Rav Suri, Stanford University - A connectionist approach to value based decision making

Abstract: Connectionist approaches involving neural network modeling have been broadly and successfully applied in many areas of cognitive psychology including language, memory, learning and perception. However they have been infrequently applied in 'hot' psychological processes that feature affect and motivation. In this talk I will propose an Interactive Activation and Competition (IAC) neural network for value based decision making. This IAC neural network will provide a single algorithmic framework that applies to goal-directed, habitual, and pavlovian decision making. It will model various valuation related variables such as feature generalization, stimulus benefits, action costs, physiological needs, goal schemas, attentional effects, action readiness, and reciprocity. I will use the network to simulate a variety of empirical data from the decision making and motivation literatures. Overall, I suggest that decisions related to aspects of life that humans most care about may be described via non-homuncular processes in which activation is the only currency.