Error Correction as a Structurally Scarce Public Good--Extremal Aggregation, Effective Multiplicity, and the Decoupling of Alarm Channels

The world population is roughly eighty times that of the Axial Age, and the literate population several orders of magnitude larger, yet civilisation-scale reconception of frameworks has not scaled with it. We show that this is not a paradox but an arithmetic necessity, and that the standard explanations (over-specialisation, secularisation, moral decline) all misidentify the operator involved. We model a civilisation’s capacity to survive a regime change by a survival margin S = B/|g| − ln T/D, in which the time to ruin B/|g| must exceed the sequential-detection delay ln T/D (Page–Lorden), where T is the tolerated false-alarm threshold — behaviourally, the cost of admitting one might be wrong — and D is the per-observation Kullback–Leibler distinguishability of the new regime from the old. Aggregating to organisations gives two extremal operators, T_org = max Tᵢ and D_eff = max Dᵢ, whose actionability is opposite: D_eff can be raised by anyone from any position provided the channel is open, whereas T_org can be lowered only by whoever occupies the arg-max position. Three consequences follow. (i) Because max grows logarithmically, and because under equicorrelation ρ the effective multiplicity is N_eff = N^(1−ρ), a population of 10⁹ sharing 90% of its paradigm has the effective diversity of about eight independent observers; population is nearly irrelevant, correlation is nearly everything. (ii) For a mobile innovator facing J independent centres, the operator inverts from max to min, giving adoption delay τ/(J·D) rather than τ·ln N/D — exit beats voice by a factor J·ln N. (iii) Because pain is generated on the payoff channel while framework obsolescence occurs on the meaning channel, and because these two regime shifts coincided in the Axial Age but are decoupled under material satiation, the modern condition is not an absence of seers but a misconnected alarm. We derive a conjunctive (non-compensatory) criterion for the corrector role, show why any weighted-sum score is mathematically wrong, give five falsifiable predictions, and state the conditions under which the whole account should be withdrawn.