What Should an Agent Know Not to Fail in Persuasion?

This paper presents strategies and conditions for non-failing persuasion using a dialogue model using argumentation. A novel concept of the predicted knowledge of the other agent participating in the dialogue is introduced. In the dialogue model, an agent’s knowledge is updated as the dialogue proceeds; an argumentation framework is constructed from the current knowledge; and only the content of an acceptable argument can be offered as the next move. In this paper, a modified dialogue model is proposed in which the next move is determined using predicted knowledge and a strategy that navigates a non-failing persuasive argumentation is presented. Conditions under which persuasion never fails using this strategy when the prediction is equivalent to the actual knowledge of an opponent are described. Moreover, what the predicted knowledge should contain for non-failing persuasion are discussed. The introduction of predicted knowledge improves the formulation of real dialogue.