The Tree of Agreement
The is–ought relocation rested its whole weight on a single verb — converge. An “ought,” that essay argued, is what valuing agents converge on as their context widens. But convergence cannot simply be assumed; it has to be explained — why agents who start far apart should come together at all, rather than drift forever apart. That explanation is the tree of agreement, and it is the most exposed claim in the book. The other foundations are interpretations; this one makes a prediction, and a prediction can be wrong. So the essay runs in two registers at once: a defense of the tree, and an honest test of it — which means stating plainly what it promises and what it does not.
The picture, and what it is
Picture the disagreements of moral life as the canopy of a tree. At the tips — the leaves — agents are far apart: this culture eats the animal, that one venerates it; this faction calls the tax just, that one calls it theft. These leaf-level quarrels are structurally the shallowest part of the tree, yet they are the ones lived most viscerally — which is exactly why moral disagreement feels total when it is not: the fights we feel hardest are the ones furthest from the root. Trace any leaf inward and it joins a twig; the twig joins a branch; the branch, a limb. The claim of the tree is that as you trace inward — from applied judgments toward the values beneath them, and from those toward the conditions of being any valuing agent at all — the branches thicken and converge: more agents share each deeper junction, until at the root they share the most general thing of all, the bare fact of being living, embodied, mutually dependent creatures modeling one world.
The reframe that matters most comes first, because without it the tree is only a picture. In AoM, seeking agreement is not a separate activity from increasing coherence — it is the same activity, seen between agents instead of within one. To widen an agent’s context is to take in more of reality and more perspectives; to reach agreement with another is to integrate their perspective into a shared one. These are one motion. So the tree is not an ornament hung beside the master thesis; it is that thesis projected onto the social — coherence over a widening context, drawn between people rather than inside a single mind. That identity is what the rest of the essay has to protect, because every serious objection to the tree is at bottom an objection to whether the projection holds.
The parts are well-precedented
Almost every component of the tree is already established, and saying so is part of the defense: the picture is assembled from well-tested materials, not conjured.
That values are hierarchical and structured — a fine-grained order of what matters rather than a flat list — is mapped from several directions: Schwartz’s cross-culturally validated circle of basic values, Rokeach’s value survey, Maslow’s hierarchy, and Jonathan Haidt’s Moral Foundations Theory. That the structure is shared at depth and variable at the surface is Haidt’s explicit formula — the foundations “universal in their intension, culturally determined in their extensional occurrences” — and it has deep empirical company in Donald Brown’s Human Universals and Frans de Waal’s work on the evolutionary roots of moral sentiment. And the engine of convergence is older still: Robert Aumann’s agreement theorem — that rational agents with common priors and common knowledge cannot agree to disagree — is the formal backbone; Peirce’s convergence of inquiry in the limit and Habermas’s rational consensus under ideal speech are its philosophical kin; Yudkowsky’s coherent extrapolated volition is its recent decision-theoretic cousin.
None of these is AoM’s to claim. What AoM does is fuse them into one structure — which is where both its contribution and its exposure lie.
The paradigm case: the Golden Rule
No abstraction makes the tree as vivid as its most famous instance — and a reader will have been thinking it for pages: this sounds a great deal like the Golden Rule. It does, and the resemblance is not an embarrassment but the best single piece of evidence the tree has. Reciprocity in its golden form — do not do to others what you would not have done to you; do as you would be done by — was set down independently by Confucius, by Hillel, by Jesus, by the authors of the Mahabharata, by traditions whose doctrines agree on almost nothing else and never met to compare notes. That so many unrelated peoples should arrive at the same rule is the tree’s prediction made flesh: distinct constructions of one kind of creature, converging below the level where their doctrines fight, on the recognition that the self across the table is the same kind of thing as the self in the chair. The Golden Rule is re-convergence below doctrine, caught in a single sentence.
So the framework does not argue against the Golden Rule; it claims it — and then does the thing the maxim alone cannot, which is to repair and extend it. Three moves separate the framework from the wise old rule it inherits.
First, it is a compass, not a code. As a direction — stand where the other stands; let your circle take them in — the Golden Rule is the whole framework in a breath. As a fixed rule cranked for verdicts it seizes up, in just the way Shaw enjoyed: do not do unto others as you would that they should do unto you — their tastes may not be the same. The framework keeps the compass and lets the code go.
Second, it repairs the projection. Shaw’s needle finds the real flaw — the rule sends you to consult your own wants and visit them on someone who may not share them. Perspectival realism forbids exactly that: you do not get to treat the other as a mirror; widening the context means taking in their world and what they value, not stamping yours onto them. The cleanest case is the one every parent knows. A small child wants sweets for dinner and no bedtime. The Golden Rule read straight is no help — you are not a child, and would not want a child’s terms. The Platinum Rule, treat others as they want to be treated, is worse than no help, since it hands the evening to the four-year-old. What actually guides the parent is neither projection nor deference but a coherence wide enough to hold both the child’s present and the self the child is becoming — the framework’s move exactly, and one the Golden Rule cannot reach on its own.
Third, it widens without flattening. This is the reading a thoughtful reader most often arrives at, and most often carries one step too far: that as morality matures, the “others” and the “you” of the rule expand until the out-group is folded wholly into an ever-larger “us.” The widening is real — it is the whole arrow — but the folding is not total, and must not be. What widens is the circle of standing and concern, not a demand that everyone become the same; a gradient of care stays legitimate, and difference is protected rather than dissolved — the coherent pluralism of Part Two once more. “Others become us” is right if it means their standing enters our context; it is the wrong turn if it means their difference disappears, which would be coherence bought by narrowing — the counter-dynamic in the Golden Rule’s own robes.
Seen this way, the Golden Rule is the tree’s root glimpsed from inside a single relationship: the near-universal evidence that convergence below doctrine is real, and, taken as a compass rather than a code, the first rough draft of the framework itself. What the book adds is the reading that keeps the draft from breaking — whom to include, whose wants to consult, and how far to widen the circle without erasing what is inside it.
Not Parfit’s mountain
The convergence intuition has a famous modern champion, and the tree has to be set carefully against it. In On What Matters, Derek Parfit argued that the best versions of Kantian ethics, contractualism, and rule-consequentialism converge on the same verdicts — “climbing the same mountain on different sides.” It is the premier convergence thesis in recent philosophy, and a reader will reasonably ask whether the tree is just Parfit in arboreal dress.
It is not, and the difference is the whole metaphysics. Parfit was a realist: his three traditions converge because there is an objective moral truth at the summit that each is ascending toward. The tree has no summit and no truth at its center. What sits at the root is not a moral fact that agreement discovers but a shared condition — the common evolutionary and physical inheritance that gives distinct agents enough in common to converge at all. Parfit’s agents converge on what is true; AoM’s converge from a shared origin. Same upward intuition, opposite explanation: Parfit climbs toward a fact, the tree traces back toward a root. Keeping that straight is what lets AoM use the convergence intuition without taking on the realism it usually travels with.
The objection that runs the other way
The serious threat to the tree is not skepticism about convergence in general; it is a body of work showing that agreement, in practice, is found in the opposite place from where the tree puts it.
John Rawls’s overlapping consensus is the central case. A stable, just society, Rawls argued, does not rest on citizens sharing a comprehensive doctrine — a deep account of the good and the true. It rests on their agreeing to shared political principles for their own, divergent, deeper reasons: the Catholic, the secular liberal, and the Kantian endorse the same constitution from different foundations. Cass Sunstein sharpened the practical version into “incompletely theorized agreements”: people who cannot agree on the deep questions can often agree on a particular outcome, or a mid-level rule, precisely by bracketing the depths — “go shallow,” he advises, “when parties cannot agree deep.”
The challenge this poses is exact and severe. Both Rawls and Sunstein locate workable agreement at the shallow end — outcomes, mid-level principles — and treat deep agreement, on comprehensive worldviews, as the hardest thing of all. That is the precise inverse of the tree’s claim that agreement increases as you go deeper. If they are right, the tree is upside down.
Re-convergence below doctrine
The answer is not to deny Rawls and Sunstein — their observation is correct — but to see exactly which level it is about, and to give up a claim the tree never needed: monotonicity.
The tempting reading of the tree is that agreement rises steadily the deeper you go — more at every step inward, right to the root. That reading is false, and AoM should not defend it. What Rawls, Sunstein, and Haidt have located is real, and it is deep-ish: the fiercest disagreements live at the level of comprehensive doctrines — religions, ideologies, worldviews — and at the level of how people weight the moral foundations (Haidt’s finding that liberals and conservatives differ not in which foundations they have but in how much each one counts). That is not a leaf-level quarrel over a single tax; it is structural, and it sits well down the trunk.
But it is not the root. AoM’s root is deeper than doctrine — below culture and ideology entirely, at the shared evolutionary and physical substrate: that we are mortal, embodied, vulnerable, mutually dependent creatures who must model one world in order to act in it. The honest shape of the agreement curve, then, is not a steady rise but a valley: high agreement at the physical root (no human tradition disputes that pain is to be reckoned with, or that the future matters to the living), low agreement through the cultural-doctrinal middle where worldviews clash hardest, and low again out at the applied leaves. The tree’s real claim is not “more agreement right to the root.” It is re-convergence below doctrine — that beneath the level where worldviews fight, there is a deeper layer on which they rest on common ground again, because they are the constructions of the same kind of creature in the same world.
This concedes everything Rawls and Sunstein actually showed — deep-doctrine disagreement is real, and it peaks — while preserving what the tree needs: a root below the doctrines, where convergence returns. It is also the most exposed of AoM’s original moves, and is best flagged as such. The claim that the curve turns back upward below doctrine is a prediction, and whether the empirical record bears it out is exactly the kind of question that could confirm, qualify, or break it.
A root that is not bedrock
A second objection is quieter but cuts deeper, because it comes from inside AoM. To speak of values “rooted in reality” sounds foundationalist — as if there were a privileged base of fixed truths underwriting everything above. But AoM is anti-foundationalist nearly everywhere else: perspectival realism denies any view from nowhere, and the framework is coherentist about how values hold together. A foundationalist root would contradict all of that.
The resolution is to be exact about what sits at the root. It is not a layer of fixed moral truths, a bedrock of oughts that agreement uncovers. It is the shared causal, physical, and evolutionary substrate — and its role is the role Aumann gives a common prior: not a conclusion agents reach but a starting point they already hold in common, which is what makes their convergence possible at all. Two agents with no common prior have no reason ever to converge; two agents who are both mortal, embodied creatures modeling one world share an enormous prior whether they like it or not. One honest qualification belongs here: Aumann’s theorem is strictly about agents converging on beliefs under shared information, not on values, so applying it to value-convergence leans on a claim made earlier — that valuing is the same kind of situated modeling as knowing, perspectival realism carried from fact to value. Grant that, and beliefs and values are near enough the same sort of thing for the common-prior logic to carry; withhold it, and Aumann is a suggestive analogy rather than a structural backbone. The root, so understood, posits no moral bedrock and grants no privileged base of value. It only names what distinct perspectives have in common as perspectives on one reality — which is perspectival realism’s own claim, now doing social work. The branches stay coherentist; the root is a shared origin, not a foundation.
Why the root bends toward agreement, not war
A Hobbesian will press here, and the pressure is fair: shared vulnerability, by itself, does not bend toward agreement. Two starving animals share the whole biological prior — the need for food, the aversion to pain — and it sets them fighting over the carcass, not converging on a value. If the root were only shared exposure, the cynic would be right that it predicts war as readily as peace.
What gives the root its convergent pull is a second thing the prior contains, and it is structural, not sentimental: over a widening context — more agents, repeated encounters, longer horizons — coordination produces synergy, a surplus that zero-sum conflict simply forgoes. Cooperation is not a moral preference draped over biology; past a certain scale it is the stronger survival strategy, because two agents who integrate can do what neither can do alone. The major transitions of evolution are this fact made visible — genes into cells, cells into organisms, organisms into societies — each a wider coordination reaching something the smaller scale could not. So the root is not bare vulnerability; it is vulnerability in a world where widening coordination is generative, and that is what tilts convergence toward agreement rather than toward Hobbes’s war.
Two honesties keep this from overclaiming. The tilt is a tendency under widening, not an automatic law: in a narrow enough frame — two animals, one carcass, no tomorrow — competition still wins, and that narrow frame is exactly where the counter-dynamic and the multipolar trap live, which is why so much of the book’s practical half is spent on escaping them. And the surplus is available, not guaranteed to be seized; the arrow is the drive that reaches for it, not a promise that it is always reached. Evolution does not hand us agreement. It makes agreement the stronger strategy as the context widens — and leaves the widening to us.
What the tree promises, and what it does not
Every convergence result the tree draws on holds only under heavy idealization. Aumann’s theorem needs common priors and common knowledge — agents who fully share what they know. Habermas’s consensus needs an ideal speech situation free of distortion and power. Peirce’s truth waits at the limit of inquiry. Yudkowsky’s volition is extrapolated, idealized past anything an actual person holds. Real agents have none of these in full: priors diverge, knowledge is private and partial, speech is bent by power, and no inquiry reaches its limit.
So the tree has to promise carefully. It does not promise that agents will agree, or that convergence is guaranteed, or that a single consensus waits at the bottom for everyone. It promises a tendency — that under widening context, with more reality admitted and more perspectives integrated, there is a real and rising probability of significant congruence, because the agents share so much at the root. A pull, not a destination; the direction agreement tends to move when the context genuinely widens, not a promise that it arrives. Stated at full strength the tree would be false; stated as a tendency disciplined by the counter-dynamic, it is the honest and defensible claim — and it is the one the is–ought relocation actually needs, which asked only that convergence be the thing “ought” tracks, not that it be everywhere achieved.
What is genuinely the tree’s own
With the parts acknowledged and the objections met, the contribution can be stated at its true size — smaller than “a new theory of moral agreement,” realer than “a metaphor.”
Three things are AoM’s. The first is the unified structure: one figure fusing the hierarchy of values, the phylogenetic and physical rooting, the convergence-toward-the-root account of agreement, and the Aumann backbone into a single object, where before there were four separate literatures. The second is the probabilistic gradient — reading agreement as tracking depth-and-probability in a physically grounded tree, so that “deeper” means both “more general”1 and “more probably shared,” a quantitative-flavored framing the component traditions do not state. The third is the identity already named: agreement as the interpersonal face of coherence over a widening context, which is what turns the tree from an assembled picture into a structural part of one framework. The parts are precedented; the integration, the gradient, and the re-convergence-below-doctrine curve are the original — and exposed — contribution.
A falsifiable claim, and a guarded core
It is worth ending where the prior-art work began: this is the framework’s most testable claim — and being testable is a virtue, not a vulnerability: it means the framework says something the world could actually check. The re-convergence curve is a prediction about how human — and other — agents actually behave as their contexts widen, and predictions answer to evidence. It is entirely possible that the disagreement literature, examined closely, shows the valley never turns back upward — that some value-conflicts are irreducible all the way to the root. AoM wagers that the shared substrate is deep enough, often enough, that re-convergence below doctrine is real; it does not pretend that is a proof.
Two things keep the claim from being reckless. It promises only a tendency, never a guaranteed consensus, so it is not refuted by the mere persistence of disagreement — only by the absence of the deep pull. And it has the counter-dynamic as its discipline: where convergence is reached by narrowing rather than by widening — by silencing a branch rather than integrating it — AoM does not count it as convergence at all, which sets aside exactly the false agreements that would otherwise inflate the claim.
It also helps to see how much actually rides on the wager, because the objection assumes the answer is everything, and it is not. The demand that convergence be total — that every conflict resolve at the root or the framework fail — is the view from nowhere in a critic’s coat: it asks a perspectival, process ethics to deliver the one thing perspectivalism denies, a guaranteed terminus true for all from no standpoint at all. And a theory that did promise it would be the worse for the promise, because when convergence failed it would have to suppress the holdouts to keep its word — Berlin’s totalitarian temptation exactly, and the counter-dynamic by name. So the framework neither needs nor wants completeness: where conflicts do not converge they are held as coherent pluralism, distinct and viable nodes rather than one enforced peak, and the decisive verdicts do not ride on the bet at all — the counter-dynamic condemns narrowing whether or not the branches ever merge, and the equal standing of persons rests on that structural condemnation, not on the empirical tendency. What a lost wager would cost is only the strongest form of the claim — that the valley always turns back up — not the framework around it. And that the claim can be lost to evidence at all is a virtue, not a flaw: it is one of the few moral theories that says something the world could disconfirm, and points at where to check.
What remains is a bounded, falsifiable, probabilistic claim: that creatures rooted in one reality, widening their context in good faith, tend toward common ground. The book stands on that tendency — and, unlike most of its rivals, says something the world could check.
Sources & further reading
This essay engages its literature directly rather than through the book’s per-chapter end notes. A citation-level pass is still owed (see the note above).
Value structure and shared roots. Shalom Schwartz on basic human values; Jonathan Haidt, Moral Foundations Theory; Donald Brown, Human Universals; Frans de Waal on the evolutionary roots of moral sentiment.
The Golden Rule. The cross-cultural reciprocity maxim (Confucius’ shu; Hillel; the Mahabharata; the Gospels), surveyed in Jeffrey Wattles, The Golden Rule (1996); George Bernard Shaw’s caution that “their tastes may not be the same” (Maxims for Revolutionists, 1903); and the “Platinum Rule” (Tony Alessandra) as the halfway correction.
The convergence engine. Robert Aumann, “Agreeing to Disagree” (1976); C.S. Peirce on convergence in the limit of inquiry; Jürgen Habermas on the ideal speech situation; Eliezer Yudkowsky on coherent extrapolated volition.
The realist contrast. Derek Parfit, On What Matters — the “same mountain.”
The objection that agreement is found shallow. John Rawls on overlapping consensus; Cass Sunstein on incompletely theorized agreements.
Within AoM. The Is–Ought Relocation (which leans on convergence); Foundations (perspectival realism — the one rooted reality that makes re-convergence possible); and coherent pluralism (Part 2), where the tendency-not-guarantee shape matters most: a network of networks rather than one enforced consensus.
General here means widely shared — the common condition more agents already hold — not content-rich. This runs opposite to a value’s composition: the deepest things are the simplest and the most widely held (reciprocity; the bare aversion to pain), while the great composite values built above them (justice, love) are richer in content yet more contested. Depth on this tree measures how commonly held a thing is, not how much it integrates — two different dials that happen to run opposite ways.↩︎