A note on ded κ \kappa and the part-whole principle

Yuanshan Li · 2026

Read A note on ded κ \kappa and the part-whole principle by Yuanshan Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-25290.txt.

Introduction

Set theory meets the philosophy of probability in a short note by Yuanshan Li, forthcoming in Philosophy of Science. The part whole principle says a proper part should not measure the same as the whole. Li connects that principle to a quantity built from Dedekind cuts, which count the ways a linear order can be split into two pieces with no largest element below. The quantity generalises an older cardinal characteristic studied in the foundations of probability.

Establishing the connection sharpens a result by Mancosu and Massas about generalised probability functions, which assign values in ordered structures richer than plain numbers. Li also proposes follow up questions the link suggests. The argument shows how a purely combinatorial measure of order complexity can decide which exotic probability assignments are even possible.

Catalog record

CREATORYuanshan Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE15 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.25290

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A note on ded κ \kappa and the part-whole principle Thanks: The author thanks Nicholas Ramsey, Joel David Hamkins, and Gabriel Goldberg for discussion. Forthcoming in Philosophy of Science . Yuanshan Li Abstract We establish a connection between the part-whole principle and the quantity ded κ \kappa – a generalized cardinal characteristic related to the number of Dedekind cuts of a linear order. As consequences, we improve a result of Mancosu and Massas [ 4 ] on generalized probability functions and propose some questions. 1 Introduction We recall Definitions 2.1 and 2.2 from Mancosu and Massas [ 4 ] : Definition . A generalized probability range is a tuple (V,\leq,+,0) such that: • (V,\leq) is a partial order with a least element 0 ; • (V,+,0) is a commutative monoid; • if x,y,z\in V and x<y then x+z<y+z . Definition . Let (V,\leq,+,0) be a generalized probability range. A generalized V -probability space is a tuple (\Omega,\mathcal{F},\mu) such that: • \mathcal{F}\subseteq 2^{\Omega} is an algebra; • \mu:\mathcal{F}\rightarrow V satisfies \mu(A)\geq 0 for all A\in\mathcal{F} and \mu(A\sqcup B)=\mu(A)+\mu(B) for disjoint A,B\in\mathcal{F} . In this context, the function \mu is to be called a generalized probability function, or a generalized probability. The main objects of interest in [ 4 ] are generalized probabilities that are both total and regular , defined as follows: Definition . A generalized probability \mu is total if its domain \mathcal{F}=2^{\Omega} . Definition . A generalized probability \mu is regular if \mu(A)>0 for all nonempty A\in\mathcal{F} . One of the main mathematical issues investigated in [ 4 ] is the extent to which the existence of regular total generalized probabilities can be characterized in terms of the cardinalities of the associated sets \Omega and V . Hofweber and Schindler [ 2 ] obtained the following result via an elementary compactness argument: Theorem 1 (Hofweber and Schindler) . For any \Omega there is a hyperreal field 1 1 1 Note that a hyperreal field is a model of the axioms of real closed fields, hence satisfies Definition 1. \mathbb{R}^{*} with |\mathbb{R}^{*}|\leq 2^{|\Omega|} and a regular total generalized probability function \mu:2^{\Omega}\rightarrow\mathbb{R}^{*} . Theorem 3.1 of [ 4 ] , an improvement of Corollary 1 of Pruss [ 7 ] , states the following: Theorem 2 (Mancosu and Massas) . If |V|\leq|\Omega| then there is no regular total generalized probability function \mu:2^{\Omega}\rightarrow V . Recall that the continuum hypothesis ( \mathsf{CH} ) states that the cardinality of the powerset of \omega is the first uncountable cardinal, i.e. 2^{\omega}=\aleph_{1} ; the generalized continuum hypothesis ( \mathsf{GCH} ) at a cardinal \kappa states that the cardinality of the powerset of \kappa is its successor cardinal, i.e. 2^{\kappa}=\kappa^{+} . By Theorems 1 and 2, it follows that: Corollary 3 (Mancosu and Massas) . If \mathsf{GCH} holds at |\Omega| , then there exists a regular total generalized probability function \mu:2^{\Omega}\rightarrow V iff |V|\geq 2^{|\Omega|} . Corollary 3 invites a natural question, namely whether the assumption of \mathsf{GCH} can be removed, as it is well-known that various failures of \mathsf{GCH} are consistent with \mathsf{ZFC} . Main Question 4 . Is it consistent with \mathsf{ZFC} that there is a regular total generalized probability function \mu:2^{\Omega}\rightarrow V with |V|<2^{|\Omega|} ? In this note we make some very modest clarifications on the Main Question: first, we improve Theorem 2 by weakening the hypothesis |V|\leq|\Omega| , sharpening the existing results and leading to some necessary conditions that the candidate set-theoretic models for the Main Question must satisfy. Second, this is done merely using the order structure on V , which avoids the complications introduced by the additive structure. To illustrate the content of our improvement, let us consider the special case where \Omega is countable. In this case, the Main Question reduces to the following. Question 5 . Is it consistent with \mathsf{ZFC} that there is a regular total generalized probability function \mu:2^{\omega}\rightarrow V with |V|<2^{\omega} ? By Corollary 3, there is no such function if \mathsf{CH} holds. However, Corollary 3 does not settle the question when the cardinality of continuum is large. For example, if 2^{\omega}=\aleph_{2} , is there a regular total generalized probability function for subsets of \omega whose range has cardinality \aleph_{1} ? As the reader shall see, it follows from our results in this note that the answer to Question 5 is negative. 2 Main results The part-whole principle states that if A,B are sets and A\subsetneq B , then the size of A is strictly less than the size of B . This principle is of independent interest in set theory and its philosophy. It motivates the following definition: Definition . A part-whole size function (for subsets of \kappa ) is an order homomorphism \mu:(2^{\kappa},\subsetneq)\rightarrow(V,<) , where (V,<) is a strict partial order. 2 2 2 One might think that the sizes of any two sets should be comparable, so it may appear natural to require the range of size functions to be linearly ordered. Here we don’t make this requirement to coordinate with generalized probability ranges being merely partially ordered. All results in our note remain valid if we impose linearity in both definitions. Part-whole size functions are closely related to the previous setting in the following way: Lemma 6 . If \mu is a regular total generalized probability function then its reduct \mu:(2^{\Omega},\subsetneq)\rightarrow(V,<) is a part-whole size function for subsets of \Omega . Proof. Let A\subsetneq B , so C=B\setminus A is nonempty. By regularity, \mu(C)>0 . By finite additivity, \mu(A)+\mu(C)=\mu(B) . We have for x,y,z\in V , x<y implies x+z<y+z . In particular, 0<\mu(C) implies 0+\mu(A)<\mu(C)+\mu(A)=\mu(B). Hence \mu(A)<\mu(B) . ∎ From now on, the chief objects of interest would be part-whole size functions. For convenience, we make the following convention: Convention . We shall identify the base space \Omega with a cardinal \kappa . We note that once the additive structure is removed, it already makes the proof of Theorem 2 much simpler than the proof written down in [ 4 ] , even following their existing approach, which involves the following fixed point theorem attributed to Zermelo. 3 3 3 For an informative discussion of Zermelo’s result and its context, see Kanamori [ 3 ] , where it appears as Corollary 2.3. Fact 7 (Zermelo) . For any function F:2^{\kappa}\rightarrow\kappa there is A\subsetneq B s.t. F(A)=F(B) . It readily follows that there can be no part-whole size function if |V|\leq\kappa : suppose \mu:2^{\kappa}\rightarrow V is such a function, fix an injection i:V\rightarrow\kappa . By Zermelo’s theorem applied to i\circ\mu , there is A\subsetneq B such that i(\mu(A))=i(\mu(B)) . Since i is injective, \mu(A)=\mu(B) , but A\subsetneq B , contradicting part-whole. By Lemma 6, if |V|\leq\kappa there can’t be regular total generalized probability functions as well. We now state our main result, from which we can prove a stronger version of Theorem 2 without using Zermelo’s fixed point theorem. The key observation is the following: by the part-whole principle, if \mu is a part-whole size function and A\subsetneq B , then \mu(A)<\mu(B) , in particular they are different values in V . In general, if (A_{i}:i\in I) forms a strictly increasing chain of subsets of \kappa , meaning that I is a linear order and if i<_{I}j then A_{i}\subsetneq A_{j} , we

Read the complete text