Optimization of Adaptable Decision-Making During Uncertainty

Complex decision-making problems are often shrouded in layers of uncertainty. When ambiguous evidence (e.g., from memory, or one’s senses) is compounded by dynamic changes in the environment, or varying decision-outcome contingencies (e.g., the cost of making a critical error), the optimal decision strategy can be nontrivial to identify. In general, however, changing contexts and circumstances mandate flexible adaptation of a criterion (i.e., the necessary evidence threshold for action) that optimally controls/minimizes the frequency of false positive or false negative decision errors. We refer to this strategic flexibility as criterion shifting. Our lab remains the preeminent research group dedicated to understanding both the cognitive and neurobiological mechanisms of criterion shifting behavior, including the ways in which these factors vary on a person-by-person basis.

This project will use novel experimental paradigms and methods to derive a more detailed account of these decision making behaviors and address fundamental questions about their underlying principles, malleability, and optimizability. More specifically, we will explore the factors contributing to individual limitations in adaptive decision-making and the extent to which one can learn to be a more optimal criterion shifter, either through explicit training and feedback, or via more implicit processes. One factor identified as a target for further investigation is the ability to accurately calibrate one’s confidence to one’s performance—this ‘metacognitive efficiency’ (or lack thereof) may be an important mediator through which individuals can become more adaptable (e.g., by allowing a more robust assessment of the uncertainty surrounding each judgment). Acquisition of fMRI data during performance of tasks requiring decision-making under uncertainty will further allow for robust identification of the neural bases of both confidence calibration and adaptive decision-making. We will use cutting-edge network neuroscience methods and other multivariate analysis tools to map both overlapping and unique network components that explain variability in behavior. Such a map could usefully inform future neurostimulation studies with targets for intervention.

Together, this work will therefore not only address basic science questions about the nature of brain-behavior relationships and the variability inherent therein, but will also provide quantitative commentaries on the methods we use to manipulate those relationships in an effort to produce desired behavioral outcomes.