Adaptive time-dependent decisions: behavior, cognition, and neural mechanisms
OA Version
Citation
Abstract
Decision makers can utilize temporal information in a context-dependent manner. The goal of this project was to examine such processes at the behavioral, cognitive and neural levels.
People can regulate their persistence towards uncertain prospects adaptively. This overall tendency toward adaptive calibration is accompanied by substantial individual differences. Study 1 proposed a reinforcement learning (RL) model to understand adaptive persistence behavior as it varies across individuals. I hypothesized that the RL model would be able to mimic human variation in adaptive persistence choices. Participant-level parameter fits enabled this model to account for the range of behaviors seen in three data sets involving a laboratory task that offered repeated choices between waiting for rewards and quitting (n=142, healthy young adults). As predicted, parameters of the model efficiently captured variation in multiple dimensions of behavior.
Inflexibility in calibrating persistence can make a person appear impulsive. Through two large-sample, online test-retest experiments on MTurk (n=397), Study 2 tested the hypotheses that a computerized adaptive persistence task would capture stable, trait-like variation across individuals and that this behavioral variation would correlate with self-reported impulsivity. Characteristics of individual survival curves and participant-level RL parameter fits were used to quantify distinct aspects of the task behavior. As predicted, both survival curve characteristics and RL parameters had fair to good test-retest reliability. Contrary to hypotheses, they showed little overlap with self-reported impulsivity.
Study 3 was a functional magnetic resonance imaging (fMRI) experiment investigating neural mechanisms underlying flexible, time-dependent decisions. 10 healthy young adults assessed how long ago a past event occurred and utilized this information to evaluate choice options. Consistent with the hypothesis that the brain computes context-dependent values in a distributed manner, value decoding was detected in multiple brain regions, including striatum, hippocampus, and angular gyrus. Representations of recency and value information were positioned in adjacent yet separate locations, suggesting that distinguishable brain systems preferentially engage in context processing and value computation.
Taken together, this work provides interpretable computational markers to capture trait-like variability in adaptive, time-dependent decisions and suggests such choices depend on a distributed neural system.
Description
License
Attribution-NonCommercial-NoDerivatives 4.0 International