Model-based Analysis of a Novel Variant of the Balloon Analogue Risk Task Improves Sensitivity to Risk-taking Behavior

The assessment and quantification of risk-taking behaviors is critical to parsing out individual differences in cognitive performance. The Balloon Analogue Risk Task (BART) is frequently implemented for differentiating risk-taking behaviors between different populations. We developed a new version of the BART that is able to differentiate decision makers’ (DM) behavior within a single population. Past BART studies typically utilize summary statistics that disregard the experimental context at the choice-by-choice level, grossly oversimplifying risk-taking behavior. To address this, we present a new methodology and metric for quantifying risk-taking behaviors by comparing a DM’s sequence of decisions to those of an optimal performer with knowledge of the popping probability of the balloons. We then propose and assess several possible cognitive mechanisms that may underlie the decision-making process within a generative modeling framework. This culminates in the Predictive Linear Utility Model (PLUM), which accurately predicts observed behavior. Finally, we compare a version of PLUM fit only to choices to a version fit to choices and reaction times in order to assess the utility of reaction times within this task environment.