Scientists often study hypotheses that they know to be false. This creates an interpretive problem for Bayesians because the probability assigned to a hypothesis is typically interpreted as the probability that the hypothesis is true. I argue that solving the interpretive problem requires coming up with a new semantics for Bayesian inference. I present and contrast two new semantic frameworks, and I argue that both of them support the claim that there is pervasive pragmatic encroachment on whether a (...) given Bayesian probability assignment is rational. (shrink)

We propose a Bayesian framework for the attribution of knowledge, and apply this framework to generate novel predictions about knowledge attribution for different types of “Gettier cases”, in which an agent is led to a justified true belief yet has made erroneous assumptions. We tested these predictions using a paradigm based on semantic integration. We coded the frequencies with which participants falsely recalled the word “thought” as “knew” (or a near synonym), yielding an implicit measure of conceptual activation. Our (...) experiments confirmed the predictions of our Bayesian account of knowledge attribution across three experiments. We found that Gettier cases due to counterfeit objects were not treated as knowledge (Experiment 1), but those due to intentionally-replaced evidence were (Experiment 2). Our findings are not well explained by an alternative account focused only on luck, because accidentally-replaced evidence activated the knowledge concept more strongly than did similar false belief cases (Experiment 3). We observed a consistent pattern of results across a number of different vignettes that varied the quality and type of evidence available to agents, the relative stakes involved, and surface details of content. Accordingly, the present findings establish basic phenomena surrounding people’s knowledge attributions in Gettier cases, and provide explanations of these phenomena within a Bayesian framework. (shrink)

The contributions in this volume focus on the Bayesian interpretation of natural languages, which is widely used in areas of artificial intelligence, cognitive science, and computational linguistics. This is the first volume to take up topics in Bayesian Natural Language Interpretation and make proposals based on information theory, probability theory, and related fields. The methodologies offered here extend to the target semantic and pragmatic analyses of computational natural language interpretation. Bayesian approaches to natural language semantics and (...) pragmatics are based on methods from signal processing and the causal Bayesian models pioneered by especially Pearl. In signal processing, the Bayesian method finds the most probable interpretation by finding the one that maximizes the product of the prior probability and the likelihood of the interpretation. It thus stresses the importance of a production model for interpretation as in Grice’s contributions to pragmatics or in interpretation by abduction. (shrink)

This paper presents a general-purpose update semantics for expressions of subjective uncertainty in natural language. First, a set of desiderata are established for how expressions of subjective uncertainty should behave in dynamic, update-based semantic systems; then extant implementations of expressions of subjective uncertainty in such models are evaluated and found wanting; finally, a new update semantics is proposed. The desiderata at the heart of this paper center around the contention that expressions of subjective uncertainty express beliefs which are (...) not persistent, whereas propositions express beliefs that are persistent. I argue that if we make the move of treating updates in a dynamic semantics as Bayesian updates, i.e. as conditionalization, then expressions of subjective uncertainty will behave the way we want them to without altering the way propositions behave. (shrink)

I develop a conception of expressivism according to which it is chiefly a pragmatic thesis about some fragment of discourse, one imposing certain constraints on semantics. The first half of the paper uses credal expressivism about the language of probability as a stalking-horse for this purpose. The second half turns to the question of how one might frame an analogous form of expressivism about the language of deontic modality. Here I offer a preliminary comparison of two expressivist lines. The (...) first, expectation expressivism, looks again to Bayesian modelling for inspiration: it glosses deontically modal language as characteristically serving to express decision-theoretic expectation (expected utility). The second, plan expressivism, develops the idea (due to Gibbard 2003) that this language serves to express 'plan-laden' states of belief. In the process of comparing the views, I show how to incorporate Gibbard's modelling ideas into a compositional semantics for attitudes and modals, filling a lacuna in the account. I close with the question whether and how plan expressivism might be developed with expectation-like structure. (shrink)

Many researchers want to unify probability and logic by defining logical probability or probabilistic logic reasonably. This paper tries to unify statistics and logic so that we can use both statistical probability and logical probability at the same time. For this purpose, this paper proposes the P–T probability framework, which is assembled with Shannon’s statistical probability framework for communication, Kolmogorov’s probability axioms for logical probability, and Zadeh’s membership functions used as truth functions. Two kinds of probabilities are connected by an (...) extended Bayes’ theorem, with which we can convert a likelihood function and a truth function from one to another. Hence, we can train truth functions (in logic) by sampling distributions (in statistics). This probability framework was developed in the author’s long-term studies on semantic information, statistical learning, and color vision. This paper first proposes the P–T probability framework and explains different probabilities in it by its applications to semantic information theory. Then, this framework and the semantic information methods are applied to statistical learning, statistical mechanics, hypothesis evaluation (including falsification), confirmation, and Bayesian reasoning. Theoretical applications illustrate the reasonability and practicability of this framework. This framework is helpful for interpretable AI. To interpret neural networks, we need further study. (shrink)

This paper develops connections between objective Bayesian epistemology—which holds that the strengths of an agent’s beliefs should be representable by probabilities, should be calibrated with evidence of empirical probability, and should otherwise be equivocal—and probabilistic logic. After introducing objective Bayesian epistemology over propositional languages, the formalism is extended to handle predicate languages. A rather general probabilistic logic is formulated and then given a natural semantics in terms of objective Bayesian epistemology. The machinery of objective Bayesian (...) nets and objective credal nets is introduced and this machinery is applied to provide a calculus for probabilistic logic that meshes with the objective Bayesian semantics. (shrink)

Inductive logic admits a variety of semantics (Haenni et al., 2011, Part 1). This paper develops semantics based on the norms of Bayesian epistemology (Williamson, 2010, Chapter 7). §1 introduces the semantics and then, in §2, the paper explores methods for drawing inferences in the resulting logic and compares the methods of this paper with the methods of Barnett and Paris (2008). §3 then evaluates this Bayesian inductive logic in the light of four traditional critiques (...) of inductive logic, arguing (i) that it is language independent in a key sense, (ii) that it admits connections with the Principle of Indifference but these connections do not lead to paradox, (iii) that it can capture the phenomenon of learning from experience, and (iv) that while the logic advocates scepticism with regard to some universal hypotheses, such scepticism is not problematic from the point of view of scientific theorising. (shrink)

This paper studies a fundamental mechanism by which conflicts between arguments are drawn from sentiments regarding acceptability of the arguments. Given sets of arguments, an inverse abstract argumentation problem seeks attack relations between arguments such that acceptability semantics interprets each argument in the sets of arguments as being acceptable in each of the attack relations. It is an inverse problem of the traditional problem we refer to as the forward abstract argumentation problem. Given an attack relation, the forward abstract (...) argumentation problem seeks sets of arguments such that acceptability semantics interprets each argument in the sets of arguments as being acceptable in the attack relation. We give a probabilistic model of argumentation-theoretic inference. It is a generative model formalising the process by which acceptability semantics interprets acceptability of arguments in a given attack relation. We show that it gives a broad view of solutions to the forward and inverse abstract argumentation problems. Specifically, solutions to the inverse and forward abstract argumentation problems are shown to be equivalent to a maximum likelihood estimate and maximum likelihood prediction, respectively, which are both available with the generative model. In addition, they are shown to be special cases of the posterior distribution and the evidence, respectively, which are both obtained by probabilistic inference on the generative model. We report an experiment result and application example of the generative model in the inverse problems. (shrink)

The theory of rational decision has always been implicitly involved with subjunctive and counterfactual conditionals. “If I were to do A, this would happen; if I were to do B that would happen. ” When I have done A, I use the counterfactual: “If I had done B, the outcome would have been worse. ” Counterfactuals are handled so smoothly in decision theory and game theory that they are hardly ever explicitly discussed except in cases where they cause problems. For (...) many philosophers, Newcomb’s paradox was such a critical case ). In game theory, questions about beliefs off the equilibrium path in extensive form games and related issues of backward and forward induction provoked explicit discussion of counterfactuals , Bicchieri , Binmore , Harsanyi and Selten , Reny , Stalnaker , ). Where subjunctive conditionals cause problems, we need an explicit theory. What theory should we use? There is the selection function semantics of Stalnaker based on the idea of “most similar possible world in which the antecedent is true”, and related theories of Woodruff , Lewis , Pollock , Nute , which relax some of Stalnaker’s assumptions. (shrink)

Lexical ambiguity can be syntactic if it involves more than one grammatical category for a single word, or semantic if more than one meaning can be associated with a word. In this article we discuss the application of a Bayesian‐network model in the resolution of lexical ambiguities of both types. The network we propose comprises a parsing subnetwork, which can be constructed automatically for any context‐free grammar, and a subnetwork for semantic analysis, which, in the spirit of Fillmore's (1968) (...) case grammars, seeks to fulfill the required cases of all candidates for verb of the sentence. Solving for the highest joint probability of the variables conditioned upon the evidences to the network yields the most likely candidate with its meaning, along with its cases and respective meanings. This is achieved by fixing the values of all evidence nodes concurrently, and then performing a stochastic simulation in which the remaining nodes are updated probabilistically with a high degree of parallelism. The process of disambiguation is directed neither by the syntax nor the semantics, but rather by the interrelation between the two subnetworks. The use of a Bayesian‐network model allows us to express this interrelation between the two subnetworks and among their constituents in a rather direct and rigorous way that, in connection with the convergence properties of the stochastic simulation, reveals a very robust model. (shrink)

This paper develops a trivalent semantics for indicative conditionals and extends it to a probabilistic theory of valid inference and inductive learning with conditionals. On this account, (i) all complex conditionals can be rephrased as simple conditionals, connecting our account to Adams's theory of p-valid inference; (ii) we obtain Stalnaker's Thesis as a theorem while avoiding the well-known triviality results; (iii) we generalize Bayesian conditionalization to an updating principle for conditional sentences. The final result is a unified semantic (...) and probabilistic theory of conditionals with attractive results and predictions. (shrink)

We review the various explanations that have been offered toaccount for subjects'' behaviour in Wason ''s famous selection task. Weargue that one element that is lacking is a good understanding ofsubjects'' semantics for the key expressions involved, and anunderstanding of how this semantics is affected by the demands the taskputs upon the subject''s cognitive system. We make novel proposals inthese terms for explaining the major content effects of deonticmaterials. Throughout we illustrate with excerpts from tutorialdialogues which motivate the (...) kinds of analysis proposed. Our long termgoal is an integration of the various insights about conditionalreasoning on offer from different cognitive science methodologies. Thepurpose of this paper is to try to draw the attention of logicians andsemanticists to this area, since we believe that empirical investigationof the cognitive processes involved could benefit from semanticanalyses. (shrink)

Any inferential system in which the addition of new premises can lead to the retraction of previous conclusions is a non-monotonic logic. Classical conditional probability provides the oldest and most widely respected example of non-monotonic inference. This paper presents a semantic theory for a unified approach to qualitative and quantitative non-monotonic logic. The qualitative logic is unlike most other non- monotonic logics developed for AI systems. It is closely related to classical (i.e., Bayesian) probability theory. The semantic theory for (...) qualitative non-monotonic entailments extends in a straightforward way to a semantic theory for quantitative partial entailment relations, and these relations turn out to be the classical probability functions. (shrink)

We develop a systematic approach to quantum probability as a theory of rational betting in quantum gambles. In these games of chance the agent is betting in advance on the outcomes of several incompatible measurements. One of the measurements is subsequently chosen and performed and the money placed on the other measurements is returned to the agent. We show how the rules of rational betting imply all the interesting features of quantum probability, even in such finite gambles. These include the (...) uncertainty principle and the violation of Bell's inequality among others. Quantum gambles are closely related to quantum logic and provide a new semantics to it. We conclude with a philosophical discussion on the interpretation of quantum mechanics. (shrink)

We develop a systematic approach to quantum probability as a theory of rational betting in quantum gambles. In these games of chance, the agent is betting in advance on the outcomes of several (finitely many) incompatible measurements. One of the measurements is subsequently chosen and performed and the money placed on the other measurements is returned to the agent. We show how the rules of rational betting imply all the interesting features of quantum probability, even in such finite gambles. These (...) include the uncertainty principle and the violation of Bell's inequality among others. Quantum gambles are closely related to quantum logic and provide a new semantics for it. We conclude with a philosophical discussion on the interpretation of quantum mechanics. (shrink)

In this paper, we describe a model that learns semantic representations from the distributional statistics of language. This model, however, goes beyond the common bag‐of‐words paradigm, and infers semantic representations by taking into account the inherent sequential nature of linguistic data. The model we describe, which we refer to as a Hidden Markov Topics model, is a natural extension of the current state of the art in Bayesian bag‐of‐words models, that is, the Topics model of Griffiths, Steyvers, and Tenenbaum (...) (2007), preserving its strengths while extending its scope to incorporate more fine‐grained linguistic information. (shrink)

In 1972,Ernst Ulrich and Christine von Weizs ¨acker introduced the concept of pragmatic information with three desiderata:(i) Pragmatic information should assess the impact of a message upon its receiver;(ii)Pragmatic information should vanish in the limits of complete (non-interpretable)'novelty 'and complete 'confirmation';(iii)Pragmatic information should exhibit non-classical properties since novelty and confirmation behave similarly to Fourier pairs of complementary operators in quantum mechanics. It will be shown how these three desiderata can be naturally fulfilled within the framework of Gardenfors' dynamic semantics (...) of Bayesian belief models.(i)The meaning of a message is its impact upon the epistemic states of a cognitive agent. A pragmatic information measure can then be quanti .ed by the average information gain for the transition from a prior to a posterior state.(ii)Total novelty can be represented by the identical proposition, total con- .rmation by the logical consequence of propositions. In both cases, pragmatic information vanishes.(iii)For operators that are neither idempotent nor commuting, novelty and confirmation relative to a message sequence can be defined within Gardenfors' theory of belief revisions.The proposed approach is consistent with measures of relevance derived from statistical decision theory and it contains Bar-Hillel 's and Carnap's theory of semantic information as a special case. (shrink)

In 1972,Ernst Ulrich and Christine von Weizs ¨acker introduced the concept of pragmatic information with three desiderata: Pragmatic information should assess the impact of a message upon its receiver;Pragmatic information should vanish in the limits of complete 'novelty 'and complete 'confirmation';Pragmatic information should exhibit non-classical properties since novelty and confirmation behave similarly to Fourier pairs of complementary operators in quantum mechanics. It will be shown how these three desiderata can be naturally fulfilled within the framework of Gardenfors' dynamic semantics (...) of Bayesian belief models.The meaning of a message is its impact upon the epistemic states of a cognitive agent. A pragmatic information measure can then be quanti .ed by the average information gain for the transition from a prior to a posterior state.Total novelty can be represented by the identical proposition, total con- .rmation by the logical consequence of propositions. In both cases, pragmatic information vanishes.For operators that are neither idempotent nor commuting, novelty and confirmation relative to a message sequence can be defined within Gardenfors' theory of belief revisions.The proposed approach is consistent with measures of relevance derived from statistical decision theory and it contains Bar-Hillel 's and Carnap's theory of semantic information as a special case. (shrink)

When unmanned underwater vehicles perform tasks, the marine environment situation information perceived by their sensors is insufficient and cannot be shared; moreover, the reasoning efficiency of the situation information is not high. To deal with these problems, this paper proposes an ontology-based situation awareness information expression method, using the Bayesian network method to reason about situation information. First, the situation awareness information is determined in uncertain events when performing tasks in the marine environment. The core and application ontologies of (...) UUV situation awareness are also established. Subsequently, semantic rules are constructed, and uncertain events are identified through the corresponding semantic rules. The Jess inference engine is used to update the ontology model. Because the ontology cannot reason about uncertainty, it is transformed into the Bayesian network to reason about the impacts of uncertain events on tasks. Simulation experiments verify the effectiveness and accuracy of the situation awareness reasoning method that combines the ontology and the Bayesian network. (shrink)

Intertheoretic relations are an important topic in the philosophy of science. However, since their classical discussion by Ernest Nagel, such relations have mostly been restricted to relations between pairs of theories in the natural sciences. This paper presents a case study of a new type of intertheoretic relation that is inspired by Montague’s analysis of the linguistic syntax-semantics relation. The paper develops a simple model of this relation. To motivate the adoption of our new model, we show that this (...) model extends the scope of application of the Nagelian models and that it shares the epistemological advantages of the Nagelian model. The latter is achieved in a Bayesian framework. (shrink)

Intertheoretic relations are an important topic in the philosophy of science. However, since their classical discussion by Ernest Nagel, such relations have mostly been restricted to relations between pairs of theories in the natural sciences. In this paper, we present a model of a new type of intertheoretic relation, called 'Montague Reduction', which is assumed in Montague's framework for the analysis and interpretation of natural language syntax. To motivate the adoption of our new model, we show that this model extends (...) the scope of application of the Nagelian (or related) models, and that it shares the epistemological advantages of the Nagelian model. The latter is achieved in a Bayesian framework. (shrink)

Cognition is both empowered and limited by representations. The matrix lens model explicates tasks that are based on frequency counts, conditional probabilities, and binary contingencies in a general fashion. Based on a structural analysis of such tasks, the model links several problems and semantic domains and provides a new perspective on representational accounts of cognition that recognizes representational isomorphs as opportunities, rather than as problems. The shared structural construct of a 2 × 2 matrix supports a set of generic tasks (...) and semantic mappings that provide a unifying framework for understanding problems and defining scientific measures. Our model's key explanatory mechanism is the adoption of particular perspectives on a 2 × 2 matrix that categorizes the frequency counts of cases by some condition, treatment, risk, or outcome factor. By the selective steps of filtering, framing, and focusing on specific aspects, the measures used in various semantic domains negotiate distinct trade-offs between abstraction and specialization. As a consequence, the transparent communication of such measures must explicate the perspectives encapsulated in their derivation. To demonstrate the explanatory scope of our model, we use it to clarify theoretical debates on biases and facilitation effects in Bayesian reasoning and to integrate the scientific measures from various semantic domains within a unifying framework. A better understanding of problem structures, representational transparency, and the role of perspectives in the scientific process yields both theoretical insights and practical applications. (shrink)