong ontological readings.

5.3 The persistence connection

For an agent operating in an environment with non-zero variance and finite resources, sustained misalignment between internal model and external reality degrades prediction accuracy and increases the rate of costly errors. Over time, this reduces persistence probability.

Proposition 5.1. In environments with variance V > 0 and resource constraints R < ∞, sustained model-reality misalignment in agent S monotonically reduces P(persist(S, t + Δt)) for Δt sufficiently large.

The qualification matters: in the limit V → 0 (the dark-room limit), false models persist without disconfirmation, and the cost of misalignment is not paid. This is a known boundary condition of the framework (§8.2).

6. The Epistemic Imperative

From §5: for agents in normal environments (V > 0, R < ∞), maintaining accurate models reduces cognitive load and improves persistence probability. The "ought" associated with this is:

Proposition 6.1 (Epistemic Imperative). For S in normal environments, S ought to minimize divergence between internal model and external reality.

This follows from Proposition 4.1 and Proposition 5.1 under the relevant environmental conditions. The derivation is tight: the connection between model fidelity and persistence is structural, mediated by the cost of error.

The framework's positive normative output is this single imperative. It does not output:

A theory of inter-agent normativity (justice, rights, cooperation; see §8.4).
A closed criterion for supererogation (see §4.2c, §8.6).
A solution to trolley-type dilemmas, which turn on inter-agent normativity not developed here.
A complete decision procedure for moral dilemmas (see §8.1).

These are scope limitations of the present framework's normative output, not defects of the dissolution argument as such.

7. Positioning Against Alternatives

7.1 Cornell naturalistic moral realism

Boyd (1988), Brink (1989), Sturgeon (1985), and Railton (1986) advance naturalistic moral realism: moral properties are natural properties (broadly construed), and moral facts are facts about those properties.

The dissolution account agrees that "ought" claims, properly framed, are facts about agent-persistence-conditions in the natural world. It differs in two ways:

(a) Cornell realism typically retains universal-scope semantics for "good" and "right" and argues that these terms refer to natural properties via a causal-historical theory of reference (analogous to Kripke-Putnam for natural kinds). The dissolution account denies universal-scope semantics: "ought" is irreducibly agent-relative.

(b) Cornell realism is vulnerable to refined open-question objections (Horgan and Timmons 1991, "Moral Twin Earth"). The dissolution account is not vulnerable in the same way, because the open question is reframed as a type-error rather than treated as a substantive question about reference.

7.2 Constitutivism

Korsgaard (1996, 2009) derives normative conclusions from the constitutive aims of agency: to be an agent at all is to be committed to certain principles (the categorical imperative, in her Kantian reconstruction).

The dissolution account shares the structural move of grounding "ought" in conditions constitutive of agency. It differs in:

(a) The constitutive aim. Korsgaard locates it in the unity of agency and self-legislation of practical reason. The present account locates it in persistence under thermodynamic constraints.

(b) The normative content derived. Korsgaard derives Kantian universalizability. The present account derives only the epistemic imperative of §6.

(c) Scope of agents. Korsgaard's account targets specifically rational agents. The present account targets any self-maintaining system; the derivation of normative content scales with the complexity of the agent's persistence-relevant model.

7.3 Evolutionary ethics

Ruse (1986), Wilson (1975), and others argue that moral intuitions and norms are products of evolutionary pressures on cooperative behavior. The dissolution account agrees with the causal claim (moral capacities have evolutionary origins) but distinguishes:

(a) The evolutionary account is a genealogy of moral intuitions; it does not settle questions about their normative authority.

(b) The dissolution account is not a genealogy. It is a semantic and metaphysical account of what "ought" claims pick out. The two are compatible at different explanatory levels.

(c) The dissolution account does not commit to the claim that all moral intuitions track persistence-relevant facts. Some intuitions may be evolutionary byproducts that misfire in modern environments; the framework supplies a basis for evaluating which.

7.4 Mackie's error theory

Mackie (1977) holds that moral claims commit to objective values that do not exist, and concludes that all moral claims are false.

The dissolution account denies Mackie's first premise. Moral claims, properly framed, do not commit to frame-independent objective values. They commit to agent-relative persistence-conditions, which exist. Mackie's error theory presupposes the standard universal-scope reading of "ought" the dissolution account rejects.

8. Scope, Open Questions, Boundary Conditions

The following are explicitly unsettled or unaddressed by the present framework. They are flagged, not concealed.

8.1 Computational intractability

Even granting that "ought" facts for agent S are determinate (given S's persistence-conditions and environment), computing them in real time may exceed S's resources. The framework identifies what S ought to approximate; it does not provide a tractable algorithm. This is the bounded-rationality problem (Simon 1956) extended to ethics.

8.2 The dark-room limit

In environments with V → 0 and unlimited resources, false models persist without disconfirmation. The cost of misalignment is not paid, and the epistemic imperative loses its grounding. The framework's normative output is conditional on V > 0 and R < ∞. The dark-room case is not refuted; it is identified as the regime where the framework does not bind.

8.3 Phenomenal consciousness

The framework supports a functional account of valence (positive and negative affect as signals about persistence-trajectory) but does not explain the phenomenal character of these states. The hard problem of consciousness (Chalmers 1995) is not addressed and is not claimed to be.

8.4 Inter-agent normativity

The framework grounds individual-scale "ought" claims. It does not develop an account of:

Justice (distributive and procedural).
Rights (claim-rights, liberty-rights, immunities).
Conflict resolution under incompatible persistence-claims between agents.
Aggregation across persons.
Whether cooperation is normatively required, optional, or context-dependent for self-maintaining systems in multi-agent environments.

These are core domains of normative theory. The dissolution account, as presently developed, restricts its primary claim to the dissolution of the core is-ought distinction. Inter-agent normativity is flagged as unsolved and is treated as a domain for future work, not as a settled output of the framework.

8.5 Counterfactual and historical ought-claims

"S ought to have F'd" where S did not F requires evaluation against S's past persistence-conditions. The semantics extends straightforwardly. The epistemology of accessing past conditions is non-trivial and may yield underdetermined answers.

8.6 Supererogation

Acts that exceed what is required of S are not captured by the strict-necessity identity in §4.1. The framework can extend to include them via persistence-quality considerations (§4.2b), but the threshold between required and supererogatory remains underdetermined.

8.7 Population ethics

Questions about whether one ought to bring new agents into existence (procreation, antinatalism) require an account of how persistence-claims aggregate across actual and potential agents. The framework does not currently supply one.

8.8 Status of the underlying theoretical commitments

The framework's tighter results (§§2–4) depend on the semantic analysis of "ought" and the persistence tautology. These are defensible on their own terms. The framework's looser results (§5 and the imperative derived in §6) depend additionally on the free-energy principle as a productive research program rather than as a settled physical principle. The dissolution claim itself does not depend on the looser results; the epistemic imperative of §6 does.

9. Conclusion

The is-ought distinction, as Hume formulated it, is a correct observation about formal deductive entailment between propositions of distinct syntactic types. It is not, as commonly read, an ontological claim about a categorical gap between fact and value.

The apparent gap dissolves when the semantic structure of "ought" is correctly identified. "Ought" is an agent-relative regulatory operator. Its content presupposes an agent with persistence-conditions. Outside this presupposition, "ought" claims are type-errored, not false. Within this presupposition, "ought" claims pick out conditional necessity-relations between actions and persistence-conditions.

The apparent gap was generated by treating "ought" as if it had universal-scope semantics. Once universal-scope is dropped and agent-relativity is recognised, the gap is not bridged but shown to be an artefact of misapplied grammar.

The dissolution claim is bounded. It resolves the core is-ought distinction and grounds an epistemic imperative tightly. It does not settle inter-agent normativity, supererogation, population ethics, or the phenomenal character of moral experience. These remain open and are explicitly flagged as such.

References

Anscombe, G. E. M. (1958). Modern Moral Philosophy. Philosophy 33(124), 1–19.

Ayer, A. J. (1936). Language, Truth and Logic. Gollancz.

Boyd, R. (1988). How to Be a Moral Realist. In G. Sayre-McCord (ed.), Essays on Moral Realism. Cornell University Press.

Brink, D. (1989). Moral Realism and the Foundations of Ethics. Cambridge University Press.

Chalmers, D. (1995). Facing Up to the Problem of Consciousness. Journal of Consciousness Studies 2(3), 200–219.

Foot, P. (1958). Moral Beliefs. Proceedings of the Aristotelian Society 59, 83–104.

Foot, P. (2001). Natural Goodness. Oxford University Press.

Friston, K. (2010). The Free-Energy Principle: A Unified Brain Theory? Nature Reviews Neuroscience 11(2), 127–138.

Horgan, T. and Timmons, M. (1991). New Wave Moral Realism Meets Moral Twin Earth. Journal of Philosophical Research 16, 447–465.

Hume, D. (1739). A Treatise of Human Nature. Book III, Part I, §I.

Korsgaard, C. (1996). The Sources of Normativity. Cambridge University Press.

Korsgaard, C. (2009). Self-Constitution: Agency, Identity, and Integrity. Oxford University Press.

Mackie, J. L. (1977). Ethics: Inventing Right and Wrong. Penguin.

Moore, G. E. (1903). Principia Ethica. Cambridge University Press.

Railton, P. (1986). Moral Realism. Philosophical Review 95(2), 163–207.

Ruse, M. (1986). Taking Darwin Seriously. Blackwell.

Simon, H. (1956). Rational Choice and the Structure of the Environment. Psychological Review 63(2), 129–138.

Stevenson, C. L. (1944). Ethics and Language. Yale University Press.

Sturgeon, N. (1985). Moral Explanations. In D. Copp and D. Zimmerman (eds.), Morality, Reason and Truth. Rowman and Littlefield.

Vrij, A., Fisher, R., Mann, S., and Leal, S. (multiple studies, 2003–2012). Cognitive-load approach to deception detection.

Walczyk, J., Roper, K., Seemann, E., and Humphrey, A. (multiple studies, 2003–2014). Reaction-time asymmetries in deception.

Wilson, E. O. (1975). Sociobiology: The New Synthesis. Harvard University Press.

Written by Andraž. Developed in collaboration with Claude (Anthropic). The dissolution argument, the persistence-tautology grounding, and the agent-relative semantic framework originate with the author across earlier working papers; this paper is the consolidated statement.