Decision-making in brains and robots - the case for an interdisciplinary approach

Abstract

Reinforcement Learning describes a general method for trial-and-error learning, and it has emerged as a dominant framework both for optimal control in autonomous robots, and understanding decision-making in the brain. Despite their common roots, however, these two fields have evolved largely independently. In this perspective, we consider how each now face problems that could potentially be addressed by insights from the other, and argue that an interdisciplinary approach could greatly accelerate progress in both.