Gradually Intrusive Argumentative Agents for Diagnosis

A new model of agent for acting in processes of diagnosis where more constraints are placed on the performance in real world domains is presented. Preferences on actions and policies are used to induce a preferred course of action by the diagnosis agent. Past experience, gathered from controlled trials, enables the experience agent to offer counterarguments for improving the diagnosis process. The outcome of actions are introduced in the scenario and agent models to allow argumentation on actions, not just beliefs. We argue that the approach is general and flexible, allowing the natural integration of new knowledgeable agents, while, at the same time, offering human agents transparency of the decision process.