A Direct Reputation Model for VO Formation

We show that reputation is a basic ingredient in the Virtual Organisation (VO) formation process. Agents can use their experiences gained in direct past interactions to model other’s reputation and deciding on either join a VO or determining who is the most suitable set of partners. Reputation values are computed using a reinforcement learning algorithm, so agents can learn and adapt their reputation models of their partners according to their recent behaviour. Our approach is especially powerful if the agent participates in a VO in which the members can change their behaviour to exploit their partners. The reputation model presented in this paper deals with the questions of deception and fraud that have been ignored in current models of VO formation.