Learning to move for rewards

This thesis investigates how well people learn to move for monetary rewards using three complementary experiments. In each experiment participants are given monetary rewards that are a function of their movement choices. The thesis is therefore an investigation into human reward learning strategies when learning how to move. Chapter 2 investigates participants' efficiency for learning to move along a single dimension (p. 22). The results show that an optimal decision-theoretic model predicts participants' behavior well. It is shown that people efficiently acquire rewards, combining the computational sub-goals of exploration and exploitation in a manner that closely matches the optimal decision policy of the movement task. Chapter 3 presents an experiment that tests how well participants learn to make two-dimensional movements (p. 40). The experiment addresses how people learn what properties of their movements are responsible for the obtained reward amounts. Results show that participants are able to quickly and efficiently discover the properties of their movements that are most important for acquiring high rewards. Chapter 4 introduces an experiment that investigates how participants evaluate the potential losses and gains associated with their movement decisions (p. 54). All the movement tasks used provide participants with explicit performance feedback with respect to their movements. This introduces a cognitive component to the tasks that is not present in traditional motor control experiments. The experiment presented in Chapter 4 is designed to test whether known cognitive decision biases affect how movements for monetary rewards are made. Results reveal that participants demonstrate the much studied biases of loss aversion and diminishing returns when evaluating the potential outcomes of their movements. There is some evidence in the results that suggests that there is reduced loss aversion bias when making movement decisions when compared to the loss aversion revealed in a traditional behavioral economics decision task. Overall, this thesis provides evidence that people are extremely effective when learning to move for rewards and that optimal decision-theoretic models provide a good descriptive framework of human behavior in the studied tasks.