More Than Its Notes

Before anyone can read a group’s values, or decide anything with them, we have to say what kind of thing they even are. That’s all this essay does — and for most of its length, it says what they’re not.

The question, and why it isn’t a poll — or an average

Ask what a community values and the mind reaches almost at once for a survey. Poll the members, add up the answers, read off the result: the town values safety at seventy-one percent, tradition at fifty-eight. It’s a natural picture, and it’s the wrong one. A poll can tell you what a bunch of people, asked one at a time, will say about themselves on a Tuesday. It can’t tell you what the we they belong to values — because the we isn’t the list of its members, and its values aren’t the average of theirs.

So this essay makes one claim about what a group’s values are, before we get to how you’d find them or decide with them: a group’s values are a real, structured, living thing — the values-model of a self made of selves — and they are not an average. Get that wrong and everything after it goes wrong too. You’ll try to read a we’s values off some surface, and miss them; you’ll try to decide with them by counting votes, and misfire. This essay just describes what kind of thing they are — and why no surface gives them away; how a we knows its world, and how it decides, are the two essays after this one.

Not an average

Start with the negative claim, because it’s the one the rest leans on, and because you can show it instead of just asserting it.

Take a hung jury. As a body it has a perfectly clear character: it’s indecisive, deadlocked, can’t reach a verdict. Now look inside it. A hung jury is almost never made of indecisive people — it’s usually made of the most decisive people you could find, each so sure of a position that clashes with the others that nobody will budge. The group’s trait — indecisive — isn’t the average of its members’ trait — decisive. It’s something new, made by how those decided people are set with and against one another. Sociologists have a name for the mistake of reading the members off the group, or the group off the members: the ecological fallacy. It’s not a small statistical footnote. Average twelve decisive jurors and you conjure up a mildly decisive “mean juror” who isn’t in the room and never was.

The same lesson comes from cross-cultural psychology, from the other direction. Shalom Schwartz spent decades mapping how one person’s values are arranged, and found a surprisingly steady shape: a circle, where prizing novelty sits across from prizing safety, and looking out for yourself sits across from looking out for others — so leaning hard into one tends to cost you the one across the ring. That’s a real finding about people. But when researchers take the same data up to the level of whole societies, that personal circle doesn’t just scale up. It breaks, and a society’s values rearrange onto different axes entirely — a whole culture’s pull between fitting in and standing apart, between rank and equality, isn’t any one person’s inner circle blown up large. If the group’s shape had been the average of its members’, it would have kept their shape and just smoothed it out. It doesn’t keep their shape. It has its own.

So a we’s values aren’t its members’ values blown up large, or averaged down small. They’re the values of a different thing.

Emergent, not summed

If not an average, then what? Here the essay leans on ground the book already laid, and doesn’t lay it again. Chapter 3 built the values-model — the layered, fine-grained working picture of what matters that any valuer carries around. Chapter 5 argued that selves are made of selves: that a person is already a huge collective of cells that coordinated until a new agent woke up over them. A we is that same move, run one more time. Members coordinate — come to share enough of a values-model and a methods-model to act together — and a new valuer wakes up at the larger scale, with a model of its own.

The word for that waking-up is emergence, and it’s worth being careful about which kind we mean, because the sociologists have argued about it. One reading (Keith Sawyer’s) keeps only individuals real, and treats a group’s traits as a kind of shadow the people cast — you can’t reduce them in practice, but there’s nothing there over and above the people. Another (Dave Elder-Vass’s) says a whole made of parts arranged just so genuinely has powers the parts, scattered or rearranged, would not — that emergence isn’t our failure to do the arithmetic but a real feature of organized stuff. AoM goes with the second. A we’s values are a real capacity of members-in-relation, not just a handy name for what its members each happen to hold; and what makes the difference is the arrangement — which is also why AoM’s functionalism, its interest in what a thing does more than what it’s made of, is no problem here. The role is real. It’s the structure that unlocks it.

The plainest name for what emergence buys you is one the book has used before: putting things together opens the adjacent possible — the next thing that only becomes reachable once you’ve combined what you have. Ten people standing near each other are ten people. Ten people who coordinate can hold a value none of them could hold alone — can care about, and act on, the fate of a watershed, a language, a hundred-year project — because caring on that scale needs a memory and a reach no single body has. The we doesn’t average its members down to what they share. It builds them up into something none of them was.

Here’s the essay’s one picture, and it isn’t the average. A chord isn’t the average of its notes. Play a C, an E, and a G together and you don’t hear the pitch halfway between them; you hear a major chord — a whole, settled sound that not one of the three notes has on its own. That’s emergence, going up. But a chord does a second thing, and it’s what the next section is about: once the chord is sounding, you stop hearing that C as a bare C. You hear it as the root of this chord — its meaning bent by the whole it’s joined.

Downward causation

That second thing the chord does is the one people forget, and it’s the one this essay puts at its center. A we doesn’t only emerge from its members. It reaches back down and shapes them.

The how isn’t mysterious; Elder-Vass calls it the norm circle. A group that shares even a faint sense of standing behind some value — this is how we treat each other here; this is what we don’t do — leans on each member to fall in line, and that pressure works in two steps. First it becomes expectation: the member comes to know, below the level of deciding, what’s done and what gets answered for. Then that taken-in expectation becomes action. By the time the member “freely” keeps the norm, the freedom is real — but its shape was laid down by the group long before the moment of choosing. A we’s values aren’t filed away in a cabinet somewhere over its members’ heads; they live in the members, as leanings the we pressed into them. A group’s values are causally real for exactly this reason: they’re the kind of thing that can remake the people it’s made of.

This is a real power, and — like every real power in this book — it’s the same power that can go wrong. Downward causation is how a we coordinates: how a scattered crowd becomes a body that can act, how a rule nobody could enforce alone comes to hold. It’s also, turned the wrong way, the exact machinery of the counter-dynamic. A we can earn its coherence honestly, by widening until more of its members’ reality fits inside — or it can buy the same coherence on the cheap, by narrowing: pressing its members into agreement by walling off the doubter, the awkward fact, the out-group. From the inside, the two feel identical. The only difference is whether the circle is opening or closing — which is the whole difference between a culture and a cult. A later essay is spent on that fork. Here it’s enough to see that the power to make a we can’t be separated from the power to deform one, because they’re one power.

Made in the asking

If a we’s values are this kind of thing — emergent, structured, held in the arrangement — then the usual ways of reading them all fail the same way: they take a surface for the depth. Polling fails first, and we’ve seen why; it assumes the average. But the sneakier failures look like progress. Read a we’s values off its behavior — what it buys, clicks, and votes for, the revealed preferences an economist or a recommendation engine would tally — and you fold its lasting commitments in with its passing appetites and its worst Tuesdays, then call the mix its values; a we is no more the sum of its purchases than a person is. Read them off a stated creed instead — the mission statement, the list of nice words, be helpful, honest, and kind — and you get something slippery and endlessly deniable, a text each faction reads to mean whatever it already wanted. A value is less a slogan a we recites than a way it’s learned to pay attention — what it notices, what it weighs, what it won’t give up — and no headcount of preferences or principles holds that shape.

There’s a deeper reason the surfaces come up empty, and it’s the one that changes how the reading has to be done. A we’s values aren’t sitting finished inside it, waiting to be dug out; to a real degree they’re made in the asking. When a community actually sits with a hard question — argues it, tests one caring against another, listens — it doesn’t so much report values it already had as form them: the passing preference gets pressed against others and either hardens into something shared or falls away. Drawing the values out isn’t extraction, it’s construction — which is why it does the very thing the last section described, and remakes the people doing it, moving them from saying what I want to working out what we should. So the tools that will do this reading at scale — the survey, the citizens’ assembly, and now the AI mediators built to boil a million opinions down to one — are never neutral windows onto something that was already there. They’re part of the making. The good ones map the real, many-voiced shape and keep the minority in view; the bad ones optimize for agreement and manufacture a consensus nobody holds — the counter-dynamic again, wearing the face of a friendly summary. Which kind we build is the work of the essays ahead. Here it’s enough to see that a we’s values can never simply be found, because in the looking we’re always, in part, making.

Rooted, and converging — not fragmenting

A fair worry runs the other way now. If a we’s values are their own real thing, and members differ — and they always differ — doesn’t the we just shatter into as many models as it has people?

It doesn’t, for the reason that keeps the whole framework from flying apart: widening turns up shared ground. Members argue most, and most loudly, at the surface — over the policy, the doctrine, the case in front of them. Follow those arguments down and they thin toward things almost no member turns down: that suffering counts, that the group should be able to go on, that a promise is worth keeping. The Tree of Agreement essay makes that case in full; here it does one job. It’s why a we’s values can be one thing without being a same-for-everyone thing. The agreement lives at the center, where the shared roots are; the variety lives at the edges, where members try out what the center hasn’t settled. A we held together this way isn’t fragile for holding disagreement. The disagreement is its search space — the passing variety a monoculture throws away, and then, every time, learns the worse for having thrown away.

The honest edges

Three cautions, because the thing is real but it isn’t tidy.

First, the ecological fallacy cuts both ways, and the second way is the more dangerous. It’s just as wrong to read a member off the group — she’s from that town, so she must value this — as to read the group off its members, and the first mistake is a working definition of prejudice. A we’s values are real; they’re not a license to assume any particular member carries them.

Second, because a we’s values live in its members through that downward pressure, the pressure can be engineered. The same channel that lets a community teach its kids what matters lets a platform, a demagogue, or an algorithm manufacture a “shared value” that serves the engineer and not the we — Tinbergen’s supernormal stimulus, scaled up to a whole population. The recent capture of family health decisions by live-shopping platforms, quietly sliding themselves into the trusted spot a household’s own judgment used to hold, is a small clean case of downward causation redirected by an outside hand. That a we’s values are built doesn’t make them arbitrary — but it does make them attackable, which is one more reason the how-to essays after this one matter.

Third — planted here, and left for later — even a we whose members are each perfectly reasonable can produce a group judgment that no member holds. List and Pettit’s discursive dilemma is the sharp version: add up a rational group’s votes one question at a time, and you can land on a group position every single member would reject. It’s the deepest sign yet that a we’s stance isn’t its members’ stances added together — and it’s the doorway into the hardest problem of all, how a we should actually decide. That door is a later essay’s to open.

What a we values

Pull it together. A we’s values are real, not just a way of talking; structured, not a flat list; emergent, not added up; downward-shaping, not inert; rooted-and-converging, not shattered; and, being a values-model like any other, never finished — a model still being corrected by what happens, or else a we that’s stopped listening and started to narrow. That’s the kind of thing a group’s values are.

Which is exactly why the work that comes next is hard in the particular ways it is. You can’t read a thing like this off any surface — a section here has said why, and why the honest reading is itself a kind of making. And you can’t decide with a thing like this by counting votes, because counting runs straight into the dilemma we just planted; deciding well together is the craft of a later essay. This essay has done only the first, most easily skipped thing — to say, before the how-to arrives, that there’s really something there: a valuer, not an average, and worth the trouble of reading right.

Sources & further reading

This essay leans on its sources directly, rather than through the book’s per-chapter endnotes. A citation-level pass is still owed.

Not an average (the two demonstrations). The ecological fallacy and the hung-jury illustration — the classic statements in the levels-of-analysis literature (Zito; W. S. Robinson), with Daphna Oyserman and Ayşe Uskul on societal- versus individual-level processes. The individual-versus-cultural value structure: Shalom Schwartz, “Universals in the Content and Structure of Values” (1992), and his culture-level theory of values, with the cross-cultural work showing the individual circle doesn’t survive the lift to the collective level.

Emergence and downward causation. Émile Durkheim on social facts as emergent realities with causal power over individuals; R. Keith Sawyer, Social Emergence: Societies as Complex Systems (2005), for non-reductive individualism; Dave Elder-Vass, The Causal Power of Social Structures (2010), for causal-powers holism and the norm circle.

Group agency and the discursive dilemma. Christian List and Philip Pettit, Group Agency: The Possibility, Design, and Status of Corporate Agents (2011), and their “Group Agency and Supervenience.” (The dilemma is planted here and taken up in the collective-decision essay.)

Reading a we’s values — why the surfaces fail, and the making. On deliberation as forming rather than mining shared values: Jasper Kenter and colleagues on deliberative valuation. On values as thick, structured attentional policies rather than thin preferences or slogans: the Meaning Alignment Institute (Joe Edelman, Oliver Klingefjord). On AI-mediated elicitation and its double edge — proportionally mapping a population’s views versus manufacturing a false consensus: generative social choice (Ariel Procaccia and colleagues) and the “Habermas Machine” (Google DeepMind), with Isaiah Berlin’s value pluralism as the warning against optimizing deliberation for agreement.

Selves made of selves (the biology). Michael Levin’s work on how coordinating cells compose agents at larger scales — his cognitive light cone, the reach in space and time of the largest goal a system can pursue; read alongside Chapter 5 — Selves Made of Selves. AoM says the same reach in its own words — the widening context and the reach of concern.

Engineered values. Niko Tinbergen, The Study of Instinct (1951), on supernormal stimuli; and the socio-technical literature on algorithmic capture of household decision-making.

Within the framework: Chapter 3 — What Matters (the values-model this essay doesn’t re-teach); Chapter 5 — Selves Made of Selves (the emergent we); The Tree of Agreement (why we converge below doctrine); Held Together, Held Apart and Coherent Pluralism (the counter-dynamic, and coherence at the center with room at the edges); and the companion essays to come — Farther Than One Can See (how a we knows its world) and The Course We Carve (how it decides).