Heterogeneous Behavior in the Iowa Gambling Task: A Clustering Approach

Iowa Gambling Task is a widely used tool that involves the evaluation of decision-making ability while playing a strategic gambling game. The assessment of the game results has been analyzed using tools such as inferential statistics, knowledge-based systems, and machine-learning classification techniques. Hence, the purpose of this research is to examine a publicly available subset of data from over 300 non-clinical participants who have performed the Iowa Gambling Task (IGT), using K-means to identify performance-related subgroups during IGT. The findings suggest that while the samples were initially reported as homogeneous, there are significant differences observed based on their database origin. Furthermore, the application of the K-means algorithm revealed numerous distinct subgroups of behavior in relation to the credits earned during the test. Thus, the analysis of this feature may suggest previously unconsidered information that could help us to obtain a more detailed view of the behavior in decision making, in both non-clinical and patients with decision making disorders.