Incremental Bayesian Category Learning From Natural Language

Models of category learning have been extensively studied in cognitive science and primarily tested on perceptual abstractions or artificial stimuli. In this paper, we focus on categories acquired from natural language stimuli, that is, words (e.g., chair is a member of the furniture category). We present a Bayesian model that, unlike previous work, learns both categories and their features in a single process. We model category induction as two interrelated subproblems: (a) the acquisition of features that discriminate among categories, and (b) the grouping of concepts into categories based on those features. Our model learns categories incrementally using particle filters, a sequential Monte Carlo method commonly used for approximate probabilistic inference that sequentially integrates newly observed data and can be viewed as a plausible mechanism for human learning. Experimental results show that our incremental learner obtains meaningful categories which yield a closer fit to behavioral data compared to related models while at the same time acquiring features which characterize the learned categories. (An earlier version of this work was published in Frermann and Lapata 2014 .)     

Three Models of Sequential Belief Updating on Uncertain Evidence

Jeffrey updating is a natural extension of Bayesian updating to cases where the evidence is uncertain. But, the resulting degrees of belief appear to be sensitive to the order in which the uncertain evidence is acquired, a rather un-Bayesian looking effect. This order dependence results from the way in which basic Jeffrey updating is usually extended to sequences of updates. The usual extension seems very natural, but there are other plausible ways to extend Bayesian updating that maintain order-independence. I will explore three models of sequential updating, the usual extension and two alternatives. I will show that the alternative updating schemes derive from extensions of the usual rigidity requirement, which is at the heart of Jeffrey updating. Finally, I will establish necessary and sufficient conditions for order-independent updating, and show that extended rigidity is closely related to these conditions.     

Bayesian hypothesis testing for psychologists: A tutorial on the Savage–Dickey method

In the field of cognitive psychology, the p-value hypothesis test has established a stranglehold on statistical reporting. This is unfortunate, as the p-value provides at best a rough estimate of the evidence that the data provide for the presence of an experimental effect. An alternative and arguably more appropriate measure of evidence is conveyed by a Bayesian hypothesis test, which prefers the model with the highest average likelihood. One of the main problems with this Bayesian hypothesis test, however, is that it often requires relatively sophisticated numerical methods for its computation. Here we draw attention to the Savage–Dickey density ratio method, a method that can be used to compute the result of a Bayesian hypothesis test for nested models and under certain plausible restrictions on the parameter priors. Practical examples demonstrate the method’s validity, generality, and flexibility.     

Inexact Knowledge without Improbable Knowing

Abstract

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.     

Exemplars, prototypes, similarities, and rules in category representation: an example of hierarchical bayesian analysis

This article demonstrates the potential of using hierarchical Bayesian methods to relate models and data in the cognitive sciences. This is done using a worked example that considers an existing model of category representation, the Varying Abstraction Model (VAM), which attempts to infer the representations people use from their behavior in category learning tasks. The VAM allows for a wide variety of category representations to be inferred, but this article shows how a hierarchical Bayesian analysis can provide a unifying explanation of the representational possibilities using 2 parameters. One parameter controls the emphasis on abstraction in category representations, and the other controls the emphasis on similarity. Using 30 previously published data sets, this work shows how inferences about these parameters, and about the category representations they generate, can be used to evaluate data in terms of the ongoing exemplar versus prototype and similarity versus rules debates in the literature. Using this concrete example, this article emphasizes the advantages of hierarchical Bayesian models in converting model selection problems to parameter estimation problems, and providing one way of specifying theoretically based priors for competing models.     

Order Matters! Influences of Linear Order on Linguistic Category Learning

Linguistic category learning has been shown to be highly sensitive to linear order, and depending on the task, differentially sensitive to the information provided by preceding category markers (premarkers, e.g., gendered articles) or succeeding category markers (postmarkers, e.g., gendered suffixes). Given that numerous systems for marking grammatical categories exist in natural languages, it follows that a better understanding of these findings can shed light on the factors underlying this diversity. In two discriminative learning simulations and an artificial language learning experiment, we identify two factors that modulate linear order effects in linguistic category learning: category structure and the level of abstraction in a category hierarchy. Regarding category structure, we find that postmarking brings an advantage for learning category diagnostic stimulus dimensions, an effect not present when categories are non-confusable. Regarding levels of abstraction, we find that premarking of super-ordinate categories (e.g., noun class) facilitates learning of subordinate categories (e.g., nouns). We present detailed simulations using a plausible candidate mechanism for the observed effects, along with a comprehensive analysis of linear order effects within an expectation-based account of learning. Our findings indicate that linguistic category learning is differentially guided by pre- and postmarking, and that the influence of each is modulated by the specific characteristics of a given category system.     

Modelling mechanisms with causal cycles

Mechanistic philosophy of science views a large part of scientific activity as engaged in modelling mechanisms. While science textbooks tend to offer qualitative models of mechanisms, there is increasing demand for models from which one can draw quantitative predictions and explanations. Casini et al. (Theoria 26(1):5–33, 2011) put forward the Recursive Bayesian Networks (RBN) formalism as well suited to this end. The RBN formalism is an extension of the standard Bayesian net formalism, an extension that allows for modelling the hierarchical nature of mechanisms. Like the standard Bayesian net formalism, it models causal relationships using directed acyclic graphs. Given this appeal to acyclicity, causal cycles pose a prima facie problem for the RBN approach. This paper argues that the problem is a significant one given the ubiquity of causal cycles in mechanisms, but that the problem can be solved by combining two sorts of solution strategy in a judicious way.     

On an Intuitionistic Modal Logic

In this paper we consider an intuitionistic variant of the modal logic S4 (which we call IS4). The novelty of this paper is that we place particular importance on the natural deduction formulation of IS4— our formulation has several important metatheoretic properties. In addition, we study models of IS4— not in the framework of Kirpke semantics, but in the more general framework of category theory. This allows not only a more abstract definition of a whole class of models but also a means of modelling proofs as well as provability.     

La balance optimisme/pessimisme : détermination des styles d’expectations envers le futur en se basant sur la bidimensionnalité du Life Orientation Test–Revised

IntroductionSince the 1980s, two major conceptions of the representation of optimism and pessimism have been disputed: a unidimensional structure and a bidimensional structure.ObjectiveThe bidimensional properties of the LOT-R in French are further explored in order to determine the styles of expectations towards the future according to the levels of optimism and pessimism.MethodA study carried out on a sample of 913 adults from France (72.7% women, M = 41.14 years) proposes (1) to analyze the factorial structure of the instrument, the relationship between the two constructs as well as their separability, (2) to explore the styles of expectations according to the levels of positive and negative expectations by latent profile analysis, (3) to study the influence of socio-demographic factors (age, sex, socio-professional category) on cognitive expectations about the future.ResultsThe confirmatory factor analyzes establish the replicability of the bidimensional latent structure of the instrument across age and gender groups, with the separability between the two constructs increasing with advancing age. The latent profile analysis supports the existence of three styles of expectations towards the future within the sample: the optimistic style, the mixed style and the pessimistic style. The MIMIC model demonstrates that increasing age and socio-professional category are determinants of levels of optimism and pessimism as well as of expectation styles.ConclusionThis research makes it possible to use the styles of expectations towards the future which empirically models the balance between the levels of optimism and pessimism.     

Exemplar models as a mechanism for performing Bayesian inference

Probabilistic models have recently received much attention as accounts of human cognition. However, most research in which probabilistic models have been used has been focused on formulating the abstract problems behind cognitive tasks and their optimal solutions, rather than on mechanisms that could implement these solutions. Exemplar models are a successful class of psychological process models in which an inventory of stored examples is used to solve problems such as identification, categorization, and function learning. We show that exemplar models can be used to perform a sophisticated form of Monte Carlo approximation known as importance sampling and thus provide a way to perform approximate Bayesian inference. Simulations of Bayesian inference in speech perception, generalization along a single dimension, making predictions about everyday events, concept learning, and reconstruction from memory show that exemplar models can often account for human performance with only a few exemplars, for both simple and relatively complex prior distributions. These results suggest that exemplar models provide a possible mechanism for implementing at least some forms of Bayesian inference.     

Bayesian Estimation of the DINA Q matrix

Cognitive diagnosis models are partially ordered latent class models and are used to classify students into skill mastery profiles. The deterministic inputs, noisy “and” gate model (DINA) is a popular psychometric model for cognitive diagnosis. Application of the DINA model requires content expert knowledge of a Q matrix, which maps the attributes or skills needed to master a collection of items. Misspecification of Q has been shown to yield biased diagnostic classifications. We propose a Bayesian framework for estimating the DINA Q matrix. The developed algorithm builds upon prior research (Chen, Liu, Xu, & Ying, in J Am Stat Assoc 110(510):850–866, 2015) and ensures the estimated Q matrix is identified. Monte Carlo evidence is presented to support the accuracy of parameter recovery. The developed methodology is applied to Tatsuoka’s fraction-subtraction dataset.     

Motivating Williamson's Model Gettier Cases

Abstract

Williamson has a strikingly economical way of showing how justified true belief can fail to constitute knowledge: he models a class of Gettier cases by means of two simple constraints. His constraints can be shown to rely on some unstated assumptions about the relationship between reality and appearance. These assumptions are epistemologically non-trivial but can be defended as plausible idealizations of our actual predicament, in part because they align well with empirical work on the metacognitive dimension of experience.     

Locally Bayesian learning with applications to retrospective revaluation and highlighting

A scheme is described for locally Bayesian parameter updating in models structured as successions of component functions. The essential idea is to back-propagate the target data to interior modules, such that an interior component's target is the input to the next component that maximizes the probability of the next component's target. Each layer then does locally Bayesian learning. The approach assumes online trial-by-trial learning. The resulting parameter updating is not globally Bayesian but can better capture human behavior. The approach is implemented for an associative learning model that first maps inputs to attentionally filtered inputs and then maps attentionally filtered inputs to outputs. The Bayesian updating allows the associative model to exhibit retrospective revaluation effects such as backward blocking and unovershadowing, which have been challenging for associative learning models. The back-propagation of target values to attention allows the model to show trial-order effects, including highlighting and differences in magnitude of forward and backward blocking, which have been challenging for Bayesian learning models.     

Link between early maladaptive schemas and defense mechanisms

IntroductionDefense mechanism and early maladaptive schemas are two concepts distorting the perception of reality.ObjectiveThe aim of this study was to explore the link between two reality-distorting concepts from two theoretical models: early maladaptive schemas from the cognitive and behavioral model and defense mechanisms based on the psychoanalytic model.MethodTwo hundred thirty-two non-clinical participants completed the Defense Style Questionnaire and the Young Schema Questionnaire (short version). Then a Bravais Pearson correlation analysis connecting these two concepts, and a multiple regressions analysis using early maladaptive schemas as predictors for defense style mechanisms levels were conducted.ResultsThe results indicate that 2 early maladaptive schema domains (i.e. other-directedness as well as over-vigilance and inhibition) predict the frequency of use of the neurotic defense mechanism, and 3 schema domains (i.e. disconnection and rejection, impaired autonomy and performances as well as impaired limits) predict the frequency of use of the immature defense mechanism.ConclusionTo conclude, two psychological concepts based on two different theoretical models (psychoanalytic and cognitive and behavioral therapy) seem to share an important link justifying the use of integrative therapies such as schema therapy.     

Dynamic Bayesian Network Modeling of Game-Based Diagnostic Assessments

Abstract

Digital games offer an appealing environment for assessing student proficiencies, including skills and misconceptions in a diagnostic setting. This paper proposes a dynamic Bayesian network modeling approach for observations of student performance from an educational video game. Drawing from and advancing methods in dynamic Bayesian networks, cognitive diagnostic modeling, and analysis of process data, a Bayesian approach to model construction, calibration, and use in facilitating inferences about students on the fly is described, and implemented in the context of an educational video game.     

Simple use of BIC to Assess Model Selection Uncertainty: An Illustration using Mediation and Moderation Models

Abstract

The Bayesian information criterion (BIC) has been used sometimes in SEM, even adopting a frequentist approach. Using simple mediation and moderation models as examples, we form posterior probability distribution via using BIC, which we call the BIC posterior, to assess model selection uncertainty of a finite number of models. This is simple but rarely used. The posterior probability distribution can be used to form a credibility set of models and to incorporate prior probabilities for model comparisons and selections. This was validated by a large scale simulation and results showed that the approximation via the BIC posterior is very good except when both the sample sizes and magnitude of parameters are small. We applied the BIC posterior to a real data set, and it has the advantages of flexibility in incorporating prior, addressing overfitting problems, and giving a full picture of posterior distribution to assess model selection uncertainty.     

The dynamics of fidelity over the time course of long-term memory

Bayesian models of cognition assume that prior knowledge about the world influences judgments. Recent approaches have suggested that the loss of fidelity from working to long-term (LT) memory is simply due to an increased rate of guessing (e.g. Brady, Konkle, Gill, Oliva, & Alvarez, 2013). That is, recall is the result of either remembering (with some noise) or guessing. This stands in contrast to Bayesian models of cognition while assume that prior knowledge about the world influences judgments, and that recall is a combination of expectations learned from the environment and noisy memory representations. Here, we evaluate the time course of fidelity in LT episodic memory, and the relative contribution of prior category knowledge and guessing, using a continuous recall paradigm. At an aggregate level, performance reflects a high rate of guessing. However, when aggregate data is partitioned by lag (i.e., the number of presentations from study to test), or is un-aggregated, performance appears to be more complex than just remembering with some noise and guessing. We implemented three models: the standard remember-guess model, a three-component remember-guess model, and a Bayesian mixture model and evaluated these models against the data. The results emphasize the importance of taking into account the influence of prior category knowledge on memory.     

Categorization as causal reasoning⋆

A theory of categorization is presented in which knowledge of causal relationships between category features is represented in terms of asymmetric and probabilistic causal mechanisms. According to causal‐model theory, objects are classified as category members to the extent they are likely to have been generated or produced by those mechanisms. The empirical results confirmed that participants rated exemplars good category members to the extent their features manifested the expectations that causal knowledge induces, such as correlations between feature pairs that are directly connected by causal relationships. These expectations also included sensitivity to higher‐order feature interactions that emerge from the asymmetries inherent in causal relationships. Quantitative fits of causal‐model theory were superior to those obtained with extensions to traditional similarity‐based models that represent causal knowledge either as higher‐order relational features or “prior exemplars” stored in memory.     

Contrast effects in typicality judgements: A hierarchical Bayesian approach

We examine the influence of contrast categories on the internal graded membership structure of everyday concepts using computational models proposed in the artificial category learning tradition. In particular, the generalized context model (Nosofsky, 1986), which assumes that only members of a given category contribute to the typicality of a category member, is contrasted to the similarity–dissimilarity generalized context model (SD-GCM; Stewart & Brown, 2005), which assumes that members of other categories are also influential in determining typicality. The models are compared in a hierarchical Bayesian framework in their account of the typicality gradient of five animal categories and six artefact categories. For each target category, we consider all possible relevant contrast categories. Three separate issue are examined: (a) whether contrast effects can be found, (b) which categories are responsible for these effects, and (c) whether more than one category influences the typicality. Results indicate that the internal category structure is codetermined by dissimilarity towards potential contrast categories. In most cases, only a single contrast category contributed to the typicality. The present findings suggest that contrast effects might be more widespread than has previously been assumed. Further, they stress the importance of characteristics particular of everyday concepts, which require careful consideration when applying computational models of representation of the artificial category learning tradition to everyday concepts.