Temporal difference learning is favored for rewards, but not punishments, in simulations and human behavior.

Adam Morris, Fiery Cushman

Browse the full CogSci paper archive.