Towards 3D Virtual Electronic Institutions

In this paper we discuss the relationship between autonomous agents and their principals within an Electronic Institution. We propose the use of 3D virtual environments to allow humans to establish trust on the behaviour of their agents and override the agent’s decision process whenever they consider it necessary. Each avatar in the 3D virtual environment can be manipulated both by an autonomous agent and its principal. This approach provides the possibility of continuous co-learning. On the one hand, the autonomous agent is always active, and when the human is driving the avatar the autonomous part learns from the observation of the behaviour of the human. On the other hand, autonomous agents can assist humans with learning the structure of the Electronic Institution. Moreover, they can provide advice to a principle on the basis of the information the agent gathered from external sources or from the observation of other participants’ behaviour. 3D virtual environments in this case offer not only an immersive experience but also add new degrees of freedom to user modelling and co-learning.