Expectations, Learning and the Kalman Filter