Computerized agents versus human agents in finding core coalition in glove games. (September 2022)

• Record Type: Journal Article
• Authors: Grigoryan, Gayane; Etemadidavan, Sheida; Collins, Andrew J
• Is Part Of: Simulation. Volume 98:Number 9(2022)
• Page Start: 807
• Page End: 821
• Publication Date: 2022-09
• Subjects: Agent-based simulation -- agent-based model -- cooperative game theory -- hybrid model -- human subject experiment; Computer simulation -- Periodicals; 003.3
• DOI: 10.1177/00375497221093652
• Languages: English
• ISSNs: 0037-5497

Abstract

One of the challenges for agent-based modeling is being able to incorporate human behavior. Human behavior is a multifaceted phenomenon, with strategic coalition formation being one form. A hybrid agent-based modeling approach, called ABMSCORE, has been derived to emulate strategic group formation. In this paper, we describe a simulation experiment to compare the ABMSCORE with actual human behavior. The comparison criterion is the respective rates of finding an ideal coalition. In our experimental design, we go to great lengths to ensure the similarity of the scenarios in the two trial types: trials with computerized agents only and trials involving human participants when one of the computerized agents is replaced by an actual human. We did this to limit the number of possible extraneous variables introduced into the experimental system. The scenario considered is the glove game, a standard cooperative game that has been previously used in human experiments. Our results indicate that the ABMSCORE model produces similar rates of finding the ideal coalition as the human players; however, there are some limitations. This research provides evidence for using the ABMSCORE modeling approach to model human strategic coalition formation in agent-based models.