An Interaction Approach for Agents Working against the User’s Intentions without Losing Likeability

With the support of an agent, a person can take desirable actions (e.g., eco-friendly or healthy actions). However, an agent’s suggestion of a desirable action does not always match the action that a person wants to take. When such a conflict arises, the user may object to the suggestion and the agent’s likeability may decrease, resulting in the user stopping using the agent. In this study, we propose a novel interaction approach, in which an agent takes a desirable action without the user’s permission and then reports that it took the action (action-before-report interaction). We investigated the acceptability and changes in likeability for agents with three interaction approaches (i.e., action-before-report, action-after-report, and instruction-only). In our experiments, participants (users) watched a video in which an agent takes an action employing one of the three interaction approaches. Then, the users answered their degree of acceptability or likeability of the agent. Results showed that there was no significant difference in the acceptability of agents. However, the likeability of an agent adopting the proposed interaction approach decreased less compared to that taking the instruction-only approach. The results suggest that an agent is less likely to lose likeability when employing the proposed approach.