Harsanyi claimed that his Aggregation and Impartial Observer Theorems provide a justification for utilitarianism. This claim has been strongly resisted, notably by Sen and Weymark, who argue that while Harsanyi has perhaps shown that overall good is a linear sum of individuals’ von Neumann-Morgenstern utilities, he has done nothing to establish any con- nection between the notion of von Neumann-Morgenstern utility and that of well-being, and hence that utilitarianism does not follow. The present article defends Harsanyi against the (...) Sen-Weymark cri- tique. I argue that, far from being a term with precise and independent quantitative content whose relationship to von Neumann-Morgenstern utility is then a substantive question, terms such as ‘well-being’ suffer (or suffered) from indeterminacy regarding precisely which quantity they refer to. If so, then (on the issue that this article focuses on) Harsanyi has gone as far towards defending ‘utilitarianism in the original sense’ as could coherently be asked. (shrink)

I argue that prioritarianism cannot be assessed in abstraction from an account of the measure of utility. Rather, the soundness of this view crucially depends on what counts as a greater, lesser, or equal increase in a person’s utility. In particular, prioritarianism cannot accommodate a normatively compelling measure of utility that is captured by the axioms of John von Neumann and Oskar Morgenstern’s expected utility theory. Nor can it accommodate a plausible and elegant generalization of this theory that has (...) been offered in response to challenges to von Neumann and Morgenstern. This is, I think, a theoretically interesting and unexpected source of difficulty for prioritarianism, which I explore in this article. (shrink)

The paper summarizes expected utility theory, both in its original von Neumann-Morgenstern version and its later developments, and discusses the normative claims to rationality made by this theory.

People with the kind of preferences that give rise to the St. Petersburg paradox are problematic---but not because there is anything wrong with infinite utilities. Rather, such people cannot assign the St. Petersburg gamble any value that any kind of outcome could possibly have. Their preferences also violate an infinitary generalization of Savage's Sure Thing Principle, which we call the *Countable Sure Thing Principle*, as well as an infinitary generalization of von Neumann and Morgenstern's Independence axiom, which we call (...) *Countable Independence*. In violating these principles, they display foibles like those of people who deviate from standard expected utility theory in more mundane cases: they choose dominated strategies, pay to avoid information, and reject expert advice. We precisely characterize the preference relations that satisfy Countable Independence in several equivalent ways: a structural constraint on preferences, a representation theorem, and the principle we began with, that every prospect has a value that some outcome could have. (shrink)

In his classic book “the Foundations of Statistics” Savage developed a formal system of rational decision making. The system is based on (i) a set of possible states of the world, (ii) a set of consequences, (iii) a set of acts, which are functions from states to consequences, and (iv) a preference relation over the acts, which represents the preferences of an idealized rational agent. The goal and the culmination of the enterprise is a representation theorem: Any preference relation that (...) satisfies certain arguably acceptable postulates determines a (finitely additive) probability distribution over the states and a utility assignment to the consequences, such that the preferences among acts are determined by their expected utilities. Additional problematic assumptions are however required in Savage's proofs. First, there is a Boolean algebra of events (sets of states) which determines the richness of the set of acts. The probabilities are assigned to members of this algebra. Savage's proof requires that this be a σ-algebra (i.e., closed under infinite countable unions and intersections), which makes for an extremely rich preference relation. On Savage's view we should not require subjective probabilities to be σ-additive. He therefore finds the insistence on a σ-algebra peculiar and is unhappy with it. But he sees no way of avoiding it. Second, the assignment of utilities requires the constant act assumption: for every consequence there is a constant act, which produces that consequence in every state. This assumption is known to be highly counterintuitive. The present work contains two mathematical results. The first, and the more difficult one, shows that the σ-algebra assumption can be dropped. The second states that, as long as utilities are assigned to finite gambles only, the constant act assumption can be replaced by the more plausible and much weaker assumption that there are at least two non-equivalent constant acts. The second result also employs a novel way of deriving utilities in Savage-style systems -- without appealing to von Neumann-Morgenstern lotteries. The paper discusses the notion of “idealized agent" that underlies Savage's approach, and argues that the simplified system, which is adequate for all the actual purposes for which the system is designed, involves a more realistic notion of an idealized agent. (shrink)

Although expected utility theory has proven a fruitful and elegant theory in the finite realm, attempts to generalize it to infinite values have resulted in many paradoxes. In this paper, we argue that the use of John Conway's surreal numbers shall provide a firm mathematical foundation for transfinite decision theory. To that end, we prove a surreal representation theorem and show that our surreal decision theory respects dominance reasoning even in the case of infinite values. We then bring our theory (...) to bear on one of the more venerable decision problems in the literature: Pascal's Wager. Analyzing the wager showcases our theory's virtues and advantages. To that end, we analyze two objections against the wager: Mixed Strategies and Many Gods. After formulating the two objections in the framework of surreal utilities and probabilities, our theory correctly predicts that (1) the pure Pascalian strategy beats all mixed strategies, and (2) what one should do in a Pascalian decision problem depends on what one's credence function is like. Our analysis therefore suggests that although Pascal's Wager is mathematically coherent, it does not deliver what it purports to, a rationally compelling argument that people should lead a religious life regardless of how confident they are in theism and its alternatives. (shrink)

In this paper, we examine how a research institution’s social structure and academic opinion leaders’ presence shaped the early adoption of a scientific innovation. Our case considers the early engagement of mathematical economists at the Cowles Commission with John von Neumann and Oskar Morgenstern’s Theory of Games and Economic Behavior. We argue that scholars with administrative leadership functions who were not only scientifically but also organizationally central – in our case the director of research Jacob Marschak – played a (...) crucial role in promoting the adoption of the Theory of Games. We support our argument with a scientometric analysis of all acknowledgments made in 488 papers published from 1944 to 1955 in the two main paper series at Cowles. We apply blockmodeling techniques to the acknowledgments network to reconstruct the formal and informal social structure at Cowles at the time. Our findings suggest that studies of the early adoption of scientific theories benefit from complementary perspectives on the role of academic leaders and scientists. Our study reveals the importance of formal and informal social structures and the agenda of administrative leaders in these structures to explain their adoption. (shrink)

The Humean conception of the self consists in the belief-desire model of motivation and the utility-maximizing model of rationality. This conception has dominated Western thought in philosophy and the social sciences ever since Hobbes’ initial formulation in Leviathan and Hume’s elaboration in the Treatise of Human Nature. Bentham, Freud, Ramsey, Skinner, Allais, von Neumann and Morgenstern and others have added further refinements that have brought it to a high degree of formal sophistication. Late twentieth century moral philosophers such as (...) Rawls, Brandt, Frankfurt, Nagel and Williams have taken it for granted, and have made use of it to supply metaethical foundations for a wide variety of normative moral theories. But the Humean conception of the self also leads to seemingly insoluble problems about moral motivation, rational final ends, and moral justification. Can it be made to work? (shrink)

Recenze: Robert ZIMMER - Martin Morgenstern, Gespräche mit Hans Albert. Münster: LIT Verlag 2011, 168 s.

One striking observation in the history of rational choice models is that those models have not only been used in economics but spread widely across the social and behavioral sciences. How do such model transfers proceed? By closely studying the early efforts to transfer such models by William Riker – a major protagonist in pushing the adoption of game theoretic models in political science – this article examines the transfer process as one of ‘translation’ by which abstract and mathematical rational (...) choice models were constructed and modified such that they applied to a specific target system in a new domain. In this paper, we argue that to overcome a set of challenges that hampered the straightforward transfer of game theoretic models into political science, Riker complemented theoretical and conceptual modifications of von Neumann and Morgenstern’s game schemes with the use of narratives to turn them into applicable and testable models. As such, those narratives played a crucial role in enabling their transfer and ultimately facilitated the applicability of game theoretic models in political science. (shrink)

The bi-polar confrontation between the Soviet Union and the USA involved many leading game theorists from both sides of the Iron Curtain: Oskar Morgenstern, John von Neumann, Michael Intriligator, John Nash, Thomas Schelling and Steven Brams from the United States and Nikolay Vorob’ev, Leon A. Petrosyan, Elena B. Yanovskaya and Olga N. Bondareva from the Soviet Union. The formalization of game theory took place prior to the Cold War but the geopolitical confrontation hastened and shaped its evolution. In our (...) article we outline four similarities and differences between Western GT and Soviet GT: 1) the Iron Curtain resulted in a lagged evolution of GT in the Soviet Union; 2) Soviet GT focused more on operations research and issues of centralized planning; 3) the contemporary Western view on Soviet GT was biased and Soviet contributions, including works on dynamic stability, non-emptiness of the core and many refinements, suggest that Soviet GT was able to catch up to the Western level relatively fast; 4) international conferences, including Vilnius, 1971, fostered interaction between Soviet game theorists and their Western colleagues. In general, we consider the Cold War to be a positive environment for GT in the West and in the Soviet Union. (shrink)

This is a series of lectures on formal decision theory held at the University of Bayreuth during the summer terms 2008 and 2009. It largely follows the book from Michael D. Resnik: Choices. An Introduction to Decision Theory, 5th ed. Minneapolis London 2000 and covers the topics: -/- Decisions under ignorance and risk Probability calculus (Kolmogoroff Axioms, Bayes' Theorem) Philosophical interpretations of probability (R. v. Mises, Ramsey-De Finetti) Neuman-Morgenstern Utility Theory Introductory Game Theory Social Choice Theory (Sen's Paradox of (...) Liberalism, Arrow's Theorem) . (shrink)

Whereas many others have scrutinized the Allais paradox from a theoretical angle, we study the paradox from an historical perspective and link our findings to a suggestion as to how decision theory could make use of it today. We emphasize that Allais proposed the paradox as a normative argument, concerned with ‘the rational man’ and not the ‘real man’, to use his words. Moreover, and more subtly, we argue that Allais had an unusual sense of the normative, being concerned not (...) so much with the rationality of choices as with the rationality of the agent as a person. These two claims are buttressed by a detailed investigation – the first of its kind – of the 1952 Paris conference on risk, which set the context for the invention of the paradox, and a detailed reconstruction – also the first of its kind – of Allais’s specific normative argument from his numerous but allusive writings. The paper contrasts these interpretations of what the paradox historically represented, with how it generally came to function within decision theory from the late 1970s onwards: that is, as an empirical refutation of the expected utility hypothesis, and more specifically of the condition of von Neumann–Morgenstern independence that underlies that hypothesis. While not denying that this use of the paradox was fruitful in many ways, we propose another use that turns out also to be compatible with an experimental perspective. Following Allais’s hints on ‘the experimental definition of rationality’, this new use consists in letting the experiment itself speak of the rationality or otherwise of the subjects. In the 1970s, a short sequence of papers inspired by Allais implemented original ways of eliciting the reasons guiding the subjects’ choices, and claimed to be able to draw relevant normative consequences from this information. We end by reviewing this forgotten experimental avenue not simply historically, but with a view to recommending it for possible use by decision theorists today. (shrink)

A ansiedade gerada pela possível criação de inteligência artificial geral, algo profetizado desde a fundação do campo de pesquisa (i.e., Dartmouth's Summer Research Project on Artificial Intelligence) (MCCARTHY et al., 1955), é algo comumente investigado dentro dos círculos transhumanistas e singularistas (KURZWEIL, 2005; CHALMERS, 2010; TEGMARK, 2016; 2017; CORRÊA; DE OLIVEIRA, 2021). Por exemplo, em seu livro “Superintelligence: Paths, Dangers, Strategies”, Nick Bostrom (2014) apresenta uma série de argumentos para justificar a ideia de que precisamos ser cautelosos em relação a (...) criação de inteligência artificial avançada ou inteligência artificial geral. Contudo, muito dos argumentos levantados, em suas versões originais (i.e., convergência instrumental e ortogonalidade (OMOHUNDRO, 2008), são baseados em noções “vagas” de conceitos complexos (i.e., “inteligência”, “racionalidade”, “agência”). O objetivo deste resumo (apresentação) será introduzir o leitor/ouvinte ao campo de Segurança da Inteligência Artificial. Nesse campo, questões como o “Problema de controle” (i.e., evitar construir inadvertidamente uma inteligência artificial geral que prejudicará seus criadores) e o “Problema de Alinhamento” (i.e., criar sistemas de inteligência artificial que permaneçam seguros mesmo quando atuando de forma autônoma no ambiente real), que são questões muitas vezes debatidas apenas dentro da Filosofia ou Ficção Científica, ganham um aspecto muito mais real e formal (AMODEI et al., 2016, HUBINGER et al., 2019). A criação de alguns dos modelos de inteligência artificial mais avançados já criados (e.g., AlphaGo, GPT-3) (SILVER el al., 2016; BROWN et al., 2020), está ligado a utilização de dois paradigmas da área de Aprendizagem de Máquina, i.e., Aprendizagem Profunda e Aprendizagem por Reforço. Dentro deste contexto, podemos definir “um sistema autônomo inteligente” como “um sistema que implementa uma política ótima (tecnicamente "ϵ-greedy optimal") para alguma função de recompensa” (ORSEAU et al., 2018). Podemos também definir tais sistemas como “maximizadores de utilidade esperada” (VON NEUMANN; MORGENSTERN, 1944). Um sistema de IA maximizando uma função de recompensa (i.e., objetivo) especificada erroneamente pode vir a causar danos reais quando interagindo no ambiente real, seja por erros de especificação (i.e., Efeitos Colaterais Negativos, Hackeamento de Recompensa, Interrupção Segura), ou quando o domínio de treinamento é diferente do domínio de implantação (i.e., Exploração Segura e Mudança de Domínio). Ao mesmo tempo, especificar objetivos complexos (e.g., preferências e valores humanos) é uma tarefa extremamente complexa, e quando visamos a possibilidade de inteligência artificial avançada agindo no mundo real, é imperativo que tais sistemas estejam: (1) alinhados com nossos objetivos; e (2) desenvolvidos com um design que evite os problemas citados acima. Contudo, tais problemas permanecem em aberto. Por fim, gostaria de ressaltar que no cerne desta problemática, não temos apenas problemas técnicos de aprendizagem de máquina, mas uma importante questão de cunho “Moral”: “O que é a coisa ‘certa” a fazer”? “Qual o objetivo correto a ser perseguido”? “Como podemos especificar isso”? Perguntas como essas prometem um fértil campo para pesquisas interdisciplinares envolvendo inteligência artificial, aprendizagem de máquina, ciências da computação e filosofia moral. (shrink)

This monographic chapter explains how expected utility (EU) theory arose in von Neumann and Morgenstern, how it was called into question by Allais and others, and how it gave way to non-EU theories, at least among the specialized quarters of decion theory. I organize the narrative around the idea that the successive theoretical moves amounted to resolving Duhem-Quine underdetermination problems, so they can be assessed in terms of the philosophical recommendations made to overcome these problems. I actually follow Duhem's (...) recommendation, which was essentially to rely on the passing of time to make many experiments and arguments available, and evebntually strike a balance between competing theories on the basis of this improved knowledge. Although Duhem's solution seems disappointingly vague, relying as it does on "bon sens" to bring an end to the temporal process, I do not think there is any better one in the philosophical literature, and I apply it here for what it is worth. In this perspective, EU theorists were justified in resisting the first attempts at refuting their theory, including Allais's in the 50s, but they would have lacked "bon sens" in not acknowledging their defeat in the 80s, after the long process of pros and cons had sufficiently matured. This primary Duhemian theme is actually combined with a secondary theme - normativity. I suggest that EU theory was normative at its very beginning and has remained so all along, and I express dissatisfaction with the orthodox view that it could be treated as a straightforward descriptive theory for purposes of prediction and scientific test. This view is usually accompanied with a faulty historical reconstruction, according to which EU theorists initially formulated the VNM axioms descriptively and retreated to a normative construal once they fell threatened by empirical refutation. From my historical study, things did not evolve in this way, and the theory was both proposed and rebutted on the basis of normative arguments already in the 1950s. The ensuing, major problem was to make choice experiments compatible with this inherently normative feature of theory. Compability was obtained in some experiments, but implicitly and somewhat confusingly, for instance by excluding overtly incoherent subjects or by creating strong incentives for the subjects to reflect on the questions and provide answers they would be able to defend. I also claim that Allais had an intuition of how to combine testability and normativity, unlike most later experimenters, and that it would have been more fruitful to work from his intuition than to make choice experiments of the naively empirical style that flourished after him. In sum, it can be said that the underdetermination process accompanying EUT was resolved in a Duhemian way, but this was not without major inefficiencies. To embody explicit rationality considerations into experimental schemes right from the beginning would have limited the scope of empirical research, avoided wasting resources to get only minor findings, and speeded up the Duhemian process of groping towards a choice among competing theories. (shrink)