Toward a Probabilistic Model of Trust in Agent Societies

Abstract

The literature about trust in societies of agents collects a huge number of works that analyse almost any facets of this concept from nearly every point of view. Nevertheless, an accepted and stable formal model of trust in agent societies is lacking. In this paper, we address this remarkable flaw of the current research by introducing a probabilistic model of trust capable of capturing two-party interactions, either direct of mediated by a Guarantor. Some interesting properties of this model are demonstrated and the final result of this work is an estimation (upper-bound) of the improvements that we expect from the inclusion of a Guarantor in a two-party interaction. In details, after an introductory section, Section 2 provides the foundations of our model and quantifies the increment of the utility that agents perceive because of the mediation of a Guarantor. Then, Section 3 deals with the decision-making strategies of rational agents and it shows a worst-case specialization of our model that justifies why agents are more likely choosing Guarantor-mediated interactions. Section 4 describes the overall results of this work in terms of bounds and evaluation of performances of the effects of mediation in interactions. Finally, Section 5 summarizes the main outcome of this work and outline some future lines of development.