Modeling and Analysis of Affective Influences on Human Experience, Prediction, Decision Making, and Behavior

Subjective and affective elements are well-known to influence human decision making. This dissertation presents a theoretical and empirical framework on how human decision makers’ subjective experience and affective prediction influence their choice behavior under uncertainty, frames and emotions. The framework extends and integrates existing theories of prospect theory (PT) and reinforcement learning (RL), drawing on a growing literature illuminating the role of affect in decision making and the neural underpinnings of human decision behavior. The proposed Affective-Cognitive (AC) model extends Prospect Theory (PT)-based subjective value functions to model human experienced-utility and predictedutility functions. The AC model assumes that the shapes (or parameters) of these subjective value functions dynamically vary with the decision maker’s affective states in sequential decision making. Human decision-making experiments were conducted to infer how people adjust the parameters (i.e., shape and reference point) of their experienced-utility and predicted-utility functions in sequential decision-making situations involving incidental affective states (e.g., anger, fear, economic fear) and task-related confidence. I constructed a new model combining measures to evaluate risk preferences: behavioral choices, selfreported experience, self-reported predicted utility, self-reported confidence. The analysis results show how domain uncertainty, framing, and emotional state of decision makers influence their subjective experience and discriminability, affective prediction, optimal decisions and exploratory regulation. I found empirically that there were significant interaction effects of framing and emotion on risk preferences: negative emotions made people more risk-averse in the face of gains. When it comes to losses, anger made people more risk-averse and fear more risk-seeking. I also characterized how gender and emotion influence confidence and exploratory choice behavior. The theoretical analysis nicely supports empirical findings from human experiments. The new model provides a theory that better explains and simulates human behavior under uncertainty, frames and emotions. Thesis Supervisor: Rosalind W. Picard Title: Professor of Media Arts and Sciences, Program in Media Arts and Sciences, MIT