BEYOND HOMOGENOUS DECISION-MAKING MODELS: ROLE OF BRAIN AREAS INTERACTION AND HETEROGENEITY

Most existing mechanistic models of decision making consider microcircuits within a homogenous brain area and fail to consider the distributed nature of this cognitive process. More specifically, these models do not take into account: a) how different types of information might influence choice in different stages of the process; and b) how interactions and the heterogeneity between circuits across different brain areas might be essential for the process. In this dissertation, I present a combination of experimental and modeling approaches to study distributed nature of the decision-making by examining both the distributed processing of information as well as the interaction between and heterogeneity of circuits across different brain areas. First, in a value-based perceptual decision-making task in humans, we studied the stages within the decision-making processes in which reward and sensory information exert their influences on the process. We used behavioral analyses and modeling to show that, in contrast to sensory information, reward information does not influence early stages of the decision making. Instead, it exerts its influence in the later stages of the process by biasing response towards the choice option with higher expected reward. Second, we re-examined data from microstimulation of the Frontal Eye Field (FEF) in monkeys during a value-based decision-making task to study the contribution of interactions within oculomotor system to choice behavior. Using different models that can account for interaction between circuits, we found that the brain areas within oculomotor system use reward information to prevent global changes to behavior in presence of perturbations in the FEF. Based on these findings, we developed new biologically plausible decision-making models that can account for the distributed nature of decision making by considering interactions and heterogeneity between circuits across different brain areas. The simulation results highlight the critical roles of heterogeneity and distributed processing in generating robust and stable behavior and enhanced sensitivity to the sensory information