Adapting to Users
Previous research in HRI has demonstrated the benefit of a
robot adapting to its users. In prior work in which a robot
provided cooking help to participants in a kitchen, researchers
found that adaptive dialogue—in which the robot adapted
the content of its speech depending on the expertise of the
user—improved information exchange and social relations,
especially when users were under time pressure [47]. Previous
research with children investigated the use of adaptive
empathic behaviors (e.g., encouraging comments and offering
help) for a chess-playing robot [24]. Children responded
positively to the robot when the empathic behaviors were employed
adaptively, rather than randomly. Previous work has
also investigated the positive effect in eliciting user compliance
of matching a “playful” or a “serious” robot (conveyed
through the robot’s speech) to a playful or serious task [14].
Adapting to user personality has been more widely studied
in HCI. For example, previous research shows that computer
interfaces can be manipulated to exhibit an extroverted or introverted
personality through the use of language, pictures,
and sounds, and that introverted users will perform tasks faster
when using introverted software [36]. This result follows from the theory of similarity-attraction which predicts that a person
will be attracted more to others who match their personalities
than to those who mismatch [22]. In the same way, matching
the personality of a synthesized voice, expressed through pitch,
prosody, and so on, to user personality positively affects users’
feelings of social presence, especially in extroverts [22]. Emotion
matching is also important in computer interfaces. In a
study of emotional speech generation for car interfaces, it was
found that when the emotion expressed by the car voice (energetic
or subdued) matches the emotional state of the driver
(happy or upset), drivers have fewer accidents, attend more
to the road, and speak more to the car [31]. Our research in
socially assistive robots parallels these efforts by following
similarity-attraction theory to match the gaze behaviors of the
robot to the personality of the user.