Development of Novel Generative Models of Sequentially Dependent Decision Making in Risk-Ambiguous Environments

In this dissertation, I present multiple generative cognitive models of sequentially dependent decision-making in risk-ambiguous environments. Cognitive models provide a means by which we can test theorized latent mechanisms that give rise to observed behavior. Utilizing variants of the Balloon Analogue Risk Task (BART), we investigate risk-related decision-making when the likelihood of success and failure is unknown to participants. We present a novel performance metric for the BART that compares participant performance to optimal performance, eliminating the issue of data censorship associated with traditional performance metrics. We evaluate a model of BART decision-making and use it to generate representations of subjective risk, reward, and value in order to relate these model-predicted time series with hemodynamic response via fMRI. We then investigate asymmetries in behavior in the contexts of risk taking versus risk avoidance.