382
V. Vouloutsi et al.
interaction while taking a selfie. In contrast, the introverted robot’s behaviour
results in larger proximate distances (by refraining from approaching others),
exhibits smaller and less exaggerated gestures and acts as a “follower” (passive)
during the task. To ensure that the personality models display sufficiently interpretable social sensorimotor contingencies, we conducted an observational study
of selfie-taking in humans (experiment 1) along with a video pilot evaluating
the robot’s behaviour (experiment 2). The results of both informed the main
experiment (experiment 3).
The Robot. To evaluate the robot’s personality, we used the Pepper robot
(SoftBank Mobile Corp.), which is an autonomous humanoid robot, 1,21m high
that is equipped with speakers, microphones, cameras, and laser and bumper
sensors. To take a selfie, the robot was equipped with a GoPro Hero7 camera
mounted on the robot using a 3D printed “selfie stick” (Fig. 1, left). Experiments took place in a controlled environment, namely the eXperience Induction
Machine (XIM), an immersive space constructed as an experimental platform to
study human behaviour in complex but ecologically valid situations [13].
Robot Control. For the robot to successfully achieve the task of taking a
selfie,a variety of capabilities, such as human recognition, navigation, obstacle and collision detection is necessary. The robot’s control is based on the
Distributed Adaptive Control (DAC) architecture [14], a biologically grounded
cognitive architecture which characterises the link between mind, brain and
behaviour. The robot’s reactive behaviour is based on self-regulation system
of homeostatic and allostatic control [15] and the personality of the robot arises
from the interplay of these subsystems [16]. The control system has been tested
in a variety of HRI scenarios [7] and provides autonomous behaviour in dynamic
environments. Here, we focus on the robot’s personality, as expressed by sensorimotor contingencies, using only the reactive layer of the DAC architecture.
The robot randomly explores an area and looks for humans. Once humans are
detected, the get attention behaviour is triggered followed by communicate intention. If amount of people is not sufficient (in the case the robot got the attention
of only one person), then the gather more people behaviour is triggered. Once
humans stand behind the robot to take the selfie (as detected by the robot’s
laser sensors), the coordinate/take the selfie behaviour is triggered for the taking
of the selfie.
Experiment 1 - Selfie Taking in Humans. To create a baseline in the act
of taking a selfie in group settings, we observed human behaviour in a task
that emulated the final experiment. We asked participants (“selfie-takers”) to
approach a group of people to take a group selfie, using only sensorimotor information (gestures) without any verbalisation. First, participants completed a 10item Big Five survey [11] to determine their personality and then performed
the task of taking a selfie with a group. After the completion of the task, we
V. Vouloutsi et al.
interaction while taking a selfie. In contrast, the introverted robot’s behaviour
results in larger proximate distances (by refraining from approaching others),
exhibits smaller and less exaggerated gestures and acts as a “follower” (passive)
during the task. To ensure that the personality models display sufficiently interpretable social sensorimotor contingencies, we conducted an observational study
of selfie-taking in humans (experiment 1) along with a video pilot evaluating
the robot’s behaviour (experiment 2). The results of both informed the main
experiment (experiment 3).
The Robot. To evaluate the robot’s personality, we used the Pepper robot
(SoftBank Mobile Corp.), which is an autonomous humanoid robot, 1,21m high
that is equipped with speakers, microphones, cameras, and laser and bumper
sensors. To take a selfie, the robot was equipped with a GoPro Hero7 camera
mounted on the robot using a 3D printed “selfie stick” (Fig. 1, left). Experiments took place in a controlled environment, namely the eXperience Induction
Machine (XIM), an immersive space constructed as an experimental platform to
study human behaviour in complex but ecologically valid situations [13].
Robot Control. For the robot to successfully achieve the task of taking a
selfie,a variety of capabilities, such as human recognition, navigation, obstacle and collision detection is necessary. The robot’s control is based on the
Distributed Adaptive Control (DAC) architecture [14], a biologically grounded
cognitive architecture which characterises the link between mind, brain and
behaviour. The robot’s reactive behaviour is based on self-regulation system
of homeostatic and allostatic control [15] and the personality of the robot arises
from the interplay of these subsystems [16]. The control system has been tested
in a variety of HRI scenarios [7] and provides autonomous behaviour in dynamic
environments. Here, we focus on the robot’s personality, as expressed by sensorimotor contingencies, using only the reactive layer of the DAC architecture.
The robot randomly explores an area and looks for humans. Once humans are
detected, the get attention behaviour is triggered followed by communicate intention. If amount of people is not sufficient (in the case the robot got the attention
of only one person), then the gather more people behaviour is triggered. Once
humans stand behind the robot to take the selfie (as detected by the robot’s
laser sensors), the coordinate/take the selfie behaviour is triggered for the taking
of the selfie.
Experiment 1 - Selfie Taking in Humans. To create a baseline in the act
of taking a selfie in group settings, we observed human behaviour in a task
that emulated the final experiment. We asked participants (“selfie-takers”) to
approach a group of people to take a group selfie, using only sensorimotor information (gestures) without any verbalisation. First, participants completed a 10item Big Five survey [11] to determine their personality and then performed
the task of taking a selfie with a group. After the completion of the task, we