Greaves and Wallace argue that conditionalization maximizes expected accuracy. In this paper I show that their result only applies to a restricted range of cases. I then show that the update procedure that maximizes expected accuracy in general is one in which, upon learning P, we conditionalize, not on P, but on the proposition that we learned P. After proving this result, I provide further generalizations and show that much of the accuracy-first epistemology program is committed to KK-like iteration (...) principles and to the existence of a class of propositions that rational agents will be certain of if and only if they are true. (shrink)

According to Bayesian epistemology, the epistemically rational agent updates her beliefs by conditionalization: that is, her posterior subjective probability after taking account of evidence X, pnew, is to be set equal to her prior conditional probability pold(·|X). Bayesians can be challenged to provide a justification for their claim that conditionalization is recommended by rationality—whence the normative force of the injunction to conditionalize? There are several existing justifications for conditionalization, but none directly addresses the idea that conditionalization (...) will be epistemically rational if and only if it can reasonably be expected to lead to epistemically good outcomes. We apply the approach of cognitive decision theory to provide a justification for conditionalization using precisely that idea. We assign epistemic utility functions to epistemically rational agents; an agent’s epistemic utility is to depend both upon the actual state of the world and on the agent’s credence distribution over possible states. We prove that, under independently motivated conditions, conditionalization is the unique updating rule that maximizes expected epistemic utility. (shrink)

At the heart of the Bayesianism is a rule, Conditionalization, which tells us how to update our beliefs. Typical formulations of this rule are underspecified. This paper considers how, exactly, this rule should be formulated. It focuses on three issues: when a subject’s evidence is received, whether the rule prescribes sequential or interval updates, and whether the rule is narrow or wide scope. After examining these issues, it argues that there are two distinct and equally viable versions of (...) class='Hi'>Conditionalization to choose from. And which version we choose has interesting ramifications, bearing on issues such as whether Conditionalization can handle continuous evidence, and whether Jeffrey Conditionalization is really a generalization of Conditionalization. (shrink)

Why should a Bayesian bother performing an experiment, one the result of which might well upset his own favored credence function? The Ramsey-Good theorem provides a decision theoretic answer. Provided you base your decision on expected utility, and the the experiment is cost-free, performing the experiment and then choosing has at least as much expected utility as choosing without further ado. Furthermore, doing the experiment is strictly preferable just in case at least one possible outcome of the experiment could alter (...) your choice of options. That is, you should pursue the information just in case it could make a difference to what you do. The problem remains, however, why one should do an experiment when nothing of much practical consequence hangs in the balance. One possibility here is to invoke the pure cognitive utility of states. It is shown that one constraint on cognitive value (what I call cogency) guarantees the value of learning: from a cognitive point of view you are justified in performing an experiment if the outcome of the experiment could make a different to what you think. A bonus of the basic framework is that it provides a new decision theoretic argument for the rationality of conditionalization. (shrink)

The applicability of Bayesian conditionalization in setting one’s posterior probability for a proposition, α, is limited to cases where the value of a corresponding prior probability, PPRI(α|∧E), is available, where ∧E represents one’s complete body of evidence. In order to extend probability updating to cases where the prior probabilities needed for Bayesian conditionalization are unavailable, I introduce an inference schema, defeasible conditionalization, which allows one to update one’s personal probability in a proposition by conditioning on a proposition (...) that represents a proper subset of one’s complete body of evidence. While defeasible conditionalization has wider applicability than standard Bayesian conditionalization (since it may be used when the value of a relevant prior probability, PPRI(α|∧E), is unavailable), there are circumstances under which some instances of defeasible conditionalization are unreasonable. To address this difficulty, I outline the conditions under which instances of defeasible conditionalization are defeated. To conclude the article, I suggest that the prescriptions of direct inference and statistical induction can be encoded within the proposed system of probability updating, by the selection of intuitively reasonable prior probabilities. (shrink)

Seeing a red hat can (i) increase my credence in the hat is red, and (ii) introduce a negative dependence between that proposition and po- tential undermining defeaters such as the light is red. The rigidity of Jeffrey Conditionalization makes this awkward, as rigidity preserves inde- pendence. The picture is less awkward given ‘Holistic Conditionalization’, or so it is claimed. I defend Jeffrey Conditionalization’s consistency with underminable perceptual learning and its superiority to Holistic Conditionalization, arguing that (...) the latter is merely a special case of the former, is itself rigid, and is committed to implausible accounts of perceptual con- firmation and of undermining defeat. (shrink)

A vexing question in Bayesian epistemology is how an agent should update on evidence which she assigned zero prior credence. Some theorists have suggested that, in such cases, the agent should update by Kolmogorov conditionalization, a norm based on Kolmogorov’s theory of regular conditional distributions. However, it turns out that in some situations, a Kolmogorov conditionalizer will plan to always assign a posterior credence of zero to the evidence she learns. Intuitively, such a plan is irrational and easily Dutch (...) bookable. In this paper, we propose a revised norm, Kolmogorov-Blackwell conditionalization, which avoids this problem. We prove a Dutch book theorem and converse Dutch book theorem for this revised norm, and relate our results to those of Rescorla (2018). (shrink)

Bayesian Conditionalization is a widely used proposal for how to update one’s beliefs upon the receipt of new evidence. This is in part because of its attention to the totality of one’s evidence, which often includes facts about what one’s new evidence is and how one has come to have it. However, an increasingly popular position in epistemology holds that one may gain new evidence, construed as knowledge, without being in a position to know that one has gained this (...) evidence. These are cases of KK-Failure, cases where one knows p but is not in a position to know that one knows p. This paper assumes that certain KK-Failures are possible and argues that Conditionalization goes wrong in those cases. (shrink)

Van Fraassen famously endorses the Principle of Reflection as a constraint on rational credence, and argues that Reflection is entailed by the more traditional principle of Conditionalization. He draws two morals from this alleged entailment. First, that Reflection can be regarded as an alternative to Conditionalization – a more lenient standard of rationality. And second, that commitment to Conditionalization can be turned into support for Reflection. Van Fraassen also argues that Reflection implies Conditionalization, thus offering a (...) new justification for Conditionalization. I argue that neither principle entails the other, and thus neither can be used to motivate the other in the way van Fraassen says. There are ways to connect Conditionalization to Reflection, but these connections depend on poor assumptions about our introspective access, and are not tight enough to draw the sorts of conclusions van Fraassen wants. Upon close examination, the two principles seem to be getting at two quite independent epistemic norms. (shrink)

One can have no prior credence whatsoever (not even zero) in a temporally indexical claim. This fact saves the principle of conditionalization from potential counterexample and undermines the Elga and Arntzenius/Dorr arguments for the thirder position and Lewis' argument for the halfer position on the Sleeping Beauty Problem, thereby supporting the double-halfer position. -/- .

Conditionalization is a widely endorsed rule for updating one’s beliefs. But a sea of complaints have been raised about it, including worries regarding how the rule handles error correction, changing desiderata of theory choice, evidence loss, self-locating beliefs, learning about new theories, and confirmation. In light of such worries, a number of authors have suggested replacing Conditionalization with a different rule — one that appeals to what I’ll call “ur-priors”. But different authors have understood the rule in different (...) ways, and these different understandings solve different problems. In this paper, I aim to map out the terrain regarding these issues. I survey the different problems that might motivate the adoption of such a rule, flesh out the different understandings of the rule that have been proposed, and assess their pros and cons. I conclude by suggesting that one particular batch of proposals, proposals that appeal to what I’ll call “loaded evidential standards”, are especially promising. (shrink)

Probabilism in epistemology does not have to be of the Bayesian variety. The probabilist represents a person''s opinion as a probability function; the Bayesian adds that rational change of opinion must take the form of conditionalizing on new evidence. I will argue that this is the correct procedure under certain special conditions. Those special conditions are important, and instantiated for example in scientific experimentation, but hardly universal. My argument will be related to the much maligned Reflection Principle (van Fraassen, 1984, (...) 1995), and partly inspired by the work of Brian Skyrms (1987). (shrink)

I show that David Lewis’s principal principle is not preserved under Jeffrey conditionalization. Using this observation, I argue that Lewis’s reason for rejecting the desire as belief thesis and Adams’s thesis applies also to his own principal principle. 1 Introduction2 Adams’s Thesis, the Desire as Belief Thesis, and the Principal Principle3 Jeffrey Conditionalization4 The Principal Principles Not Preserved under Jeffrey Conditionalization5 Inadmissible Experiences.

Are counterfactuals with true antecedents and consequents trivially true? The principle of Conjunction Conditionalization →) is highly controversial. Many philosophers view it as an attractive feature of Lewis’s semantics for counterfactuals that it can easily be modified to avoid this principle. However, Walters and Williams beg to differ. They argue that Conjunction Conditionalization is an indispensable ingredient of any Lewisian semantics, since CC is entailed by standard Lewisian theorems and a plausible semantic claim about irrelevant semifactuals. If this (...) is true, the entire tradition of revisionist counterfactual semantics is misguided, and so are many philosophical theories in which counterfactuals play a role. We argue, in defense of the revisionist tradition, that Walters and Williams’ ‘plausible semantic claim’ is in fact anything but plausible. It turns out to entail semantic principles far more controversial than Conjunction Conditionalization. (shrink)

Conditionalization is one of the central norms of Bayesian epistemology. But there are a number of competing formulations, and a number of arguments that purport to establish it. In this paper, I explore which formulations of the norm are supported by which arguments. In their standard formulations, each of the arguments I consider here depends on the same assumption, which I call Deterministic Updating. I will investigate whether it is possible to amend these arguments so that they no longer (...) depend on it. As I show, whether this is possible depends on the formulation of the norm under consideration. (shrink)

Colin Howson (1995 ) offers a counter-example to the rule of conditionalization. I will argue that the counter-example doesn't hit its target. The problem is that Howson mis-describes the total evidence the agent has. In particular, Howson overlooks how the restriction that the agent learn 'E and nothing else' interacts with the de se evidence 'I have learnt E'.

Accuracy-based arguments for conditionalization and probabilism appear to have a significant advantage over their Dutch Book rivals. They rely only on the plausible epistemic norm that one should try to decrease the inaccuracy of one’s beliefs. Furthermore, conditionalization and probabilism apparently follow from a wide range of measures of inaccuracy. However, we argue that there is an under-appreciated diachronic constraint on measures of inaccuracy which limits the measures from which one can prove conditionalization, and none of the (...) remaining measures allow one to prove probabilism. That is, among the measures in the literature, there are some from which one can prove conditionalization, others from which one can prove probabilism, but none from which one can prove both. Hence at present, the accuracy-based approach cannot underwrite both conditionalization and probabilism. (shrink)

It has frequently been argued recently that Jeffrey Conditionalization (JC) does not permit undermining. For JC to be inapplicable in cases where the evidence could be undermined would severely compromise JC’s range. However, this paper contends that the argument fails to show that JC cannot accommodate undermining. This response turns on using the proper partition to capture the direct impact of our evidence in redistributing our credences.

We present a new argument for the claim that in the Sleeping Beauty problem, the probability that the coin comes up heads is 1/3. Our argument depends on a principle for the updating of probabilities that we call ‘generalized conditionalization’, and on a species of generalized conditionalization we call ‘synchronic conditionalization on old information’. We set forth a rationale for the legitimacy of generalized conditionalization, and we explain why our new argument for thirdism is immune to (...) two attacks that Pust (Synthese 160:97–101, 2008) has leveled at other arguments for thirdism. (shrink)

Accuracy-based arguments for conditionalization and probabilism appear to have a significant advantage over their Dutch Book rivals. They rely only on the plausible epistemic norm that one should try to decrease the inaccuracy of one's beliefs. Furthermore, it seems that conditionalization and probabilism follow from a wide range of measures of inaccuracy. However, we argue that among the measures in the literature, there are some from which one can prove conditionalization, others from which one can prove probabilism, (...) and none from which one can prove both. Hence at present, the accuracy-based approach cannot underwrite both conditionalization and probabilism. (shrink)

In “Generalized Conditionalization and the Sleeping Beauty Problem,” Anna Mahtani and I offer a new argument for thirdism that relies on what we call “generalized conditionalization.” Generalized conditionalization goes beyond conventional conditionalization in two respects: first, by sometimes deploying a space of synchronic, essentially temporal, candidate-possibilities that are not “prior” possibilities; and second, by allowing for the use of preliminary probabilities that arise by first bracketing, and then conditionalizing upon, “old evidence.” In “Beauty and Conditionalization: (...) Reply to Horgan and Mahtani,” Joel Pust replies to the Horgan/Mahtani argument, raising several objections. In my view his objections do not undermine the argument, but they do reveal a need to provide several further elaborations of it—elaborations that I think are independently plausible. In this paper I will address his objections, by providing the elaborations that I think they prompt. (shrink)

Recently, several epistemologists have defended an attractive principle of epistemic rationality, which we shall call Ur-Prior Conditionalization. In this essay, I ask whether we can justify this principle by appealing to the epistemic goal of accuracy. I argue that any such accuracy-based argument will be in tension with Evidence Externalism, i.e., the view that agent's evidence may entail non-trivial propositions about the external world. This is because any such argument will crucially require the assumption that, independently of all empirical (...) evidence, it is rational for an agent to be certain that her evidence will always include truths, and that she will always have perfect introspective access to her own evidence. This assumption is in tension with Evidence Externalism. I go on to suggest that even if we don't accept Evidence Externalism, the prospects for any accuracy-based justification for Ur-Prior Conditionalization are bleak. (shrink)

It has been argued that if the rigidity condition is satisfied, a rational agent operating with uncertain evidence should update her subjective probabilities by Jeffrey conditionalization or else a series of bets resulting in a sure loss could be made against her. We show, however, that even if the rigidity condition is satisfied, it is not always safe to update probability distributions by JC because there exist such sequences of non-misleading uncertain observations where it may be foreseen that an (...) agent who updates her subjective probabilities by JC will end up nearly certain that a false hypothesis is true. We analyze the features of JC that lead to this problem, specify the conditions in which it arises and respond to potential objections. (shrink)

Epistemic decision theory produces arguments with both normative and mathematical premises. I begin by arguing that philosophers should care about whether the mathematical premises (1) are true, (2) are strong, and (3) admit simple proofs. I then discuss a theorem that Briggs and Pettigrew (2020) use as a premise in a novel accuracy-dominance argument for conditionalization. I argue that the theorem and its proof can be improved in a number of ways. First, I present a counterexample that shows that (...) one of the theorem’s claims is false. As a result of this, Briggs and Pettigrew’s argument for conditionalization is unsound. I go on to explore how a sound accuracy-dominance argument for conditionalization might be recovered. In the course of doing this, I prove two new theorems that correct and strengthen the result reported by Briggs and Pettigrew. I show how my results can be combined with various normative premises to produce sound arguments for conditionalization. I also show that my results can be used to support normative conclusions that are stronger than the one that Briggs and Pettigrew’s argument supports. Finally, I show that Briggs and Pettigrew’s proofs can be simplified considerably. (shrink)

This paper shows that any view of future contingent claims that treats such claims as having indeterminate truth values or as simply being false implies probabilistic irrationality. This is because such views of the future imply violations of reflection, special reflection and conditionalization.

In “Generalized Conditionalization and the Sleeping Beauty Problem,” Anna Mahtani and I offer a new argument for thirdism that relies on what we call “generalized conditionalization.” Generalized conditionalization goes beyond conventional conditionalization in two respects: first, by sometimes deploying a space of synchronic, essentially temporal, candidate-possibilities that are not “prior” possibilities; and second, by allowing for the use of preliminary probabilities that arise by first bracketing, and then conditionalizing upon, “old evidence.” In “Beauty and Conditionalization: (...) Reply to Horgan and Mahtani,” Joel Pust replies to the Horgan/mahtani argument, raising several objections. In my view his objections do not undermine the argument, but they do reveal a need to provide several further elaborations of it—elaborations that I think are independently plausible. In this paper I will address his objections, by providing the elaborations that I think they prompt. (shrink)

Conditionalization is an intuitive and popular epistemic principle. By contrast, the Reflection principle is well known to have some very unappealing consequences. But van Fraassen argues that Conditionalization entails Reflection, so that proponents of Conditionalization must accept Reflection and its consequences. Van Fraassen also argues that Reflection implies Conditionalization, thus offering a new justification for Conditionalization. I argue that neither principle entails the other, and thus neither can be used to motivate the other in the (...) way van Fraassen says. I also propose a replacement for Reflection that accounts for the intuitions that made Reflection appealing, but doesn’t lead to Reflection’s bad consequences. (shrink)

A theory of evidential probability is developed from two assumptions:(1) the evidential probability of a proposition is its probability conditional on the total evidence;(2) one's total evidence is one's total knowledge. Evidential probability is distinguished from both subjective and objective probability. Loss as well as gain of evidence is permitted. Evidential probability is embedded within epistemic logic by means of possible worlds semantics for modal logic; this allows a natural theory of higher-order probability to be developed. In particular, it is (...) emphasized that it is sometimes uncertain which propositions are part of one's total evidence; some surprising implications of this fact are drawn out. (shrink)

David Chalmers has recently argued that Bayesian conditionalization is a constraint on conceptual constancy, and that this constraint, together with “standard Bayesian considerations about evidence and updating,” is incompatible with the Quinean claim that every belief is rationally revisable. Chalmers’s argument presupposes that the sort of conceptual constancy that is relevant to Bayesian conditionalization is the same as the sort of conceptual constancy that is relevant to the claim that every belief is rationally revisable. To challenge this presupposition (...) I explicate a sort of “conceptual role” constancy that a rational subject could take to be necessary and sufficient for a rule of Bayesian conditionalization to govern her belief updating, and show that a rational subject may simultaneously commit herself to updating her beliefs in accord with such a rule and accept the claim that every belief is rationally revisable. (shrink)

I show that David Lewis’s principal principle is not preserved under Jeffrey conditionalization. Using this observation, I argue that Lewis’s reason for rejecting the desire as belief thesis and Adams’s thesis applies also to his own principal principle. 1 Introduction2 Adams’s Thesis, the Desire as Belief Thesis, and the Principal Principle3 Jeffrey Conditionalization4 The Principal Principles Not Preserved under Jeffrey Conditionalization5 Inadmissible Experiences.

This paper employs epistemic logic to investigate the philosophical foundations of Bayesian updating in belief revision. By Bayesian updating, we understand the tenet that an agent's degrees of belief—assumed to be encoded as a probability distribution—should be revised by conditionalization on the agent's total knowledge up to that time. A familiar argument, based on the construction of a diachronic Dutch book, purports to show that Bayesian updating is the only rational belief-revision policy. We investigate the conditions under which the (...) premises of this argument might be satisfied. Specifically, we consider the case of an artificial agent whose language (of thought) features a modal operator TK, where TK ψ has the interpretation “My total knowledge is ψ”. We show that every proposition of the form TK ψis epistemically categorical: it determines, for every proposition φ in the agent's language, whether the agent knows that φ. We argue that, for certain artificial agents employing such a language, the diachronic Dutch book argument for Bayesian updating is on firm ground. (shrink)

This discussion note examines a recent argument for the principle that any counterfactual with true components is itself true. That argument rests upon two widely accepted principles of counterfactual logic to which the paper presents counterexamples. The conclusion speculates briefly upon the wider lessons that philosophers should draw from these examples for the semantics of counterfactuals.

Are counterfactuals with true antecedents and consequents automatically true? That is, is Conjunction Conditionalization: if (X & Y), then (X > Y) valid? Stalnaker and Lewis think so, but many others disagree. We note here that the extant arguments for Conjunction Conditionalization are unpersuasive, before presenting a family of more compelling arguments. These arguments rely on some standard theorems of the logic of counterfactuals as well as a plausible and popular semantic claim about certain semifactuals. Denying Conjunction (...) class='Hi'>Conditionalization, then, requires rejecting other aspects of the standard logic of counterfactuals, or else our intuitive picture of semifactuals. (shrink)

Expected accuracy arguments have been used by several authors (Leitgeb and Pettigrew, and Greaves and Wallace) to support the diachronic principle of conditionalization, in updates where there are only finitely many possible propositions to learn. I show that these arguments can be extended to infinite cases, giving an argument not just for conditionalization but also for principles known as ‘conglomerability’ and ‘reflection’. This shows that the expected accuracy approach is stronger than has been realized. I also argue that (...) we should be careful to distinguish diachronic update principles from related synchronic principles for conditional probability. (shrink)

Bayesian decision theory can be viewed as the core of psychological theory for idealized agents. To get a complete psychological theory for such agents, you have to supplement it with input and output laws. On a Bayesian theory that employs strict conditionalization, the input laws are easy to give. On a Bayesian theory that employs Jeffrey conditionalization, there appears to be a considerable problem with giving the input laws. However, Jeffrey conditionalization can be reformulated so that the (...) problem disappears, and in fact the reformulated version is more natural and easier to work with on independent grounds. (shrink)

I argue that three arguments for conditionalization -- the Diachronic Dutch Book, the expected-accuracy maximization argument from Greaves and Wallace, and the accuracy-dominance argument from Briggs and Pettigrew -- can all be improved by narrowing their focus. I suggest alternative, targeted arguments which better identify the flaw involved in non-conditionalizing updates.

Philosophers investigating the interpretation and use of conditional sentences have long been intrigued by the intuitive correspondence between the probability of a conditional `if A, then C' and the conditional probability of C, given A. Attempts to account for this intuition within a general probabilistic theory of belief, meaning and use have been plagued by a danger of trivialization, which has proven to be remarkably recalcitrant and absorbed much of the creative effort in the area. But there is a strategy (...) for avoiding triviality that has been known for almost as long as the triviality results themselves. What is lacking is a straightforward integration of this approach in a larger framework of belief representation and dynamics. This paper discusses some of the issues involved and proposes an account of belief update by conditionalization. (shrink)

Recently, Ernest Sosa (2007) has proposed two novel solutions to the problem of dream skepticism. In the present paper, I argue that Sosa’s first solution falls prey to what I will refer to as the conditionality problem, i.e., the problem of only establishing a conditional—in this case, if x, then I am awake, x being a placeholder for a condition incompatible with dreaming—in a context where it also needs to be established that we can know that the antecedent holds, and (...) as such can infer the consequent, i.e., I am awake. Sosa’s second solution, in terms of so-called reflective knowledge, is shown to land him in the dilemma of either facing yet another conditionality problem, or violating an internalist constraint that he explicitly grants the skeptic with respect to what kind of factors can be legitimately invoked in our account of how we may know the relevant antecedent. For these reasons, I conclude that Sosa has not solved the problem of dream skepticism. (shrink)