Computational and Predictive Trust Game Model for Switching Process in fMRI

Game theory paradigms allow analyzing experimentally through simulating the cognitive processes required for social cognition, such as perception, learning and decision-making, and actions. In this paper, the trust game paradigm, which is an extensively used task, is studied for measuring the trust. First, the trust game is designed as an fMRI task and then the trust data is acquired by running the task with n=80 subjects. Additionally, a computational mathematical model is proposed to model the subject-specific decision-making behavior by learning the previous game trials and then computing the difference between the actual and predicted gain, i.e., the prediction error (PE). The model is evaluated on two groups: depressed participants and healthy controls. The performance of the model is evaluated with the root mean square errors (RMSE) metric. The evaluation results show that even if the decision behavior of the participants exhibits different patterns, the proposed model can converge the actual gain fairly well with a low RMSE value. © 2022 IEEE.