Human concept learning presents a version of the classic problem of induction, which is made particularly difficult by the combination of two requirements: the need to learn from a rich (i.e. nested and overlapping) vocabulary of possible concepts and the need to be able to generalize concepts reasonably from only a few positive examples. I begin this thesis by considering a simple number concept game as a concrete illustration of this ability. On this task, human learners can with reasonable confidence lock in on one out of a billion billion billion logically possible concepts, after seeing only four positive examples of the concept, and can generalize informatively after seeing just a single example. Neither of the two classic approaches to inductive inference -- hypothesis testing in a constrained space of possible rules and computing similarity to the observed examples -- can provide a complete picture of how people generalize concepts in even this simple setting. This thesis prop...

Cosmides and Tooby (1996) increased performance using a frequency rather than probability frame on a problem known to elicit base-rate neglect. Analogously, Gigerenzer (1994) claimed that the conjunction fallacy disappears when formulated in terms of frequency rather than the more usual single-event probability. These authors conclude that a module or algorithm of mind exists that is able to compute with frequencies but not probabilities. The studies reported here found that base-rate neglect could also be reduced using a clearly stated single-event probability frame and by using a diagram that clarified the critical nested-set relations of the problem; that the frequency advantage could be eliminated in the conjunction fallacy by separating the critical statements so that their nested relation was opaque; and that the large effect of frequency framing on the two problems studied is not stable. Facilitation via frequency is a result of clarifying the probabilistic interpretation of the...

An important research tradition in the cognitive psychology of reasoning--called the heuristics and biases approach--has firmly established that people’s responses often deviate from the performance considered normative on many reasoning tasks. For example, people assess probabilities incorrectly, they display confirmation bias, they test hypotheses inefficiently, they violate the axioms of utility theory, they do not properly calibrate degrees of belief, they overproject their own opinions onto others, they display illogical framing effects, they uneconomically honor sunk costs, they allow prior knowledge to become implicated in deductive reasoning, and they display numerous other information processing biases (for summaries of the large literature, see

The authors provide evidence that people typically evaluate conditional probabilities by subjectively partitioning the sample space into n interchangeable events, editing out events that can be eliminated on the basis of conditioning information, counting remaining events, then reporting probabilities as a ratio of the number of focal to total events. Participants ’ responses to conditional probability problems were influenced by irrelevant information (Study 1), small variations in problem wording (Study 2), and grouping of events (Study 3), as predicted by the partition–edit–count model. Informal protocol analysis also supports the authors ’ interpretation. A 4th study extends this account from situations where events are treated as interchangeable (chance and ignorance) to situations where participants have information they can use to distinguish among events (uncertainty). People are often called on to make judgments of conditional likelihood. For example, a patient might try to assess the likelihood of having a disease, given a positive test result; a litigant might try to estimate the odds of prevailing in court, given a piece of damning evidence. Over the last 30 years, psychologists have shown that people typically judge conditional probabilities using a

Abstract. Subjective Bayesian inference is unsuitable as an ideal for learning strategies to approximate, as the arbitrariness in prior probabilities makes claims to Bayesian learning too easily vulnerable to inductive skepticism. An objective Bayesian approach, which determines priors by maximizing information entropy, runs into insurmountable difficulties in conditions where no definite background theory is available. However, this lack of background knowledge makes the maximum entropy argument directly applicable to the process of drawing samples from a population. As a result, evidence can be seen not just as eliminating a number of incompatible hypotheses out of an infinity of possibilities, but as being representative of the true state of affairs. Hence inductive skepticism can be avoided, as demonstrated by a resolution of Goodman’s ‘grue ’ paradox. This leads to a clearer understanding of the vital role abductive processes and tools like simple generalization play in learning. Keywords: Goodman’s paradox, induction, inductive skepticism, statistical inference, Bayesian learning, maximum entropy, machine learning, abduction, generalization

THE THEORY OF evolution by natural selection has revolutionary implications for understanding the design of the human mind and brain, as Darwin himself was the first to recognize (Darwin, 1859). Indeed, a principled understanding of the network of causation that built the functional architecture of the human species offers the possibility of transforming the study of humanity into a natural science capable of precision and rapid progress. Yet, nearly a century and a half after The Origin of Species was published, the psychological, social, and behavioral sciences remain largely untouched by these implications, and many of these disciplines continue to be founded on assumptions evolutionarily informed researchers know to be false (Pinker, 2002; Tooby & Cosmides, 1992). Evolutionary psychology is the long-forestalled scientific attempt to assemble out of the disjointed, fragmentary, and mutually contradictory human disciplines a single, logically integrated research framework for the psychological, social, and behavioral sciences—a framework that not only incorporates the evolutionary sciences on a full and equal basis, but that systematically works out all of the revisions in existing belief and research practice that such a synthesis requires (Tooby & Cosmides, 1992). The long-term scientific goal toward which evolutionary psychologists are working is the mapping of our universal human nature. By this, we mean the construction of a set of empirically validated, high-resolution models of the evolved mechanisms that collectively constitute universal human nature. Because the evolved function of a psychological mechanism is computational—to regulate behavior and the body adaptively in response to informational inputs—such a model consists of a description of the functional circuit logic or information

The way people respond to the chance that an unlikely event will occur depends on how the event is described. We propose that people attach more weight to unlikely events when they can easily generate or imagine examples in which the event has occurred or will occur than when they cannot. We tested this idea in two experiments with mock jurors using written murder scenarios. The results suggested that jurors attach more weight to the defendant's claim that an incriminating DNA match is merely coincidental when it is easy for them to imagine other individuals whose DNA would also match than when it is not easy for them to imagine such individuals. We manipulated the difficulty of imagining such examples by varying the description of the DNA-match statistic. Some of the variations that influenced the jurors were normatively irrelevant.

www.guilford.com

When and why do people avoid unknown probabilities in decisions under uncertainty? Testing some predictions from optimal foraging theory

ABSTRACT Cognitive scientists need theoretical guidance that is grounded in something beyond intuition. They need evolu-tionary biology's "adaptationist program": a research strategy in which theories of adaptive function are key inferential tools, used to identify and investigate the design of evolved systems. Using research on how humans reason about social exchange, the authors will (1) illustrate how theories of adaptive function can generate detailed and highly testable hypotheses about the design of computational machines in the human mind and (2) review research that tests for the presence of these machines. This research suggests that the human computational architec-ture contains an expert system designed for reasoning about cooperation for mutual benefit, with a subroutine specialized for cheater detection. Natural competences