UC San Diego

Intelligent agents often need to make actions with uncertain consequences under changing environment, and to modify those actions adaptively according to ongoing sensory processing and changing task demands. The ability to cancel or modify planned actions according to changing task conditions is known as inhibitory control, and thought to be an important aspect of human cognitive function. Inhibitory control has been studied extensively using the stop-signal experiment. Although a few models and experiments attempted to explain the subject's behavioral result,much work was still needed to understand the underlying mechanism.

Using Bayesian inference, hidden Markov model and stochastic control theory, this dissertation proposes new model and experimental investigations to attain a more comprehensive understanding of the underlying mechanism of human decision making process in inhibitory control. We demonstrate how human's reaction time, previously thought of as a random quantity, is highly correlated with model simulated predictive belief state. More specifically, the model and data enable us to provide strong evidence that Go process and Stop process are highly dependent, in contrast to being independent, as previously proposed. Our new proposed model can not only cover the behavior data but also the neural data. Finally, by applying our model to the clinical data, we discover the behavior and neural difference between methamphetamine-dependent individuals and comparison group, indicating that the model simulated quantity could be served as a biomarker to predict substance dependent user.