Patterns to Measure and Utilize Trust in Multi-agent Systems

Abstract

This paper introduces three patterns to measure and utilize trust values in multi-agent systems (MAS). They allow a software engineer to incorporate mechanisms to gauge the benefit of interactions with other agents into the modeling and design process. In particular, patterns for trust from direct observation, reputation, and Trusted Communities are described. Each pattern contains the elements necessary to incorporate trust into a MAS and the collaboration required between the elements to make use of the measured values. The application of the patterns is demonstrated with an example from the domain of energy management systems.