living together · metaphor 61 of 100

The will of the people is missing its verb.

"The will of the people" is a sentence missing its verb: want, under which counting? Kenneth Arrow proved that no method of merging individual rankings into a collective one can satisfy a short list of obviously fair requirements at once. The people's will is not discovered by elections; it is partly constituted by the rules.

A club of nine friends ranks three restaurants. Count first choices only, and Italian wins. Same nine ballots, but eliminate the weakest and re-count — a runoff — and Thai wins. Same ballots again, scored by position the way sports leagues score seasons, and the winner is Sushi — a place almost nobody put first, but nobody put last. No one changed their mind between counts. Only the arithmetic of "we" changed.

The reflex is to ask which count is the right one, the others defective drafts of it. Arrow's theorem, proved in 1951, says the reflex has no object: the good counting, the one meeting every reasonable fairness requirement at once, does not exist. Below: one small electorate, five honest counts, and the places where "we" comes apart in your hands.

The preference lab every ballot editable — click any row to move it up one rank

voters · 9
candidates
skin
Nine ballots, untouched, about to be counted five ways at once.

The counting rack the same ballots through five aggregation rules — live

Conventions, disclosed: approval assumes each voter approves their top two; all ties break toward the earlier-listed candidate, and the runoff eliminates the later-listed on a tie — every method needs some such rule, and that too is a choice. The splitting-profile button runs an honest randomized search over real ballots on this page; nothing is faked.

The missing verb

"The people have spoken" — through which function?

Each ballot above is a complete little fact: one person's actual ordering of the options. The trouble starts at the step everyone treats as clerical — adding them up. Aggregation is an operation, and the operation has choices in it. Count only first loves, or weigh whole rankings? Eliminate losers and re-ask, or compare everyone to everyone? Each choice is defensible. Each produces a different "we" from identical "I"s.

So the phrase the people have spoken is always quietly elliptical. The sentence borrows the grammar of a single speaker for a crowd that has no voice until a rule lends it one. The rack above is that ellipsis, expanded: five lendings, side by side. When they agree, the ellipsis is harmless. When they split, there was never a fact of the matter waiting underneath — the counting was doing constitutional work all along.

The cycle maker

The majority that beats itself.

The deeper cut is that majority rule can disagree with itself. Below, every arrow is a computed head-to-head majority on your current ballots. With ordinary, sincere preferences — three modest blocs, nobody scheming — the arrows can close into a loop: Italian beats Thai, Thai beats Sushi, Sushi beats Italian. Rock, paper, scissors. Whatever wins, a majority of these same people prefers something else.

Pairwise majorities computed from the ballots above — arrows point from winner to loser

Build a cycle from three ordinary blocs

3
3
3

The enabling ballots are unexotic. No bloc is irrational; each has a perfectly coherent ranking. The incoherence lives nowhere in the individuals — it emerges in the sum. This is Condorcet's paradox, the seed Arrow grew into a theorem: transitivity, the minimal sanity of preferring A to C if you prefer A to B and B to C, is a property individuals can have that their majorities can simply lack.

The impossible checklist

Four fair demands. Pick three.

Each of Arrow's conditions is something you'd insist on before trusting any group decision. The theorem: for three or more options, no ranking method satisfies all four. Every rule on the rack is a decision about which one to live without — watch each axiom checked live against the current ballots.

axiom 01 · unanimity

If everyone agrees, the group agrees.

When every single ballot ranks one option above another, the collective verdict must too. The easy axiom — all five methods here honor it. Unanimity only constrains the case where there was nothing to decide.

axiom 02 · independence (IIA)

Irrelevant options shouldn't decide.

The group's choice between two options should depend only on how voters rank those two — not on who else is on the menu. Plurality, runoff, Borda and approval all fail this: the spoiler, formalized.

axiom 03 · a coherent verdict

The group's ranking must be an ordering.

The collective preference should be transitive and complete — no loops, no shrugs. Pairwise majority rule honors the other axioms and fails exactly here: the cycle maker above is its failure mode, computed.

axiom 04 · non-dictatorship

No single voter is the function.

There must be no person whose ballot is the outcome regardless of everyone else's. One rule passes the other three axioms perfectly — install voter ♛01 as dictator. That it survives every other test is the theorem's darkest joke.

methodunanimityiiacoherent ordernon-dictator
plurality
runoff (IRV)
Borda
approval (top-2)
pairwise majority
dictatorship

Every row has its ✗ — Arrow says every row must. (Approval is judged here as a ranking rule under the disclosed top-two convention; ✓ entries hold up to the tie-break fine print.)

What to try

Sixty seconds of constitutional crisis.

  1. Run the rack on the classic. Load the restaurant club: three verdicts from nine ballots. Then click one voter's second choice up to first and watch which verdicts move — and which don't care.
  2. Build a cycle. Load the cycle, or set the three bloc sliders yourself and deal. Watch the arrows close into a loop and the pairwise cell of the rack give up. Unbalance one bloc and watch coherence return.
  3. Do the spoiler flip. Load the spoiler, then switch candidates 4 → 3. Tapas never had a chance of winning — yet deleting it hands plurality from Thai back to Italian. Then switch 3 → 4 on any profile and watch a newcomer sometimes decide a race it isn't in.
  4. Hunt a grand split. Press find a splitting profile and let the randomized search grind through thousands of real electorates until four counting rules crown four different winners — or load the preset one it already found.

What it does — and doesn't — say

No perfect method is not "no difference."

Read precisely, the theorem says only this: no ranking method satisfies unanimity, independence, coherence and non-dictatorship at once, given three or more options. It does not say all methods are equally bad, or that democracy is a sham. Methods differ in which failures they accept, how often, at whose expense — plurality falls to spoilers constantly; Borda is gameable by burying a rival; runoffs can punish a candidate for gaining support; pairwise majority is immaculate until it cycles. Choosing among these failure profiles is substantive politics — a decision about which distortions your community can survive.

The theorem's real target is a manner of speaking: the habit of treating "what the people want" as a preexisting object that elections merely photograph, and any odd result as camera shake. There is no photograph. There is a family of portraits, each honest, none neutral.

Living downstream of the theorem

Whoever picks the counting picks, a little, the winner.

If the outcome depends on the rule, then rule-choosers and agenda-setters hold real power that never appears on a ballot: the chair who decides whether to vote pairwise or all-at-once, the party that designs the primary. Constitutional attention to these meta-rules — who may change the counting, and how hard it is — is where much of the actual sovereignty lives.

Downstream too is a certain humility about the word mandate. A winner under one counting was often a loser under another that no one would call unfair; underdetermination is the normal case, not the scandal. And the theorem scales all the way down to the household: a family "deciding" on dinner by whoever speaks first, or by veto, or by taking turns, is running an aggregation rule with its own failure profile. Most resentment about "we never do what I want" is about the counting — unchosen, unexamined, and somebody's.

The mapping

Mathematics ↔ life.

MathematicsLife
individual rankingsWhat each person actually wants, in order — coherent, sincere, complete.
the aggregation rule fThe constitution, bylaw, or household habit that manufactures a "we" out of the "I"s.
disagreeing verdictsThe same people, several different wills — none more real than the counting that produced it.
the Condorcet cycleThe majority that beats itself: every possible decision, opposed by most of the deciders.
an IIA violationThe spoiler, the wrecker candidacy, the menu trick — the option that can't win but can choose.
Arrow's axiomsFairness demands each obviously right, provably unable to hold hands all at once.

Where the metaphor tears

Three honest failures.

Rankings are a narrow pipe.

Arrow governs methods fed on rankings. Richer inputs — cardinal scores, as in range voting — escape its letter, since "how much" carries information an ordering drops. But they inherit its cousins: Gibbard–Satterthwaite says essentially any non-dictatorial method over three or more outcomes rewards strategic lying somewhere. The exits from Arrow are real; none leads outside the building.

Most Tuesdays aren't the worst case.

The theorem is about what can't be guaranteed, not about what usually happens. Small electorates with broadly aligned or single-peaked preferences — most committees, most families, most Tuesdays — rarely hit the pathologies, and the rack above often agrees five-for-five on random ballots. Treating every vote as secretly arbitrary because a worst case exists is as wrong as treating the worst case as impossible because today went fine.

Impossibility is not the strongman's alibi.

"No counting is perfect, so let one decisive person decide" fails the theorem's own test: non-dictatorship is one of the axioms, not a casualty of them — the dictator "solves" Arrow only the way amputation solves a fracture. The theorem indicts naive talk about a pre-political popular will; it does not indict collective self-rule. If anything it dignifies it: the rules are worth arguing about like they matter, because they are the part of the people's will the people can actually write.