A novel approach for agents who work against the user's intentions without losing likeability

Abstract

With the support of an agent, a person may be able to behave in a desirable manner. However, the action that a person wants to take does not always match the agent's suggestion. Then, the agent may be antagonized by the user and the agent's likeability may decrease. As a result, users may stop using the agent. In this study, we propose a method of suggestion, in which an agent reports after taking an action without user's permission, in order not to decrease the agent’s likeability even if it is against the user's will. In the experiment, participants were shown an agent that instructs the user to act against their will, an agent that proposes to act on behalf of the user, and the proposed agent. The participants were asked to report the amount of change in their level of acceptance of agents and agents’ likeability. There was no significant difference in the degree of acceptance between the agents. However, the likeability of the agent did not decrease compared to the agent who instructed the agent to act.