Improving Bayesian Reasoning What Works And Why

We confess that the first part of our title is somewhat of a misnomer. Bayesian reasoning is a normative approach to probabilistic belief revision and, as such, it is in need of no improvement. Rather, it is the typical individual whose reasoning and judgments often fall short of the Bayesian ideal who is the focus of improvement. What have we learnt from over a half-century of research and theory on this topic that could explain why people are often non-Bayesian? Can Bayesian reasoning be facilitated, and if so why? These are the questions that motivate this Frontiers in Psychology Research Topic. Bayes' theorem, named after English statistician, philosopher, and Presbyterian minister, Thomas Bayes, offers a method for updating one’s prior probability of an hypothesis H on the basis of new data D such that P(H|D) = P(D|H)P(H)/P(D). The first wave of psychological research, pioneered by Ward Edwards, revealed that people were overly conservative in updating their posterior probabilities (i.e., P(D|H)). A second wave, spearheaded by Daniel Kahneman and Amos Tversky, showed that people often ignored prior probabilities or base rates, where the priors had a frequentist interpretation, and hence were not Bayesians at all. In the 1990s, a third wave of research spurred by Leda Cosmides and John Tooby and by Gerd Gigerenzer and Ulrich Hoffrage showed that people can reason more like a Bayesian if only the information provided takes the form of (non-relativized) natural frequencies. Although Kahneman and Tversky had already noted the advantages of frequency representations, it was the third wave scholars who pushed the prescriptive agenda, arguing that there are feasible and effective methods for improving belief revision. Most scholars now agree that natural frequency representations do facilitate Bayesian reasoning. However, they do not agree on why this is so. The original third wave scholars favor an evolutionary account that posits human brain adaptation to natural frequency processing. But almost as soon as this view was proposed, other scholars challenged it, arguing that such evolutionary assumptions were not needed. The dominant opposing view has been that the benefit of natural frequencies is mainly due to the fact that such representations make the nested set relations perfectly transparent. Thus, people can more easily see what information they need to focus on and how to simply combine it. This Research Topic aims to take stock of where we are at present. Are we in a proto-fourth wave? If so, does it offer a synthesis of recent theoretical disagreements? The second part of the title orients the reader to the two main subtopics: what works and why? In terms of the first subtopic, we seek contributions that advance understanding of how to improve people’s abilities to revise their beliefs and to integrate probabilistic information effectively. The second subtopic centers on explaining why methods that improve non-Bayesian reasoning work as well as they do. In addressing that issue, we welcome both critical analyses of existing theories as well as fresh perspectives. For both subtopics, we welcome the full range of manuscript types.

This book focuses on how statistical reasoning works and on training programs that can exploit people's natural cognitive capabilities to improve their statistical reasoning. Training programs that take into account findings from evolutionary psychology and instructional theory are shown to have substantially larger effects that are more stable over time than previous training regimens. The theoretical implications are traced in a neural network model of human performance on statistical reasoning problems. This book apppeals to judgment and decision making researchers and other cognitive scientists, as well as to teachers of statistics and probabilistic reasoning.

Modern Mathematics is constructed rigorously through proofs, based on truths, which are either axioms or previously proven theorems. Thus, it is par excellence a model of rational inquiry. Links between Cognitive Psychology and Mathematics Education have been particularly strong during the last decades. Indeed, the Enlightenment view of the rational human mind that reasons, makes decisions and solves problems based on logic and probabilities, was shaken during the second half of the twentieth century. Cognitive psychologists discovered that humans' thoughts and actions often deviate from rules imposed by strict normative theories of inference. Yet, these deviations should not be called "errors": as Cognitive Psychologists have demonstrated, these deviations may be either valid heuristics that succeed in the environments in which humans have evolved, or biases that are caused by a lack of adaptation to abstract information formats. Humans, as the cognitive psychologist and economist Herbert Simon claimed, do not usually optimize, but rather satisfice, even when solving problem. This Research Topic aims at demonstrating that these insights have had a decisive impact on Mathematics Education. We want to stress that we are concerned with the view of bounded rationality that is different from the one espoused by the heuristics-and-biases program. In Simon’s bounded rationality and its direct descendant ecological rationality, rationality is understood in terms of cognitive success in the world (correspondence) rather than in terms of conformity to content-free norms of coherence (e.g., transitivity).

A practical introduction perfect for final-year undergraduate and graduate students without a solid background in linear algebra and calculus.

Widely considered to be the most comprehensive and accessible textbook in the field of Cognitive Psychology Emphasis on applied cognition with ‘in the real world’ case studies and examples Comprehensive companion website including access to Primal Pictures’ interactive 3D atlas of the brain, test simulations of key experiments, multiple choice questions, glossary flashcards and instructor PowerPoint slides Simple, clear pedagogy in every chapter to highlight key terms, case studies and further reading Updated references throughout the textbook to reflect the latest research

This undergraduate textbook reviews psychological research in the major areas of reasoning and thinking: deduction, induction, hypothesis testing, probability judgement, and decision making. It also covers the major theoretical debates in each area, and devotes a chapter to one of the liveliest issues in the field: the question of human rationality. Central themes that recur throughout the book include not only rationality, but also the relation between normative theories such as logic, probability theory, and decision theory, and human performance, both in experiments and in the world outside the laboratory. No prior acquaintance with formal systems is assumed, and everyday examples are used throughout to illustrate technical and theoretical points. The book differs from others in the market firstly in the range of material covered: other tend to focus primarily on on either reasoning or thinking. It is also the first student-level text to survey an imporatant new theoretical perspective, the information-gain or rational analysis approach, and to review the rationality debate from the standpoint of psuchological research in a wide range of areas.

This book describes an approach to understanding, modeling, and improving the probabilistic reasoning of ordinary adults, comparing their reasoning to that of "experts." For specialists in judgment and decision making and all cognitive scientists.

This book provides a multi-level introduction to Bayesian reasoning (as opposed to OC conventional statisticsOCO) and its applications to data analysis. The basic ideas of this OC newOCO approach to the quantification of uncertainty are presented using examples from research and everyday life. Applications covered include: parametric inference; combination of results; treatment of uncertainty due to systematic errors and background; comparison of hypotheses; unfolding of experimental distributions; upper/lower bounds in frontier-type measurements. Approximate methods for routine use are derived and are shown often to coincide OCo under well-defined assumptions! OCo with OC standardOCO methods, which can therefore be seen as special cases of the more general Bayesian methods. In dealing with uncertainty in measurements, modern metrological ideas are utilized, including the ISO classification of uncertainty into type A and type B. These are shown to fit well into the Bayesian framework.