The measure of affective decision making: Modulatory circuitry as interface between emotion and decision

Abstract

Decision making is influenced by modulatory processes that enable coordinated responses to environmental and emotional contexts. The measurement of modulatory processes is typically performed via biophysical metrics which carry only partial information on the unobserved processes. We provide an alternative, data-driven, methodology for the targeted measurement of the impact of modulatory processes on decisions. We apply directed dimensionality reduction to a large set of biometric measures including galvanic skin response, heart rate, pupilometry, facial emotion, and electroencephalography, to extract information predictive of human behavior in a standard two-alternative forced-choice decision making paradigm. Using a pre-existing model of decisions in this domain (i.e., the drift diffusion model) affords the ability to specify how the inferred modulatory process informs interpretable decision parameters. We validate this method with model comparisons together with cross validation. This method can be adapted to arbitrary decision domains to investigate how emotional state interacts with decision processes. We find an unexpected correlation between decision parameters, drift rate and decision threshold, when using this latent state extraction procedure not otherwise found when investigating behavioral responses alone. We interpret the correlation in parameters as evidence of their being both influenced by a common upstream modulatory process. We then systematically relax the constraints of the drift diffusion model and performed logistic regression to extract within trial weights on external information. We found that confidence acts as an internal representation of information reliability and adapts integration time to offset conditions of low information gain. Taken together, these findings support the interpretation that emotional state modulates decision making processes.