ALCOVE: An exemplar-based connectionist model of category learning.

ALCOVE (attention learning covering map) is a connectionist model of category learning that incorporates an exemplar-based representation (D. L. Medin and M. M. Schaffer, 1978; R. M. Nosofsky, 1986) with error-driven learning (M. A. Gluck and G. H. Bower, 1988; D. E. Rumelhart et al, 1986). ALCOVE selectively attends to relevant stimulus dimensions, is sensitive to correlated dimensions, can account for a form of base-rate neglect, does not suffer catastrophic forgetting, and can exhibit 3-stage (U-shaped) learning of high-frequency exceptions to rules, whereas such effects are not easily accounted for by models using other combinations of representation and learning method. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Base-rate effects in category learning: A comparison of parallel network and memory storage-retrieval models.

Exemplar-memory and adaptive network models were compared in application to category learning data, with special attention to base rate effects on learning and transfer performance. Subjects classified symptom charts of hypothetical patients into disease categories, with informative feedback on learning trials and with the feedback either given or withheld on test trials that followed each fourth of the learning series. The network model proved notably accurate and uniformly superior to the exemplar model in accounting for the detailed course of learning; both the parallel, interactive aspect of the network model and its particular learning algorithm contribute to this superiority. During learning, subjects' performance reflected both category base rates and feature (symptom) probabilities in a nearly optimal manner, a result predicted by both models, though more accurately by the network model. However, under some test conditions, the data showed substantial base-rate neglect, in agreement with M. A. Gluck and G. H. Bower (see record 1989-00340-001). (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Everyday base rates (sex stereotypes): Potent and resilient.

Undergraduate Ss studied photographs of students and estimated the heights of the pictured models. Contrary to reports of base-rate neglect, sex stereotypes regarding height (the implicit recognition that men are normally taller than women) significantly affected these estimates, even when the targets' actual height was statistically controlled. Base rates were especially influential when information about targets was ambiguous, that is when targets were pictured seated. These base-rate effects were robust, remaining significant and substantial despite efforts to lessen their magnitude. Attempts to reduce base-rate effects by encouraging Ss to strive for accuracy, discouraging their reliance on the target's sex (as a cue), or offering cash rewards for accuracy did not succeed. Informing Ss that for the sample to be judged, sex would not predict targets' heights attenuated the base-rate effect, although it remained highly significant. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The effects of judge's roles on probability judgment

A number of studies on probability judgment have reported that people often tend to neglect the base-rate in the Cab Problem (Bar-Hillel, 1980). To explain the base-rate neglect, the present research hypothesized that subjects presented with the Cab Problem focused on the case information in the problem, because the subjects were asked to estimate probability based on that particular case. In 'two experiments, subjects' individuated or general interpretation of the Cab Problem was manipulated by instructing the subjects to play either responsible or neutral judge's role, respectively. The results suggested that in the responsible judge condition the case information had large impact on subjects' judgment, whereas in the neutral judge condition it did not. These results implied that the subject's base-rate neglect arose from taking the individuated interpretation.     

Representing the task in Bayesian reasoning: Comment on Lovett and Schunn (1999).

The RCCL model (M. C. Lovett and C. D. Schunn, 1999; see record 1999-05245-001) produces predictions that are non-novel or that do not truly spring from its principles. However, it offers the valuable insight that learning processes may affect the selection of both representations and strategies within those representations, and points the way to possible theoretical progress on implicit and explicit control. The authors' account of base-rate neglect under direct experience is compared with RCCL, and it is concluded that learning-based models allow for tests that are not fostered by representation-based models. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Base rates in category learning.

Previous researchers have discovered perplexing inconsistencies in how people appear to utilize category base rates when making category judgments. In particular, D. L. Medin and S. M. Edelson (see record 1988-31640-001) found an inverse base-rate effect, in which participants tended to select a rare category when tested with a combination of conflicting cues, and M. A. Gluck and G. H. Bower (see record 1989-00340-001) reported apparent base-rate neglect, in which participants tended to select a rare category when tested with a single symptom for which objective diagnosticity was equal for all categories. This article suggests that common principles underlie both effects. First, base-rate information is learned and consistently applied to all training and testing cases. Second, the crucial effect of base rates is to cause frequent categories to be learned before rare categories so that the frequent categories are encoded by their typical features and the rare categories are encoded by their distinctive features. Four new experiments provide evidence consistent with those principles. The principles are formalized in a new connectionist model that can rapidly shift attention to distinctive features. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

On the relative independence of thinking biases and cognitive ability.

In 7 different studies, the authors observed that a large number of thinking biases are uncorrelated with cognitive ability. These thinking biases include some of the most classic and well-studied biases in the heuristics and biases literature, including the conjunction effect, framing effects, anchoring effects, outcome bias, base-rate neglect, "less is more" effects, affect biases, omission bias, myside bias, sunk-cost effect, and certainty effects that violate the axioms of expected utility theory. In a further experiment, the authors nonetheless showed that cognitive ability does correlate with the tendency to avoid some rational thinking biases, specifically the tendency to display denominator neglect, probability matching rather than maximizing, belief bias, and matching bias on the 4-card selection task. The authors present a framework for predicting when cognitive ability will and will not correlate with a rational thinking tendency. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Conversational conventions, order of information acquisition, and the effect of base rates and individuating information on social judgments.

Numerous studies have shown that people underuse base-rate information in making social judgments and rely instead almost exclusively on individuating information. Seven studies reported here demonstrate that this occurred partly because most past studies gave Ss base-rate information before giving them individuating information. A recency effect in the use of base-rate and individuating information is demonstrated using a set of reasoning problems of varying character. The recency effect is shown to be the result of Ss' inferences (based on conversational conventions) that the experimenter believes that Ss should rely most on the piece of information presented last. Additional evidence discredits the hypothesis that the recency effect is due to heightened availability of more recently acquired information in memory. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Thirty categorization results in search of a model.

One category structure dominated in the shift toward exemplar-based theories of categorization. Given the theoretical burden on this category structure, the authors reanalyzed 30 of its uses over 20 years in 8 articles. The authors suggest 4 conclusions. (1) This category structure may encourage exemplar-memorization processes because of its poor structure, the learning difficulties it causes, and its small, memorizable exemplar sets. Its results may only generalize narrowly. (2) Exemplar models have an advantage in fitting these 30 data sets only because they reproduce a performance advantage for training items. Other models fit equally well if granted this capacity. (3) A simpler exemplar process than assumed by exemplar models suffices to explain these data sets. (4) An important qualitative result predicted by exemplar theory is not found overall and possibly should not even be expected. The authors conclude that the data produced by this category structure do not clearly support exemplar theory. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Base rates can affect individual predictions.

356 college students in 4 experiments learned base-rate information and then participated in a predictive social judgment task. In all experiments, predictive behavior was significantly affected by the relevant base rates. In Exp II, for example, Ss in a high base-rate condition were more likely to predict that a given student planned to enroll for further training than were Ss in a low base-rate condition. The base-rate variable had a greater influence than might have been anticipated from the Ss' confidence ratings or from their replies to a series of postexperimental questions concerning the cues they had relied on in formulating their predictions. Nonetheless, the base-rate data were substantially less influential than a Bayesian model would require. In contrast to D. Kahneman and A. Tversky's (1973) analysis, individuating information that reflected a high (vs low) degree of representativeness did not determine the magnitude of the base-rate effect. (22 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

High-level reasoning and base-rate use: Do we need cue-competition to explain the inverse base-rate effect?

Previous accounts of the inverse base-rate effect (D. L. Medin & S. M. Edelson, see record 1988-31640-001) have revolved around the concept of cue-competition. In this article, the authors propose that high-level reasoning in the form of an eliminative inference mechanism may contribute to the effect. A quantitative implementation of this idea demonstrates that it has the power by itself to produce the pattern of base-rate effects in the Medin and Edelson (1988) design. Four predictions are derived that contradict the predictions by attention to distinctive input (ADIT; J. K. Kruschke, see record 1996-02680-001), up to date the most successful account of the inverse base-rate effect. Results from 3 experiments disconfirm the predictions by ADIT and demonstrate the importance of high-level reasoning in designs of the Medin and Edelson kind. Implications for the interpretation of the inverse base-rate effect and the attention-shifting mechanisms presumed by ADIT are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Detecting symptom exaggeration in combat veterans using the MMPI–2 symptom validity scales: A mixed group validation.

Although validity scales of the Minnesota Multiphasic Personality Inventory–2 (MMPI–2; J. N. Butcher, W. G. Dahlstrom, J. R. Graham, A. Tellegen, & B. Kaemmer, 1989) have proven useful in the detection of symptom exaggeration in criterion-group validation (CGV) studies, usually comparing instructed feigners with known patient groups, the application of these scales has been problematic when assessing combat veterans undergoing posttraumatic stress disorder (PTSD) examinations. Mixed group validation (MGV) was employed to determine the efficacy of MMPI–2 exaggeration scales in compensation-seeking (CS) and noncompensation-seeking (NCS) veterans. Unlike CGV, MGV allows for a mix of exaggerating and nonexaggerating individuals in each group, does not require that the exaggeration versus nonexaggerating status of any individual be known, and can be adjusted for different base-rate estimates. MMPI–2 responses of 377 male veterans were examined according to CS versus NCS status. MGV was calculated using 4 sets of base-rate estimates drawn from the literature. The validity scales generally performed well (adequate sensitivity, specificity, and efficiency) under most base-rate estimations, and most produced cutoff scores that showed adequate detection of symptom exaggeration, regardless of base-rate assumptions. These results support the use of MMPI–2 validity scales for PTSD evaluations in veteran populations, even under varying base rates of symptom exaggeration. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

An associative framework for probability judgement: An application to biases.

Three experiments show that understanding of biases in probability judgment can be improved by extending the application of the associative-learning framework. In Experiment 1, the authors used M. A. Gluck and G. H. Bower's (1988) diagnostic-learning task to replicate apparent base-rate neglect and to induce the conjunction fallacy in a later judgment phase as a by-product of the conversion bias. In Experiment 2, the authors found stronger evidence of the conversion bias with the same learning task. In Experiment 3, the authors changed the diagnostic-learning task to induce some conjunction fallacies that were not based on the conversion bias. The authors show that the conjunction fallacies obtained in Experiment 3 can be explained by adding an averaging component to M. A. Gluck and G. H. Bower's model. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Can Sample-Specific Simulations Help Detect Low Base-Rate Taxonicity?

The authors examined the role of the sample-specific simulations (SSS; A. M. Ruscio & J. Ruscio, 2002; J. Ruscio & A. M. Ruscio, 2004) procedure in detecting low base-rate taxa that might otherwise prove elusive. The procedure preserved key distributional characteristics for moderate to high base-rate taxa, but it performed inadequately for low base-rate taxa, with low to moderate nuisance covariance, decreasing indicator validity, and increasing average nuisance covariance. Visual inspection of the curves produced by the SSS procedure, combined with the expectation that the inchworm test should produce a clear peak for low base-rate taxonic samples, led to inappropriate conclusions about the suitability of data. The authors speculate that examination of low base-rate taxa using SSS may require other algorithms for generating comparison distributions, consideration of the shape of the whole curve, and/or relaxation of the expectation that a clear peak will emerge with the inchworm test. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Base-rate and payoff effects in multidimensional perceptual categorization

The optimality of multidimensional perceptual categorization performance with unequal base rates and payoffs was examined. In Experiment 1, observers learned simultaneously the category structures and base rates or payoffs. Observers showed conservative cutoff placement when payoffs were unequal and extreme cutoff placement when base rates were unequal. In Experiment 2, observers were trained on the category structures before the base-rate or payoff manipulation. Simultaneous base-rate and payoff manipulations tested the hypothesis that base-rate information and payoff information are combined independently. Observers showed (a) small suboptimalities in base-rate and payoff estimation, (b) no qualitative differences across base-rate and payoff conditions, and (c) support for the hypothesis that base-rate and payoff information is combined independently. Implications for current theories of base-rate and payoff learning are discussed.     

Beware of samples! A cognitive-ecological sampling approach to judgment biases.

A cognitive-ecological approach to judgment biases is presented and substantiated by recent empirical evidence. Latent properties of the environment are not amenable to direct assessment but have to be inferred from empirical samples that provide the interface between cognition and the environment. The sampling process may draw on the external world or on internal memories. For systematic reasons (proximity, salience, and focus of attention), the resulting samples tend to be biased (selective, skewed, or conditional on information search strategies). Because people lack the metacognitive ability to understand and control for sampling constraints (predictor sampling, criterion sampling, selective-outcome sampling, etc.), the sampling biases carry over to subsequent judgments. Within this framework, alternative accounts are offered for a number of judgment biases, such as base-rate neglect, confirmation bias, illusory correlation, pseudocontingency, Simpson's paradox, outgroup devaluation, and pragmatic-confusion effects. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Limitations of exemplar models of multi-attribute probabilistic inference.

Observers were presented with pairs of objects varying along binary-valued attributes and learned to predict which member of each pair had a greater value on a continuously varying criterion variable. The predictions from exemplar models of categorization were contrasted with classic alternative models, including generalized versions of a "take-the-best" model and a weighted-additive model, by testing structures in which interactions between attributes predicted the magnitude of the criterion variable. Under typical training conditions, observers showed little sensitivity to the attribute interactions, thereby challenging the predictions from the exemplar models. In a condition involving highly extended training, observers eventually learned the relations between the attribute interactions and the criterion variable. However, an analysis of the observers' response times for making their paired-comparison decisions also challenged the exemplar model predictions. Instead, it appeared that most observers recoded the interacting attributes into emergent configural cues. They then applied a set of hierarchically organized rules based on the priority of the cues to make their decisions. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Wanted: A New Psychology of Exemplars.

Present-day exemplar theory faces difficult challenges, and important questions have arisen about the kinds of exemplar effects and processes that are empirically supported, and about the kind of exemplar theory that could still be constructive. One question concerns whether exemplar generalization in memory and categorization is broad and collective - extending to many related exemplars stored in memory - or whether it is focused and singular - extending only to highly similar (nearly identical) exemplars. The present article considers this continuum from broad to narrow generalization. I demonstrate that in prominent memory and category tasks - tasks in which exemplar theory predicts broad generalization - generalization i.s in psychological reality very tightly focused. These demonstrations could ground a new, productive exemplar theory that is true to psychological process as humans conduct themselves in memory and category tasks. This new psychology may actually reprise the traditional exemplar theory that predated our sophisticated, mathematical exemplar models. (PsycINFO Database Record (c) 2010 APA, all rights reserved)