Desirable Gambles Research Articles

Desirable gambles based on pairwise comparisons

This paper proposes a model for imprecise probability information based on bounds on probability ratios, instead of bounds on events. This model is studied in the language of coherent sets of desirable gambles, which provides an elegant mathematical formulation and a more expressive power. The paper provides methods to check avoiding sure loss and coherence, and to compute the natural extension. The relationships with other formalisms such as imprecise multiplicative preferences, the constant odd ratio model, or comparative probability are analyzed.

Independent natural extension for choice functions

We introduce an independence notion for choice functions, which we call ‘epistemic independence’ following the work by De Cooman et al. [17] for lower previsions, and study it in a multivariate setting. This work is a continuation of earlier work of one of the authors [29], and our results build on the characterization of choice functions in terms of sets of binary preferences recently established by De Bock and De Cooman [11]. We obtain the many-to-one independent natural extension in this framework. Given the generality of choice functions, our expression for the independent natural extension is the most general one we are aware of, and we show how it implies the independent natural extension for sets of desirable gambles, and therefore also for less expressive imprecise-probabilistic models. Once this is in place, we compare this concept of epistemic independence to another independence concept for choice functions proposed by Seidenfeld [28], which De Bock and De Cooman [2] have called S-independence. We show that neither is more general than the other.

Exposing some points of interest about non-exposed points of desirability

We study the representation of sets of desirable gambles by sets of probability mass functions. Sets of desirable gambles are a very general uncertainty model, that may be non-Archimedean, and therefore not representable by a set of probability mass functions. Recently, Cozman (2018) has shown that imposing the additional requirement of even convexity on sets of desirable gambles guarantees that they are representable by a set of probability mass functions. Already more that 20 years earlier, Seidenfeld et al. (1995) gave an axiomatisation of binary preferences—on horse lotteries, rather than on gambles—that also leads to a unique representation in terms of sets of probability mass functions. To reach this goal, they use two devices, which we will call ‘SSK–Archimedeanity’ and ‘SSK–extension’. In this paper, we will make the arguments of Seidenfeld et al. (1995) explicit in the language of gambles, and show how their ideas imply even convexity and allow for conservative reasoning with evenly convex sets of desirable gambles, by deriving an equivalence between the SSK–Archimedean natural extension, the SSK–extension, and the evenly convex natural extension.

Joint desirability foundations of social choice and opinion pooling

We develop joint foundations for the fields of social choice and opinion pooling using coherent sets of desirable gambles, a general uncertainty model that allows to encompass both complete and incomplete preferences. This leads on the one hand to a new perspective of traditional results of social choice (in particular Arrow’s theorem as well as sufficient conditions for the existence of an oligarchy and democracy) and on the other hand to using the same framework to analyse opinion pooling. In particular, we argue that weak Pareto (unanimity) should be given the status of a rationality requirement and use this to discuss the aggregation of experts’ opinions based on probability and (state-independent) utility, showing some inherent limitation of this framework, with implications for statistics. The connection between our results and earlier work in the literature is also discussed.

Graphoid properties of concepts of independence for sets of probabilities

We examine several concepts of independence associated with (1) credal sets, understood as sets of probability measures, (2) sets of full conditional probabilities, (3) sets of lexicographic probabilities, and (4) sets of desirable gambles. Concepts of independence are evaluated with respect to the graphoid properties they satisfy, as these properties capture important abstract features of “independence”. We emphasize the analysis of sets of probability measures as this is a popular formalism, looking at versions of epistemic, confirmational, and type-5 independence that are based on regular conditioning, as well as complete and strong independence. We then examine analogous concepts of independence for sets of full conditional probabilities, sets of lexicographic probabilities, and sets of desirable gambles.

Believing Probabilistic Contents: On the Expressive Power and Coherence of Sets of Sets of Probabilities

Moss argues that rational agents are best thought of not as having degrees of belief in various propositions. Instead, they are best thought of as having beliefs in probabilistic contents. Probabilistic contents are sets of probability functions. Probabilistic belief states, in turn, can then be modelled by sets of probabilistic contents, or sets of sets of probability functions. We argue that this Mossean framework is of considerable interest quite independently of its role in Moss’s account of probabilistic knowledge or her semantics for epistemic modals and probability operators. It is an extremely general model of uncertainty. Indeed, it is at least as general and expressively powerful as every other current imprecise probability framework, including lower probabilities, lower previsions, sets of probabilities, sets of desirable gambles and choice functions. In addition, we partially answer an important question that Moss leaves open, viz., why should rational agents have consistent probabilistic beliefs? We will show that a large range of failures of consistency will lead you to be Dutch bookable.

On the convexity of preferences in decisions and games under (quasi-)convex/concave imprecise probability correspondences

The Shafer and Sonnenshein convexity of preferences is a key property in game theory. Previous research has shown that, in case of decisions under uncertainty, the compliance with this property (jointly) depends on the concavity/convexity of the imprecise probabilistic model with respect to the decision variable and on the attitudes towards imprecision of the decision maker. The present paper deepens the analysis by looking at set-valued imprecise probabilistic models that encompass sets of probability distributions and sets of almost desirable gambles. Moreover, it is shown that the required Shafer and Sonnenshein convexity property is obtained also in case the imprecise probability correspondences satisfy quasi-concavity/convexity with respect to the decision variable so that the set of admissible probabilistic models is significantly broadened.It is well known that sets of probability distributions and sets of almost desirable gambles are general models of representation of uncertainty that are connected to each other; moreover, they are both related to another model known as lower expectation. Therefore, the second part of this work explores the links between the (quasi-)concavity/convexity properties across the three different models so as to understand to what extent the Shafer and Sonnenshein convexity results hold.

The contradiction between belief functions: Its description, measurement, and correction based on generalized credal sets

In the paper, we formalize the notion of contradiction between belief functions: we argue that belief functions are not contradictory if they provide non-contradictory models for decision-making. To elaborate on this idea, we take the decision rule from imprecise probabilities and show that sources of information described by belief functions are not contradictory iff the intersection of corresponding credal sets is not empty. We demonstrate that evidential conjunctive and disjunctive rules fit with this idea and they are justified in a probabilistic setting. In the case of contradictory sources of information, we analyze possible conjunctions and show how the result can be described by generalized credal sets. Based on generalized credal sets, we propose a measure of contradiction between information sources and find its axiomatics. We show how the contradiction correction can be produced based on generalized credal sets and how it can be done on sets of surely desirable gambles.

Evaluating betting odds and free coupons using desirability

In the UK betting market, bookmakers often offer a free coupon to new customers. These free coupons allow the customer to place extra bets, at lower risk, in combination with the usual betting odds. We are interested in whether a customer can exploit these free coupons in order to make a sure gain, and if so, how the customer can achieve this. To answer this question, we evaluate the odds and free coupons as a set of desirable gambles for the bookmaker.We show that we can use the Choquet integral to check whether this set of desirable gambles incurs sure loss for the bookmaker, and hence, results in a sure gain for the customer. In the latter case, we also show how a customer can determine the combination of bets that make the best possible gain, based on complementary slackness.As an illustration, we look at some actual betting odds in the market and find that, without free coupons, the set of desirable gambles derived from those odds avoids sure loss. However, with free coupons, we identify some combinations of bets that customers could place in order to make a guaranteed gain.

Sum-of-squares for bounded rationality

In the gambling foundation of probability theory, rationality requires that a subject should always (never) find desirable all nonnegative (negative) gambles, because no matter the result of the experiment the subject never (always) decreases her money. Evaluating the nonnegativity of a gamble in infinite spaces is a difficult task. In fact, even if we restrict the gambles to be polynomials in Rn, the problem of determining nonnegativity is NP-hard. The aim of this paper is to develop a computable theory of desirable gambles. Instead of requiring the subject to desire all nonnegative gambles, we only require her to desire gambles for which she can efficiently determine the nonnegativity (in particular sum-of-squares polynomials). We refer to this new criterion as bounded rationality.

Independent natural extension for infinite spaces

We define and study the independent natural extension of two local uncertainty models for the general case of infinite spaces, using the frameworks of sets of desirable gambles and conditional lower previsions. In contrast to Miranda and Zaffalon [16], we adopt Williams-coherence instead of Walley-coherence. We show that our notion of independent natural extension always exists—whereas theirs does not—and that it satisfies various convenient properties, including factorisation and external additivity. The strength of these properties depends on the specific type of epistemic independence that is adopted. In particular, epistemic event-independence is shown to outperform epistemic atom-independence. Finally, the cases of lower expectations, expectations, lower probabilities and probabilities are obtained as special instances of our general definition. By applying our results to these instances, we demonstrate that epistemic independence is indeed epistemic, and that it includes the conventional notion of independence as a special case.

Improved linear programming methods for checking avoiding sure loss

We review the simplex method and two interior-point methods (the affine scaling and the primal-dual) for solving linear programming problems for checking avoiding sure loss, and propose novel improvements. We exploit the structure of these problems to reduce their size. We also present an extra stopping criterion, and direct ways to calculate feasible starting points in almost all cases. For benchmarking, we present algorithms for generating random sets of desirable gambles that either avoid or do not avoid sure loss. We test our improvements on these linear programming methods by measuring the computational time on these generated sets. We assess the relative performance of the three methods as a function of the number of desirable gambles and the number of outcomes. Overall, the affine scaling and primal-dual methods benefit from the improvements, and they both outperform the simplex method in most scenarios. We conclude that the simplex method is not a good choice for checking avoiding sure loss. If problems are small, then there is no tangible difference in performance between all methods. For large problems, our improved primal-dual method performs at least three times faster than any of the other methods.

Axiomatising Incomplete Preferences through Sets of Desirable Gambles

We establish the equivalence of two very general theories: the first is the decision-theoretic formalisation of incomplete preferences based on the mixture independence axiom; the second is the theory of coherent sets of desirable gambles (bounded variables) developed in the context of imprecise probability and extended here to vector-valued gambles. Such an equivalence allows us to analyse the theory of incomplete preferences from the point of view of desirability. Among other things, this leads us to uncover an unexpected and clarifying relation: that the notion of `state independence'---the traditional assumption that we can have separate models for beliefs (probabilities) and values (utilities)---coincides with that of `strong independence' in imprecise probability; this connection leads us also to propose much weaker, and arguably more realistic, notions of state independence. Then we simplify the treatment of complete beliefs and values by putting them on a more equal footing. We study the role of the Archimedean condition---which allows us to actually talk of expected utility---, identify some weaknesses and propose alternatives that solve these. More generally speaking, we show that desirability is a valuable alternative foundation to preferences for decision theory that streamlines and unifies a number of concepts while preserving great generality. In addition, the mentioned equivalence shows for the first time how to extend the theory of desirability to imprecise non-linear utility, thus enabling us to formulate one of the most powerful self-consistent theories of reasoning and decision-making available today.

Lexicographic choice functions

We investigate a generalisation of the coherent choice functions considered by Seidenfeld et al. [35], by sticking to the convexity axiom but imposing no Archimedeanity condition. We define our choice functions on vector spaces of options, which allows us to incorporate as special cases both Seidenfeld et al.'s [35] choice functions on horse lotteries and also pairwise choice—which is equivalent to sets of desirable gambles [29]—, and to investigate their connections.We show that choice functions based on sets of desirable options (gambles) satisfy Seidenfeld's convexity axiom only for very particular types of sets of desirable options, which are exactly those that are representable by lexicographic probability systems that have no non-trivial Savage-null events. We call them lexicographic choice functions. Finally, we prove that these choice functions can be used to determine the most conservative convex choice function associated with a given binary relation.

Robust Bayesian analysis in partially ordered plausibility calculi

In Robust Bayesian analysis one attempts to avoid the ‘Dogma of Precision’ in Bayesian analysis by entertaining a set of probability distributions instead of exactly one. The algebraic approach to plausibility calculi is inspired by Cox's and Jaynes' analyses of plausibility assessment as a logic of uncertainty. In the algebraic approach one is not so much interested in different ways to prove that precise Bayesian probability is inevitable but rather in how different sets of assumptions are reflected in the resulting plausibility calculus. It has repeatedly been pointed out that a partially ordered plausibility domain is more appropriate than a totally ordered one, but it has not yet been completely resolved exactly what such domains can look like. One such domain is the natural robust Bayesian representation, an indexed family of probabilities.We show that every plausibility calculus embeddable in a partially ordered ring is equivalent to a subring of a product of ordered fields, i.e., the robust Bayesian representation is universal under our assumptions, if extended rather than standard probability is used.We also show that this representation has at least the same expressiveness as coherent sets of desirable gambles with real valued payoffs, for a finite universe.

Conditioning, updating and lower probability zero

We discuss the issue of conditioning on events with probability zero within an imprecise-probabilistic setting, where it may happen that the conditioning event has lower probability zero, but positive upper probability. In this situation, two different conditioning rules are commonly used: regular extension and natural extension. We explain the difference between them and discuss various technical and computational aspects. Both conditioning rules are often used to update an imprecise belief model after receiving the information that some event O has occurred, simply by conditioning on O , but often little argumentation is given as to why such an approach would make sense. We help to address this problem by providing a firm foundational justification for the use of natural and regular extension as updating rules. Our results are presented in three different, closely related frameworks: sets of desirable gambles, lower previsions, and sets of probabilities. What makes our justification especially powerful is that it avoids making some of the unnecessary strong assumptions that are traditionally adopted. For example, we do not assume that lower and upper probabilities provide bounds on some ‘true’ probability mass function, on which we can then simply apply Bayes's rule. Instead a subject's lower probability for an event O is taken to be the supremum betting rate at which he is willing to bet on O , and his upper probability is the infimum betting rate at which he is willing to take bets on O ; we do not assume the existence of a fair betting rate that lies in between these bounds. • We provide an introduction to four different imprecise probability frameworks. • We discuss the issue of conditioning on events with (lower) probability zero. • We study two specific conditioning rules: natural and regular extension. • We justify the use of natural and regular extension as updating rules. • Our justifications do not require an assumption of ideal precision.

Coherent Predictive Inference under Exchangeability with Imprecise Probabilities

Coherent reasoning under uncertainty can be represented in a very general manner by coherent sets of desirable gambles. In a context that does not allow for indecision, this leads to an approach that is mathematically equivalent to working with coherent conditional probabilities. If we do allow for indecision, this leads to a more general foundation for coherent (imprecise-)probabilistic inference. In this framework, and for a given finite category set, coherent predictive inference under exchangeability can be represented using Bernstein coherent cones of multivariate polynomials on the simplex generated by this category set. This is a powerful generalisation of de Finetti's Representation Theorem allowing for both imprecision and indecision. We define an inference system as a map that associates a Bernstein coherent cone of polynomials with every finite category set. Many inference principles encountered in the literature can then be interpreted, and represented mathematically, as restrictions on such maps. We discuss, as particular examples, two important inference principles: representation insensitivity—a strengthened version of Walley's representation invariance—and specificity. We show that there is an infinity of inference systems that satisfy these two principles, amongst which we discuss in particular the skeptically cautious inference system, the inference systems corresponding to (a modified version of) Walley and Bernard's Imprecise Dirichlet Multinomial Models (IDMM), the skeptical IDMM inference systems, and the Haldane inference system. We also prove that the latter produces the same posterior inferences as would be obtained using Haldane's improper prior, implying that there is an infinity of proper priors that produce the same coherent posterior inferences as Haldane's improper one. Finally, we impose an additional inference principle that allows us to characterise uniquely the immediate predictions for the IDMM inference systems.

Representation theorems for partially exchangeable random variables

We provide representation theorems for both finite and countable sequences of finite-valued random variables that are considered to be partially exchangeable. In their most general form, our results are presented in terms of sets of desirable gambles, a very general framework for modelling uncertainty. Its key advantages are that it allows for imprecision, is more expressive than almost every other imprecise-probabilistic framework and makes conditioning on events with (lower) probability zero non-problematic. We translate our results to more conventional, although less general frameworks as well: lower previsions, linear previsions and probability measures. The usual, precise-probabilistic representation theorems for partially exchangeable random variables are obtained as special cases.

Credal networks under epistemic irrelevance: The sets of desirable gambles approach

We present a new approach to credal networks, which are graphical models that generalise Bayesian networks to deal with imprecise probabilities. Instead of applying the commonly used notion of strong independence, we replace it by the weaker, asymmetrical notion of epistemic irrelevance. We show how assessments of epistemic irrelevance allow us to construct a global model out of given local uncertainty models, leading to an intuitive expression for the so-called irrelevant natural extension of a credal network. In contrast with Cozman [4], who introduced this notion in terms of credal sets, our main results are presented using the language of sets of desirable gambles. This has allowed us to derive some remarkable properties of the irrelevant natural extension, including marginalisation properties and a tight connection with the notion of independent natural extension. Our perhaps most important result is that the irrelevant natural extension satisfies a collection of epistemic irrelevancies that is induced by AD-separation, an asymmetrical adaptation of d-separation. Both AD-separation and the induced collection of irrelevancies are shown to satisfy all graphoid properties except symmetry.

Comments on “Statistical reasoning with set-valued information: Ontic vs. epistemic view” by Inés Couso and Didier Dubois

I discuss some aspects of the distinction between ontic and epistemic views of sets as representation of imprecise or incomplete information. In particular, I consider its implications on imprecise probability representations: credal sets and sets of desirable gambles. It is emphasized that the interpretation of the same mathematical object can be different depending on the point of view from which this element is considered. In the case of a fuzzy information on a random variable, it is possible to define a possibility distribution on the simplex of probability distributions. I add some comments about the properties of this possibility distribution.