Often, when one faces a choice between alternative actions, there are reasons both for and against each alternative. On one way of understanding these words, what one “ought to do all things considered (ATC)” is determined by the totality of these reasons. So, these reasons can somehow be “combined” or “aggregated” to yield an ATC verdict on these alternatives. First, various assumptions about this sort of aggregation of reasons are articulated. Then it is shown that these assumptions allow for the (...) proof of a “Reasons Aggregation Theorem” – parallel to John Harsanyi’s 1955 “Social Aggregation Theorem”. All reasons for action are grounded in reason-providing values; and, for every such reason-providing value, there is in principle a way of measuring how strongly this value counts against each alternative. The theorem tells us that the appropriate measure of how much reason ATC there is against each alternative is simply a weighted sum of the strengths with which these values count against that alternative; the agent ought not ATC to perform an action iff there is ATC more reason against it than against some available alternative. (shrink)

This paper provides an exposition and defence of Lewis' theory of radical interpretation. The first part explains what Lewis' theory was; the second part explains what it wasn't, and in so doing addresses a number of common objections that arise as a result of widespread myths and misunderstandings about how Lewis' theory is supposed to work.

Abstract:One view of practical reasoning is that it involves the weighing of reasons. It is not clear, however, how the weights of reasons combine, especially given the logical and substantive relations among different reasons. Nor is it clear how the weighing of reasons relates to decision theoretic maximization of expected value. This paper constructs a formal model of reasons and their weight in order to shed light on these issues. The model informs philosophical debates about reasons, such as the question (...) of whether reasons or values are more fundamental and the opposing doctrines of atomism and holism about reasons. (shrink)

The main aim of this book is to introduce the topic of limited awareness, and changes in awareness, to those interested in the philosophy of decision-making and uncertain reasoning.

Consequentialists often assume rational monism: the thesis that options are always made rationally permissible by the maximization of the selfsame quantity. This essay argues that consequentialists should reject rational monism and instead accept rational pluralism: the thesis that, on different occasions, options are made rationally permissible by the maximization of different quantities. The essay then develops a systematic form of rational pluralism which, unlike its rivals, is capable of handling both the Newcomb problems that challenge evidential decision theory and the (...) unstable problems that challenge causal decision theory. (shrink)

In Richard Bradley’s book, Decision Theory with a Human Face, we have selected two themes for discussion. The first is the Bolker-Jeffrey theory of decision, which the book uses throughout as a tool to reorganize the whole field of decision theory, and in particular to evaluate the extent to which expected utility theories may be normatively too demanding. The second theme is the redefinition strategy that can be used to defend EU theories against the Allais and Ellsberg paradoxes, a strategy (...) that the book by and large endorses, and even develops in an original way concerning the Ellsberg paradox. We argue that the BJ theory is too specific to fulfil Bradley’s foundational project and that the redefinition strategy fails in both the Allais and Ellsberg cases. Although we share Bradley’s conclusion that EU theories do not state universal rationality requirements, we reach it not by a comparison with BJ theory, but by a comparison with the non-EU theories that the paradoxes have heuristically suggested. (shrink)

Lewis proved a Dutch book theorem for Conditionalization. The theorem shows that an agent who follows any credal update rule other than Conditionalization is vulnerable to bets that inflict a sure loss. Lewis’s theorem is tailored to factive formulations of Conditionalization, i.e. formulations on which the conditioning proposition is true. Yet many scientific and philosophical applications of Bayesian decision theory require a non-factive formulation, i.e. a formulation on which the conditioning proposition may be false. I prove a Dutch book theorem (...) tailored to non-factive Conditionalization. I also discuss the theorem’s significance. (shrink)

Decision theorists have attempted to accommodate several violations of decision theory’s axiomatic requirements by modifying how agents’ choice options are individuated and formally represented. In recent years, prominent authors have worried that these modifications threaten to trivialize decision theory, make the theory unfalsifiable, impose overdemanding requirements on decision theorists, and hamper decision theory’s internal coherence. In this paper, I draw on leading descriptive and normative works in contemporary decision theory to address these prominent concerns. In doing so, I articulate and (...) assess several different criteria for individuating and formally representing agents’ choice options. (shrink)

The impossibility theorem of Dekel, Lipman and Rustichini has been thought to demonstrate that standard state-space models cannot be used to represent unawareness. We first show that Dekel, Lipman and Rustichini do not establish this claim. We then distinguish three notions of awareness, and argue that although one of them may not be adequately modeled using standard state spaces, there is no reason to think that standard state spaces cannot provide models of the other two notions. In fact, standard space (...) models of these forms of awareness are attractively simple. They allow us to prove completeness and decidability results with ease, to carry over standard techniques from decision theory, and to add propositional quantifiers straightforwardly. (shrink)

In their recently published book Nudge (2008) Richard H. Thaler and Cass R. Sunstein (T&S) defend a position labelled as ‘libertarian paternalism’. Their thinking appeals to both the right and the left of the political spectrum, as evidenced by the bedfellows they keep on either side of the Atlantic. In the US, they have advised Barack Obama, while, in the UK, they were welcomed with open arms by the David Cameron's camp (Chakrabortty 2008). I will consider the following questions. What (...) is Nudge? How is it different from social advertisement? Does Nudge induce genuine preference change? Does Nudge build moral character? Is there a moral difference between the use of Nudge as opposed to subliminal images to reach policy objectives? And what are the moral constraints on Nudge? (shrink)

Causal decision theory (CDT) is the best theory of rational choice now available.2 I intend to provide some support for that claim in part I of this paper by responding to two criticisms of CDT. The first criticism says that CDT is superfluous, since it does no better in the problems that matter than does evidential decision theory (EDT) at recommending correct choices. A second criticism says that CDT by itself is flawed: according to this view, there are problems in (...) which CDT makes bad recommendations, unless it is supplemented with an additional deliberation mechanism, either involving metatickles or screening for ratifiable choices.3 I will argue in response to the first criticism that CDT is genuinely better than any EDT of the most sophisticated sort: there are problems where CDT gives better recommendations, and they are problems that do matter. (shrink)

The dominant practice of AI alignment assumes (1) that preferences are an adequate representation of human values, (2) that human rationality can be understood in terms of maximizing the satisfaction of preferences, and (3) that AI systems should be aligned with the preferences of one or more humans to ensure that they behave safely and in accordance with our values. Whether implicitly followed or explicitly endorsed, these commitments constitute what we term apreferentistapproach to AI alignment. In this paper, we characterize (...) and challenge the preferentist approach, describing conceptual and technical alternatives that are ripe for further research. We first survey the limits of rational choice theory as a descriptive model, explaining how preferences fail to capture the thick semantic content of human values, and how utility representations neglect the possible incommensurability of those values. We then critique the normativity of expected utility theory (EUT) for humans and AI, drawing upon arguments showing how rational agents need not comply with EUT, while highlighting how EUT is silent on which preferences are normatively acceptable. Finally, we argue that these limitations motivate a reframing of the targets of AI alignment: Instead of alignment with the preferences of a human user, developer, or humanity-writ-large, AI systems should be aligned with normative standards appropriate to their social roles, such as the role of a general-purpose assistant. Furthermore, these standards should be negotiated and agreed upon by all relevant stakeholders. On this alternative conception of alignment, a multiplicity of AI systems will be able to serve diverse ends, aligned with normative standards that promote mutual benefit and limit harm despite our plural and divergent values. (shrink)

The rule to maximize expected utility is intended for decisions where options involve risk. In those decisions the decision maker's attitude toward risk is important, and the rule ought to take it into account. Allais's and Ellsberg's paradoxes, however, suggest that the rule ignores attitudes toward risk. This suggestion is supported by recent psychological studies of decisions. These studies present a great variety of cases where apparently rational people violate the rule because of aversion or attraction to risk. Here I (...) attempt to resolve the issue concerning expected utility and risk. I distinguish two versions of the rule to maximize expected utility. One adopts a broad interpretation of the consequences of an option and has great intuitive appeal. The other adopts a narrow interpretation of the consequences of an option and seems to have certain technical and practical advantages. I contend that the version of the rule that interprets consequences narrowly does indeed neglect attitudes toward risk. That version of the rule excludes the risk involved in an option from the consequences of the option and, contrary to what is usually claimed, cannot make up for this exclusion through adjustments in probability and utility assignments. I construct a new, general argument that establishes this in a rigorous way. On the other hand, I contend that the version of the rule that interprets consequences broadly takes account of attitudes toward risk by counting the risk involved in an option among the consequences of the option. I rebut some objections to this version of the rules, in particular, the objection that the rule lacks practical interest. Drawing upon the literature on 'mean-risk' decision rules, I show that this version of the rule can be used to solve some realistic decision problems. (shrink)

In two curiously neglected papers, David Lewis claims to reduce to absurdity the supposition (commonly labeled DAB) that (some) desires are belief-like. My aim in this paper is to explain the significance of this claim and rebut the proof.

In Tsuji 1997 the concept of Jeffrey-Keynes algebras was introduced in order to construct a paraconsistent theory of decision under uncertainty. In the present paper we show that these algebras can be used to develop a theory of decision under uncertainty that measures the degree of belief on the quasi (or partial) truth of the propositions. As applications of this new theory of decision, we use it to analyze Popper's paradox of ideal evidence and to indicate a possible way of (...) formalizing Keynes' theory of economic action. (shrink)

This paper defends two related claims about belief. First, the claim that unlike numerical degrees of belief, comparative beliefs are primitive and psychologically real. Second, the claim that the fundamental norm of Probabilism is not that numerical degrees of belief should satisfy the probability axioms, but rather that comparative beliefs should satisfy certain constraints.

The central problem of the dissertation concerns the possibility of a distinction between truth-evaluable and non-truth-evaluable utterances of a natural language. The class of truth-evaluable utterances includes assertions, con. ectures and other kinds of speech act susceptible of truth evaluation. The class of non-truth-evaluable utterances includes commands, exhortations, wishes i.e. utterances not evaluated as being true or false. The problem is placed in the context of radical interpretation theory and it shown that it is a substantial problem of Davidson‘s early (...) theory of radical interpret at ion. I consider the possibility of distinguishing between locutionary and illocutionary act in uttering a sentence and its significance in the present project. I discuss the suggestion that the mood of the verb of the sentence signifies the required distinction between truth-evaluable utterances and non-truth-evaluable ones. I argue that no criterion for the distinction based on the mood of the verb is adequate. The solution that I propose to the problem of classifylng truth-evaluable utterances appeals to mental states. The view that grounds this line of inquiry is that the truth-evaluability of an utterance is a characteristic of it exclusively relevant to the doxastic dimension of the speaker’s mind. I discuss the constraints that the nature of radical interpretation puts upon the way we construe the notion of belief. I propose that a possible classification of mental states into doxastic and non-doxastic that would result in a classification of utterances into truth-evaluable and non-truthevaluable ones can be given by an elaborated version of a decision theoretic scheme. I suggest that a decision theoretic scheme based on a decision theory that, like Savage’s theory, grants independence axioms is a better candidate to offer a solution to the central problem of the dissertation than a scheme based on a non- standard decision theory such as Richard Jeffrey’s. I conclude by showing that the proposal I make satisfies the constraints I have considered and that it can be accommodated by a radical interpretation theory. (shrink)

The notion of a proposition as a set of possible worlds or states occupies central stage in probability theory, semantics and epistemology, where it serves as the fundamental unit both of information and meaning. But this fact should not blind us to the existence of prospects with a different structure. In the paper I examine the use of random variables—in particular, proposition-valued random variables— in these fields and argue that we need a general account of rational attitude formation with respect (...) to them. (shrink)

This article examines Becker's thesis that the hypothesis that choices maximize expected utility relative to fixed and universal tastes provides a general framework for the explanation of behaviour. Three different models of preference revision are presented and their scope evaluated. The first, the classical conditioning model, explains all changes in preferences in terms of changes in the information held by the agent, holding fundamental beliefs and desires fixed. The second, the Jeffrey conditioning model, explains them in terms of changes in (...) both the information held by the agent and changes in her prior beliefs, holding her fundamental desires fixed. The final model, that of generalized conditioning, allows for explanations in terms of changes in the values of all three variables. Key Words: preference change • decision theory • probability • desirability • attitude change. (shrink)

We propose a model of an agent’s probability and utility that is a compromise between Savage (The foundations of statistics, Wiley, 1954) and Jeffrey (The Logic of Decision, McGraw Hill, 1965). In Savage’s model the probability–utility pair is associated with preferences over acts which are assignments of consequences to states. The probability is defined on the state space, and the utility function on consequences. Jeffrey’s model has no consequences, and both probability and utility are defined on the same set of (...) propositions. The probability–utility pair is associated with a desirability relation on propositions. Like Savage we assume a set of consequences and a state space. However, we assume that states are comprehensive, that is, each state describes a consequence, as in Aumann (Econometrica 55:1–18, 1987). Like Jeffrey, we assume that the agent has a preference relation, which we call desirability, over events, which by definition involves uncertainty about consequences. For a given probability and utility of consequences, the desirability relation is presented by conditional expected utility, given an event. We axiomatically characterize desirability relations that are represented by a probability–utility pair. We characterize the family of all the probability–utility pairs that represent a given desirability relation. (shrink)

Suppose we have a group of Bayesian agents, and suppose that theywould like for their group as a whole to be a Bayesian agent as well. Moreover, suppose that thoseagents want the probabilities and utilities attached to this group agent to be aggregated from theindividual probabilities and utilities in reasonable ways. Two ways of aggregating their individual data areavailable to them, viz., ex ante aggregation and ex post aggregation. The former aggregatesexpected utilities directly, whereas the latter aggregates probabilities and utilities (...) separately.A number of recent formal results show that both approaches have problematic implications. This studydiscusses the philosophical issues arising from those results. In this process, I hope to convincethe reader that these results about Bayesian aggregation are highly significant to decision theorists, butalso of immense interest to theorists working in areas such as ethics and political philosophy. (shrink)

This paper reconstructs and evaluates the representation theorem presented by Ramsey in his essay 'Truth and Probability', showing how its proof depends on a novel application of Hölder's theory of measurement. I argue that it must be understood as a solution to the problem of measuring partial belief, a solution that in many ways remains unsurpassed. Finally I show that the method it employs may be interpreted in such a way as to avoid a well known objection to it due (...) to Richard Jeffrey. (shrink)

In The Logic of Decision Richard Jeffrey defends a version of expected utility theory that advises agents to choose acts with an eye to securing evidence for thinking that desirable results will ensue. Proponents of "causal" decision theory have argued that Jeffrey's account is inadequate because it fails to properly discriminate the causal features of acts from their merely evidential properties. Jeffrey's approach has also been criticized on the grounds that it makes it impossible to extract a unique probability/utility representation (...) from a sufficiently rich system of preferences (given a zero and unit for measuring utility). The existence of these problems should not blind us to the fact that Jeffrey's system has advantages that no other decision theory can match: it can be underwritten by a particularly compelling representation theorem proved by Ethan Bolker; and it has a property called partition invariance that every reasonable theory of rational choice must possess. I shall argue that the non-uniqueness problem can be finessed, and that it is impossible to adequately formulate causal decision theory, or any other, without using Jeffrey's theory as one's basic analysis of rational desire. (shrink)

In this paper Richard Jeffrey's 'Logic of Decision' is extended by examination of agents' attitudes to the sorts of possibilities identified by indicative conditional sentences. An expression for the desirability of conditionals is proposed and, along with Adams' thesis that the probability of a conditional equals the conditional probability of its antecedent given its consequent, is defended by informally deriving it from Jeffrey's notion of desirability and some weak constraints on rational preference for conditional possibilities. Finally a statement is given (...) of a representation theorem establishing the conditions under which a rational agent's preferences for conditionals determines the existence of unique measures (up to choice of scale) of her degrees of belief and desire. (shrink)

With characteristic incisiveness Georg Henrik von Wright identifies pro­ haireticIogic (i. e. the logic of preference) as the core of a general theory of value concepts. Essentially, this nucleus involves the logical study of acts from the point of view of their preferability. 1 (italics added) Though the term prohairesis is found in Plato, as well as in Aristotle's treatment of the relations of preference, it is von Wright who introduces this word into contemporary analytical philoso­ phy, and succinctly specifies (...) the philosophical dimensions it encompasses. The above emphasis upon the philosophical study of the formalization of preferences is a matter of utmost importance for understanding the type of in­ quiry this investigation attempts to initiate. Over the past one hundred years the literature on general theories of subjective utility has become massive, where one considers the work done in psychometrics, econometrics, statistical theories, probability theories, etc., etc. Histories in these areas are strong in tracing various evolutions in the development of the concept of preference in decision-making. However, what has not been investigated with sustained at­ tention are the fundamentally philosophical inquiries into the formalization of preference-relations. (shrink)

Isaac Levi and I have different views of probability and decision making. Here, without addressing the merits, I will try to answer some questions recently asked by Levi (1985) about what my view is, and how it relates to his.

Subjective probability considered as a logic of partial belief succumbs to three fundamental fallacies. These concern the representation of preference via expectation, the measurability of partial belief, and the normalization of belief.

This paper provides new foundations for Bayesian Decision Theory based on a representation theorem for preferences defined on a set of prospects containing both factual and conditional possibilities. This use of a rich set of prospects not only provides a framework within which the main theoretical claims of Savage, Ramsey, Jeffrey and others can be stated and compared, but also allows for the postulation of an extended Bayesian model of rational belief and desire from which they can be derived as (...) special cases. The main theorem of the paper establishes the existence of a such a Bayesian representation of preferences over conditional prospects, i.e. the existence of a pair of real-valued functions respectively measuring the agent’s degrees of belief and desire and which satisfy the postulated rationality conditions on partial belief and desire. The representation of partial belief is shown to be unique and that of partial desire, unique up to a linear transformation. (shrink)

This paper investigates the role of conditionals in hypothetical reasoning and rational decision making. Its main result is a proof of a representation theorem for preferences defined on sets of sentences (and, in particular, conditional sentences), where an agent’s preference for one sentence over another is understood to be a preference for receiving the news conveyed by the former. The theorem shows that a rational preference ordering of conditional sentences determines probability and desirability representations of the agent’s degrees of belief (...) and desire that satisfy, in the case of non-conditional sentences, the axioms of Jeffrey’s decision theory and, in the case of conditional sentences, Adams’ expression for the probabilities of conditionals. Furthermore, the probability representation is shown to be unique and the desirability representation unique up to positive linear transformation. (shrink)

In this paper I discuss some of the mathematics behind an often quoted existence theorem from Richard Jeffrey's The Logic of Decision (Jeffrey 1990) in order to pose several new questions about the meaning and value of that mathematics for decision theory.

Changing preferencesis a phenomenonoften invoked but rarely properlyaccounted for. Throughout the history of the social sciences, researchers have come against the possibility that their subjects’ preferenceswere affected by the phenomenato be explainedor by otherfactorsnot taken into accountin the explanation.Sporadically, attempts have been made to systematically investigate these in uences, but none of these seems to have had a lasting impact. Today we are still not much further with respect to preference change than we were at the middle of the last (...) century. This anthology hopes to provide a new impulse for research into this important subject. In particular, we have chosen two routes to amplify this impulse. First, we stress the use of modellingtechniquesfamiliar from economicsand decision theory. Instead of constructing complex, all-encompassing theories of preference change, the authors of this volume start with very simple, formal accounts of some possible and hopefully plausible mechanism of preference change. Eventually, these models may nd their way into larger, empirically adequate theories, but at this stage, we think that the most importantwork lies in building structure.Secondly,we stress the importance of interdisciplinary exchange. Only by drawing together experts from different elds can the complex empirical and theoretical issues in the modelling of preference change be adequately investigated. (shrink)

In this thesis I investigate the behaviour of uncertainty about vague matters. It is a fairly common view that vagueness involves uncertainty of some sort. However there are many fundamental questions about this kind of uncertainty that are left open. Could you be genuinely uncertain about p when there is no matter of fact whether p? Could you remain uncertain in a vague proposition even if you knew exactly which possible world obtained? Should your degrees of belief be probabilistically coherent? (...) Should your beliefs in the vague be fixed by your beliefs in the precise? Could one in principle tell what credences a person has in the vague? (shrink)