Reinforcement Learning and Attention

How does the brain decide what information to prioritize in complex, high-dimensional environments? Using eye-tracking and multivariate decoding of fMRI data, I measured participants’ attention as they performed a trial-and-error learning task with multidimensional stimuli (Leong, Radulescu et al., Neuron, 2017). I found behavioral and neural evidence that attention constrains learning and decision-making to a subset of environmental dimensions predicting reward, while learned values bias attentional selection towards dimensions relevant to current task demands. The dynamic modulation of attention was associated with increased activity in a frontoparietal network, which also exhibited enhanced connectivity with the ventromedial prefrontal cortex when participants sustained attention to the same dimension over multiple trials. This work provides new understanding on how attention and reward processing influence each other to facilitate learning and decision-making.