10.1101/295022

Reference-point centering and range-adaptation enhance human reinforcement learning at the cost of irrational preferences