In systems in which many agents interact with each other and a high fluctuation of interaction partners can be expected, direct trust based on personal experiences is often too limited. Instead, a reputation system that aggregates trust values created by the agents or collects experiences and transforms them with a trust metric can be used to allow agents to gauge previously unknown partners. Such a system thus helps to spread information about other agents in the system and can be a basis for trustworthy interaction even in cases in which individual pairings are unique occurences.

It is recommended to use existing reputation system architectures whenever possible. Such architectures can be broadly distinguished into centralised and decentralised approaches. In the former case, the reputation system has a central manager that gathers and aggregates data and is the reference for all agents in the system. In the latter case, a number of such agents exist or reputation is calculated in a fully decentralised fashion as a form of gossipping.