Reinforcement learning: Reverse-engineering the behavior of pigeons
Understanding how the brain works is in principles just like understanding any other complex mechanism. Human brains and human beings are hard to work with, so the lower animals provide a better starting place. There are two parts to the problem: How do the pieces function individually (neuroscience)? And, How does the system work as a whole (experimental psychology)? We work on the second problem by devising, and testing experimentally, simple dynamic models for the behavior of pigeons in learning tasks. I will give examples from recent work on reinforcement learning.