In order to account for how children can generalize words beyond a very limited set of labeled examples, Bloom's proposal of word learning requires two extensions: a better understanding of the “general learning and memory abilities” involved, and a principled framework for integrating multiple conflicting constraints on word meaning. We propose a framework based on Bayesian statistical inference that meets both of those needs.

Surprise—operationalized as looking time—has a long history in developmental research, providing a window into the perception and cognition of infants. Recently, however, a number of developmental researchers have considered infants’ and children's surprise in its own right. This article reviews empirical evidence and computational models of complex statistical inferences underlying surprise, and discusses how these findings relate to the role that surprise appears to play as a catalyst for learning.

I make two points in this commentary on Carey (2009). First, it may be too soon to conclude that core cognition is innate. Recent advances in computational cognitive science and developmental psychology suggest possible mechanisms for developing inductive biases. Second, there is another possible answer to Fodor's challenge – if concepts are merely mental tokens, then cognitive scientists should spend their time on developing a theory of belief fixation instead.

Statistical inference is often justified by long-run properties of the sampling distributions, such as the repeated sampling rationale. These are frequentist justifications of statistical inference. I argue, in line with existing philosophical literature, but against a widespread image in empirical science, that these justifications are flawed. Then I propose a novel interpretation of probability in statistics, the artefactual interpretation. I believe that this interpretation is able to bridge the gap between statistical probability calculations and (...) class='Hi'>rational decisions on the basis of observed data. The artefactual interpretation is able to justify statistical inference without making any assumptions about probability in the material world. (shrink)

This paper addresses the problem that Bayesian statistical inference cannot accommodate theory change, and proposes a framework for dealing with such changes. It first presents a scheme for generating predictions from observations by means of hypotheses. An example shows how the hypotheses represent the theoretical structure underlying the scheme. This is followed by an example of a change of hypotheses. The paper then presents a general framework for hypotheses change, and proposes the minimization of the distance between hypotheses (...) as a rationality criterion. Finally the paper discusses the import of this for Bayesian statistical inference. (shrink)

After some general remarks about the interrelation between philosophical and statistical thinking, the discussion centres largely on significance tests. These are defined as the calculation of p-values rather than as formal procedures for ‘acceptance‘ and ‘rejection‘. A number of types of null hypothesis are described and a principle for evidential interpretation set out governing the implications of p- values in the specific circumstances of each application, as contrasted with a long-run interpretation. A number of more complicated situ- ations are (...) discussed in which modification of the simple p-value may be essential. (shrink)

Although both philosophers and scientists are interested in how to obtain reliable knowledge in the face of error, there is a gap between their perspectives that has been an obstacle to progress. By means of a series of exchanges between the editors and leaders from the philosophy of science, statistics and economics, this volume offers a cumulative introduction connecting problems of traditional philosophy of science to problems of inference in statistical and empirical modelling practice. Philosophers of science and (...) scientific practitioners are challenged to reevaluate the assumptions of their own theories - philosophical or methodological. Practitioners may better appreciate the foundational issues around which their questions revolve and thereby become better 'applied philosophers'. Conversely, new avenues emerge for finally solving recalcitrant philosophical problems of induction, explanation and theory testing. (shrink)

Modern scientific cosmology pushes the boundaries of knowledge and the knowable. This is prompting questions on the nature of scientific knowledge. A central issue is what defines a 'good' model. When addressing global properties of the Universe or its initial state this becomes a particularly pressing issue. How to assess the probability of the Universe as a whole is empirically ambiguous, since we can examine only part of a single realisation of the system under investigation: at some point, data will (...) run out. We review the basics of applying Bayesian statistical explanation to the Universe as a whole. We argue that a conventional Bayesian approach to model inference generally fails in such circumstances, and cannot resolve, e.g., the so-called 'measure problem' in inflationary cosmology. Implicit and non-empirical valuations inevitably enter model assessment in these cases. This undermines the possibility to perform Bayesian model comparison. One must therefore either stay silent, or pursue a more general form of systematic and rational model assessment. We outline a generalised axiological Bayesian model inference framework, based on mathematical lattices. This extends inference based on empirical data (evidence) to additionally consider the properties of model structure (elegance) and model possibility space (beneficence). We propose this as a natural and theoretically well-motivated framework for introducing an explicit, rational approach to theoretical model prejudice and inference beyond data. (shrink)

When do data provide good evidence for a hypothesis, evidence that warrants an inference to the hypothesis? Standard answers either reject the legitimacy of induction or else allow warranted inference from data to hypothesis when there are suitable relationships between and among the data and hypotheses. The severity account rejects all of these, maintaining instead that the good evidence relation concerns not only relations between data and hypotheses but also the methods for obtaining the data and the sensitivity (...) of these methods to detecting falsehood. So, for example, the severity account allows two rival hypotheses to fit the data equally well, even to the point of having the same prior and posterior probabilities relative to those data, while the data nonetheless favour one hypothesis over the other by virtue of those data providing a severe test of one hypothesis and not the other. The contributions to Error and Inference variously trace the genesis of the severity account from statistics-oriented sciences, elaborate upon the account’s details, apply the account to various philosophical issues, and offer challenges to the account’s adequacy and replies to major criticisms. The editors have developed the severity account in detail elsewhere, and this anthology is an effort …. (shrink)

This paper addresses a common challenge with computational cognitive models: identifying parameter values that are both theoretically plausible and generate predictions that match well with empirical data. While computational models can offer deep explanations of cognition, they are computationally complex and often out of reach of traditional parameter fitting methods. Weak methodology may lead to premature rejection of valid models or to acceptance of models that might otherwise be falsified. Mathematically robust fitting methods are, therefore, essential to the progress of (...) computational modeling in cognitive science. In this article, we investigate the capability and role of modern fitting methods—including Bayesian optimization and approximate Bayesian computation—and contrast them to some more commonly used methods: grid search and Nelder–Mead optimization. Our investigation consists of a reanalysis of the fitting of two previous computational models: an Adaptive Control of Thought—Rational model of skill acquisition and a computational rationality model of visual search. The results contrast the efficiency and informativeness of the methods. A key advantage of the Bayesian methods is the ability to estimate the uncertainty of fitted parameter values. We conclude that approximate Bayesian computation is (a) efficient, (b) informative, and (c) offers a path to reproducible results. (shrink)

[Correction Notice: An erratum for this article was reported in Vol 109 of Psychological Review. Due to circumstances that were beyond the control of the authors, the studies reported in "Models of Ecological Rationality: The Recognition Heuristic," by Daniel G. Goldstein and Gerd Gigerenzer overlap with studies reported in "The Recognition Heuristic: How Ignorance Makes Us Smart," by the same authors and with studies reported in "Inference From Ignorance: The Recognition Heuristic". In addition, Figure 3 in the Psychological Review (...) article was originally published in the book chapter and should have carried a note saying that it was used by permission of Oxford University Press.] One view of heuristics is that they are imperfect versions of optimal statistical procedures considered too complicated for ordinary minds to carry out. In contrast, the authors consider heuristics to be adaptive strategies that evolved in tandem with fundamental psychological mechanisms. The recognition heuristic, arguably the most frugal of all heuristics, makes inferences from patterns of missing knowledge. This heuristic exploits a fundamental adaptation of many organisms: the vast, sensitive, and reliable capacity for recognition. The authors specify the conditions under which the recognition heuristic is successful and when it leads to the counter-intuitive less-is-more effect in which less knowledge is better than more for making accurate inferences. (shrink)

Deborah Mayo's view of science is that learning occurs by severely testing specific hypotheses. Mayo expounded this thesis in her (1996) Error and the Growth of Experimental Knowledge (EGEK). This volume consists of a series of exchanges between Mayo and distinguished philosophers representing competing views of the philosophy of science. The tone of the exchanges is lively, edifying and enjoyable. Mayo's error-statistical philosophy of science is critiqued in the light of positions which place more emphasis on large-scale theories. The (...) result clarifies Mayo's account and highlights her contribution to the philosophy of science -- in particular, her contribution to the philosophy of those sciences that rely heavily on statistical analysis. The second half of the volume considers the application (or extension) of an error-statistical philosophy of science to theory testing in economics, causal modelling and legal epistemology. The volume also includes a contribution to the frequentist philosophy of statistics written by Mayo in collaboration with Sir David Cox. (shrink)

How do human infants learn the causal dependencies between events? Evidence suggests that this remarkable feat can be achieved by observation of only a handful of examples. Many computational models have been produced to explain how infants perform causal inference without explicit teaching about statistics or the scientific method. Here, we propose a spiking neuronal network implementation that can be entrained to form a dynamical model of the temporal and causal relationships between events that it observes. The network uses (...) spike-time dependent plasticity, long-term depression, and heterosynaptic competition rules to implement Rescorla–Wagner-like learning. Transmission delays between neurons allow the network to learn a forward model of the temporal relationships between events. Within this framework, biologically realistic synaptic plasticity rules account for well-known behavioral data regarding cognitive causal assumptions such as backwards blocking and screening-off. These models can then be run as emulators for state inference. Furthermore, this mechanism is capable of copying synaptic connectivity patterns between neuronal networks by observing the spontaneous spike activity from the neuronal circuit that is to be copied, and it thereby provides a powerful method for transmission of circuit functionality between brain regions. (shrink)

This article argues that a particular notion of rationality, more exactly a specific notion of legitimate inference, is presupposed by much work in the social sciences to their detriment. The author describes the notion of rationality he has in mind, explains why it is misguided, identifies where and how it affects social research, and illustrates why that research is weaker as a result. The notion of legitimate inference the author has in mind is one that believes inferences are (...) guided by principles that are formal, universal and a priori—that is, that make no substantive, domain-specific empirical assumptions about the world. The author briefly provides a variety of reasons to be skeptical about such principles. Those reasons extend from broad philosophical considerations to quasi-empirical evidence. The author then argues that this notion of inference is involved with both the practice and the content of the social sciences in various ways. In terms of practice, this notion of inference lies behind the way statistical testing is generally done in the social sciences. In terms of content, the author argues that this notion of rationality is invoked, for example, in game theory and in rational expectations macroeconomics. In all cases, use of the formal rationality notion leads social scientists to put more faith in their results than is warranted and thus is an obstacle to doing better social science. (shrink)

MaxEnt inference algorithm and information theory are relevant for the time evolution of macroscopic systems considered as problem of incomplete information. Two different MaxEnt approaches are introduced in this work, both applied to prediction of time evolution for closed Hamiltonian systems. The first one is based on Liouville equation for the conditional probability distribution, introduced as a strict microscopic constraint on time evolution in phase space. The conditional probability distribution is defined for the set of microstates associated with the (...) set of phase space paths determined by solutions of Hamilton’s equations. The MaxEnt inference algorithm with Shannon’s concept of the conditional information entropy is then applied to prediction, consistently with this strict microscopic constraint on time evolution in phase space. The second approach is based on the same concepts, with a difference that Liouville equation for the conditional probability distribution is introduced as a macroscopic constraint given by a phase space average. We consider the incomplete nature of our information about microscopic dynamics in a rational way that is consistent with Jaynes’ formulation of predictive statistical mechanics, and the concept of macroscopic reproducibility for time dependent processes. Maximization of the conditional information entropy subject to this macroscopic constraint leads to a loss of correlation between the initial phase space paths and final microstates. Information entropy is the theoretic upper bound on the conditional information entropy, with the upper bound attained only in case of the complete loss of correlation. In this alternative approach to prediction of macroscopic time evolution, maximization of the conditional information entropy is equivalent to the loss of statistical correlation, and leads to corresponding loss of information. In accordance with the original idea of Jaynes, irreversibility appears as a consequence of gradual loss of information about possible microstates of the system. (shrink)

Statistical model specification and validation raise crucial foundational problems whose pertinent resolution holds the key to learning from data by securing the reliability of frequentist inference. The paper questions the judiciousness of several current practices, including the theory-driven approach, and the Akaike-type model selection procedures, arguing that they often lead to unreliable inferences. This is primarily due to the fact that goodness-of-fit/prediction measures and other substantive and pragmatic criteria are of questionable value when the estimated model is statistically (...) misspecified. Foisting one's favorite model on the data often yields estimated models which are both statistically and substantively misspecified, but one has no way to delineate between the two sources of error and apportion blame. The paper argues that the error statistical approach can address this Duhemian ambiguity by distinguishing between statistical and substantive premises and viewing empirical modeling in a piecemeal way with a view to delineate the various issues more effectively. It is also argued that Hendry's general to specific procedures does a much better job in model selection than the theory-driven and the Akaike-type procedures primary because of its error statistical underpinnings. (shrink)

Error statistics is an important methodological view in philosophy of statistics and philosophy of science that can be applied to scientific experiments such as clinical trials. In this paper, I raise two potential issues for ES when it comes to guiding, and explaining early stopping of randomized controlled trials : ES provides limited guidance in cases of early unfavorable trends due to the possibility of trend reversal; ES is silent on how to prospectively control error rates in experiments requiring multiple (...) interim analyses. The method of conditional power, together with a rationing principle for RCTs, can assist ES in addressing such issues. (shrink)

The replication crisis has prompted many to call for statistical reform within the psychological sciences. Here we examine issues within Frequentist statistics that may have led to the replication crisis, and we examine the alternative—Bayesian statistics—that many have suggested as a replacement. The Frequentist approach and the Bayesian approach offer radically different perspectives on evidence and inference with the Frequentist approach prioritising error control and the Bayesian approach offering a formal method for quantifying the relative strength of evidence (...) for hypotheses. We suggest that rather than mere statistical reform, what is needed is a better understanding of the different modes of statistical inference and a better understanding of how statistical inference relates to scientific inference. (shrink)

Abductive reasoning assigns special status to the explanatory power of a hypothesis. But how do people make explanatory judgments? Our study clarifies this issue by asking: How does the explanatory power of a hypothesis cohere with other cognitive factors? How does probabilistic information affect explanatory judgments? In order to answer these questions, we conducted an experiment with 671 participants. Their task was to make judgments about a potentially explanatory hypothesis and its cognitive virtues. In the responses, we isolated three constructs: (...) Explanatory Value, Rational Acceptability, and Entailment. Explanatory judgments strongly cohered with judgments of causal relevance and with a sense of understanding. Furthermore, we found that Explanatory Value was sensitive to manipulations of statistical relevance relations between hypothesis and evidence, but not to explicit information about the prior probability of the hypothesis. These results indicate that probabilistic information about statistical relevance is a strong determinant of Explanatory Value. More generally, our study suggests that abductive and probabilistic reasoning are two distinct modes of inference. (shrink)

We provide statistical inference for measures of predictive success. These measures are frequently used to evaluate and compare the performance of different models of individual and group decision making in experimental and revealed preference studies. We provide a brief illustration of our findings by comparing the predictive success of different revealed preference tests for models of intertemporal decision making. This demonstrates that it is possible to compare the predictive success of different models in a statistically meaningful way.

An accessible and illuminating exploration of the conceptual basisof scientific and statistical inference and the practical impact this has on conducting psychological research. The book encourages a critical discussion of the different approaches and looks at some of the most important thinkers and their influence.

Much research on cognitive development focuses either on early-emerging domain-specific knowledge or domain-general learning mechanisms. However, little research examines how these sources of knowledge interact. Previous research suggests that young infants can make inferences from samples to populations (Xu & Garcia, 2008) and 11- to 12.5-month-old infants can integrate psychological and physical knowledge in probabilistic reasoning (Teglas, Girotto, Gonzalez, & Bonatti, 2007; Xu & Denison, 2009). Here, we ask whether infants can integrate a physical constraint of immobility into a (...) class='Hi'>statistical inference mechanism. Results from three experiments suggest that, first, infants were able to use domain-specific knowledge to override statistical information, reasoning that sometimes a physical constraint is more informative than probabilistic information. Second, we provide the first evidence that infants are capable of applying domain-specific knowledge in probabilistic reasoning by using a physical constraint to exclude one set of objects while computing probabilities over the remaining sets. (shrink)

One of Ian Hacking's earliest publications, this book showcases his early ideas on the central concepts and questions surrounding statistical reasoning. He explores the basic principles of statistical reasoning and tests them, both at a philosophical level and in terms of their practical consequences for statisticians. Presented in a fresh twenty-first-century series livery, and including a specially commissioned preface written by Jan-Willem Romeijn, illuminating its enduring importance and relevance to philosophical enquiry, Hacking's influential and original work has been (...) revived for a new generation of readers. (shrink)

We analyze the quantum mechanical measuring process from the standpoint of information theory. Statistical inference is used in order to define the most likely state of the measured system that is compatible with the readings of the measuring instrument and the a priori information about the correlations between the system and the instrument. This approach has the advantage that no reference to the time evolution of the combined system need be made. It must, however, be emphasized that the (...) result is to be interpreted as the statistically inferred state of the original system rather than the state of the system after measurement. The phenomenon of “reduction of states” appears in this light as a consequence of incomplete information rather than the physical interaction between measured system and measuring instrument. (shrink)

Frequentist inference typically is described in terms of hypothetical repeated sampling but there are advantages to an interpretation that uses a single random sample. Contemporary examples are given that indicate probabilities for random phenomena are interpreted as classical probabilities, and this interpretation of equally likely chance outcomes is applied to statistical inference using urn models. These are used to address Bayesian criticisms of frequentist methods. Recent descriptions of p-values, confidence intervals, and power are viewed through the lens (...) of classical probability based on a single random sample from the population. (shrink)

The major competing statistical paradigms share a common remarkable but unremarked thread: in many of their inferential applications, different probability interpretations are combined. How this plays out in different theories of inference depends on the type of question asked. We distinguish four question types: confirmation, evidence, decision, and prediction. We show that Bayesian confirmation theory mixes what are intuitively “subjective” and “objective” interpretations of probability, whereas the likelihood-based account of evidence melds three conceptions of what constitutes an “objective” (...) probability. (shrink)

Although written in Japanese, what the statistical inference is philosophically investigated.

Scientists and Bayesian statisticians often study hypotheses that they know to be false. This creates an interpretive problem because the Bayesian probability of a hypothesis is supposed to represent the probability that the hypothesis is true. I investigate whether Bayesianism can accommodate the idea that false hypotheses are sometimes approximately true or that some hypotheses or models can be closer to the truth than others. I argue that the idea that some hypotheses are approximately true in an absolute sense is (...) hard to square with Bayesianism, but that the notion that some hypotheses are comparatively closer to the truth than others can be made compatible with Bayesianism, and that this provides an adequate and potentially useful solution to the interpretive problem. Finally, I compare my ``verisimilitude'' solution to the interpretive problem with a ``counterfactual'' solution recently proposed by Jan Sprenger. (shrink)

NHSTP is embedded in the research of “cognitive science.” Its use is based on unstated assumptions about the practices of sampling, “operationalizing,” and using group data. NHSTP has facilitated both research and theorizing – research findings of limited interest – diverse theories that seldom complement one another. Alternative methods are available for data acquisition and analysis, and for assessing the “truth- value” of generalizations.

In May of 1973 we organized an international research colloquium on foundations of probability, statistics, and statistical theories of science at the University of Western Ontario. During the past four decades there have been striking formal advances in our understanding of logic, semantics and algebraic structure in probabilistic and statistical theories. These advances, which include the development of the relations between semantics and metamathematics, between logics and algebras and the algebraic-geometrical foundations of statistical theories (especially in the (...) sciences), have led to striking new insights into the formal and conceptual structure of probability and statistical theory and their scientific applications in the form of scientific theory. The foundations of statistics are in a state of profound conflict. Fisher's objections to some aspects of Neyman-Pearson statistics have long been well known. More recently the emergence of Baysian statistics as a radical alternative to standard views has made the conflict especially acute. In recent years the response of many practising statisticians to the conflict has been an eclectic approach to statistical inference. Many good statisticians have developed a kind of wisdom which enables them to know which problems are most appropriately handled by each of the methods available. The search for principles which would explain why each of the methods works where it does and fails where it does offers a fruitful approach to the controversy over foundations. (shrink)

The main thesis of the paper is that in the case of modern statistics, the differences between the various concepts of models were the key to its formative controversies. The mathematical theory of statistical inference was mainly developed by Ronald A. Fisher, Jerzy Neyman, and Egon S. Pearson. Fisher on the one side and Neyman–Pearson on the other were involved often in a polemic controversy. The common view is that Neyman and Pearson made Fisher's account more stringent mathematically. (...) It is argued, however, that there is a profound theoretical basis for the controversy: both sides held conflicting views about the role of mathematical modelling. At the end, the influential programme of Exploratory Data Analysis is considered to be advocating another, more instrumental conception of models. Introduction Models in statistics—‘of what population is this a random sample?’ The fundamental lemma Controversy about models Exploratory data analysis as a model-critical approach. (shrink)

Abstract Based on recalling two characteristic features of Bayesian statistical inference in commutative probability theory, a stability property of the inference is pointed out, and it is argued that that stability of the Bayesian statistical inference is an essential property which must be preserved under generalization of Bayesian inference to the non?commutative case. Mathematical no?go theorems are recalled then which show that, in general, the stability can not be preserved in non?commutative context. Two possible (...) interpretations of the impossibility of generalization of Bayesian statistical inference to the non?commutative case are offered, none of which seems to be completely satisfying. (shrink)

The rational status of the Bayesian calculus for revising likelihoods is compromised by the common but still unfamiliar phenomenon of information distortion. This bias is the distortion in the evaluation of a new datum toward favoring the currently preferred option in a decision or judgment. While the Bayesian calculus requires the independent combination of the prior probability and a new datum, information distortion invalidates such independence (because the prior influences the datum). Although widespread, information distortion has not generally been (...) recognized. First, individuals are not aware when they themselves commit this bias. In addition, it is often hidden in more obvious suboptimal phenomena. Finally, the Bayesian calculus is usually explained only with undistortable data like colored balls drawn randomly. Partly because information distortion is unrecognized by the individuals exhibiting it, no way has been devised for eliminating it. Partial reduction is possible in some situation such as presenting all data simultaneously rather than sequentially with revision after each datum. The potential dangers of information distortion are illustrated for three professional revision tasks: forecasting, predicting consumer choices from internet data, and statistical inference from experimental results. The optimality of the Bayesian calculus competes with people’s natural desire that their belief systems remain coherent in the face of new data. Information distortion provides this coherence by biasing those data toward greater agreement with the currently preferred position – but at the cost of Bayesian optimality. (shrink)