Subjective Probability, Misc

It is widely held that the influence of risk on rational decisions is not entirely explained by the shape of an agent’s utility curve. Buchak (Erkenntnis, 2013, Risk and rationality, Oxford University Press, Oxford, in press) presents an axiomatic decision theory, risk-weighted expected utility theory (REU), in which decision weights are the agent’s subjective probabilities modified by his risk-function r. REU is briefly described, and the global applicability of r is discussed. Rabin’s (Econometrica 68:1281–1292, 2000) calibration theorem strongly suggests that (...) plausible levels of risk aversion cannot be fully explained by concave utility functions; this provides motivation for REU and other theories. But applied to the synchronic preferences of an individual agent, Rabin’s result is not as problematic as it may first appear. Theories that treat outcomes as gains and losses (e.g. prospect theory and cumulative prospect theory) account for risk sensitivity in a way not available to REU. Reference points that mark the difference between gains and losses are subject to framing, many instances of which cannot be regarded as rational. However, rational decision theory may recognize the difference between gains and losses, without endorsing all ways of fixing the point of reference. In any event, REU is a very interesting theory. (shrink)

In his account of probable reasoning, Poincare used the concept, or at least the language, of conventions. In particular, he claimed that the prior probabilities essential for inverse probable reasoning are determined conventionally. This paper investigates, in the light of Poincare's well known claim about the conventionality of metric geometry, what this could mean, and how it is related to other views about the determination of prior probabilities. Particular attention is paid to the similarities and differences between Poincare's conventionalism as (...) it applies to probabilities and de Finetti's subjectivism. The aim of the paper is to suggest that in accounts of the development of ideas about probable reasoning, particularly those customarily described as Bayesian, Poincare's discussion deserves more attention than it has so far received. (shrink)

The new paradigm in the psychology of reasoning adopts a Bayesian, or prob- abilistic, model for studying human reasoning. Contrary to the traditional binary approach based on truth functional logic, with its binary values of truth and falsity, a third value that represents uncertainty can be introduced in the new paradigm. A variety of three-valued truth table systems are available in the formal literature, including one proposed by de Finetti. We examine the descriptive adequacy of these systems for natural language (...) indicative condi- tionals and bets on conditionals. Within our framework the so-called “defective” truth table, in which participants choose a third value when the antecedent of the indicative conditional is false, becomes a coherent response. We show that only de Finetti’s system has a good descriptive fit when uncer- tainty is the third value. (shrink)

Decision Theory and Rationality offers a challenging new interpretation of a key theoretical tool in the human and social sciences. This accessible book argues, contrary to orthodoxy in politics, economics, and management science, that decision theory cannot provide a theory of rationality.

This paper provides an account of what it is to have faith in a proposition p, in both religious and mundane contexts. It is argued that faith in p doesn’t require adopting a degree of belief that isn’t supported by one’s evidence but rather it requires terminating one’s search for further evidence and acting on the supposition that p. It is then shown, by responding to a formal result due to I.J. Good, that doing so can be rational in a (...) number of circumstances. If expected utility theory is the correct account of practical rationality, then having faith can be both epistemically and practically rational if the costs associated with gathering further evidence or postponing the decision are high. If a more permissive framework is adopted, then having faith can be rational even when there are no costs associated with gathering further evidence. (shrink)

The objects of credence are the entities to which credences are assigned for the purposes of a successful theory of credence. I use cases akin to Frege's puzzle to argue against referentialism about credence : the view that objects of credence are determined by the objects and properties at which one's credence is directed. I go on to develop a non-referential account of the objects of credence in terms of sets of epistemically possible scenarios.

This research examined whether youth's forecasted risk taking is best predicted by a compensatory (namely, subjective expected utility) or non-compensatory (e.g., single-factor) model. Ninety youth assessed the importance of perceived benefits, importance of perceived drawbacks, subjective probability of benefits, and subjective probability of drawbacks for 16 risky behaviors clustered evenly into recreational and health/safety domains. In both domains, there was strong support for a noncompensatory model in which only the perceived importance of the benefits of engaging in a risky behavior (...) predicted youths' forecasted engagement in risky behavior. The study overcomes earlier methodological weaknesses by fully decomposing participants' assessments into importance and probability aspects for both benefits and drawbacks. As such, the findings provide clear evidence in support of a boundedrationality perspective on youth decision making regarding risk taking. (shrink)

A group is often construed as a single agent with its own probabilistic beliefs (credences), which are obtained by aggregating those of the individuals, for instance through averaging. In their celebrated contribution “Groupthink”, Russell et al. (2015) apply the Bayesian paradigm to groups by requiring group credences to undergo a Bayesian revision whenever new information is learnt, i.e., whenever the individual credences undergo a Bayesian revision based on this information. Bayesians should often strengthen this requirement by extending it to non-public (...) or even private information (learnt by not all or just one individual), or to non-representable information (not corresponding to an event in the algebra on which credences are held). I propose a taxonomy of six kinds of `group Bayesianism', which differ in the type of information for which Bayesian revision of group credences is required: public representable information, private representable information, public non-representable information, and so on. Six corresponding theorems establish exactly how individual credences must (not) be aggregated such that the resulting group credences obey group Bayesianism of any given type, respectively. Aggregating individual credences through averaging is never permitted. One of the theorems – the one concerned with public representable information – is essentially Russell et al.'s central result (with minor corrections). (shrink)

This article presents a comparative theory of subjective argument strength simple enough for application. Using the axioms and corollaries of the theory, anyone with an elementary knowledge of logic and probability theory can produce an at least minimally rational ranking of any set of arguments according to their subjective strength, provided that the arguments in question are descriptive ones in standard form. The basic idea is that the strength of argument A as seen by person x is a function of (...) three factors: x's degree of belief in the premisses of A; x's degree of belief in the conclusion of A under the assumption that all premisses of A are true; and x's belief in the conclusion of A under the assumption that not all premisses of A are true. (shrink)

Bishop Butler, [Butler, 1736], said that probability was the very guide of life. But what interpretations of probability can serve this function? It isn’t hard to see that empirical views won’t do, and many recent writers-for example John Earman, who has said that Bayesianism is “the only game in town”-have been persuaded by various dutch book arguments that only subjective probability will perform the function required. We will defend the thesis that probability construed in this way offers very little guidance, (...) dutch book arguments notwithstanding. We will sketch a way out of the impasse. (shrink)

One response to the problem of logical omniscience in standard possible worlds models of belief is to extend the space of worlds so as to include impossible worlds. It is natural to think that essentially the same strategy can be applied to probabilistic models of partial belief, for which parallel problems also arise. In this paper, I note a difficulty with the inclusion of impossible worlds into probabilistic models. Under weak assumptions about the space of worlds, most of the propositions (...) which can be constructed from possible and impossible worlds are in an important sense inexpressible; leaving the probabilistic model committed to saying that agents in general have at least as many attitudes towards inexpressible propositions as they do towards expressible propositions. If it is reasonable to think that our attitudes are generally expressible, then a model with such commitments looks problematic. (shrink)

The standard representation theorem for expected utility theory tells us that if a subject’s preferences conform to certain axioms, then she can be represented as maximising her expected utility given a particular set of credences and utilities—and, moreover, that having those credences and utilities is the only way that she could be maximising her expected utility. However, the kinds of agents these theorems seem apt to tell us anything about are highly idealised, being always probabilistically coherent with infinitely precise degrees (...) of belief and full knowledge of all a priori truths. Ordinary subjects do not look very rational when compared to the kinds of agents usually talked about in decision theory. In this paper, I will develop an expected utility representation theorem aimed at the representation of those who are neither probabilistically coherent, logically omniscient, nor expected utility maximisers across the board—that is, agents who are frequently irrational. The agents in question may be deductively fallible, have incoherent credences, limited representational capacities, and fail to maximise expected utility for all but a limited class of gambles. (shrink)

Decision-theoretic representation theorems have been developed and appealed to in the service of two important philosophical projects: in attempts to characterise credences in terms of preferences, and in arguments for probabilism. Theorems developed within the formal framework that Savage developed have played an especially prominent role here. I argue that the use of these ‘Savagean’ theorems create significant difficulties for both projects, but particularly the latter. The origin of the problem directly relates to the question of whether we can have (...) credences regarding acts currently under consideration and the consequences which depend on those acts; I argue that such credences are possible. Furthermore, I argue that attempts to use Jeffrey’s non-Savagean theorem in the service of these two projects may not fare much better. (shrink)

Frank Ramsey's ‘Truth and Probability’ sketches a proposal for the empirical measurement of credences, along with a corresponding set of axioms for a representation theorem intended to characterize the preference conditions under which this measurement process is applicable. There are several features of Ramsey's formal system which make it attractive and worth developing. However, in specifying his measurement process and his axioms, Ramsey introduces the notion of an ethically neutral proposition, the assumed existence of which plays a key role throughout (...) Ramsey's system. A number of later representation theorems have also appealed to ethically neutral propositions. The notion of ethical neutrality has often been called into question — in fact, there seem to be good reasons to suppose that no ethically neutral propositions exist. In this paper, I present several new, Ramsey-inspired representation theorems that avoid any appeal to ethical neutrality. These theorems preserve the benefits of Ramsey's system, without paying the cost of ethical neutrality. (shrink)

Comparativism is the view that comparative beliefs (e.g., believing p to be more likely than q) are more fundamental than partial beliefs (e.g., believing p to some degree x), with the latter explicable as theoretical constructs designed to facilitate reasoning about patterns within systems of comparative beliefs that exist under special conditions. In this paper, I fi rst outline several varieties of comparativism, including two `Ramseyan' varieties which generalise the standard `probabilistic' approaches. I then provide a general critique that applies (...) to any and all comparativist views. Ultimately, there are too many things that we ought to be able to say about partial beliefs that comparativism renders unintelligible. Moreover, there are alternative ways to account for the measurement of belief that need not face the same expressive limitations. (shrink)

We present a linear formalism which makes explicit and precise the confirming effect of independent multiple observers and repeated trials on composite ratings, taking as parameters quantitative estimates of the subjective inputs discussed. -/- Note that the subjective probability used here is so used to study the past not predict the future and is rather limited to what has been called in artificial intelligence "certainty factors," which are arbitrary, or, more well-known, the arbitrary values ascribed to predicates in fuzzy "logic." (...) This is clear because the scales themselves are arbitrary and of the radio-button type. (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)

Starting from a recent paper by S. Kaufmann, we introduce a notion of conjunction of two conditional events and then we analyze it in the setting of coherence. We give a representation of the conjoined conditional and we show that this new object is a conditional random quantity, whose set of possible values normally contains the probabilities assessed for the two conditional events. We examine some cases of logical dependencies, where the conjunction is a conditional event; moreover, we give the (...) lower and upper bounds on the conjunction. We also examine an apparent paradox concerning stochastic independence which can actually be explained in terms of uncorrelation. We briefly introduce the notions of disjunction and iterated conditioning and we show that the usual probabilistic properties still hold. (shrink)

In a series of recent papers, Timothy Williamson has argued for the surprising conclusion that there are cases in which you know a proposition in spite of its being overwhelmingly improbable given what you know that you know it. His argument relies on certain formal models of our imprecise knowledge of the values of perceptible and measurable magnitudes. This paper suggests an alternative class of models that do not predict this sort of improbable knowing. I show that such models are (...) motivated by independently plausible principles in the epistemology of perception, the epistemology of estimation, and concerning the connection between knowledge and justified belief. (shrink)

Theories of consciousness divide over whether perceptual consciousness is rich or sparse in specific representational content and whether it requires cognitive access. These two issues are often treated in tandem because of a shared assumption that the representational capacity of cognitive access is fairly limited. Recent research on working memory challenges this shared assumption. This paper argues that abandoning the assumption undermines post-cue-based “overflow” arguments, according to which perceptual conscious is rich and does not require cognitive access. Abandoning it also (...) dissociates the rich/sparse debate from the access question. The paper then explores attempts to reformulate overflow theses in ways that don’t require the assumption of limited capacity. Finally, it discusses the problem of relating seemingly non-probabilistic perceptual consciousness to the probabilistic representations posited by the models that challenge conceptions of cognitive access as capacity-limited. (shrink)

Behaviour norms are considered for decision trees which allow both objective probabilities and uncertain states of the world with unknown probabilities. Terminal nodes have consequences in a given domain. Behaviour is required to be consistent in subtrees. Consequentialist behaviour, by definition, reveals a consequence choice function independent of the structure of the decision tree. It implies that behaviour reveals a revealed preference ordering satisfying both the independence axiom and a novel form of sure-thing principle. Continuous consequentialist behaviour must be expected (...) utility maximizing. Other plausible assumptions then imply additive utilities, subjective probabilities, and Bayes' rule. (shrink)

We propose and investigate an Analogy Principle in the context of Unary Inductive Logic based on a notion of support by structural similarity which is often employed to motivate scientific conjectures.

We propose an Analogy Principle in the context of Unary Inductive Logic and characterize the probability functions which satisfy it. In particular in the case of a language with just two predicates the probability functions satisfying this principle correspond to solutions of Skyrmsʼ ‘Wheel of Fortune’.

We propose two principles of inductive reasoning related to how observed information is handled by conditioning, and justify why they may be said to represent aspects of rational reasoning. A partial classification is given of the probability functions which satisfy these principles.

Kyburg’s opposition to the subjective Bayesian theory, and in particular to its advocates’ indiscriminate and often questionable use of Dutch Book arguments, is documented and much of it strongly endorsed. However, it is argued that an alternative version, proposed by both de Finetti at various times during his long career, and by Ramsey, is less vulnerable to Kyburg’s misgivings. This is a logical interpretation of the formalism, one which, it is argued, is both more natural and also avoids other, widely-made (...) objections to Bayesianism. (shrink)

Automated reasoning about uncertain knowledge has many applications. One difficulty when developing such systems is the lack of a completely satisfactory integration of logic and probability. We address this problem directly. Expressive languages like higher-order logic are ideally suited for representing and reasoning about structured knowledge. Uncertain knowledge can be modeled by using graded probabilities rather than binary truth-values. The main technical problem studied in this paper is the following: Given a set of sentences, each having some probability of being (...) true, what probability should be ascribed to other (query) sentences? A natural wish-list, among others, is that the probability distribution (i) is consistent with the knowledge base, (ii) allows for a consistent inference procedure and in particular (iii) reduces to deductive logic in the limit of probabilities being 0 and 1, (iv) allows (Bayesian) inductive reasoning and (v) learning in the limit and in particular (vi) allows confirmation of universally quantified hypotheses/sentences. We translate this wish-list into technical requirements for a prior probability and show that probabilities satisfying all our criteria exist. We also give explicit constructions and several general characterizations of probabilities that satisfy some or all of the criteria and various (counter) examples. We also derive necessary and sufficient conditions for extending beliefs about finitely many sentences to suitable probabilities over all sentences, and in particular least dogmatic or least biased ones. We conclude with a brief outlook on how the developed theory might be used and approximated in autonomous reasoning agents. Our theory is a step towards a globally consistent and empirically satisfactory unification of probability and logic. (shrink)

Subjective probability plays an increasingly important role in many fields concerned with human cognition and behavior. Yet there have been significant criticisms of the idea that probabilities could actually be represented in the mind. This paper presents and elaborates a view of subjective probability as a kind of sampling propensity associated with internally represented generative models. The resulting view answers to some of the most well known criticisms of subjective probability, and is also supported by empirical work in neuroscience and (...) behavioral psychology. The repercussions of the view for how we conceive of many ordinary instances of subjective probability, and how it relates to more traditional conceptions of subjective probability, are discussed in some detail. (shrink)

The inductive reliability of Bayesian methods is explored. The first result presented shows that for any solvable inductive problem of a general type, there exists a subjective prior which yields a Bayesian inductive method that solves the problem, although not all subjective priors give rise to a successful inductive method for the problem. The second result shows that the same does not hold for computationally bounded agents, so that Bayesianism is "inductively incomplete" for such agents. Finally a consistency proof shows (...) that inductive agents do not need to disregard inductive failure on sets of subjective probability 0 in order to be ideally rational. Together the results reveal the inadequacy of the subjective Bayesian norms for scientific methodology. (shrink)

When can a Bayesian investigator select an hypothesis H and design an experiment (or a sequence of experiments) to make certain that, given the experimental outcome(s), the posterior probability of H will be lower than its prior probability? We report an elementary result which establishes sufficient conditions under which this reasoning to a foregone conclusion cannot occur. Through an example, we discuss how this result extends to the perspective of an onlooker who agrees with the investigator about the statistical model (...) for the data but who holds a different prior probability for the statistical parameters of that model. We consider, specifically, one-sided and two-sided statistical hypotheses involving i.i.d. Normal data with conjugate priors. In a concluding section, using an "improper" prior, we illustrate how the preceding results depend upon the assumption that probability is countably additive. (shrink)

This paper takes two tasks. The one is elaborating on the relationship of inductive logic with decision theory to which later Carnap planned to apply his system (§§1-7); this is a surveying side of this article. The other is revealing the property of our prediction of the future, subjectivity (§§8-11); this is its philosophical aspect. They are both discussed under the name of belief in causation. Belief in causation is a kind of “degree of belief” born about the causal effect (...) of the action. As such, it admits of the analysis by inductive logic. (shrink)

We discuss the cable guy paradox, both as an object of interest in its own right and as something which can be used to illuminate certain issues in the theories of rational choice and belief. We argue that a crucial principle—The Avoid Certain Frustration (ACF) principle—which is used in stating the paradox is false, thus resolving the paradox. We also explain how the paradox gives us new insight into issues related to the Reflection principle. Our general thesis is that principles (...) that base your current opinions on your current opinions about your future opinions need not make reference to the particular times in the future at which you believe you will have those opinions, but they do need to make reference to the particular degrees of belief you believe you will have in the future. (shrink)

In Pure Inductive Logic, the rational principle of Predicate Exchangeability states that permuting the predicates in a given language L and replacing each occurrence of a predicate in an L-sentence phi according to this permutation should not change our belief in the truth of phi. In this paper we study when a prior probability function w on a purely unary language L satisfying Predicate Exchangeability also satisfies the principle of Unary Language Invariance.

We extend the framework of Inductive Logic to Second Order languages and introduce Wilmers' Principle, a rational principle for probability functions on Second Order languages. We derive a representation theorem for functions satisfying this principle and investigate its relationship to the first order principles of Regularity and Super Regularity.

We give a unified account of some results in the development of Polyadic Inductive Logic in the last decade with particular reference to the Principle of Spectrum Exchangeability, its consequences for Instantial Relevance, Language Invariance and Johnson's Sufficientness Principle, and the corresponding de Finetti style representation theorems.

There is a long-standing disagreement in the philosophy of probability and Bayesian decision theory about whether an agent can hold a meaningful credence about an upcoming action, while she deliberates about what to do. Can she believe that it is, say, 70% probable that she will do A, while she chooses whether to do A? No, say some philosophers, for Deliberation Crowds Out Prediction (DCOP), but others disagree. In this paper, we propose a valid core for DCOP, and identify terminological (...) causes for some of the apparent disputes. (shrink)

We investigate the conflict between the ex ante and ex post criteria of social welfare in a new framework of individual and social decisions, which distinguishes between two sources of uncertainty, here interpreted as an objective and a subjective source respectively. This framework makes it possible to endow the individuals and society not only with ex ante and ex post preferences, as is usually done, but also with interim preferences of two kinds, and correspondingly, to introduce interim forms of the (...) Pareto principle. After characterizing the ex ante and ex post criteria, we present a first solution to their conflict that extends the former as much possible in the direction of the latter. Then, we present a second solution, which goes in the opposite direction, and is also maximally assertive. Both solutions translate the assumed Pareto conditions into weighted additive utility representations, and both attribute to the individuals common probability values on the objective source of uncertainty, and different probability values on the subjective source. We discuss these solutions in terms of two conceptual arguments, i.e., the by now classic spurious unanimity argument and a novel informational argument labelled complementary ignorance. The paper complies with the standard economic methodology of basing probability and utility representations on preference axioms, but for the sake of completeness, also considers a construal of objective uncertainty based on the assumption of an exogeneously given probability measure. JEL classification: D70; D81. (shrink)

In this paper, a concept of chance is introduced that is compatible with deterministic physical laws, yet does justice to our use of chance-talk in connection with typical games of chance. We take our cue from what Poincaré called "the method of arbitrary functions," and elaborate upon a suggestion made by Savage in connection with this. Comparison is made between this notion of chance, and David Lewis' conception.

B.K.Matilal, and earlier J.F.Staal, have suggested a reading of the `Nyaya five limb schema' (also sometimes referred to as the Indian Schema or Hindu Syllogism) from Gotama's Nyaya-Sutra in terms of a binary occurrence relation. In this paper we provide a rational justification of a version of this reading as Analogical Reasoning within the framework of Polyadic Pure Inductive Logic.

We investigate the relative probabilistic support afforded by the combination of two analogies based on possibly different, structural similarity within the context of Pure Inductive Logic and under the assumption of Language Invariance. We show that whilst repeated analogies grounded on the same structural similarity only strengthen the probabilistic support this need not be the case when combining analogies based on different structural similarities. That is, two analogies may provide less support than each would individually.

We investigate the relative probabilistic support afforded by the combination of two analogies based on possibly different, structural similarity (as opposed to e.g. shared predicates) within the context of Pure Inductive Logic and under the assumption of Language Invariance. We show that whilst repeated analogies grounded on the same structural similarity only strengthen the probabilistic support this need not be the case when combining analogies based on different structural similarities. That is, two analogies may provide less support than each would (...) individually. (shrink)

Pure Inductive Logic is the study of rational probability treated as a branch of mathematical logic. This monograph, the first devoted to this approach, brings together the key results from the past seventy years, plus the main contributions of the authors and their collaborators over the last decade, to present a comprehensive account of the discipline within a single unified context.