We study the wisdom of the crowd in three sequential decision‐making tasks: the Balloon Analogue Risk Task (BART), optimal stopping problems, and bandit problems. We consider a behavior‐based approach, using majority decisions to determine crowd behavior and show that this approach performs poorly in the BART and bandit tasks. The key problem is that the crowd becomes progressively more extreme as the decision sequence progresses, because the diversity of opinion that underlies the wisdom of the crowd is lost. We also (...) consider model‐based approaches to each task. This involves inferring cognitive models for each individual based on their observed behavior, and using these models to predict what each individual would do in any possible task situation. We show that this approach performs robustly well for all three tasks and has the additional advantage of being able to generalize to new problems for which there are no behavioral data. We discuss potential applications of the model‐based approach to real‐world sequential decision problems and discuss how our approach contributes to the understanding of collective intelligence. (shrink)

We consider a class of dynamic tournaments in which two contestants are faced with a choice between two courses of action. The first is a riskless option (“hold”) of maintaining the resources the contestant already has accumulated in her turn and ceding the initiative to her rival. The second is the bolder option (“roll”) of taking the initiative of accumulating additional resources, and thereby moving ahead of her rival, while at the same time sustaining a risk of temporary setback. We (...) study this tournament in the context of a jeopardy race game (JRG), extend the JRG to $N > 2$ contestants, and construct its equilibrium solution. Compared to the equilibrium solution, the results of three experiments reveal a dysfunctional bias in favor of the riskless option. This bias is substantially mitigated when the contestants are required to commit in advance how long to pursue the risky course of action. (shrink)

Figner, Mackinlay, Wilkening, and Weber (2009) developed the Columbia Card Task (CCT) to measure risk-taking attitudes. This tool consists of two versions: in the COLD version the decision maker needs to state in advance how many cards (out of 32) they want to turn over (so called static risk taking), in the HOT version they have the possibility of turning over all 32 cards one-by-one until they decide to finish (dynamic risk taking). We argue that the HOT version confounds an (...) individual’s willingness to accept risk with their beliefs in trend continuation vs. trend reversal in a prognostic task. In two experimental studies we show that people believing in trend continuation (momentum subjects) turn over more cards than those believing in trend reversal (contrarians) in the HOT version of the task. However, this is not the case in the COLD version. Thus, we provide evidence that, when considered as a dynamic risk propensity measure, the number of turned over cards in the HOT version of the CCT is a contaminated measure and reflects two phenomena: (1) risk preference and (2) the decision-maker’s belief in trend continuation. We speculate that other dynamic risk taking measures can also be biased by a momentum strategy. (shrink)

This paper addresses a common challenge with computational cognitive models: identifying parameter values that are both theoretically plausible and generate predictions that match well with empirical data. While computational models can offer deep explanations of cognition, they are computationally complex and often out of reach of traditional parameter fitting methods. Weak methodology may lead to premature rejection of valid models or to acceptance of models that might otherwise be falsified. Mathematically robust fitting methods are, therefore, essential to the progress of (...) computational modeling in cognitive science. In this article, we investigate the capability and role of modern fitting methods—including Bayesian optimization and approximate Bayesian computation—and contrast them to some more commonly used methods: grid search and Nelder–Mead optimization. Our investigation consists of a reanalysis of the fitting of two previous computational models: an Adaptive Control of Thought—Rational model of skill acquisition and a computational rationality model of visual search. The results contrast the efficiency and informativeness of the methods. A key advantage of the Bayesian methods is the ability to estimate the uncertainty of fitted parameter values. We conclude that approximate Bayesian computation is (a) efficient, (b) informative, and (c) offers a path to reproducible results. (shrink)

The Iowa Gambling Task (IGT) is widely used to study decision‐making within healthy and psychiatric populations. However, the complexity of the IGT makes it difficult to attribute variation in performance to specific cognitive processes. Several cognitive models have been proposed for the IGT in an effort to address this problem, but currently no single model shows optimal performance for both short‐ and long‐term prediction accuracy and parameter recovery. Here, we propose the Outcome‐Representation Learning (ORL) model, a novel model that provides (...) the best compromise between competing models. We test the performance of the ORL model on 393 subjects' data collected across multiple research sites, and we show that the ORL reveals distinct patterns of decision‐making in substance‐using populations. Our work highlights the importance of using multiple model comparison metrics to make valid inference with cognitive models and sheds light on learning mechanisms that play a role in underweighting of rare events. (shrink)