A Bayesian analysis of human decision-making on bandit problems

The bandit problem is a dynamic decision-making task that is simply described, well-suited to controlled laboratory study, and representative of a broad class of real-world problems. In bandit problems, people must choose between a set of alternatives, each with different unknown reward rates, to maximize the total reward they receive over a fixed number of trials. A key feature of the task is that it challenges people to balance the exploration of unfamiliar choices with the exploitation of familiar ones. We use a Bayesian model of optimal decision-making on the task, in which how people balance exploration with exploitation depends on their assumptions about the distribution of reward rates. We also use Bayesian model selection measures that assess how well people adhere to an optimal decision process, compared to simpler heuristic decision strategies. Using these models, we make inferences about the decision-making of 451 participants who completed a set of bandit problems, and relate various measures of their performance to other psychological variables, including psychometric assessments of cognitive abilities and personality traits. We find clear evidence of individual differences in the way the participants made decisions on the bandit problems, and some interesting correlations with measures of general intelligence.     

Learning, risk attitude and hot stoves in restless bandit problems

This research examines decisions from experience in restless bandit problems. Two experiments revealed four main effects. (1) Risk neutrality: the typical participant did not learn to become risk averse, a contradiction of the hot stove effect. (2) Sensitivity to the transition probabilities that govern the Markov process. (3) Positive recency: the probability of a risky choice being repeated was higher after a win than after a loss. (4) Inertia: the probability of a risky choice being repeated following a loss was higher than the probability of a risky choice after a safe choice. These results can be described with a simple contingent sampler model, which assumes that choices are made based on small samples of experiences contingent on the current state.     

A general solution for a class of weakly constrained linear regression problems

This paper contains a globally optimal solution for a class of functions composed of a linear regression function and a penalty function for the sum of squared regression weights. Global optimality is obtained from inequalities rather than from partial derivatives of a Lagrangian function. Applications arise in multidimensional scaling of symmetric or rectangular matrices of squared distances, in Procrustes analysis, and in ridge regression analysis. The similarity of existing solutions for these applications is explained by considering them as special cases of the general class of functions addressed.The author is obliged to Henk Kiers and Willem Heiser for helpful comments.     

The design and evaluation of a worksharing system for experimental group living

Failure to share household chores equitably may be a major cause of the high failure rate of experimental group-living arrangements. A behavioral approach to worksharing based on a point system was implemented in one experimental group and its most important components experimentally evaluated. Experiment I showed that awarding credits produced more work than not awarding credits. Experiment II showed that making credits contingent on the outcome of a detailed inspection produced more work than awarding credits noncontingently. Experiment III demonstrated that awarding rent reductions contingent on credit earnings produced more work than awarding rent reductions noncontingently. Other evaluative data suggest that the resulting living arrangement is cheaper, more effective, and more satisfactory to the residents when compared to the most popular alternative living arrangements.     

Optimal design of parity tests

Researchers often test a null hypothesis of no ability in the populattion (the so-called “parity” hypothesis) using a single, forced-choice question with k alternatives. In this study a result is presented which should help researchers select the number of alternatives that maximizes the statistical power of the parity hypothesis test. Also the conditions under which it is always beneficial to add alternatives to the test are derived. Finally, the derived result is used to compare several popular parity test designs. The results show that the frequently used triangle test is optimal under a very broad range of plausible conditions.     

Optimal training design for procedural motor skills: a review and application to laparoscopic surgery

This literature review covers the choices to consider in training complex procedural, perceptual and motor skills. In particular, we focus on laparoscopic surgery. An overview is provided of important training factors modulating the acquisition, durability, transfer, and efficiency of trained skills. We summarize empirical studies and their theoretical background on the topic of training complex cognitive and motor skills that are pertinent to proficiency in laparoscopic surgery. The overview pertains to surgical simulation training for laparoscopy, but also to training in other demanding procedural and dexterous tasks, such as aviation, managing complex systems and sports. Evidence-based recommendations are provided for facilitating efficiency in laparoscopic motor skill training such as session spacing, adaptive training, task variability, part-task training, mental imagery and deliberate practice.     

The optimal experimental design for multiple alternatives perceptual search

Perceptual bias is inherent to all our senses, particularly in the form of visual illusions and aftereffects. However, many experiments measuring perceptual biases may be susceptible to nonperceptual factors, such as response bias and decision criteria. Here, we quantify how robust multiple alternative perceptual search (MAPS) is for disentangling estimates of perceptual biases from these confounding factors. First, our results show that while there are considerable response biases in our four-alternative forced-choice design, these are unrelated to perceptual biases estimates, and these response biases are not produced by the response modality (keyboard vs. mouse). We also show that perceptual bias estimates are reduced when feedback is given on each trial, likely due to feedback enabling observers to partially (and actively) correct for perceptual biases. However, this does not impact the reliability with which MAPS detects the presence of perceptual biases. Finally, our results show that MAPS can detect actual perceptual biases and is not a decisional bias towards choosing the target in the middle of the candidate stimulus distribution. In summary, researchers conducting a MAPS experiment should use a constant reference stimulus, but consider varying the mean of the candidate distribution. Ideally, they should not employ trial-wise feedback if the magnitude of perceptual biases is of interest.     

Optimal sample sizes for the design of reliability studies: Power consideration

Intraclass correlation coefficients are used extensively to measure the reliability or degree of resemblance among group members in multilevel research. This study concerns the problem of the necessary sample size to ensure adequate statistical power for hypothesis tests concerning the intraclass correlation coefficient in the one-way random-effects model. In view of the incomplete and problematic numerical results in the literature, the approximate sample size formula constructed from Fisher’s transformation is reevaluated and compared with an exact approach across a wide range of model configurations. These comprehensive examinations showed that the Fisher transformation method is appropriate only under limited circumstances, and therefore it is not recommended as a general method in practice. For advance design planning of reliability studies, the exact sample size procedures are fully described and illustrated for various allocation and cost schemes. Corresponding computer programs are also developed to implement the suggested algorithms.     

Optimal trade-offs of multiple factors in transfer pricing problems

Transfer pricing problems have been extensively researched by a number of scholars. It is well recognized that transfer pricing problems have a multiple-criteria (objectives) feature and can be formulated as a model of multiple-criteria linear programming. However, few methods have the capability of dealing with all possible optimal trade-offs of multiple criteria in optimal solutions of the models. In this paper a linear multiple-factor model is developed to provide managers with a more systematic and comprehensive scenario of all possible optimal transfer prices depending on both multiple criteria and multiple constraint levels. The trade-offs of all possible optimal transfer prices can be used as a basis for managers of a corporation to make a high-quality decision in selecting their transfer pricing systems for business competition. © 1998 John Wiley & Sons, Ltd.     

Ready-signal effects as a function of experimental design in simple reaction time

In the between-S design, 20 Ss had a visual ready signal presented on each trial; for another 20 Ss, the ready signal was always absent. The within-S design consisted of 40 Ss that experienced both ready-signal conditions in semirandom order. Two intensities of a 1,000-Hz tone were used as the response signals Ready-signal manipulation had pronounced effects under the within-S but not the between-S design irrespective of response-signal intensity     

Crossover of burnout: An experimental design

Two studies tested the hypothesis that burnout may cross over from one person to another. In Study 1, teachers were randomly exposed to a bogus newspaper article in which a colleague expressed himself negatively about his work (burnout condition), or about a topic unrelated to work (control condition). The results showed that participants' burnout (exhaustion and depersonalization) was higher in the burnout condition compared to the control condition. In Study 2, soldiers were randomly exposed to a videotape of a burned-out or an engaged colleague who was either similar in profession and status (soldier), or who had a considerably higher status (squadron leader). The results were partly consistent with those of Study 1, and confirmed the crossover of burnout (cynicism and reduced professional efficacy). In line with predictions, a significant interaction effect for cynicism revealed that the crossover of burnout is moderated by similarity with the stimulus person.     

Using quizzes to enhance summative-assessment performance in a web-based class: An experimental study

The testing effect—a mnemonic benefit of quizzing or testing—has been oft-demonstrated in the laboratory, but rarely evaluated experimentally in the classroom. In a college web-based class, using a within-subjects design, core target concepts were quizzed with multiple choice and short-answer questions, presented for reading, or not presented (no additional-exposure control). Multiple attempts on each quiz (and in the read-control) were encouraged and feedback was available after each. When exam questions were identical to earlier-seen quizzes, short answer and multiple choice quizzes enhanced exam performance over rereading targeted material. When examination questions were different but related to the quizzes, short answer and multiple-choice quizzes conferred benefits over unquizzed target content, and produced exam performance that was nominally better than rereading target material. These experimental results indicate that unsupervised on-line quizzing in a college course enhances exam performance, thereby representing an important extension of laboratory testing effects into the classroom setting.     

A multivariate analysis of an experimental design involving a complete set of latin squares

A model for an experimental design involving a complete set of Latin squares for testing the homogeneity of treatment effects was constructed and analyzed by Gourlay. In his analysis, however, if one or both of the preliminaryF-tests are significant, the analysis cannot differentiate. He then suggests the use of a less desirable test which is biased and has fewer degrees of freedom, regardless of the number of replications (the d.f. cannot be increased by increasing the replications). Further, when heterogeneity of variance occurs, Gourlay's test procedures are in general invalid. The present paper reviews Gourlay's analysis and proposes a modified test procedure.My thanks are due to Drs. R. S. Hirsch, R. L. Erdmann, and R. M. Simons of IBM Corporation for many stimulating discussions, and to Professor D. Teichroew of Stanford University for permission to refer to his paper [6] and his assistance. I also wish to thank B. A. Snyder for correcting many linguistic mistakes.     

Fractional Design Wizard: A computer program for cost-effective experimental research design

The computer program Fractional Design Wizard creates fractional factorial designs that are cost-effective and especially useful for discarding irrelevant factors from a large number of possible candidates. The program is intended for researchers who are relativelynew to the field of fractional design and who want to acquaint themselves with the use of fractions for the reduction of large experimental designs. Fractional designs allow estimation of main effects, and sometimes two-way interactions, without one’s having to examine all treatment conditions. The program needs Microsoft Windows 95 or better and 32 MB of memory. In a step-by-step fashion, the user can specify the required properties of the fractional design. When there are more valid designs, the user can generate these successively If necessary the user can go back to diminish the requirements. The output can be copied, printed, and saved. The program generates all the information that is needed for the use and interpretation of fractional designs. A help file explains the use of the program and also the purpose, the analysis, and the interpretation of fractional designs. The program, which is written in Object Pascal, is available as freeware on www.fss.uu.nl/ms/hl/fracdes.htm.