Modeling the Role of Feelings in the Iowa Gambling Task

• Published in: Decision (Washington, D.C.) D.C.), 2020-01, Vol.7 (1), p.67-89
• Main Authors: Hayes, William M., Wedell, Douglas H.
• Format: Article
• Language: English
• Subjects: Affective Valence; Choice Behavior; Decision Making; Emotions; Female; Human; Learning; Male; Physiological Arousal; Probability; Simulation
• ISSN: 2325-9965; 2325-9973
• DOI: 10.1037/dec0000116

Affective responses to anticipated and experienced outcomes have been demonstrated to predict choice in scenarios where outcomes and their probabilities are explicitly given. The relationship between feelings and choice is less well established for decision problems that require learning from experience. However, converging lines of research suggest that feelings may mediate choices in the Iowa Gambling Task (IGT), a widely studied experience-based task. To investigate this relationship further, 72 participants completed a computerized version of the IGT in which they rated their subjective valence and arousal for each deck after making a choice on each trial. Reinforcement-learning (RL) models were used to examine the role of feelings as both a response to outcomes and an input to choice. Results showed that participants' valence ratings following each choice outcome were best characterized by RL models that included a prospect theory utility function and a decay learning function, with parameters for these functions varying with task performance. On average, models based on previous-trial feelings and pure RL models were equally able to predict the next choice for moderate and high performers, but low performers' previous-trial feelings were unable to predict their next choice. These results were consistent with the somatic marker hypothesis proposal that feelings guide choices for individuals who perform well in the IGT, but that low performers seem to ignore their feelings when making choices.