Learning to cooperate: decision making in cooperative and competitive social exchanges. (360G-Wellcome-082094_Z_07_Z)

This project aims to investigate how people learn to cooperate in strategic social interactions and, more specifically, the neural processes underlying this capacity. The general hypothesis I will test is that the neural processes underlying learning and decision-making in single-agent situations are extendable to multi-agent contexts, and that they can be used to understand action selection in game theoretic paradigms involving competitive and cooperative options. The project will involve pr oposing and developing theoretical models of decision-making applicable to tasks such as the sequential Prisoners Dilemma and Public Goods games. In particular, these models will be based on the computational framework provided by reinforcement learning (Year 1). The validity of these models will then be tested experimentally, using a combination of statistical model fitting and model-based functional magnetic resonance imaging (Years 2 and 3). The emphasis will be on understanding the role of l earning in these tasks, such as how reputations are formed and the impact of observation in games that involve punishment. More generally, this should lead to a theoretical understanding of how humans actually solve the game-theoretic problems (algorithmically) and how this is implemented in the brain.