Why I am not an objective Bayesian; some reflections prompted by Rosenkrantz

Summary The objective Bayesian program has as its fundamental tenet (in addition to the three Bayesian postulates) the requirement that, from a given knowledge base a particular probability function is uniquely appropriate. This amounts to fixing initial probabilities, based on relatively little information, because Bayes' theorem (conditionalization) then determines the posterior probabilities when the belief state is altered by enlarging the knowledge base. Moreover, in order to reconstruct orthodox statistical procedures within a Bayesian framework, only privileged ignorance probability functions will work.To serve all these ends objective Bayesianism seeks additional principles for specifying ignorance and partial information probabilities. H. Jeffreys' method of invariance (or Jaynes' modification thereof) is used to solve the former problem, and E. Jaynes' rule of maximizing entropy (subject to invariance for continuous distributions) has recently been thought to solve the latter. I have argued that neither policy is acceptable to a Bayesian since each is inconsistent with conditionalization. Invariance fails to give a consistent representation to the state of ignorance professed. The difficulties here parallel familiar weaknesses in the old Laplacean principle of insufficient reason. Maximizing entropy is unsatisfactory because the partial information it works with fails to capture the effect of uncertainty about related nuisance factors. The result is a probability function that represents a state richer in empirical content than the belief state targeted for representation. Alternatively, by conditionalizing on information about a nuisance parameter one may move from a distribution of lower to higher entropy, despite the obvious increase in information available.Each of these two complaints appear to me to be symptoms of the program's inability to formulate rules for picking privileged probability distributions that serve to represent ignorance or near ignorance. Certainly the methods advocated by Jeffreys, Jaynes and Rosenkrantz are mathematically convenient idealizations wherein specified distributions are elevated to the roles of ignorance and partial information distributions. But the cost that goes with the idealization is a violation of conditionalization, and if that is the ante that we must put up to back objective Bayesianism then I propose we look for a different candidate to earn our support.³¹     

Can we rationally learn to coordinate?

In this paper we examine the issue whether individual rationality considerations are sufficient to guarantee that individuals will learn to coordinate. This question is central in any discussion of whether social phenomena (read: conventions) can be explained in terms of a purely individualistic approach. We argue that the positive answers to this general question that have been obtained in some recent work require assumptions which incorporate some convention. This conclusion may be seen as supporting the viewpoint of institutional individualism in contrast to psychological individualism.     

How can an expert system help in choosing the optimal decision?

We introduce a rationality principle for a preference relation on an arbitrary set of lotteries. Such a principle is a necessary and sufficient condition for the existence of an expected utility agreeing with . The same principle also guarantees a rational extension of the preference relation to any larger set of lotteries. When the extended relation is unique with respect to the alternatives under consideration, the decision maker does not need a numerical evaluation in order to make a choice. Such a rationality condition needs little information in order to be applied, and its verification amounts to solving a linear system.The present research is supported by the Research Contract of CNR (Research National Council) 1989 and 1990 Decision Models under uncertainty and risk, for expert systems with incomplete and revisable information.     

On Playing Fair: Professor Binmore on Game Theory and the Social Contract

This paper critically reviews Ken Binmores non- utilitarian and game theoretic solution to the Arrow problem. Binmores solution belongs to the same family as Rawls maximin criterion and requires the use of Nash bargaining theory, empathetic preferences, and results in evolutionary game theory. Harsanyi has earlier presented a solution that relies on utilitarianism, which requires some exogenous valuation criterion and is therefore incompatible with liberalism. Binmores rigorous demonstration of the maximin principle for the first time presents a real alternative to a utilitarian solution.     

A formulation of the determinism hypothesis

The author tries to formulate what a determinist believes to be true. The formulation is based on some concepts defined in a systems-theoretical manner, mainly on the concept of an experiment over the sets A ^m (a set of m-tuples of input values) and B ⁿ (a set of n-tuples of output values) in the time interval (t ₁, ..., t _k ) (symbolically E[t ₁,..., t _k , A ^m , B ⁿ ]), on the concept of a behavior of the system S ^m,n (=(A ^m , B ⁿ )) on the basis of the experiment E[t ₁, ..., t _k , A ^m , B ⁿ ] and, indeed, on the concept of deterministic behavior .... The resulting formulation of the deterministic hypothesis shows that this hypothesis expresses a belief that we always could find some hidden parameters.     

Statistical analysis for negotiation support

In this paper we provide an overview of the issues involved in using statistical analysis to support the process of international negotiation. We will illustrate how the approach can contribute to a negotiator's understanding and control of the interactions that occur during the course of a negotiation. The techniques are suited to the analysis of data collected from ongoing discussions and moves made by the parties. The analyses are used to illuminate influences and processes as they operate in particular cases or in negotiations in general. They do not identify a best strategy or outcome from among alternatives suggested either from theoretical assumptions about rationality and information-processing (see Munier and Rullière's paper in this issue), from personal preference structures (see Spector's paper in this issue), or from a rule-based modeling system (see Kersten's paper in this issue). This distinction should be evident in the discussion to follow, organized into several sections: From Empirical to Normative Analysis; Statistical Analysis for Situational Diagnosis; Time-Series Analysis of Cases, and Knowledge as Leverage Over the Negotiation Process. In a final section, we consider the challenge posed by attempts to implement these techniques with practitioners.     

Dominance and Efficiency in Multicriteria Decision under Uncertainty

This paper proposes several concepts of efficient solutions for multicriteria decision problems under uncertainty. We show how alternative notions of efficiency may be grounded on different decision contexts, depending on what is known about the Decision Maker's (DM) preference structure and probabilistic anticipations. We define efficient sets arising naturally from polar decision contexts. We investigate these sets from the points of view of their relative inclusions and point out some particular subsets which may be especially relevant to some decision situations.     

Notes on the so-called incompleteness problem and on the proposed alternative concept of rational behavior

In reply to McClennen, the paper argues that his criticism is based on a mistaken assumption about the meaning of rationality postulates, to be called the Implication Principle. Once we realize that the Implication Principle has no validity, McClennen's criticisms of what he calls the Reductio Argument and what he calls the Incentive Argument fall to the ground. The rest of the paper criticizes the rationality concept McClennen proposes in lieu of that used by orthodox game theory. It is argued that McClennen's concept is inconsistent with the behavior of real-life intelligent egoists; it is incompatible with the way payoffs are defined in game theory; and it would be highly dangerous as a practical guide to human behavior.The author is indebted to the National Science Foundation for financial support trough Grant GS-3222, administered through the Center for Research in Management Science, University of California, Berkeley.     

Joint Beliefs in Conflictual Coordination Games

The traditional solution concept for noncooperative game theory is the Nash equilibrium, which contains an implicit assumption that players probability distributions satisfy t probabilistic independence. However, in games with more than two players, relaxing this assumption results in a more general equilibrium concept based on joint beliefs (Vanderschraaf, 1995). This article explores the implications of this joint-beliefs equilibrium concept for two kinds of conflictual coordination games: crisis bargaining and public goods provision. We find that, using updating consistent with Bayes rule, players beliefs converge to equilibria in joint beliefs which do not satisfy probabilistic independence. In addition, joint beliefs greatly expand the set of mixed equilibria. On the face of it, allowing for joint beliefs might be expected to increase the prospects for coordination. However, we show that if players use joint beliefs, which may be more likely as the number of players increases, then the prospects for coordination in these games declines vis-à-vis independent beliefs.     

Ordinal utility models of decision making under uncertainty

This paper studies two models of rational behavior under uncertainty whose predictions are invariant under ordinal transformations of utility. The quantile utility model assumes that the agent maximizes some quantile of the distribution of utility. The utility mass model assumes maximization of the probability of obtaining an outcome whose utility is higher than some fixed critical value. Both models satisfy weak stochastic dominance. Lexicographic refinements satisfy strong dominance.The study of these utility models suggests a significant generalization of traditional ideas of riskiness and risk preference. We define one action to be riskier than another if the utility distribution of the latter crosses that of the former from below. The single crossing property is equivalent to a minmax spread of a random variable. With relative risk defined by the single crossing criterion, the risk preference of a quantile utility maximizer increases with the utility distribution quantile that he maximizes. The risk preference of a utility mass maximizer increases with his critical utility value.     

Endogenous risk and protection premiums

Endogenous risk implies an individual perceives he can influence the likelihood that a state of nature will occur. To add structure to endogenous risk models, I define a protection premium for reduced uncertainty about protection efficiency when a stochastic variable enters the probability functionp(x) rather than the utility function. For a binary lottery, a measure of aversion of uncertain protection efficiency(x) =-p(x)/p(x) is defined to unambiguously determine the effects of increased risk on an individual's voluntary contribution to public good supply earmarked to reduce the probability of an undesirable state. Finally, I examine the protection premium in ann-state discrete lottery and when uncertainty exists in both the probability and utility function.     

An incompleteness problem in Harsanyi's general theory of games and certain related theories of non-cooperative games

The theory of games recently proposed by John C. Harsanyi in A General Theory of Rational Behavior in Game Situations, (Econometrica, Vol. 34, No. 3) has one anomalous feature, viz., that it generates for a special class of non-cooperative games solutions which are not equilibrium points. It is argued that this feature of the theory turns on an argument concerning the instability of weak equilibrium points, and that this argument, in turn, involves appeal to an unrestricted version of a postulate subsequently included in the theory in restricted form. It is then shown that if this line of reasoning is permitted, then one must, by parity of reasoning, permit another instability argument. But, if both of these instability arguments are permitted in the construction of the theory, the resultant theory must be incomplete, in the sense that there will be simple non-cooperative games for which such a theory cannot yield solutions. This result is then generalized and shown to be endemic to all theories which have made the equilibrium condition central to the treatment of non-cooperative games. Some suggestions are then offered concerning how this incompleteness problem can be resolved, and what one might expect concerning the postulate structure and implications of a theory of games which embodies the revisions necessitated by a resolution of this problem.This research was supported by a grant to the author from the City University of New York Faculty Research Award Program.     

Constructs and empirical basis in theories of economic behavior

Theories of economic behavior often use as-if-languages: for example, analytical sentences or definitions are used as if they were synthetic and factual-normative theoretical constructs are used as if they were empirical concepts. Such as-if-languages impede the acquisition of knowledge and are apt to encourage the wrong assessment of actual research strategies. The author's criticism is first leveled at revealed-preference theory. In this theory observed behavior is often understood in an empirical sense although it is a pure theoretical construct. Another example can be found in von Mises' representations on marketing behavior: here theoretical valuations are used to achieve a spurious streamlining of reality. Result: Scientists should not ogle with reality if they have nothing to say about it.     

Similarity and preferences in the space of simple lotteries

A. Rubinstein's similarity-relation approach to decision making under uncertainty is extended by introducing a new concept of correlated similarity relations. It is shown that every expected utility preference is consistent with the -difference similarity relation on the prize space and some correlated similarity relations on the probability space, and that a similarity relation on the prize space and correlated similarity relations on the probability space overdetermine the preference (in A. Rubinstein's sense).Earlier versions were presented at the Edwin Smart Symposium on Games and Economic Behavior, Ohio State University, 1989, the ESEM 1989 Meeting, the EEA 1989 Meeting, the Workshop on Rational and Boundedly Rational Principles of Strategic Behavior, Universität Bielefeld, 1989, and the ASSET 1991 Meeting. We wish to thank the comments received from the participants in those meetings and the stimulating suggestions of S. Ch. Kolm, R. D. Luce, and an anonymous referee. Financial support from the Universidad del Pais Vasco-Euskal Herriko Unibertsitatea, DGCYT (PS 87-0039) and Gobierno de Navarra are gratefully acknowledged.     

A Note on the Computation of the Mean Random Consistency Index of the Analytic Hierarchy Process (Ahp)

In this paper, we use Saaty's Eigenvector Method and the Power Method as well as =1, 2, , 9, 1/2, 1/3, , 1/9} and -={1,2, ,9,1, 1/2, ,1/9} as the sets from which the pairwise comparison judgments are assigned at random to examine the variation in the values determined for the mean random consistency index. By extensive simulation analysis, we found that both methods produce the same values for the mean random consistency random index. Also, we found that the reason for producing two different sets of values is the use of vs. - and not the selection of the Power Method vs. Saaty's Eigenvector Method.     

Expected utility, μ-σ preferences,and linear distribution classes: A further result

This article is an extension of Meyer and Sinn's results on the representation of arbitrary von Neumann-Morgenstern functions in - space when the probability distributions to be compared belong to a linear distribution class. It shows that, when absolute risk aversion decreases, stays constant, or increases not too fast, an increase in , given , increases the indifference curve slope: increased riskiness increases the required marginal compensation for risk when risk is measured by the standard deviation of wealth or income.I gratefully acknowledge useful comments by Wolfgang Buchholz, Kai Konrad, Jack Meyer, Hans Schneeweiß, Dominique Thon, and Bengt Arne Wickström.     

On the foundations of mean-variance analyses

Let (, ) and (, ) be mean-standard deviation pairs of two probability distributions on the real line. Mean-variance analyses presume that the preferred distribution depends solely on these pairs, with primary preference given to larger mean and smaller variance. This presumption, in conjunction with the assumption that one distribution is better than a second distribution if the mass of the first is completely to the right of the mass of the second, implies that (, ) is preferred to (, ) if and only if either > or ( = and < ), provided that the set of distributions is sufficiently rich. The latter provision fails if the outcomes of all distributions lie in a finite interval, but then it is still possible to arrive at more liberal dominance conclusions between (, ) and (, ).This research was supported by the Office of Naval Research.     

Experiments with cooperative 2 × 2 games

Nash's solution of a two-person cooperative game prescribes a coordinated mixed strategy solution involving Pareto-optimal outcomes of the game. Testing this normative solution experimentally presents problems in as much as rather detailed explanations must be given to the subjects of the meaning of threat strategy, strategy mixture, expected payoff, etc. To the extent that it is desired to test the solution using naive subjects, the problem arises of imparting to them a minimal level of understanding about the issue involved in the game without actually suggesting the solution.Experiments were performed to test the properties of the solution of a cooperative two-person game as these are embodied in three of Nash's four axioms: Symmetry, Pareto-optimality, and Invariance with respect to positive linear transformations. Of these, the last was definitely discorroborated, suggesting that interpersonal comparison of utilities plays an important part in the negotiations.Some evidence was also found for a conjecture generated by previous experiments, namely that an externally imposed threat (penalty for non-cooperation) tends to bring the players closer together than the threats generated by the subjects themselves in the process of negotiation.     

Prisoner's dilemma from a moral point of view

In a recent issue of this journal, C. L. Sheng claims to havesolved andexplained the Prisoner's Dilemma (PD) by studying it from a moral point of view - i.e., by assuming that each player feels sympathy for the other. Sheng does not fully clarify this claim, but there is textual evidence that his point is this: PD's arise only for agents who feel little or no sympathy for each other; they cannot arise in the presence of a high degree of reciprocal sympathy. A high degree of such sympathysolves the PD in that it prevents PD's from arising, and a low degree of itexplains the PD in that it provides an essential condition for the occurrence of that game. This thesis is false, as some examples show. These examples are important; they prevent us from underestimating the problem posed by the PD.     

Limited role of entropy in information economics

Information transmitted is defined as the amount by which added evidence (or message received) diminishes uncertainty. The latter is characterized by some properties intuitively suggested by this word and possessed by conditional entropy, a parameter of the posterior probability distribution. However, conditional entropy shares these properties with some other concave symmetric functions on the probability space.Moreover, a given transmission channel (or, in the context of statistical inference, a given experiment) yields a higher maximum expected benefit than anotherto any user if and only ifall concave functions of the posterior probability vector have higher values for the former channel (or experiment). Hence one information system (channel, experiment) may be preferable to another for a given user although its transmission rate, in entropy terms, is lower.But only entropy has the economically relevant property of measuring, in the limit, the expected length of efficiently coded messages sent in long sequences. Thus, while irrelevant to the value (maximum expected benefit) of an information system and to the costs of observing, estimating, and deciding, entropy formulas are indeed relevant to the cost of communicating, i.e., of storing, coding and transmitting messages.Revised version of a paper presented at the 5th Colloquium on Optimization Techniques (International Federation for Information Processing, Rome, May 1973), Lecture Notes series of the Springer Verlag, Heidelberg and New York. The author wishes to thank Springer Verlag for reprint permission. Acknowledgements of support are due to the Alexander von Humboldt Foundation (Special Program for the Cooperation of Research Institutes of F.R.G. and U.S.A.), the U.S. Office of Naval Research and the Institute for Economic Theory, University of Bonn.