Emotion-based reward prediction, valuation, and learning

Abstract

This thesis aims to prove how the emotion functions in the human decision-making in various conditions and to provide general framework to guide future neuroscientific and neuroimaging experiments by making use of theoretical neuroscience and functional Magnetic Resonance Imaging (fMRI). Decision-making is regarded as a process of maximizing the subjective desirability of the outcomes from chosen actions. Despite this simple nature of optimization, in reality, decision-making can come in varieties of mixtures and/or sole combination of trial-and-error, trade-offs, and short cuts. The work centers on the notion of how human forms the neural representation of the world and come to choose appropriate actions in the face of rewards and punishments. Here, numbers of important topics are discussed that involve the notion of reward-related learning and reward prediction in a number of dimensions. Here I propose emotion based reward prediction, valuation, and learning which replaces the reinforcement signal with an emotional cue, which can be interpreted as a cognitive assessment of the present state in light of goals and attentions. Using an application of Temporal Difference (TD) model in neuroimaging data from an event related fMRI, it was possible to draw conclusions from neural correlates with response profile that are consistent with a specific computational process. The computational model-based regression analysis, reveals that an emotion based prediction-error in emotion-based (disappointment and elation) reward prediction and reward learning are under separate neural systems. The results broadly support the notion that emotion-based prediction of reward and reward-related learning are closely related and that different neural systems are involved in emotion-based reward processing and learning.