Response error and processing biases in confidence judgment

Previous studies showed that random error can explain overconfidence effects typically observed in the literature. One of these studies concluded that, after accounting for random error effects in the data, there is little support for cognitive‐processing biases in confidence elicitation. In this paper, we investigate more closely the random error explanation for overconfidence. We generated data from four models of confidence and then estimated the magnitude of random error in the data. Our results show that, in addition to the true magnitude of random error specified in the simulations, the error estimates are influenced by important cognitive‐processing biases in the confidence elicitation process. We found that random error in the response process can account for the degree of overconfidence found in calibration studies, even when that overconfidence is actually caused by other factors. Thus, the error models say little about whether cognitive biases are present in the confidence elicitation process. Copyright © 2008 John Wiley & Sons, Ltd.     

Data‐driven approaches to information access

This paper summarizes three lines of research that are motivated by the practical problem of helping users find information from external data sources, most notably computers. The application areas include information retrieval, text categorization, and question answering. A common theme in these applications is that practical information access problems can be solved by analyzing the statistical properties of words in large volumes of real world texts. The same statistical properties constrain human performance, thus we believe that solutions to practical information access problems can shed light on human knowledge representation and reasoning.     

A cost-benefit analysis of demand for food.

Laboratory studies of consumer demand theory require assumptions regarding the definition of price in the absence of a medium of exchange (money). In this study we test the proposition that the fundamental dimension of price is a cost-benefit ratio expressed as the effort expended per unit of food value consumed. Using rats as subjects, we tested the generality of this "unit price" concept by varying four dimensions of price: fixed-ratio schedule, number of food pellets per fixed-ratio completion, probability of reinforcement, and response lever weight or effort. Two levels of the last three factors were combined in a 2 x 2 x 2 design giving eight groups. Each group was studied under a series of six FR schedules. Using the nominal values of all factors to determine unit price, we found that grams of food consumed plotted as a function of unit price followed a single demand curve. Similarly, total work output (responses x effort) conformed to a single function when plotted in terms of unit price. These observations provided a template for interpreting the effects of biological factors, such as brain lesions or drugs, that might alter the cost-benefit ratio.     

Changes in functional response units with briefly delayed reinforcement

In two experiments, key-peck responding of pigeons was compared under variable-interval schedules that arranged immediate reinforcement and ones that arranged unsignaled delays of reinforcement. Responses during the nominal unsignaled delay periods had no effect on the reinforcer presentations. In Experiment 1, the unsignaled delays were studied using variable-interval schedules as baselines. Relative to the immediate reinforcement condition, 0.5-s unsignaled delays decreased the duration of the reinforced interresponse times and increased the overall frequency of short (<0.5-s) interresponse times. Longer, 5.0-s unsignaled delays increased the duration of the reinforced interresponse times and decreased the overall frequency of the short interresponse times. In Experiment 2, similar effects to those of Experiment 1 were obtained when the 0.5-s unsignaled delays were imposed upon a baseline schedule that explicitly arranged reinforcement of short interresponse times and therefore already generated a large number of short interresponse times. The results support earlier suggestions that the unsignaled 0.5-s delays change the functional response unit from a single key peck to a multiple key-peck unit. These findings are discussed in terms of the mechanisms by which contingencies control response structure in the absence of specific structural requirements.     

Concurrent-schedule performance: Effects of relative and overall reinforcer rate

Six pigeons were trained to respond on two keys, each of which provided reinforcers on an arithmetic variable-interval schedule. These concurrent schedules ran nonindependently with a 2-s changeover delay. Six sets of conditions were conducted. Within each set of conditions the ratio of reinforcers available on the two alternatives was varied, but the arranged overall reinforcer rate remained constant. Each set of conditions used a different overall reinforcer rate, ranging from 0.22 reinforcers per minute to 10 reinforcers per minute. The generalized matching law fit the data from each set of conditions, but sensitivity to reinforcer frequency (a) decreased as the overall reinforcer rate decreased for both time allocation and response allocation based analyses of the data. Overall response rates did not vary with changes in relative reinforcer rate, but decreased with decreases in overall reinforcer rate. Changeover rates varied as a function of both relative and overall reinforcer rates. However, as explanations based on changeover rate seem unable to deal with the changes in generalized matching sensitivity, discrimination accounts of choice may offer a more promising interpretation.     

Canonical analysis of two convex polyhedral cones and applications

Canonical analysis of two convex polyhedral cones consists in looking for two vectors (one in each cone) whose square cosine is a maximum. This paper presents new results about the properties of the optimal solution to this problem, and also discusses in detail the convergence of an alternating least squares algorithm. The set of scalings of an ordinal variable is a convex polyhedral cone, which thus plays an important role in optimal scaling methods for the analysis of ordinal data. Monotone analysis of variance, and correspondence analysis subject to an ordinal constraint on one of the factors are both canonical analyses of a convex polyhedral cone and a subspace. Optimal multiple regression of a dependent ordinal variable on a set of independent ordinal variables is a canonical analysis of two convex polyhedral cones as long as the signs of the regression coefficients are given. We discuss these three situations and illustrate them by examples.     

A method of optimal scaling for multivariate ordinal data and its extensions

This paper develops a method of optimal scaling for multivariate ordinal data, in the framework of a generalized principal component analysis. This method yields a multidimensional configuration of items, a unidimensional scale of category weights for each item and, optionally, a multidimensional configuration of subjects. The computation is performed by alternately solving an eigenvalue problem and executing a quasi-Newton projection method. The algorithm is extended for analysis of data with mixed measurement levels or for analysis with a combined weighting of items. Numerical examples and simulations are provided. The algorithm is discussed and compared with some related methods.Earlier results of this research appeared in Saito and Otsu (1983). The authors would like to acknowledge the helpful comments and encouragement of the editor.     

Functional measurement approach to self-estimation in multiattribute evaluation

Multiattribute analysis depends on measurement of values and weights. Unless these measures reflect the decision maker's true values and weights, the multiattribute formula may put a less-preferred alternative in first place. To avoid such disordinality requires stringent measurement conditions: First, the values and weights must be on linear (equal interval) or ratio (known zero) scales. Second, these scales must satisfy a condition of common unit across disparate attribute dimensions. Most methods of range adjustment beg both of these measurement questions. Functional measurement theory can solve both problems and so can be useful in multiattribute analysis. Past work has established the operation of a general cognitive algebra as an empirical reality. The averaging model, in particular, makes possible the definition and estimation of weights and values as distinct psychological parameters. It can also solve the problem of common unit. Cognitive algebra thus provides a grounded theoretical foundation on which to develop self-estimation methodology, in which decision makers provide direct estimates of their values and weights. The logic is straightforward. Functional measurement can analyze global judgments to obtain validated psychological scales. These scales may then be used as validational criteria for the self-estimates. Procedures to eliminate biases in the self-estimates can thus be tested and refined in well-learned multiattribute tasks, such as judgments of meals, in which global judgments are trustworthy. Once developed, such self-estimation procedures may be used with some confidence for general multiattribute analysis. A number of studies from 20-odd years of work on the theory of information integration are summarized to show good, although not unmixed promise for self-estimation.     

Factor invariance of the Millon Clinical Multiaxial Inventory

Several factor analyses of the Millon Clinical Multiaxial Inventory (MCMI) have resulted in very similar solutions. Interpretation of this consistency is hampered by the fact that the 20 scales of the inventory share items. Overlapping items cause the scales to be linearly dependent and may create structure in the interscale correlation matrix which is separate from the subject response patterns. A factor analysis was performed on the matrix of item-overlap coefficients which describes the underlying artifactual structure of the instrument. Data from two new subject samples were factor analyzed and compared to previously published studies. Similarity coefficients among factors across studies were calculated.     

Effects of response requirement and alcohol on human aggressive responding.

Nine men participated in two experiments to determine the effects of increased response requirement and alcohol administration on free-operant aggressive responding. Two response buttons (A and B) were available. Pressing Button A was maintained by a fixed-ratio 100 schedule of point presentation. Subjects were instructed that completion of each fixed-ratio 10 on Button B resulted in the subtraction of a point from a fictitious second subject. Button B presses were defined as aggressive because they ostensibly resulted in the presentation of an aversive stimulus to another person. Aggressive responses were engendered by a random-time schedule of point loss and were maintained by initiation of intervals free of point loss. Instructions attributed these point losses to Button B presses of the fictitious other subject. In Experiment 1, increasing the ratio requirement on Button B decreased the number of ratios completed in 4 of 5 subjects. In Experiment 2, the effects of placebo and three alcohol doses (0.125, 0.25, and 0.375 g/kg) were determined when Button B presses were maintained at ratio values of 20, 40 and 80. Three subjects who reduced aggressive responding with increasing fixed-ratio values reduced aggressive responding further at higher alcohol doses. One subject who did not reduce aggressive responding with increasing fixed-ratio values increased aggressive responding at the highest alcohol dose. The results of this study support suggestions that alcohol alters aggressive behavior by reducing the control of competing contingencies.