The Is–Ought Relocation
Foundations argued that a constructed morality doesn’t have to be an arbitrary one — and made good on that nearly everywhere, except at one spot it named and then walked past. That spot is the oldest tripwire in moral philosophy: the step from “is” to “ought.” A reader who’s granted everything so far can still ask where an “ought” comes from, if it’s neither read off the facts nor written into the structure of things. Answering that is the job of this essay. It’s the hardest point the framework has to hold, and the one most worth holding cleanly, because every normative claim in the book leans on it.
The move: relocate, not derive
Start with what the framework does not do, because the usual caricature attacks a move it never makes. AoM does not derive an ought from an is. It runs no inference from “here’s how the world is” to “therefore this is good”; it never claims that because the world is a certain way, some thing is good as a result. Hume’s observation — that no “ought” follows by logic from any pile of “is”es — AoM grants in full. It simply declines the project Hume was warning against.
What it offers instead is a constructivist account of what an “ought” is. Normativity isn’t a fact sitting on the far side of Hume’s gap, waiting to be reached by some clever inference; it’s constituted by what valuing agents converge on as the context of their meaning-making widens. There’s no fixed ought-fact standing across the gap, so the gap doesn’t need to be crossed. What there is, on this account, is growing agreement on what ought to be done, within a growing context — and that process isn’t evidence for the ought; it’s what “ought” amounts to.
One word controls how much this essay has to prove. AoM does not dissolve the is–ought gap — it doesn’t claim Hume was confused, or that the gap was an illusion all along. That would be a bigger and needless fight. AoM relocates: it leaves the gap exactly where Hume drew it, declines to run anything across it, and puts normativity somewhere else entirely — in the convergence, not in a fact to be inferred. “Relocate” is the honest verb, and a far smaller target than “dissolve.”
Good company
This is a recognized strategy with a long and respectable lineage, and naming the lineage is part of the defense: it shows the burden has been carried before, by philosophers nobody accuses of sleight of hand.
The closest relative is Jürgen Habermas’s discourse ethics, on which a norm is valid insofar as everyone affected could agree to it under progressively less distorted, more inclusive discussion. “Less distorted, more inclusive” is very nearly AoM’s “widening context,” reached from another tradition. T.M. Scanlon’s contractualism is a cousin: an act is wrong if it violates principles no one could reasonably reject. Sharon Street and Christine Korsgaard supply the metaethical core already met in Foundations — normative truths constituted from agents’ evaluative standpoints rather than discovered as mind-independent facts. C.S. Peirce contributes the shape of the idea a century early: truth as what inquiry converges on in the limit, here turned from factual inquiry to normative. And Robert Aumann supplies the formal backbone — the result that rational agents who share priors and honestly pool what they know cannot, in the limit, agree to disagree.
AoM isn’t the first to put normativity in convergence under idealized, widening discussion. Its contribution isn’t the move but what it couples the move to — a structural criterion and a counter-dynamic — which is what the rest of this essay is about.
Two horns, owned
Two consequences of the move have to be said out loud, because a critic who finds them unstated will present them as things the framework failed to notice.
The first: the relocation is parasitic on the anti-realism defended earlier. It only works because AoM has already denied that there’s any fixed, agent-independent ought-fact. Grant the moral realist his mind-independent moral order and the relocation has nothing to recommend it — you’d just derive your oughts straight from that order. So this essay isn’t a free-standing solution to the is–ought problem; it inherits the whole burden of the process metaethics that Foundations carries. Said plainly: AoM meets Hume only for those who’ve already followed it out of moral realism.
The second: even with no derivation, one bridge premise is still there, and it’s the decisive one — what agents converge on under a widening context is what they ought to do. This premise is constitutive, not inferential: AoM isn’t inferring the ought from the convergence, it’s defining the ought as your standing relative to the convergence. That’s a real commitment with a real downside, and the downside should be named. A realist will say AoM has simply changed the subject — swapped the genuine, binding “ought” for a description of what cooperative valuers tend toward. The reply is that there was never a coherent realist subject there to change: the “genuine binding ought” the realist misses is exactly the fixed moral fact Foundations argued no one can locate or know. That reply is sound, but it’s a reply that has to be made, not assumed; the argument for it is the entire anti-realist case, not anything in this paragraph.
The real threat is Moore’s, not Hume’s
Here’s the distinction the whole defense turns on, and the one careless versions of this argument miss: the objection with real teeth is Moore’s, not Hume’s.
Once normativity is relocated, Hume’s gap stops being a threat — there’s no derivation to fault, because none is attempted. But G.E. Moore’s open-question argument survives the relocation and aims straight at it. Grant everything: grant that agents converge on X as their context widens without limit. Moore’s question — but is X good? — still seems perfectly intelligible. That the question stays askable suggests “good” can’t simply mean “what agents converge on,” because if it did the question would be as empty as “X is what they converge on, but is it what they converge on?” And the worry isn’t abstract: history is full of agents converging on stable, coordinated, durable arrangements that were monstrous. Convergence by itself isn’t the same as goodness, and a theory that says it is hasn’t answered Moore — it has ignored him.
So the real work isn’t getting past Hume. It’s meeting Moore: saying why this convergence, under these conditions, is the thing “ought” names — in a way that doesn’t just stipulate the answer.
The answer: a structural criterion and the counter-dynamic
AoM’s answer to Moore is the counter-dynamic, and it’s what lets the framework put normativity in convergence without blessing every awful consensus history has ever produced. Not all agreement counts. Agreement reached by widening the context — letting in more of reality, more affected perspectives — is the convergence “ought” tracks. Agreement reached by narrowing the context — shutting out the inconvenient fact, silencing the dissenting party, hardening the doctrine against doubt — is the counter-dynamic, the formal mark of moral regress. The monstrous consensuses are, without exception, the second kind: coherent because they’ve shrunk what they’ll look at. Moore’s awful-but-stable arrangements aren’t counterexamples to the criterion; they’re exactly what the criterion exists to disqualify.
But this answer walks right up to a circularity, and the essay has to walk through it rather than around it. If “genuine” widening just means “the morally good kind of widening,” then the counter-dynamic has quietly smuggled a prior standard of goodness back in, and the relocation fails — it would be defining the good in terms of the good. So the single most important requirement in the whole is–ought response is this: context-expansion must be defined structurally, never morally. Expansion is more of reality taken in, more perspectives and agents integrated, models selected for viability — for what actually works — with perceived consequences fed back and the models revised. Every term there is structural; not one of them is a value word. Hand the criterion to an observer who’d never heard of morality and they could apply it — counting what a system lets in and what it shuts out, what it integrates and what it suppresses — without ever once consulting a notion of the good. Because the definition of expansion contains no “ought,” the criterion resting on it doesn’t beg the question against Moore. That one discipline — keep “expansion” value-free — is what the entire relocation stands on, and every statement of the criterion anywhere in the book has to be audited for a value word that’s slipped in unnoticed.
One term in that definition is the obvious place a critic will pry, and it’s worth tightening before they do. Viability doesn’t mean mere persistence. A regime that lasts for centuries by walling out dissent is durable, but it isn’t viable in the sense the criterion uses, because viability is conjunctive with the rest — it names the persistence of a system that keeps integrating feedback from the affected, not the survival of one that endures by refusing it. Durability won by narrowing is exactly what the counter-dynamic disqualifies: a stable tyranny registers as contraction, not expansion, on the structural criterion alone, before any value word is even spoken.
When the agent is a group
The most elegant part of the account, and the most dangerous, is what happens when the agents in question aren’t individuals but groups.
Treat the move from individual to collective agency as itself an increase in context — which it is, since a “we” integrates more perspectives than any single member — and something striking follows: context-expansion and agreement-convergence turn out to be two descriptions of one process. As individual perspectives integrate into a coherent collective agent, the context widens and agreement rises, together, by construction. The tree of agreement, the levels of agency, and the relocation of “ought” turn out to be one mechanism seen from three angles. That’s a genuine unification, not a restatement.
The danger sits in the very same place. A group is the prime habitat of the counter-dynamic, because there’s an easy way to raise a group’s apparent agreement: suppress the members who disagree. Coerced consensus, groupthink, the mob, Rousseau’s “general will” that overrides the actual wills of actual people — each is a rise in group-level unanimity that’s really a narrowing of context, agreement produced by exclusion rather than by integration. So the unification only holds under a sharp distinction: the agreement of the members isn’t the same as the unified stance of the group as an agent, and AoM needs the former. A consensus reached by throwing the dissenters out of the room isn’t the convergence “ought” tracks; it’s the counter-dynamic wearing the mask of agreement. The counter-dynamic is precisely the test that tells the two apart — which collective agreements widen the context and which collapse it — and the group-agency move has to be fastened to that test at every step, or it becomes the most respectable-looking way for the whole framework to go wrong.
The three commitments
Put it all together and the response is three commitments, worth setting out in order, because each one disarms a different objection.
First, a constitutive bridge, owned as such: “ought” is constituted by convergence under a widening context — a definition of what normativity is, not a derivation of it from facts. This is what meets Hume; there’s no illicit inference, because there’s no inference.
Second, a structural, value-free definition of expansion: integration, inclusion, viability, feedback — no value term anywhere in it. This is what meets Moore; the criterion can say why some convergence counts and some doesn’t without appealing to a prior good, so the open question doesn’t reopen.
Third, the counter-dynamic as gatekeeper: only agreement under genuine, structurally-defined, non-narrowing expansion counts. This is what disqualifies mere agreement, coerced consensus, and the stable monstrosities — the cases that sink cruder convergence theories.
Meet Hume by not needing him; meet Moore by the counter-dynamic and the structural criterion together; and do it without pretending the anti-realist groundwork was skipped. That’s the whole of the relocation.
What no proof can do
One demand is left that the relocation can’t meet — and the first thing to see is that no ethics can meet it either. Even fully assembled, the relocation is a principled account of what “ought” names — not a proof that compels someone who doesn’t already care. Picture the agent comfortable in a narrow context: viable, coherent inside his walls, uninterested in widening them. AoM can tell him that what he’s doing is the counter-dynamic, that his coherence is the kind it calls regress, that he’s set himself against the direction the word “moral” points in. What it can’t do is derive, from premises he already accepts, a contradiction that forces him to care about the wider context. Its answer to why ought I widen? is constitutive — to be an agent at all is to construct and answer for value, and to defy the convergence is to saw at the coherence that lets you go on being one — but that answer works by showing him what he is, not by cornering him on pain of logic.
There’s more to say to him than that, though — not a proof, but a consequence he may not have reckoned with. To wall off his context isn’t a neutral choice with only morality on the other side of it; it’s a structurally self-limiting one. The generativity an agent can reach — new capacities, new solutions, the synergy that only ever arises from combining perspectives and powers — is exactly what a narrowed context forecloses. The agent who refuses to widen isn’t merely turning down a moral invitation; he’s choosing a more fragile, less capable, less generative existence, closing off the very openness that new value comes from. AoM can’t compel him by logic. But it can show him, in terms he already cares about — capability, resilience, reach — that he’s chosen the weaker strategy. That’s persuasion by consequence, not by proof; and for someone who’s decided not to care about widening, it’s the strongest thing any honest ethics can offer.
This isn’t a weakness peculiar to AoM. No naturalist or constructivist ethics compels the determined egoist, and the theories that claim to do so have done it by quietly slipping the very ought-fact AoM declines to posit back in. AoM sets the “ought” on ground anyone can inspect — and accepts that standing on that ground, rather than in the narrow comfort of a shrunk world, is finally something an agent does, not something it can be argued into against its will.
Sources & further reading
This essay engages its literature directly rather than through the book’s per-chapter end notes.
The gap and the open question. David Hume, A Treatise of Human Nature, Book III (the is–ought passage); G.E. Moore, Principia Ethica (the open-question argument).
Convergence and discourse kin. Jürgen Habermas on discourse ethics and the ideal speech situation; T.M. Scanlon, What We Owe to Each Other; C.S. Peirce on convergence in the limit of inquiry; Robert Aumann, “Agreeing to Disagree” (1976) — the formal backbone.
The constructivist core. Sharon Street, “A Darwinian Dilemma for Realist Theories of Value”; Christine Korsgaard, Self-Constitution — both defended at length in the Foundations essay.
Within AoM. The convergence mechanism in the tree-of-agreement note; and agent-relative standing, which this relocation underwrites.