Logo image
Back
Beyond dichotomies in reinforcement learning
Journal article   Peer reviewed

Beyond dichotomies in reinforcement learning

Anne G. E. Collins and Jeffrey Cockburn
Nature reviews. Neuroscience, Vol.21(10), pp.576-586
10/01/2020
DOI: 10.1038/s41583-020-0355-6
PMCID: PMC7800310
PMID: 32873936

View Online

Abstract

Reinforcement learning (RL) is a framework of particular importance to psychology, neuroscience and machine learning. Interactions between these fields, as promoted through the common hub of RL, has facilitated paradigm shifts that relate multiple levels of analysis in a singular framework (for example, relating dopamine function to a computationally defined RL signal). Recently, more sophisticated RL algorithms have been proposed to better account for human learning, and in particular its oft-documented reliance on two separable systems: a model-based (MB) system and a model-free (MF) system. However, along with many benefits, this dichotomous lens can distort questions, and may contribute to an unnecessarily narrow perspective on learning and decision-making. Here, we outline some of the consequences that come from overconfidently mapping algorithms, such as MB versus MF RL, with putative cognitive processes. We argue that the field is well positioned to move beyond simplistic dichotomies, and we propose a means of refocusing research questions towards the rich and complex components that comprise learning and decision-making. Reinforcement learning has been suggested to come in two flavours: model-free and model-based. In this Perspective, Collins and Cockburn explain why viewing reinforcement learning through this dichotomous lens is not always accurate or helpful, and suggest paths forward.
 Life Sciences & Biomedicine Neurosciences Neurosciences & Neurology Science & Technology

Details

Metrics

15 Record Views
88 Times Cited - Web of Science
Logo image