if the real policy questions are class purefied
they become matters not of what but of how fast
to say rat ex is a stand in for learning
skips the policy question at the heart of modeling and simulating
well just how fast as learners are these agents bub ?
