The Tragedy of the Confidence Commons Optimal Belief-Update Rate, Adaptive Overconfidence, and the Tail Index as Commons

Twenty-five centuries of moral exhortation have not made people better calibrated. This paper offers an explanation: the person being exhorted has not made a mistake. In adversarial environments the individually optimal belief bias Δ★ may be strictly positive while the collectively optimal bias is zero. That is the structure of a commons tragedy. The work of this paper is to write each of its three components as a computable object. (i) "How fast should one update" has a unique optimum. If the truth performs a random walk with per-step variance v and observations carry noise variance s², the steady-state mean squared error of exponentially weighted updating is L(α) = α·s²/2 + v/α, minimised uniquely at α★ = √(2v/s²). α → 1 is being led by noise (the self-abasing type); α → 0 is a frozen belief (the Ah-Q type). These are not two ailments but the two ends of one U-shaped curve. At the optimum the two failure modes contribute exactly equally — this is the first-order condition of "the golden mean". In particular v = 0 gives α★ = 0: in a genuinely stationary world, not updating is optimal; hence "you must update" is a theorem whose premise is that the world is non-stationary. (ii) Defining rigidity as the reciprocal of the update rate, k ≡ 1/α, gives L(k) = s²/(2k) + v·k with optimal rigidity k★ = s/√(2v), at which the "shield" cost and the "brittleness" cost are exactly equal; and even when tuned optimally the system retains an irreducible loss L(k★) = s·√(2v). Corollary: the faster the environment drifts, the softer the optimal dogma; and modernity, by raising both v and s, raises the minimum loss itself — the crisis of meaning is not "the rigidity is mis-set" but "whatever the rigidity, the floor is rising". (iii) New in this version: writing that U-shape as an object with a discriminant. Expanding at α★ as V′(g) = β₂g + β₃g² + β₄g³ gives closed forms for all three coefficients, a discriminant Δ = −23v²/α★⁸ < 0 identically, and an exact parameter-free ratio R = 9/32. Δ < 0 means this U-shape has no fold: both ends are diminishing returns, with no barrier, no second well and no hysteresis. And R is bounded by 27/88 across the whole family of power reparametrisations, so this conclusion does not depend on the identification "rigidity is the reciprocal of the update rate". The asymmetry also has an exact closed form: at relative deviation u, the excess loss on the too-fast side stands to that on the too-slow side exactly as (1−u)/(1+u). (iv) Individually optimal overconfidence is converted, through the single arrow "belief determines position size", into multiplicative volatility for the population. Under Kesten dynamics each agent's bet fraction f_i determines a tail index κ★(f_i), and the tail index of a mixture is governed by its heaviest component (Gnedenko–Kolmogorov, the single-big-jump principle): κ★_pop = min_i κ★( f_i ) Not the average. The minimum. This is a worse commons than the standard one: there, harm is the sum of each herder's grazing; here, harm is the grazing of the single most aggressive herder, and everyone else's restraint is pushed out of the tail by the single-big-jump principle.