Formal Power Series Representations in Probability and Expected Utility Theory

Arthur Paul Pedersen † and Samuel Allen Alexander ‡ · 2025

Read Formal Power Series Representations in Probability and Expected Utility Theory by Arthur Paul Pedersen † and Samuel Allen Alexander ‡ (2025) 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-2508-00294.txt.

Introduction

Arthur Paul Pedersen and Samuel Allen Alexander build a theory of preference that drops several classic demands. Their 2025 paper advances a coherence standard for preferences, modeled on de Finetti's foundation for probability. Orthodox theory insists on transitivity, boundedness, and smooth continuity; this one requires none of them.

Two results carry the paper. Any preference system meeting the coherence test can be extended to a complete one. And any complete, coherent system can be represented by utility values drawn from an ordered field that extends the real numbers. That second result flows from the paper's centerpiece, which at once extends a theorem of Holder and sharpens Hahn's embedding theorem. The authors frame formal power series as the language for these representations.

Catalog record

CREATORArthur Paul Pedersen † and Samuel Allen Alexander ‡
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-00294

Browse the full library · About the Tower of Babel

Full text

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

Formal Power Series Representations in Probability and Expected Utility Theory Thanks: † Department of Computer Science, The City University of New York. ‡ Quantitative Analytics Unit, U.S. Securities & Exchange Commission Arthur Paul Pedersen † and Samuel Allen Alexander ‡ Abstract . We advance a general theory of coherent preference that surrenders restrictions embodied in orthodox doctrine. This theory enjoys the property that any preference system admits extension to a complete system of preferences, provided it satisfies a certain coherence requirement analogous to the one de Finetti advanced for his foundations of probability. Unlike de Finetti’s theory, the one we set forth requires neither transitivity nor Archimedeanness nor boundedness nor continuity of preference. This theory also enjoys the property that any complete preference system meeting the standard of coherence can be represented by utility in an ordered field extension of the reals. Representability by utility is a corollary of this paper’s central result, which at once extends Hölder’s Theorem and strengthens Hahn’s Embedding Theorem. 1. Introduction This paper shows that any complete system of preferences admits a utility representation in an ordered field extension of the real numbers, subject to a certain standard of coherence analogous to one for gambling that de Finetti 1931 , de Finetti 1937 , de Finetti 1974a , de Finetti 1974b advanced for his theory of probability. In so doing, this paper provides a full, self-contained proof of a principal result due to one of us but for space considerations could only be stated without proof in a prior publication ( Pedersen 2014 ) . More generally, this paper provides for a unified treatment of numerical probability and expected utility that relaxes standards imposed by the orthodox canon to which belong the staple theories of de Finetti 1974a , de Finetti 1974b , von Neumann and Morgenstern 1947 , Savage 1954 , and Anscombe and Aumann 1963 , while making room for adherence to standards whose transgression is all but prescribed by the orthodox canon. 1 1 1 Among orthodox standards at issue are continuity requirements, Archimedean postulates, topological conditions, boundedness constraints, cardinality restrictions, and the like. Relaxing such requirements makes room for adherence to criteria requiring respect for forms of monotonicity or dominance preservation. For more on the challenges the unified theory addresses, see exposition by Pedersen and Arló-Costa 2012 , Pedersen 2013 , Pedersen 2014 . An abridged technical orientation to the result stated but unproven by Pedersen 2014 will be given given in § 3 . See footnote 2 . For the purpose of illustration by comparatively direct and elementary means, this paper focuses on de Finetti’s foundations of probability and expected utility, thereby sidestepping a detour through, for example, mixture spaces, convex sets, cones, affine spaces, and so on, in order to formulate and apply this paper’s result to extending the theories of de Finetti 1974a , de Finetti 1974b , von Neumann and Morgenstern 1947 , Anscombe and Aumann 1963 , and Savage 1954 . Routine arguments show how to used to to in this connection, see, e.g., Stone 1949 , Hausner 1954 , Gudder 1977 , Gudder 1978 , Gudder and Schroeck 1980 , Mongin 2001 ). The unified treatment of numerical probability and expected utility representations is a basic application of this paper’s central result. In what follows, we show that any totally ordered linear space \mathbb{V} is order isomorphic to a subspace of a lexicographically ordered field \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) of formal power series s=\sum_{\gamma}s_{\gamma}\epsilon^{\gamma} with coefficients s_{\gamma} in \Rset and exponents \gamma in a totally ordered Abelian group \Gamma for which (\gamma:s_{\gamma}\neq 0) forms a well-ordered subset of \Gamma . The operations of addition and multiplication agree with the familiar operations for power series subject to the usual requirement that \epsilon^{\alpha}\epsilon^{\beta}=\epsilon^{\alpha+\beta} for \alpha,\beta\in\Gamma , while the lexicographic order on \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) is determined by requiring a power series s to be positive when the first exponent \gamma at which the real coefficient s_{\gamma} does not vanish to be positive. So equipped, the field \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) is a totally ordered extension of the real number system. When \Gamma is the trivial group (i.e., \Gamma=\{0\} ), the field \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) is isomorphic to the field \Rset , in which case both \mathbb{V} and \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) are Archimedean in the sense that each pair (a_{1},a_{2}) of non-zero elements satisfies n|a_{1}|>|a_{2}| for some n\in\Nset , where |a|\coloneqq\max\{a,-a\} . Thus, a corollary of this paper’s central result is Hölder’s Theorem ( Hölder 1901 ) , a cornerstone of modern measurement theory. When the totally ordered Abelian group \Gamma is non-trivial (i.e., \Gamma\neq\{0\} ), the field \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) is a proper extension of the field of real numbers, in which case both \mathbb{V} and \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) are non-Archimedean . This paper’s central result therefore strengthens Hahn’s Embedding Theorem ( Hahn 1927 ) . In fact, the field \Rset\bigl(\bigl(\,\epsilon^{\Gamma}\,\bigr)\bigr) is, up to order-isomorphism, the smallest totally ordered field extension \Fset of \Rset to include (an order-isomorphic copy of) the linear space \mathbb{V} that is Archimedean complete in the sense that it admits no proper extension \Fset^{\prime} for which each element a^{\prime}\in\Fset^{\prime}\setminus\Fset has the same order of magnitude as some a\in\Fset — that is, such that n|a|>|a^{\prime}| and n|a^{\prime}|>|a| for some n\in\Nset . The rest of this paper is organized as follows. After briefly laying out preliminary terminology and notation in § 2 , attention is turned in § 3 to this paper’s fundamental results on coherence and discusses their connection to pioneering work of de Finetti. Examples given in § 4 shed light on the scope and power of our theory of coherence, whereupon § 5 develops critical ideas and conventions for this paper’s central result. Concluding the paper in § 6 is a brief discussion of related literature. 2. Preliminaries Throughout we will be concerned with real-valued maps on a common nonempty set \Omega of states — called a state space — corresponding to a collection of mutually exclusive and collectively exhaustive hypotheses. A function g:\Omega\to\Rset , called a gamble , specifies the numerical outcome to be obtained from g in any given state. A gamble only assuming the values 0 or 1 is called an indicator function . Each subset E of \Omega , called an event , corresponds to a unique indicator function, denoted by \mathbf{1}_{E} , such that for every \omega\in\Omega : \displaystyle\mathbf{1}_{E}(\omega)\hskip 10.00002pt \displaystyle=\hskip 10.00002pt\begin{cases}1&\mbox{if }\omega\in E;\\ 0&\mbox{otherwise.}\end{cases} Each indicator function likewise corresponds to a unique event from \Omega . Following ordinary convention, an event E will be identified with its indicator function \mathbf{1}_{E} when the context is clear. A gamble g is said to be bounded if its image is a subset of \bigl[\,-n,n\,\bigr] for some nonnegative integer n ; a gamble that fails to be bounded is said to be unbounded . Observe that any gamble assuming only finitely many values is bounded; thus any linear combination of indicators functions is bounded. Observe in addition that a gamble is unbounded only if the underlying state space is infinite. In this paper, we presume gamble outcomes are expressed in units of a linear utility scale that has been determined in advance. We also presume individual gambles can be combined and rescaled in accordance with pointwise arithmetic operations. Thus, given gambles f and g on state

Read the complete text