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

Abstract

Read online

Computational models of learning have proved largely successful in characterising potentialmechanisms which allow humans to make decisions in uncertain and volatile contexts. We reporthere findings that extend existing knowledge and show that a modified reinforcement learningmodel which differentiates between prior reward and punishment can provide the best fit tohuman behaviour in decision making under uncertainty. More specifically, we examined thefit of our modified reinforcement learning model to human behavioural data in a probabilistictwo-alternative decision making task with rule reversals. Our results demonstrate that this modelpredicted human behaviour better than a series of other models based on reinforcement learningor Bayesian reasoning. Unlike the Bayesian models, our modified reinforcement learning modeldoes not include any representation of rule switches. When our task is considered purely as amachine learning task, to gain as many rewards as possible without trying to describe humanbehaviour, the performance of modified reinforcement learning and Bayesian methods is similar.Others have used various computational models to describe human behaviour in similar tasks,however, we are not aware of any who have compared Bayesian reasoning with reinforcementlearning modified to differentiate rewards and punishments.

Keywords

• Decision Making
• reinforcement learning
• uncertainty
• volatility
• Bayesian learning
• Reward and punishment