Only tangentially related, but there is some cool work on using reinforcement learning to learn an optimal tax rate. See: “AI Economist”.