Hausdorff stability transfers.
ProvedJokeSurpriseUniqueness.HumorScale.lipschitz_hausdorffaether-catalogapplications
Hausdorff stability transfers. Every humor scale is 2 · scale-Lipschitz for
the Hausdorff distance between setups: robustness of the measurement is an axiomatic
consequence, not an artefact of the range formula.
theorem JokeSurpriseUniqueness.HumorScale.lipschitz_hausdorff(S T : Finset ℝ) (hS : S.Nonempty) (hT : T.Nonempty) :
|V.toFun (S.min' hS) (S.max' hS) - V.toFun (T.min' hT) (T.max' hT)|
≤ 2 * V.scale * hausdorffDist (S : Set ℝ) (T : Set ℝ) := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Applications/JokeSurpriseUniqueness.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Applications/JokeSurpriseUniqueness.lean import Mathlib import Definitions.Def_Applications_JokeSurpriseStability import Definitions.Def_Applications_JokeSurpriseUniqueness /-! # Why *this* invariant? A uniqueness theorem for surprise The two preceding files of this thread take the formula `humor S = max' S - min' S` (equivalently `Metric.diam S`, by `JokeSurpriseStability.humor_eq_diam`) as given, and study its algebra (`Applications.JokeSurpriseAlgebra`, `Applications.JokeColimitUniversality`) and its stability (`Applications.JokeSurpriseStability`). The obvious objection is that the formula is *chosen*: any monotone functional on setups would produce a plausible-looking theory. This file removes the choice. We axiomatise what a *humor scale* must satisfy — * **position blindness**: a joke is not funnier for being told about larger numbers (translation invariance); * **staged telling**: telling a joke in two consecutive stages accumulates the surprise of the stages (concatenation additivity); * **monotonicity**: widening the gap between the extreme readings cannot reduce surprise — and prove that these three axioms pin the invariant down **completely**: * `HumorScale.eq_scale_mul` : every humor scale is `V m M = c · (M - m)` with `c = V 0 1`; * `HumorScale.scale_nonneg` : the constant is nonnegative; * `HumorScale.eq_scale_mul_humor` : on setups, every humor scale is a nonnegative multiple of the catalog's `humor`; * `HumorScale.ext_of_scale_eq` : a humor scale is determined by its value on the unit gap — the theory has exactly one degree of freedom, the choice of unit. Consequently every theorem of the thread transfers automatically to *any* admissible invariant, without re-proof: * `HumorScale.submodular` : every humor scale is submodular; * `HumorScale.lipschitz_hausdorff` : every humor scale is `2c`-Lipschitz for the Hausdorff distance between setups. The technical heart is `cauchy_of_monotone`: a monotone solution of the Cauchy functional equation on `[0, ∞)` is linear. It is proved from scratch (rational dilations `g (k s) = k g s`, the floor sandwich `k/n ≤ t ≤ (k+1)/n`, and an Archimedean limit), since a monotone — as opposed to continuous or measurable — Cauchy theorem is not available off the shelf. -- !-- Lab Notes -- !-- Hypothesis (H7): the range/diameter formula is not a modelling choice but is *forced* by position blindness, staged telling, and monotonicity. Hypothesis (H8): if H7 holds, the whole thread (submodularity, Hausdorff stability) is axiom-independent and transfers to every admissible invariant. Experiment: H7 was reduced to a Cauchy functional equation for `g t = V 0 t` on `[0, ∞)`. The additivity `g (s + t) = g s + g t` comes from concatenation plus translation invariance; monotonicity of `g` from the third axiom. The linearity `g t = g 1 · t` was then proved by hand: `g (k s) = k g s` by induction, hence `g (1/n) = g 1 / n` and `g (k/n) = g 1 · k / n`; sandwiching `t` between `⌊n t⌋/n` and `(⌊n t⌋ + 1)/n` gives `|g t - g 1 · t| ≤ g 1 / n` for every `n`, and the Archimedean property finishes. Non-vacuity was checked by exhibiting the range scale `rangeScale` (with `c = 1`) as a model of the axioms. Analysis: H7 and H8 both survive, and they retro-justify the earlier files: every theorem previously proved about `humor` is a theorem about *any* invariant obeying the three axioms, up to the scale factor `c`. Critique: monotonicity is essential and not decorative — without it, a Hamel-basis solution of the Cauchy equation gives a wildly discontinuous "humor scale" that is translation invariant and concatenation additive but not proportional to the range. The axioms are consistent (`rangeScale`) and, by `ext_of_scale_eq`, categorical up to the single scale parameter. Synthesis: surprise is the unique — up to choice of unit — position-blind, stage- additive, monotone measure of a setup; the range formula of the catalog is its normalisation at `c = 1`. -/ open Finset Metric JokeSurpriseAlgebra JokeSurpriseStability open JokeSurpriseUniqueness /-! ### A monotone Cauchy equation -/ /-! ### Humor scales -/ open HumorScale variable (V : HumorScale)
Formal statement
theorem JokeSurpriseUniqueness.HumorScale.lipschitz_hausdorff(S T : Finset ℝ) (hS : S.Nonempty) (hT : T.Nonempty) :
|V.toFun (S.min' hS) (S.max' hS) - V.toFun (T.min' hT) (T.max' hT)|
≤ 2 * V.scale * hausdorffDist (S : Set ℝ) (T : Set ℝ) := by sorrySource