Actually, backprop may be the answer! See: https://www.youtube.com/watch?v=dZwB5Mj-PPM Though our brains may not explicitly run on back-propagation, their reinforcement is asymptotically equivalent. I was happily surprised by this finding — Bengio also explains it in some detail here: https://www.youtube.com/watch?v=FhRW77rZUS8
