The Lot
The Roman practice of decimation — selecting every tenth soldier by lot for execution after a collective failure — is remembered as an atrocity. But the mechanism deserves a closer look, because its distinctive feature is not its cruelty but its randomness. The soldiers killed were not necessarily the ones who fled. The selection was by lot, explicitly independent of individual guilt. Polybius describes the procedure in detail: the legion assembled, one in ten drawn, the selected beaten to death by the remaining nine. The survivors received barley instead of wheat and camped outside the fortifications.
The randomness was not a deficiency in Roman jurisprudence. It was the mechanism. If the guilty could be identified and punished individually, the rest of the unit would have no structural reason to prevent the failure. Each soldier would calculate personal risk, behave accordingly, and remain indifferent to the behavior of others. By making punishment independent of individual conduct, decimation converted a blame-attribution problem into a statistics problem. Every soldier had equal probability of selection regardless of personal bravery or cowardice. The only way to reduce individual risk was collective action — preventing the triggering condition. The injustice was the mechanism.
The Vietnam War draft lottery of December 1, 1969, operated on the same structural principle. Three hundred and sixty-six birth dates were drawn from a glass container in sequence. Number one — September 14 — meant near-certain conscription. The lottery replaced a deferment system that had concentrated military service in poor and minority communities while insulating college students and those in designated occupations. The deferment system was more individually fair: it targeted those with fewer alternatives. The lottery was less individually fair: it targeted by birth date, which correlated with nothing. And the lottery produced a political effect the deferment system could not. When student deferments were eliminated in 1971, campus antiwar protest intensified dramatically. The randomness made the war personally real for demographics that had previously been structurally insulated from it. Universal vulnerability produced universal engagement. Targeted vulnerability had not.
Athenian democracy formalized this principle at the center of its government. Most public offices were filled not by election but by lot. The boule — the council of five hundred that set the Assembly's agenda — was selected randomly from citizens over thirty using the kleroterion, a stone allotment machine with slots for bronze identity tickets. Aristotle considered this distinction definitive: election was aristocratic; lot was democratic. The randomness was not a concession to administrative convenience. It was the anti-corruption mechanism. Elected officials could form factions, build patronage networks, and entrench themselves. Officials selected by lot could not. The rotation was enforced by the same structural feature that makes decimation work — the disconnection between selection and individual characteristics.
In a genetically identical population of bacteria, some fraction spontaneously enters a dormant state — metabolically inactive, non-reproducing, and resistant to antibiotics. This is stochastic phenotypic switching, a bet-hedging strategy. The cells that enter dormancy are not responding to an environmental signal. They switch randomly, driven by molecular noise in gene expression. Individually, dormancy is costly: a dormant cell produces no offspring while its neighbors divide. But when antibiotics arrive, the dormant cells survive while the reproducing majority is killed. The population persists through the random sacrifice of individual fitness.
A bacterium that could predict when antibiotics would arrive would not need bet-hedging — it would simply time its dormancy. A Roman commander who could identify the guilty would not need decimation — he would simply punish them. The randomness exists precisely because the system cannot predict or identify, and it works precisely because individual optimization is foreclosed.
Jury selection completes the pattern by inverting its direction. Decimation distributes punishment randomly. The jury distributes authority randomly. Jurors in the Anglo-American system are selected from voter rolls or driver license records, not for competence, interest, or knowledge of the case. A jury selected for relevant expertise would be a tribunal. A jury selected for sympathy would be an advocacy group. The verdict's legitimacy derives from the jurors' interchangeability — from the fact that they could be anyone. The randomness is not a substitute for a better selection process. It is the selection process, because the property being produced — democratic legitimacy, the sense that the verdict represents the community rather than a faction — is a property that targeted selection structurally cannot produce.
When individuals can optimize against a selection criterion, they will, and the collective property that depends on universal exposure disappears. The lot is not a failure to choose. It is the only kind of choosing that works.