Revising beliefs on the basis of evidence

Approaches to belief revision most commonly deal with categorical information: an agent has a set of beliefs and the goal is to consistently incorporate a new item of information given by a formula. However, most information about the real world is not categorical. In revision, one may circumvent this fact by assuming that, in some fashion or other, an agent has elected to accept a formula ?, and the task of revision is to consistently incorporate ? into its belief corpus. Nonetheless, it is worth asking whether probabilistic information and noncategorical beliefs may be reconciled with, or even inform, approaches to revision. In this paper, one such account is presented. An agent receives uncertain information as input, and its probabilities on (a finite set of) possible worlds are updated via Bayesian conditioning. A set of formulas among the noncategorical beliefs is identified as the agent’s categorical belief set. The effect of this updating on the belief set is examined with respect to its appropriateness as a revision operator. We show that few of the classical AGM belief revision postulates are satisfied by this approach. Most significantly, though not surprisingly, the success postulate is not guaranteed to hold. However it does hold after a sufficient number of iterations. As well, it proves to be the case that in revising by a formula consistent with the agent’s beliefs, revision does not correspond to expansion. Postulates for iterated revision also examined, and it proves to be the case that most such postulates also do not hold. On the other hand, limiting cases of the presented approach correspond to specific approaches to revision that have appeared in the literature.     

Bounded rationality and correlated equilibria

We study an interactive framework that explicitly allows for nonrational behavior. We do not place any restrictions on how players’ behavior deviates from rationality, but rather, on players’ higher-order beliefs about the frequency of such deviations. We assume that there exists a probability p such that all players believe, with at least probability p, that their opponents play rationally. This, together with the assumption of a common prior, leads to what we call the set of p-rational outcomes, which we define and characterize for arbitrary probability p. We then show that this set varies continuously in p and converges to the set of correlated equilibria as p approaches 1, thus establishing robustness of the correlated equilibrium concept to relaxing rationality and common knowledge of rationality. The p-rational outcomes are easy to compute, also for games of incomplete information. Importantly, they can be applied to observed frequencies of play for arbitrary normal-form games to derive a measure of rationality \(\overline{p}\) that bounds from below the probability with which any given player chooses actions consistent with payoff maximization and common knowledge of payoff maximization.     

An epistemic analysis of the Harsanyi transformation

Harsanyi (1967–68) proposed a method for transforming uncertainty over the strategy sets of players into uncertainty over their payoffs. The transformation appears to rely on an assumption that the players are rational, or, indeed, that they are rational and that there is common belief of rationality. Such an assumption would be awkward from the perspective of the epistemic program, which is often interested in the implications of irrationality or a lack of common belief of rationality. This paper shows that without common belief of rationality, such implications are not necessarily maintained under a Harsanyi transformation. The paper then shows how, with the belief-system model of Aumann and Brandenburger (1995), such implications can be maintained in the absence of common belief of rationality. Received: December 2000/Revised: February 2002     

Core Solutions in Vector-Valued Games

In this paper, we analyze core solution concepts for vector-valued cooperative games. In these games, the worth of a coalition is given by a vector rather than by a scalar. Thus, the classical concepts in cooperative game theory have to be revisited and redefined; the important principles of individual and collective rationality must be accommodated; moreover, the sense given to the domination relationship gives rise to two different theories. Although different, we show the areas which they share. This analysis permits us to propose a common solution concept that is analogous to the core for scalar cooperative games.     

Reflexive games and non-Archimedean probabilities

In reflexive games we deal with an unlimited hierarchy of cognitive pictures. Aumann’s understanding of common knowledge satisfies the classical intuition that we can appeal only to inductive sets in our reasoning about these cognitive pictures involved in reflexive games. In this paper I propose to deny this intuition and appeal to non-Archimedean probabilities in defining cognitive pictures of our reflexion. This allows us to define reflexive games of finite or infinite levels.     

Game theory with translucent players

A traditional assumption in game theory is that players are opaque to one another—if a player changes strategies, then this change in strategies does not affect the choice of other players’ strategies. In many situations this is an unrealistic assumption. We develop a framework for reasoning about games where the players may be translucent to one another; in particular, a player may believe that if she were to change strategies, then the other player would also change strategies. Translucent players may achieve significantly more efficient outcomes than opaque ones. Our main result is a characterization of strategies consistent with appropriate analogues of common belief of rationality. Common Counterfactual Belief of Rationality (CCBR) holds if (1) everyone is rational, (2) everyone counterfactually believes that everyone else is rational (i.e., all players i believe that everyone else would still be rational even if i were to switch strategies), (3) everyone counterfactually believes that everyone else is rational, and counterfactually believes that everyone else is rational, and so on. CCBR characterizes the set of strategies surviving iterated removal of minimax-dominated strategies, where a strategy \(\sigma \) for player i is minimax dominated by \(\sigma '\) if the worst-case payoff for i using \(\sigma '\) is better than the best possible payoff using \(\sigma \).     

Reduced games,consistency, and the core

This paper establishes an axiomatization of the core by means of an internal consistency property with respect to a new reduced game introduced by Moulin (1985). Given a payoff vector chosen by a solution for some game, and given a subgroup of agents, we define thereduced game as that in which each coalition in the subgroup could attain payoffs to its members only if they are compatible with the initial payoffs toall the members outside of the subgroup. The solution isconsistent if it selects the same payoff distribution for the reduced game as initially. We show that consistency together with individual rationality characterizes the core of both transferable and non-transferable utility games.     

Overriding subsuming rules

This paper is concerned with intelligent agents that are able to perform nonmonotonic reasoning, not only with, but also about general rules with exceptions. More precisely, the focus is on enriching a knowledge base Γ with a general rule that is subsumed by other rules already there. Such a problem is important because evolving knowledge needs not follow logic as it is well-known from e.g. the belief revision paradigm. However, belief revision is mainly concerned with the case that the extra information logically conflicts with Γ. Otherwise, the extra knowledge is simply doomed to extend Γ with no change altogether. The problem here is different and may require a change in Γ even though no inconsistency arises. The idea is that when a rule is to be added, it might need to override any rule that subsumes it: preemption must take place. A formalism dedicated to reasoning with and about rules with exceptions is introduced. An approach to dealing with preemption over such rules is then developed. Interestingly, it leads us to introduce several implicants concepts for rules that are possibly defeasible.     

Continuity of the value and optimal strategies when common priors change

We show that the value of a zero-sum Bayesian game is a Lipschitz continuous function of the players?? common prior belief with respect to the total variation metric on beliefs. This is unlike the case of general Bayesian games where lower semi-continuity of Bayesian equilibrium (BE) payoffs rests on the ??almost uniform?? convergence of conditional beliefs. We also show upper semi-continuity (USC) and approximate lower semi-continuity (ALSC) of the optimal strategy correspondence, and discuss ALSC of the BE correspondence in the context of zero-sum games. In particular, the interim BE correspondence is shown to be ALSC for some classes of information structures with highly non-uniform convergence of beliefs, that would not give rise to ALSC of BE in non-zero-sum games.     

Proper rationalizability in lexicographic beliefs

Proper consistency is defined by the property that each player takes all opponent strategies into account (is cautious) and deems one opponent strategy to be infinitely more likely than another if the opponent prefers the one to the other (respects preferences). When there is common certain belief of proper consistency, a most preferred strategy is properly rationalizable. Any strategy used with positive probability in a proper equilibrium is properly rationalizable. Only strategies that lead to the backward induction outcome are properly rationalizable in the strategic form of a generic perfect information game. Proper rationalizability can test the robustness of inductive procedures. Final version: December 2001     

Proper belief revision and rationalizability in dynamic games

In this paper we develop an epistemic model for dynamic games in which players may revise their beliefs about the opponents’ utility functions as the game proceeds. Within this framework, we propose a rationalizability concept that is based upon the following three principles: (1) at every instance of the game, a player should believe that his opponents are carrying out optimal strategies, (2) a player, at information set h, should not change his belief about an opponent’s relative ranking of two strategies s and s′ if both s and s′ could have led to h, and (3) the players’ initial beliefs about the opponents’ utility functions should agree on a given profile u of utility functions. Common belief in these events leads to the concept of persistent rationalizability for the profile u of utility functions. It is shown that for a given game tree with observable deviators and a given profile u of utility functions, every properly point-rationalizable strategy is a persistently rationalizable strategy for u. This result implies that persistently rationalizable strategies always exist for all game trees with observable deviators and all profiles of utility functions. We provide an algorithm that can be used to compute the set of persistently rationalizable strategies for a given profile u of utility functions. For generic games with perfect information, persistent rationalizability uniquely selects the backward induction strategy for every player.     

Epistemic characterizations of iterated deletion of inferior strategy profiles in preference-based type spaces

Bonanno (Logics and the foundations of game and decision theory, Amsterdam University Press, Amsterdam, 2008) provides an epistemic characterization for the solution concept of iterated deletion of inferior strategy profiles (IDIP) by embedding strategic-form games with ordinal payoffs in non-probabilistic epistemic models which are built on Kripke frames. In this paper, we will follow the event-based approach to epistemic game theory and supplement strategic games with type space models, where each type is associated with a preference relation on the state space. In such a framework, IDIP can be characterized by the conditions that at least one player has correct beliefs about the state of the world and that there is common belief that every player is rational, has correct beliefs about the state of the world and has strictly monotone preferences. Moreover, we shall compare the epistemic motivations for IDIP and its mixed strategy variant known as strong rationalizability (SR). Presuppose the above conditions. Whenever there is also common belief that players’ preferences are representable by some expected utility function IDIP still applies. But if there is common belief that players’ preferences are representable by some expected payoff function, then SR results.     

多选择NTU对策核心的一个非空条件及公理化

提出了\pi-均衡多选择NTU对策的概念,证明了\pi-均衡多选择NTU对策的核心非空, 定义了多选择NTU对策的非水平性质和缩减对策,给出了相容性和逆相容性等概念. 用个体合理性、单人合理性、相容性和逆相容性对非水平多选择NTU对策的核心进行了公理化.     

A formal system for fuzzy reasoning

A formal system for fuzzy reasoning is described which is capable of dealing rationally with evidence which may be inconsistent and/or involve degrees of belief. The basic idea is that the meaning of each formal sentence should be given by a certain commitment or bet associated with it. Each item of evidence is first expressed in the form of such a (hypothetical) bet, which is then written as a formal sentence in a language related to ?ukasiewicz logic. The sentences may be weighted to express the relative reliability of the various informants. A sentence is considered to “follow” from the evidence if the bet it represents can be offered by a speaker without fear of loss, on the assumption that the bets representing various items of evidence have been offered to him. A detailed account, illustrated by concrete examples, is given of the procedures by which an arbitrary sentence in common language can be translated into a formal sentence. The treatment of inconsistency, degrees of belief, and weights is illustrated by a practical example which is solved in full. It is shown that in most practical cases the computations involved in the process of formal reasoning reduce to a problem in linear programming. In the last section the relation between this system and the procedures advocated by Zadeh is examined. It is shown that, subject to certain modifications in formulas, there is general agreement in the region of overlap.     

The τ-value,The core and semiconvex games

Theτ-value for cooperativen-person games is central in this paper. Conditions are given which guarantee that theτ-value lies in the core of the game. A full-dimensional cone of semiconvex games is introduced. This cone contains the cones of convex and exact games and there is a simple formula for theτ-value for such games. The subclass of semiconvex games with constant gap function is characterized in several ways. It turns out to be an (n+1)-dimensional cone and for all games in this cone the Shapley value, the nucleolus and theτ-value coincide.     

Distributed consensus in noncooperative inventory games

This paper deals with repeated nonsymmetric congestion games in which the players cannot observe their payoffs at each stage. Examples of applications come from sharing facilities by multiple users. We show that these games present a unique Pareto optimal Nash equilibrium that dominates all other Nash equilibria and consequently it is also the social optimum among all equilibria, as it minimizes the sum of all the players’ costs. We assume that the players adopt a best response strategy. At each stage, they construct their belief concerning others probable behavior, and then, simultaneously make a decision by optimizing their payoff based on their beliefs. Within this context, we provide a consensus protocol that allows the convergence of the players’ strategies to the Pareto optimal Nash equilibrium. The protocol allows each player to construct its belief by exchanging only some aggregate but sufficient information with a restricted number of neighbor players. Such a networked information structure has the advantages of being scalable to systems with a large number of players and of reducing each player’s data exposure to the competitors.     

On the Modal Logic of Jeffrey Conditionalization

We continue the investigations initiated in the recent papers (Brown et al. in The modal logic of Bayesian belief revision, 2017; Gyenis in Standard Bayes logic is not finitely axiomatizable, 2018) where Bayes logics have been introduced to study the general laws of Bayesian belief revision. In Bayesian belief revision a Bayesian agent revises (updates) his prior belief by conditionalizing the prior on some evidence using the Bayes rule. In this paper we take the more general Jeffrey formula as a conditioning device and study the corresponding modal logics that we call Jeffrey logics, focusing mainly on the countable case. The containment relations among these modal logics are determined and it is shown that the logic of Bayes and Jeffrey updating are very close. It is shown that the modal logic of belief revision determined by probabilities on a finite or countably infinite set of elementary propositions is not finitely axiomatizable. The significance of this result is that it clearly indicates that axiomatic approaches to belief revision might be severely limited.     

The consistency principle and an axiomatization of the α-core

This paper examines the α-core of strategic games by means of the consistency principle. I provide a new definition of a reduced game for strategic games. And I define consistency (CONS) and two forms of converse consistency (COCONS and COCONS^*) under this definition of reduced games. Then I axiomatize the α-core for families of strategic games with bounded payoff functions by the axioms CONS, COCONS^*, weak Pareto optimality (WPO) and one person rationality (OPR). Furthermore, I show that these four axioms are logically independent. In proving this, I also axiomatize the α-individually rational solution by CONS, COCONS and OPR for the same families of games. Here the α-individually rational solution is a natural extension of the classical `maximin' solution. Received: June 1998/Final version: 6 July 2001     

Behaviorism and belief

The examination of now-abandoned behaviorist analysis of the concept of belief can bring to light defects in perspectives such as functionalism and physicalism that are still considered viable. Most theories have in common that they identify the holding of the belief that p by a subject S with some matter of fact in or about S that is distinct from and independent of p. In the case of behaviorism it is easy to show that this feature of the theory generates incoherence in the first-person point of view since it gives footing to the possibility that S could correctly assert “I believe that p,” (that is, “I have the complex disposition the behaviorist theory identifies with holding the belief that p”) and at the same time deny that p is the case. Parallel incoherence can be developed in the context of other philosophically popular accounts of the nature of belief.     

The logic of justified belief,explicit knowledge,and conclusive evidence

We present a complete, decidable logic for reasoning about a notion of completely trustworthy (“conclusive”) evidence and its relations to justifiable (implicit) belief and knowledge, as well as to their explicit justifications. This logic makes use of a number of evidence-related notions such as availability, admissibility, and “goodness” of a piece of evidence, and is based on an innovative modification of the Fitting semantics for Artemov?s Justification Logic designed to preempt Gettier-type counterexamples. We combine this with ideas from belief revision and awareness logics to provide an account for explicitly justified (defeasible) knowledge based on conclusive evidence that addresses the problem of (logical) omniscience.