Invariant universal averages and Bertrand's paradox: a response to Shackel
Abstract
In 2014, with D. Aerts we proposed a solution to the hard problem of Bertrand's paradox by means of a universal average over different ways of randomizing an entity. In his recent monograph, N. Shackel raises objections concerning the approximation by cellular densities, the restrictions used in their construction, the apparently trivial limit of their averages, the choice of a second-order probability law, and the dependence of the result on the naming of chords. We revisit these objections using the invariance-based formulation of the universal average. The first objection is well taken, and the approximation argument is repaired: an explicit sequence of bivalent cellular densities provides weak approximation of any prescribed probability measure, with an error bound for intervals. We show that any second-order selection law invariant under rearrangements of equal-length cells has the same uniform barycenter, independently of the particular densities and weights employed. For cellular densities, this conclusion holds exactly at every finite resolution. Thus, non-uniqueness of the second-order law does not prevent uniqueness of the resulting average within the indifferent class. Finally, we prove that no probability measure can be invariant under both these rearrangements and arbitrary changes of interval lengths. The reference geometry therefore cannot be obtained by averaging alone: it must be supplied by the physical modelization of the randomized entity. These results clarify the scope, and the limitations, of our original distinction between the easy and hard problems of Bertrand's paradox.
Related
- Universal optimality of the double-centred matrix under unitarily invariant norms for dissimilarity data
- On short averages of the Fibonacci partition function
- Polynomial configurations and pointwise averages along Piatetski--Shapiro primes
- Polynomial configurations and pointwise averages along Piatetski-Shapiro sequences
- Ehrhart polynomials of cyclic polytopes as averages of zonotope Ehrhart polynomials
BibTeX
@misc{bianchi2026invariant,
title = {{Invariant universal averages and Bertrand's paradox: a response to Shackel}},
author = {Massimiliano Sassoli de Bianchi},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.00981},
eprint = {2611.00981},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.00981},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Invariant universal averages and Bertrand's paradox: a response to Shackel
AU - Massimiliano Sassoli de Bianchi
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.00981
UR - https://arxiv.org/abs/2611.00981
AB - In 2014, with D. Aerts we proposed a solution to the hard problem of Bertrand's paradox by means of a universal average over different ways of randomizing an entity. In his recent monograph, N. Shackel raises objections concerning the approximation by cellular densities, the restrictions used in their construction, the apparently trivial limit of their averages, the choice of a second-order probability law, and the dependence of the result on the naming of chords. We revisit these objections using the invariance-based formulation of the universal average. The first objection is well taken, and the approximation argument is repaired: an explicit sequence of bivalent cellular densities provides weak approximation of any prescribed probability measure, with an error bound for intervals. We show that any second-order selection law invariant under rearrangements of equal-length cells has the same uniform barycenter, independently of the particular densities and weights employed. For cellular densities, this conclusion holds exactly at every finite resolution. Thus, non-uniqueness of the second-order law does not prevent uniqueness of the resulting average within the indifferent class. Finally, we prove that no probability measure can be invariant under both these rearrangements and arbitrary changes of interval lengths. The reference geometry therefore cannot be obtained by averaging alone: it must be supplied by the physical modelization of the randomized entity. These results clarify the scope, and the limitations, of our original distinction between the easy and hard problems of Bertrand's paradox.
ER -