AUTHOR=Duffin Elaine , Bland Amy R., Schaefer Alexandre , De Kamps Marc 

TITLE=Differential effects of reward and punishment in decision making under uncertainty: a computational study

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 8 - 2014

YEAR=2014

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2014.00030

DOI=10.3389/fnins.2014.00030

ISSN=1662-453X

ABSTRACT=Computational models of learning have proved largely successful in characterising potential<br/>mechanisms which allow humans to make decisions in uncertain and volatile contexts. We report<br/>here findings that extend existing knowledge and show that a modified reinforcement learning<br/>model which differentiates between prior reward and punishment can provide the best fit to<br/>human behaviour in decision making under uncertainty. More specifically, we examined the<br/>fit of our modified reinforcement learning model to human behavioural data in a probabilistic<br/>two-alternative decision making task with rule reversals. Our results demonstrate that this model<br/>predicted human behaviour better than a series of other models based on reinforcement learning<br/>or Bayesian reasoning. Unlike the Bayesian models, our modified reinforcement learning model<br/>does not include any representation of rule switches. When our task is considered purely as a<br/>machine learning task, to gain as many rewards as possible without trying to describe human<br/>behaviour, the performance of modified reinforcement learning and Bayesian methods is similar.<br/>Others have used various computational models to describe human behaviour in similar tasks,<br/>however, we are not aware of any who have compared Bayesian reasoning with reinforcement<br/>learning modified to differentiate rewards and punishments.<br/>