Confidence-guided updating of choice bias during perceptual decisions is a widespread behavioral phenomenon

We report a novel form of reinforcement learning during perceptual decisions: past rewards bias future perceptual choices specifically when the previous stimulus was difficult to judge , and the confidence in obtaining the reward was low. We identified this phenomenon in six datasets from four laboratories, across mice, rats and humans, and sensory modalities from olfaction and audition to vision. We show that reinforcement learning models incorporating decision confidence into th eir teaching signal explain this choice updating. Thus, reinforcement learning mechanisms are continually engaged to produce systematic adjustments of choices even in well-learned perceptual decisions in order to optimize behavior in an uncertain world.
Source: eLife - Category: Biomedical Science Tags: Neuroscience Source Type: research