Adjusting Autonomy of Agent Systems

This paper describes some mechanisms that could be useful to constrain the autonomy of interface agents. Such mechanisms should increase the acceptance and use of agents systems that several studies have demon-strated to be quite limited. Several users do not rely upon agents capabilities and competence as a result of limited trust in decision aids. Other problems, closer to a human-computer interaction view of the problem, are connected the possible loss of control on the agents actions that humans perceive as a negative feeling. Fur-thermore, users may perceive the system as something that limits their usual behavior constraining them from a free choice. In interactive applications, agents are more effective when associated with the human users to form an integrated mixed-initiative system. The pa-per describes some solutions realized in an agent sys-tem for meeting scheduling that give both agents and their users the ability to effectively perform tasks and to develop a trustful interaction. In particular, level of autonomy of the agent is decided by the user and can vary dynamically according to user needs and current operative context