IEEE Robotics & Automation Magazine - September 2020 - 76

human-human interactions and HRIs [41], [42]. This idea is
supported by functional magnetic resonance imaging results
showing that HRI is underpinned by human-human neuro-
social mechanisms [42]-[44].
Regarding HRI, this process seems to be both bottom-up
and top-down. First, perception (i.e., a bottom-up process)
activates a motor resonance process. By providing a motor
representation of the observed action through which the
observer's motor system enters a state of direct resonance with
that of the agent [45]-[47], motor resonance makes it possible
to directly and immediately understand an action performed
by others. In the same way that we cannot refrain from recog-
nizing a face or a word when we perceive it, we cannot help
but represent in motor terms the actions we perceive. During
interactions with humans, the information that leads to this
activity is integrated automatically [48], [49] and is resistant
to modulation by context [50], [51]. This system could be
used to anticipate other people's actions [52]. In the case of
robots, this automatic need to anticipate is still present [53].
However, the resonance would naturally be weaker and
more context-dependent [54]. The reason could lie in the fact
that it is more difficult to match the representation of a specif-
ic human action with the robot's actual action than it is to
match it with another human's actual action [55]. If the con-
text requires the system to interpret the movements of the
robot in terms of human characteristics, then the system
could act more forcefully to transform a weak bottom-up sig-
nal into a valid representation at the level of motor resonance
by using top-down inferences concerning robots (i.e., social
scripts, anthropomorphic inferences) [53], [56]. Due to these
bottom-up/top-down processes, robots could trigger the
same effect as humans do [3], [5] simply by being present and
could, for example, activate "humanization" processes (i.e.,
attributing to them intentions, emotions, or mental states
comparable to those of humans) [44], [57]-[59]. In other
words, the process might represent the opposite of dehuman-
ization. Furthermore, because in HRI this process is sensitive
to contextual pressures, a particular robot will not be consid-
ered to be the same in different contexts (e.g., social versus
nonsocial context) and will produce different sociocognitive
effects [6].
The Present Study
Objective
In this study, our aim was to evaluate the mediating role of
the universal social perceptual dimensions of warmth and
competence on SFI as well as the role of the intrinsic repre-
sentation of robots activated within a social HRI compared
to a nonsocial HRI. To do so, we used a verbal social-interac-
tion paradigm with a simple humanoid robot without any
facial expression to control for emotional priming effects.
Verbal social HRI is thought to increase the likelihood of
human characteristics being attributed to the robot because
of the activation of automatic social scripts that we use in
human-human interactions [10]. This design should
76

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

SEPTEMBER 2020

therefore maximize anthropomorphic inferences in only one
of two robotic conditions [6]. After the HRI induction, the
participants performed a selective attention task in the passive
presence of the robot.
Hypotheses
First, we hypothesized that, after a social HRI, robots should
trigger the same SFI on selective attention as humans do; that
is to say that robots should prompt an improvement in
response inhibition performances [3], [5], [14], [15]. Because
the inhibition process is similar in both the Stroop task and
the EFT, we hypothesized that, in the presence of the social
robot, participants should produce better performances than
in the presence of the nonsocial robot. The presence of the
social robot should result in the attentional focus being nar-
rowed to the central cue at the expense of the peripheral cues,
thereby reducing the response conflict [6], [9].
Second, we expected to observe an increased level of attri-
bution on the universal dimensions of both warmth and
competence. We also expected an increase in the intrinsic rep-
resentation of robots on essential human dimensions. More spe-
cifically, we expected the robot in the social-robot condition to
be attributed a higher level of warmth due to the social nature of
the experience, as well as to be attributed more competence due
to the more complex HRI in the social condition compared to
the simple description required in the nonsocial condition.
Regarding the intrinsic representation of robots, we expected the
participants to perceive the robots as sharing more humannature traits in the social than in the nonsocial condition.
Third, the SFI effect in the social condition should be
mediated by the inferences made about the robot. Indeed,
according to previous research showing that social interaction
promotes the attribution of human traits [6], [60], we hypoth-
esized that the more highly the participants perceived the
robot on the universal dimensions of warmth and compe-
tence used in social categorization, the more sensitive they
should be to its passive presence, thus resulting in a higher
SFI. The same phenomenon should occur as a function of the
level of perceived shared human traits.
We conducted an exploratory analysis to compare the
hypothetical mediation factor of SFI. The purpose was to iden-
tify the main determinant of this sociocognitive phenomenon.
Method
Participants
The participants were 80 French students at Université Cler-
mont Auvergne [median age, 19.32 years; standard deviation
(SD), 2.05; 68 females; 12 males] with normal or corrected-tonormal vision (27 in the alone condition, 27 in the nonsocialrobot condition, and 26 in the social-robot condition).
Sample size was determined-as recommended by Tabach-
nick and Fidell [61]-on the basis of the desired power (0.80),
alpha level (0.05), number of groups (three in the main analy-
sis), and anticipated effect size based on human-presence
effects (using between-subjects design) in a Stroop paradigm
IEEE Robotics & Automation Magazine - September 2020

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2020
