The problem of understanding trust relations in modern society is extremely interesting nowadays. In our project, we are going to use Multi-agent systems (MAS) to model and reproduce experiments from social sciences to better understand how the relation of trust is shaped in society. In order to perform such experiments it is necessary to accurately represent trust and trustworthiness in MAS. In this paper, we introduce the model and an experimental implementation of benevolence, one of the most important components of trustworthiness.