Development of robot which recognizes user's false beliefs using view estimation

It is desired that a human-symbiotic robot behaves in consideration for a difference in recognition of situations between a user and the robot itself by putting itself in the user's position. This ability will be especially strongly necessary when the user has a false belief, that is, when he or she recognizes a situation wrongly. In such a case, it is desired that the robot understands the user's false belief and behaves in consideration for his or her mistake and its consequence. We anticipate that such a “considerate” robot will be friendlier and more reliable for users. In this paper, we pro-pose a robot that can recognize a user's false belief in scenes of the maintenance of joint attention and carry out experimental subjective evaluations of impressions that the proposed robot gives to humans during a human-robot interaction to statistically test our anticipation. Participants are asked to observe the interaction and to subjectively evaluate their impressions of the proposed robot's behaviors. The experimental results show that the robot that can recognize a user's false belief can give “familiar” and “intelligent” impressions to humans. This fact is expected to be one of fundamental recommendations for designing much friendlier interactions with robots and other intelligent systems.