This book represents the first major attempt by any author to provide an integrated account of the evidence for bias in human reasoning across a wide range of disparate psychological literatures. The topics discussed involve both deductive and inductive reasoning as well as statistical judgement and inference. In addition, the author proposes a general theoretical approach to the explanations of bias and considers the practical implications for real world decision making. The theoretical stance of the book is based on a (...) distinction between preconscious heuristic processes which determine the mental representation of 'relevant' features of the problem content, and subsequent analytic reasoning processes which generate inferences and judgements. Phenomena discussed and interpreted within this framework include feature matching biases in propositional reasoning, confirmation bias, biasing and debiasing effects of knowledge on reasoning, and biases in statistical judgement normally attributed to 'availability' and 'representativeness' heuristics. In the final chapter, the practical consequences of bias for real life decision making are considered, together with various issues concerning the problem of 'debiasing'. The major approaches discussed are those involving education and training on the one hand, and the development of intelligent software and interactive decision aids on the other. (shrink)

In this study on deduction, the authors argue that people reason by imagining the relevant state of affairs, ie building an internal model of it, formulating a tentative conclusion based on this model and then searching for alternative models.

This study investigated individual differences in the belief bias effect, which is the tendency to accept conclusions because they are believable rather than because they are logically valid. It was observed that the extent of an individual's belief bias effect was unrelated to a number of measures of reasoning competence. Instead, as predicted by mental models theory, it was related to a person's ability to generate alternative representations of premises: the more alternatives a person generated, the less likely they were (...) to show a belief bias effect. In contrast to belief effects, which were predicted by the number of alternatives generated, scores on logical reasoning tasks were predicted by an individual's understanding of the concept of logical necessity. These two mediating variables, ability to generate alternatives and understanding of logical necessity, were themselves predicted by similar variables, such as cognitive motivation and abstract reasoning ability. Our findings strongly support the mental models interpretation of the belief bias effect, and suggest that individual differences in this effect are associated with a specific ability to generate alternatives, rather than with a general logical competence. (shrink)

Hempel's paradox of the ravens has to do with the question of what constitutes confirmation from a logical point of view; Wason 's selection task has been used extensively to investigate how people go about attempting to confirm or disconfirm conditional claims. This paper presents an argument that the paradox is resolved, and that people's typical performance in the selection task can be explained, by consideration of what constitutes an effective strategy for seeking evidence of the tenability of universal or (...) conditional claims in everyday life. (shrink)

Most theories of deduction have assumed that linguistic processes transduce from language into an internal representation and back again, and that non-linguistic processes are central to deduction itself. In this article it is proposed that for deduction tasks for which the necessary information is provided verbally, the heart of deduction for untrained participants involves repeatedly reencoding the problem, a type of behavior referred to here as verbal reasoning. It is shown that model theory accounts of behavior on most deduction tasks (...) are consistent with verbal reasoning and that verbal reasoning can account for detailed behavior in a single task; a computational model of syllogistic reasoning—VR—based on linguistic mechanisms is presented. VR models all of the standard phenomena, makes a number of accurate novel predictions, and fits the behavior of individual participants with an accuracy that rivals their own test–retest reliability. (shrink)