ArticlePDF Available

Abstract and Figures

The uncanny valley hypothesis suggests that a high (but not perfect) human likeness of robots is associated with feelings of eeriness. We distinguished between experience and agency as psychological representations of human likeness. In four online experiments, vignettes about a new generation of robots were presented. The results indicate that a robot’s capacity to feel (experience) elicits stronger feelings of eeriness than a robot’s capacity to plan ahead and to exert self-control (agency, Experiment 1A), which elicits more eeriness than a robot without mind (robot as tool, Experiments 1A and 1B). This effect was attenuated when the robot was introduced to operate in a nursing environment (Experiment 2). A robot’s ascribed gender did not influence the difference between the eeriness of robots introduced as experiencers, agents, or tools (Experiment 3). Additional analyses yielded some evidence for a non-linear (quadratic) effect of participants’ age on the robot mind effects.
Content may be subject to copyright.
MIND AND MACHINE 1
The Uncanny of Mind in a Machine: Humanoid Robots as Tools, Agents, and
Experiencers
*Markus Appel 1,2, David Izydorczyk 3, Silvana Weber 1,
Martina Mara 4, & Tanja Lischetzke 2
1 University of Würzburg, Germany
2 University of Koblenz-Landau, Germany
3 University of Mannheim, Germany
4 Johannes Kepler University of Linz, Austria
Accepted for publication in the journal Computers in Human Behavior
Acknowledgements
Experiment 1A was presented in form of a poster at the 11th Annual Conference for basic and
applied human-robot interaction research (HRI 2016) in Christchurch, New Zealand.
This work was supported by the German Ministry of Education and Research (BMBF), Grand
No. PLI1663, ITA 2017, awarded to the first author.
The authors thank Jan-Philipp Stein for his comments and suggestions.
Correspondence concerning this article should be addressed to Markus Appel, Human-Computer-
Media Institute, University of Würzburg, Germany, Oswald-Külpe-Weg 82, 97074 Würzburg,
Germany, Email: markus.appel@uni-wuerzburg.de
MIND AND MACHINE 2
Abstract
The uncanny valley hypothesis suggests that a high (but not perfect) human likeness of
robots is associated with feelings of eeriness. We distinguished between experience and agency
as psychological representations of human likeness. In four online experiments, vignettes about a
new generation of robots were presented. The results indicate that a robot’s capacity to feel
(experience) elicits stronger feelings of eeriness than a robot’s capacity to plan ahead and to exert
self-control (agency, Experiment 1A), which elicits more eeriness than a robot without mind
(robot as tool, Experiments 1A and 1B). This effect was attenuated when the robot was
introduced to operate in a nursing environment (Experiment 2). A robot’s ascribed gender did not
influence the difference between the eeriness of robots introduced as experiencers, agents, or
tools (Experiment 3). Additional analyses yielded some evidence for a non-linear (quadratic)
effect of participants’ age on the robot mind effects.
Keywords— uncanny valley; mind perception; experience; agency; service robots
MIND AND MACHINE 3
The Uncanny of Mind in a Machine: Humanoid Robots as Tools, Agents, and Experiencers
Humanlike robots can be a source of bewilderment and eeriness (Freud, 1919/2003;
Jentsch, 1906/1997; Mori, 1970 / Mori, MacDorman, & Kageki, 2012). Recent work suggests
that the eeriness of humanlike robots does not only depend on their visual appearance, but can
also stem from the perception of feelings and experience in a machine (Gray & Wegner, 2012;
Wegner & Gray, 2016). We follow the distinction between the ability to feel (experience) and the
ability for thought and cognition (agency) of mind perception (Gray, Gray, & Wegner, 2007) and
investigate the effects of both components of mind perception on participants’ feelings of
eeriness in response to a service robot. Importantly, we examine the role of a robot’s experience
and agency under varying boundary conditions, that is, in a nursing context (as compared to an
unspecified field of application), for female or male gendered robots, and for the spectrum of
users’ age.
1.1 Humanoid Robots and the Uncanny Valley
The number of robots built each year is on the rise (International Federation of Robotics,
2018). Most of the robots are meant to work in automotive manufacturing or chemical industries,
like they have done since the 1950s. In recent years, however, the use of robots in non-industrial
fields has increased rapidly (Graetz & Michaels, 2018; Grewal, Motyka, & Levy, 2018). Not only
are robots built to clean the carpet or mow the garden lawn, robots are envisaged to play a key
role in future sex work, military, tourism, education, and retail. The sector that has arguably
attracted most attention by the industry as well as related research is nursing and healthcare for
older people (cf. Locsin & Ito, 2018). Initial evidence points out that social robots might bring
benefits to older people, such as a reduction in loneliness or increasing social interactions
(Kachouie, Sedighadeli, Khosla, & Chu, 2014; Shibata & Wada, 2011).
MIND AND MACHINE 4
A recent multi-wave analysis of EU survey data suggests that, along with the proliferation
of robots, attitudes towards robots are getting more negative (Gnambs & Appel, 2019). Much of
the literature on the acceptance of robots, humanoid robots in particular, has been guided by the
uncanny valley hypothesis (Mori, 1970; for reviews see Kätsyri, Förger, Mäkäräinen, & Takala,
2015; Wang, Lilienfeld, & Rochat, 2015; Złotowski, Proudfoot, Yogeeswaran, & Bartneck,
2015). According to Mori’s conception, increasing the human likeness of robots (and other
anthropomorphic technologies) elicits increasing acceptance and likeability at low to moderate
levels of human likeness. However, with a further increase, as soon as a very high level of nearly
realistic human likeness is obtained, this relationship is reversed. At this point, valence drops
substantially, and the almost perfectly humanlike robot elicits a negative and irritating feeling of
eeriness among its human observers. When a robot’s design approaches perfect human likeness
even further, user responses turn positive again (cf. Kätsyri et al., 2015; Wang et al., 2015).
Whereas human likeness is often conceptualized as a feature of the visual appearance of the robot
alone, research suggests that perceptions of human-likeness are subject to all visual and
functional features of the robot, user variables, and variables of the situation in which the human-
robot-interaction takes place (e.g., Broadbent, 2017; Lischetzke, Izydorczyk, Hüller, & Appel,
2017; MacDorman, & Entezari, 2015; Mara & Appel, 2015; Piwek, McKay, & Pollick, 2014;
Rosenthal-von der Pütten & Krämer, 2015; Rosenthal-von der Pütten, & Weiss, 2015).
More broadly, these features determine users’ perceived capabilities of the robot which
translate to cognitive, affective, and behavioral responses (e.g., Hoffmann, Bock, & Rosenthal-
von der Pütten, 2018; Rosenthal-von der Pütten & Krämer, 2014; 2015). One aspect of ascribed
robot attributes that was considered to be particularly influential from early on is the mind
ascribed to a robot (e.g., Hegel, Krach, Kircher, Wrede, Sagerer, 2008).
1.2 Mind Perception
MIND AND MACHINE 5
Several lines of theory suggest that the uncanny valley is based on users’ perceptions of
human mind in a machine (Hegel et al., 2008; Gray & Wegner, 2012; Stein & Ohler, 2017;
Wegner & Gray, 2016). More specifically, two dimensions of mind perception have been
distinguished: experience and agency (Gray et al., 2007; Waytz, Gray, Epley, & Wegner, 2010).
The general idea that human likeness and its influence on eeriness might involve more than one
dimension can be found in other work (e.g., realism and prototypicality; Burleigh, Schoenherr, &
Lacroix, 2013). From a mind perception perspective, however, identifying characteristics that are
attributed to humans only (and not to machines or animals) is key to explain the uncanny valley.
Gray and colleagues presented descriptions of characters (mainly humans of different ages and
mental states, animals) and participants evaluated these characters along a list of 24 attributes.
They identified the two factors on the basis of principal components factor analyses of these
attributes. Agency involves characteristics of self-control, morality, memory, emotion
recognition, planning, communication, and thought, whereas experience is characterized by
hunger, fear, pain, pleasure, rage, desire, personality, consciousness, pride, embarrassment, and
joy (order of the terms listed represents factor loadings in Gray et al., 2007).
Connecting both mind perception dimensions to the uncanny valley phenomenon and
resulting experiences of eeriness, Gray and Wegner (2012) presented participants with either a
video of a humanoid robot that focused on its electrical components and wirings or a video of the
same robot with a focus on its humanoid face. In the latter condition, participants ascribed more
experience to the robot and reported more eeriness, whereas ascribed agency did not differ
between conditions. Experience (but not agency) predicted eeriness and mediated the effect of the
video conditions on eeriness. In a second study, participants received descriptions of a
“supercomputer” which was simply more powerful (control condition), was able to
“independently execute actions” with “self-control and the capacity to plan ahead” (agency
MIND AND MACHINE 6
condition), or was able to feel some form of “hunger, fear and other emotions” (experience
condition). Eeriness was elevated in the experience condition as compared to both the control
condition and the agency condition (eeriness in the latter two conditions was about at par). These
results suggest that experience is the dimension of human likeness that is responsible for the
eeriness elicited by humanoid robots. Agency, on the other hand, appears to be unrelated to the
negative responses to this new technology.
In a study set in a retirement village, the differential influence of experience and agency
was tested in an applied setting (Stafford, MacDonald, Jayawardena, Wegner, & Broadbent,
2014). A healthcare robot was introduced to the retirees, and the retirees’ behavior and
perceptions in response to the robot were documented. Participants ascribed higher capacity for
agency than capacity for experience to the robot, corroborating the distinction between both mind
perception dimensions in an applied context with a key target user group. In contrast to what
could be expected from the earlier results (Gray & Wegner, 2012), however, ascribed agency was
negatively related to the actual use of the robot, whereas ascribed experience was unrelated to
using the robot. Thus, in this particular setting, negative responses to the robot were associated
with perceived agency, not with perceived experience (Stafford et al., 2014). However, the
evidence regarding robot mind and eeriness gained from this study is indirect, given that users’
eeriness was not examined. Research in a related field further supports the notion that perceived
experience might not be a driving force underlying the feeling of eeriness: embodied
conversational agents that expressed emotions were preferred over a non-experiencing, neutral
counterpart (Creed, Beale, & Cowan, 2014).
A recent study on autonomous agents in a VR environment, however, supports the notion
that mind perception elicits eeriness (Stein & Ohler, 2017). The question regarding which aspect
of mind is linked to eeriness was not answered in this study, however, as the results could be
MIND AND MACHINE 7
driven by perceptions of experience, of agency, or perceptions of both. To complicate things
further, others did not identify a relationship between both mind dimensions ascribed to a robot
and the evaluation of the robot in terms of damage to humans and their identity (Ferrari,
Paladino, & Jetten, 2016).
The differences between the available empirical results highlight the need for further
empirical investigations. Thus, our first aim was to investigate the influence of experience and
agency ascribed to a robot on users’ eeriness. We predicted that a robot with experience as well
as a robot with agency would elicit eeriness among (potential) users, with experience leading to
most eeriness. Importantly, we further assumed that the influence of a robot’s mind on user
responses is crucially affected by moderating variables that determine whether a robot’s
experience, agency, or both elicits eeriness. The variables we focused on are introduced in the
following.
1.3 The Role of Context, Robot Gender, and Users’ Age
1.3.1 Nursing as a relevant context. A robot’s field of application could moderate the
influence of the mind dimensions on eeriness. Currently, elderly care and nursing are fields in
which service robots are increasingly deployed (Archibald & Barnard, 2018; Locsin & Ito, 2018).
In hospitals and retirement homes, however, emotional sensitivity and emotional intelligence are
key affordances of successful work (e.g., Cadman & Brewer, 2001). Thus, in this important field
of application, the experience of affect by robots could be perceived as rather appropriate for
dealing with the tasks at hand (Stafford et al., 2014). We therefore assumed that the differences
between the perceived eeriness of a robot with experience, a robot with agency, and a robot
without mind would decrease in a nursing context.
1.3.2 Robot gender. Gender is one of the most salient attributes of humans and the
development of humanoid robots has put questions about the impact of a robot’s gender
MIND AND MACHINE 8
representation on the agenda (e.g., Carpenter et al., 2009; Eyssel & Hegel, 2012; Reich-Stiebert
& Eyssel, 2015; Tay, Jung, & Park, 2014). Agency and experience are closely associated with
expectations regarding both genders in person perception, with women expected to show
experience, and men expected to show agency (cf. agency vs. communion, Bakan, 1966;
competence vs. warmth, Fiske, Cuddy, & Glick, 2007; dominance vs. nurturance, Wiggins &
Broughton, 1991; see Abele & Wojciszke, 2014, for an overview). People tend to respond to
computers and robots as if they responded to other humans (computers as social actors; Nass,
Steuer, & Tauber, 1994). In a seminal study on gender and robots, Eyssel and Hegel (2012)
manipulated the perceived gender of a robot and demonstrated that a masculine robot was
perceived as having more agency and less communion than a feminine robot. Prior research has
shown that a match between a robot’s gender and its task (and a match between its personality
and task) yielded more positive attitudes and favorable perceptions than a mismatch (Tay et al.,
2014). Many explanations of the uncanny valley involve a conflict between existing schemas and
expectations (cf. Kätsyri et al., 2015). Given that the male or female gender of a robot elicits
contrasting expectations regarding agency and communion or experience (Eyssel & Hegel, 2012),
the influence of agency and experience ascribed to the robot on eeriness should vary with the
robot’s gender. If the robot’s mind fit its gender (experience and female, agency and male), we
would expect lower eeriness than when the robot’s mind did not fit its gender.
1.3.3 Age. With respect to the user, age could be a variable to moderate the impact of
robot minds. Most findings of the uncanny valley literature are based on samples of adolescents
and young adults (e.g., Bartneck, Kanda, Ishiguro, & Hagita, 2009; Burleigh et al., 2013;
Cheetham, Suter, & Jäncke, 2011; Gray & Wegner, 2012; Lischetzke et al., 2017). Some studies
suggest that the reaction towards as well as the experience with robots might change with age—
yet, the available evidence is somewhat inconclusive. For example, Liang and Lee (2017) found
MIND AND MACHINE 9
that age was a positive predictor of the fear of autonomous robots and artificial intelligence.
However, other studies found contradicting evidence. Older age groups expressed less general
anxiety towards humanoid robots in some studies (e.g., Nomura, Syrdal, & Dautenhahn, 2015),
while in another study, age was unrelated to the quality of experience with a healthcare robot
(Broadbent et al., 2010). In addition, besides the well documented age-related decline of, for
example, processing speed or working memory, findings from the research group around
Carstensen suggest that the experience of complex emotions increases with age (Carstensen et al.,
2011) and that older adults show superior cognitive performance for emotional relative to non-
emotional information (Charles, Mather, & Carstensen, 2003). In sum, age has played a
substantial role in theory and research on the uncanny valley from early on (e.g., Ishiguro, 2007),
but its impact on responses to robots with experience or agency is far from clear. Based on the
growing importance of emotional stimuli over the course of a lifetime and higher wariness
regarding new technologies (Gilly & Zeithaml, 1985; Mostaghel, 2016), older adults could show
a particularly pronounced negative reaction (high eeriness) towards a robot with experience (as
compared to a robot with agency or a control condition). In contrast, findings around the
positivity effect suggest that as people get older, they experience fewer negative emotions and
have more difficulty identifying others’ negative emotions than younger adults (Carstensen &
Mikels, 2005; Mather et al., 2004; Mather & Carstensen, 2005; Nomura & Nakao, 2010; Wong,
Cronin-Golomb, & Neargarder, 2005). Following the latter reasoning, one might expect a
different shape of the moderator effect of age, namely that older people experience less eeriness
in the experiencer condition than in the other conditions. A similar pattern of results could
emerge for participants in their late teens and early twenties, due to a particularly strong openness
for new technologies (Gnambs & Appel, 2019).
1.4 Study Overview
MIND AND MACHINE 10
Humanlike but not perfectly human robots are considered to fall into the uncanny valley
(Mori, 1970), eliciting eeriness among human observers. The aim of our series of experiments
was to examine the influence of experience and agency as theory-guided representations of
human likeness (Gray et al., 2007; Gray & Wegner, 2012). Complementing prior work in which
appearance or behavior of robots were manipulated (e.g., Hegel et al., 2008; Rosenthal-von der
Pütten & Krämer, 2015), our aim was to maximize the internal validity of our agency and
experience manipulation. To this end, we built our work on the attributes that defined agency and
experience in previous work (Gray et al., 2007).
Using vignettes of future humanoid robots, we first tested the prediction that a robot with
experience is eerier than a robot with agency, and we compared both robot mind conditions to a
control condition in which the robot was merely a tool (Experiments 1A and 1B). In addition to
eeriness, we examined the perceived femininity and masculinity of the robot (cf. Eyssel & Hegel,
2012), as well as behavioral intentions to interact with the robot. In describing the robots, we
tried to follow closely the initial operational definitions of the mind perception dimensions (Gray
et al., 2007). Importantly, the influence of agency and experience was examined under varying
boundary conditions, regarding the field of application, the robot’s gender, and users’ age. In
Experiment 2, we examined the role of the robot’s field of application on the effect of robot
minds. We contrasted a nursing context to an unspecified context. Experiment 3 focused on a
robot’s gender. We ascribed a male or a female name to the robot and investigated whether the
influence of the robot’s mind on eeriness differed between female and male robots. In a final set
of analyses, the moderating influence of users’ age was examined. Given the possibility of
nonlinear effects of age, we examined linear as well as non-linear (quadratic) effects of users’ age
on the mind perception–eeriness link.
2. Experiments 1A and 1B
MIND AND MACHINE 11
Experiments 1A and 1B were meant to investigate the effect of a robot’s mind on users.
The studies were conducted independently in two countries, the USA and Germany. We had no a
priori assumptions regarding differences between US and German samples. By conducting the
studies in two different countries (and two different languages), we sought to increase
generalizability. Both studies adhered to a similar design; hence, the method and the results
sections are presented together. Three descriptions of robots were constructed that represented a
robot with experience, a robot with agency, and a robot without mind (robot-as-tool).
Respondents’ eeriness in response to the robot served as our main dependent variable, reflecting
the assumed robot mind effects. In addition, manipulation check variables on the robot’s
experience and agency, and items on the robot’s gender and behavioral intentions in response to
the robot were included.
2.1 Method
2.1.1 Participants. The intended sample size for each study was a priori determined to be
at least 84, required to detect a medium to large effect (f = .35) for the three-group omnibus
ANOVA test (with alpha-error probability = .05, and beta-error probability = 0.20, cf. Faul,
Erdfelder, Buchner, & Lang, 2009). In Experiment 1A, members of the MTurk participant pool
from the US were invited and 101 participants completed the study. We inspected the data for
careless responding using the time spent on the survey and a control question (Meade & Craig,
2012). Based on the descriptive data distribution of the total response times for the survey, four
participants who needed less than 50 seconds to complete the questionnaire were excluded. An
additional four participants did not recognize the name of the robot at the end of the survey,
indicating low data quality.
1
These eight participants were excluded from all further analyses.
1
The results remained virtually unchanged when the participants who failed to remember the name of the
robot were included.
MIND AND MACHINE 12
The remaining sample consisted of n = 93 US residents (48 female) with an average age of M =
34.55 years (SD = 10.75, age range 19 to 64 years). In Experiment 1B, 117 members of the
German Clickworker participant pool completed the study online. The descriptive data
distribution indicated that 70 seconds appeared to be a sensitive threshold in this sample, with
less time spent signaling careless responding. Four participants needed less time to answer the
questionnaire. One participant did not recognize the name of the robot at the end of the survey.
The remaining sample consisted of 112 participants (50 female) with an average age of M =
37.41 years (SD = 10.63, age range 18 to 62 years).
2.1.2 Stimuli. The study was introduced to be about a new generation of robots. After
asking for gender and age of the participant, a short description of a humanoid robot named
Ellix’ was presented. We showed one out of three descriptions by random assignment (similar to
the descriptions of a “supercomputer” by Gray & Wegner, 2012, Study 2). One description
introduced Ellix as a tool without mind (tool condition serving as a control condition), one
introduced Ellix as a robot with agency (agency condition), and the third introduced Ellix as a
robot with experience (experience condition). The descriptions included the attributes most
characteristic of the agency and experience mind perception dimensions, as outlined by the
principle component analysis conducted by Gray and colleagues (2007). In experiment 1B, the
study material was translated into German using the committee scale translation method. The
committee approach is a procedure derived from cross-cultural research to obtain a linguistically
equivalent instrument, using a committee of bilinguals and experienced scholars fluent in both
languages (Van de Vijver & Leung, 1997). The robots were introduced as follows (tool vs. agent
vs. experiencer):
MIND AND MACHINE 13
Tool condition: “Ellix is a robot with arms and hands to carry and to grasp things. Ellix
assists people in their everyday chores. Ellix was developed with the ability to act on orders of an
individual. The user can command the robot to execute actions.”
Agency condition: “Ellix is a robot with arms and hands to carry and to grasp things. Ellix
assists people in their everyday chores. Ellix was developed with the ability of self-control,
morality, memory, and emotion recognition. Ellix has the capacity to plan ahead and to
independently execute actions.”
Experience condition: “Ellix is a robot with arms and hands to carry and to grasp things.
Ellix assists people in their everyday chores. Ellix was developed with the ability to feel some
form of hunger, fear, pain, pleasure, and other emotions. Ellix is characterized by consciousness
and personality.”
2.1.3 Measures
Perceived agency and perceived experience (manipulation check). Two items assessed
the robot’s agency as perceived by the participants (“This robot has the capacity to plan actions”;
“This robot has the capacity to exercise self-control”; Experiment 1A: α = .80; Experiment 1B: α
= .84). Both items were averaged. Two items assessed the robot’s perceived experience (“This
robot has the capacity to feel pain”; “This robot has the capacity to feel fear”; Experiment 1A: α
= .96; Experiment 1B: α = .90). An average score of the two items was built. All four items
originated from Gray and Wegner (2012) and went with a five-point scale ranging from not at all
(1) to extremely (5).
Eeriness. We again followed Gray and Wegner (2012) and measured feelings of eeriness
in response to the robot with the help of three items (“uneasy”, “unnerved”, “creeped out”). The
items went with a five-point scale ranging from not at all (1) to extremely (5). The scores of the
MIND AND MACHINE 14
three items were averaged. The reliability of this eeriness scale was good (Experiment 1A: α =
.96; Experiment 1B: α = .88).
2
Additional measures. Experiment 1A included two questions about the robot’s perceived
masculinity and femininity (“How would you describe the robot? Ellix is...”: “masculine” and
“feminine”), both to be rated independently on a five-point scale from not at all (1) to extremely
(5). Experiment 1B included three items on the behavioral intentions to interact with the robot (“I
would avoid any contact with the robot” [reverse scored]; “I can imagine that I would buy such a
robot”; “I would sign a petition to ban such robots”). Excluding the third item heightened the
scale’s reliability; thus, only the first two items were used to make up a scale (α = .71). All items
were rated on a five-point scale from not at all (1) to extremely (5).
2.1.4 Procedure. The experiment started with sociodemographic questions. Next, the
description of the robot was presented along with the eeriness scale. The additional measure
followed. The experiment closed with the perceived agency/experience (manipulation check)
items and a multiple choice-question on the name of the robot.
2.2 Results
2.2.1 Manipulation check. The results of the manipulation check indicated that the
introductions elicited the intended representations of the robot among our participants. One-way
analyses of variance revealed significant differences between the three conditions for both
agency, Experiment 1A: F(2, 90) = 25.53, p < .001, ηp2 = .36, Experiment 1B: F(2, 109) = 30.86,
p < .001, ηp2 = .36, and experience, Experiment 1A: F(2, 90) = 33.46, p < .001, ηp2 = .43,
Experiment 1B: F(2, 109) = 46.07, p < .001, ηp2 = .46. Least Significant Difference (LSD) tests
2
We acknowledge that there are other scales available to assess eeriness. We chose the present scale due
to its briefness, its face validity, unidimensionality, and translatability to German. The semantic
differential scale by Ho and MacDorman (2010; 2017), which is used by many scholars, is challenging
from a psychometric perspective, as the anchors of the scale are semantically non-opposite adjectives.
MIND AND MACHINE 15
revealed that the robot with agency (agent) was perceived to possess more agency (Experiment
1A: n = 38, M = 3.82, SD = 0.99; Experiment 1B: n = 36, M = 4.24, SD = 0.79) than the robot
introduced as a tool (Experiment 1A: n = 22, M = 1.86, SD = 0.89; Experiment 1B: n = 46, M =
2.34, SD = 1.27), ps < .001, and more agency than the robot with experience (Experiment 1A: n =
33, M = 3.03, SD = 1.13; Experiment 1B: n = 30, M = 2.90, SD = 1.12), ps < .001. Robots with
experience yielded higher experience ratings (Experiment 1A: M = 3.21, SD = 1.31; Experiment
1B: M = 3.68, SD = 1.25) than both the robot introduced as a tool (Experiment 1A: M = 1.02, SD
= 0.11; Experiment 1B: M = 1.38, SD = 0.76), ps < .001), and the agentic robot (Experiment 1A:
M = 1.83, SD = 1.02; Experiment 1B: M = 1.99, SD = 1.14), ps < .001.
2.2.2 Effect of mind on eeriness. Our main analysis focused on the eeriness ratings. As
expected, eeriness differed between conditions in Experiment 1A, F(2, 90) = 10.36, p < .001, ηp2
= .18
3
, as well as in Experiment 1B, F(2, 109) = 6.96, p = .001, ηp2 = .11. The robot with
experience was perceived to be uncanniest (Experiment 1A: M = 2.75, SD = 1.32; Experiment
1B: M = 2.46, SD = 1.04), followed by the agent (Experiment 1A: M = 2.14, SD = 1.09;
Experiment 1B: M = 2.17, SD = 0.78), while the robot introduced as a tool was perceived to be
least uncanny (Experiment 1A: M = 1.36, SD = 0.67; Experiment 1B: M = 1.70, SD = 0.89).
Follow-up comparisons (LSD tests) showed that in Experiment 1A all differences between the
groups were significant, ps < .025. In Experiment 1B, eeriness was significantly lower in the tool
condition than in the experiencer condition, p < .001, and in the agent condition, p = .02, but
there was no significant difference between the experiencer and the agent conditions, p = .19.
3
In addition to the ANOVAs reported in this manuscript, Brown-Forsythe tests were conducted which
provide more conservative estimates in case the assumptions of homoscedasticity and normality of
distributed residuals are violated (Brown & Forsythe, 1974). In all cases, the results of the Brown-
Forsythe tests were equivalent to the ANOVA results.
MIND AND MACHINE 16
2.2.3 Perceived femininity and masculinity (Experiment 1A) and behavioral
intentions (Experiment 1B). Regarding the robot’s perceived femininity and masculinity
(Experiment 1A), all robot descriptions were rated equally masculine, F(2, 90) = 1.04, p = .35,
ηp2 = .02. Significant differences were revealed regarding the robot’s supposed femininity, F(2,
90) = 4.43, p = .01, ηp2 = .09. The experiencer robot was ascribed most feminine attributes (M =
3.12, SD = 0.89), followed by the agent (M = 2.89, SD = 1.18), and the tool (M = 2.27, SD =
1.03). Follow-up comparisons (LSD tests) showed that all differences between the groups were
significant, ps < .03, except the difference between experiencer and agent, p = .36. The results
regarding behavioral intentions (Tool: M = 3.80, SD = 0.94; Experiencer: M = 3.53, SD = 0.93;
Agent: M = 3.35, SD = 1.05) revealed no significant difference between the three experimental
conditions in Experiment 1B, F(2, 109) = 2.29, p = .10, ηp2 = .04.
2.3 Discussion
The findings of both experiments provide support for the assumption that robots with
mind elicit higher eeriness than robots without mind (cf. Stein & Ohler, 2017). Our results,
obtained from two independent samples that worked on a survey in two different languages,
further support the importance of distinguishing between agency and experience as dimensions of
human and robotic minds (Gray et al., 2007). Most eeriness was elicited when robots
incorporated the ability to feel (the experience dimension of mind perception), which is in line
with prior research on computers (Gray & Wegner, 2012) and related expectations regarding
robotic technologies (Wegner & Gray, 2016). A robot that incorporated agency, however,
consistently yielded increased ratings of eeriness versus baseline as well, suggesting that agency
contributes to humanoid robots’ eeriness. This finding contradicts the results by Gray and
Wegner (2012; Study 2), who examined responses to a “supercomputer” and found that a
computer’s agentic mind did not increase eeriness scores as compared to a computer without
MIND AND MACHINE 17
mind. Complementing prior research on robot gender, we showed that ascribing experience or
agency determines the perceived femininity of the robot (cf. Eyssel & Hegel, 2012). Masculinity,
however was unaffected (Experiment 1A). Moreover, the effect on eeriness did not translate to
differences in behavioral intentions regarding the robots (Experiment 1B). After elucidating the
general effect of robot minds on eeriness, our goal was to examine whether the effect of both
mind dimensions would be attenuated in the applied setting of nursing and elderly care.
3. Experiment 2
In many relevant fields of application for service robots, emotions play a considerable role
(e.g., sex work, nursing, education). The aim of our second experiment was to replicate and
extend the findings of Experiments 1A and 1B by examining eeriness as a function of a robot’s
mind as well as of work context. We predicted that in the context of nursing, uncanny feelings
regarding the experiencer robot would be reduced, as compared to an unspecified context,
yielding a less pronounced difference between robots as tools, agents, and experiencers.
3.1 Method
3.1.1 Participants. In Experiment 2, members of the MTurk participant pool (restricted to
US residents) were recruited. Based on the more complex design and in order to be able to
identify small to medium moderation effects of the context in which the robot is presented, a
larger number of participants than in Experiments 1A and 1B was aspired. The study was
completed by 406 participants. To reduce the influence of careless responding, we excluded 15
participants who worked for less than 50 seconds on the survey, as well as five participants who
did not recognize the name of the robot at the end of the survey. The final sample consisted of
386 participants (155 female) with an average age of 33.37 years (SD = 10.42, range 19 to 74
years). Two participants did not indicate their gender or age.
MIND AND MACHINE 18
3.1.2 Stimuli. Using the same procedure as in Experiment 1, we presented descriptions of
a robot named Ellix, which was introduced as a tool, an agent, or an experiencer. For half of the
participants, we specified that the robot would fulfill tasks in the field of elderly health care,
whereas the context was unspecified (as in Experiment 1) for the other half (see Appendix for the
descriptions). In the nursing condition, the particular tasks of the robot were specified in one of
two ways, either as to carry, grasp, and to feed the elderly or as to clean the elderly and
everything around them. This variation was meant to improve the generalizability of our findings
to the actual field of nursing, and it did not affect the findings. The experiment followed a 3
(robot mind: none/tool vs. agent vs. experiencer) x 2 (context: unspecified vs. nursing) between-
subjects design.
3.1.3 Measures. The same scales as in Experiment 1A were employed. The eeriness scale
showed good reliability (α = .93). Again, we asked for the robot’s femininity and masculinity.
The items measuring agency and experience (used for the manipulation check) showed high
internal consistency (α = .78 and .94, respectively). All items went with a five-point scale ranging
from not at all (1) to extremely (5).
3.2 Results
3.2.1 Manipulation check. The results of the manipulation check indicated that the
descriptions of the robots elicited the intended representations among our participants
(descriptive statistics are reported in Table 1). Two-way analyses of variance revealed a
significant main effect of robot mind for both agency, F(2, 380) = 89.10, p < .001, ηp2 = .32, and
experience, F(2, 380) = 130.08, p < .001, ηp2 = .41. The agent was perceived to possess more
agency than both the tool and the experiencer. Robots with experience yielded higher experience
ratings than both the robot introduced as a tool and the agentic robot (all ps < .001). Work context
exerted a significant main effect on experience ratings, F(1, 380) = 12.94, p < .001, ηp2 = .03, but
MIND AND MACHINE 19
no significant main effect on agency ratings, F(1, 380) = 0.89, p = .34. The interaction between
the two factors was significant regarding the agency ratings, F(2, 380) = 3.73, p = .025, ηp2 = .02.
As Tukey post-hoc tests showed, agency ratings were larger in the agent condition than in the
experiencer condition when the context was unspecified, t (380) = 4.77, p < .001, but agency
ratings did not differ between the agent and the experiencer condition in the nursing context, t
(380) = 0.93, p = .939. No significant interaction between the two factors was found for the
experience ratings, F(2, 380) = 2.73, p = .066, ηp2 = .01.
MIND AND MACHINE 20
Table 1
Effects of robot mind and robot context on perceived eeriness, agency, experience, masculinity,
and femininity ratings: Descriptive statistics (Experiment 2)
n
Agency
M (SD)
Experience
M (SD)
Masculinity
M (SD)
Femininity
M (SD)
Context:
Unspeci-
fied
Tool
71
2.04
(1.10)
1.24 (0.81)
2.77 (1.02)
2.66
(1.07)
Agent
70
3.82
(1.04)
1.83 (0.96)
2.67 (1.13)
2.78
(1.14)
Experiencer
52
2.92
(1.10)
3.29 (1.29)
2.60 (1.18)
2.54
(1.16)
Context:
Nursing
Tool
72
2.10
(1.05)
1.20 (0.58)
2.51 (0.95)
2.99
(1.06)
Agent
61
3.57
(0.92)
1.39 (0.66)
2.18 (0.83)
3.31
(1.12)
Experiencer
60
3.40
(0.94)
2.76 (1.13)
2.15 (0.68)
3.43
(0.98)
3.2.2 Effect of mind and context on eeriness. Like in Experiment 1, our main focus was
on eeriness (see Figure 1). A two-way analysis of variance revealed a significant main effect of
the robot mind on eeriness, F(2, 380) = 15.96, p < .001, ηp2 = .08, whereas the work context
yielded no main effect, F(1, 380) = 1.68, p = .20, ηp2 = .004. Importantly, the two-way interaction
between the two factors was significant, F (2, 380) = 3.24, p = .040, ηp2 = .02. In the unspecified
context, the experiencer was perceived to be most uncanny, followed by the agent, while the
robot introduced as a tool was perceived to be least uncanny, with F(2, 380) = 16.12, p < .001,
ηp2 = .08 for the simple main effect. In the unspecified context condition, all differences between
the robot mind conditions (LSD) were significant, ps < .01. Yet, if put in a health care context,
MIND AND MACHINE 21
the simple main effect was not significant, F(2, 380) = 2.53, p = .081, ηp2 = .01, and comparisons
between single conditions indicated that the difference between experiencer and tool was
significantly different from zero (p = .026), whereas the others were not (ps > .21). Looking at
the simple main effects from the other experimental factor’s perspective, in the tool condition the
eeriness scores did not significantly differ between the unspecified and the nursing contexts F(1,
380) = 1.12, p = .291, ηp2 = .003. Likewise context did not affect eeriness when the robot with
agency was portrayed, F(1, 380) = 0.38, p = .537, ηp2 = .001. For the robot with experience,
however, eeriness was lower in the health care/nursing context than in the unspecified context,
F(1, 380) = 6.06, p = .014, ηp2 = .016.
Figure 1. Eeriness in response to robots as tools, agents, and experiencers in an unspecified and
in a nursing context (Experiment 2, means and standard errors of the mean)
Context
MIND AND MACHINE 22
3.2.3 Perceived femininity and masculinity. Regarding the robot’s supposed
masculinity, the robot mind yielded no significant main effect, F(2, 379) = 2.87, p = .06, ηp2 =
.01, whereas the work context of the robot did, F(1, 379) = 15.69, p < .001, ηp2 = .04. Robots
were ascribed less masculinity in the nursing context (M = 2.30, SD = 0.85) than in the
unspecified context (M = 2.69, SD = 1.10). There was no indication for an interaction between
the two factors, F(2, 379) = 0.52, p = .60, ηp2 = .003. Parallel results were found for femininity:
robot mind had no main effect, F(2, 379) = 1.54, p = .22, ηp2 = .008, whereas the work context of
the robot yielded a significant main effect, F(1, 379) = 27.32, p < .001, ηp2 = .07. Robots were
ascribed more femininity in the nursing context (M = 3.23, SD = 1.07) than in the unspecified
context (M = 2.67, SD = 1.12). There was no indication for an interaction between the two
factors, F(2, 379) = 2.18, p = .11, ηp2 = .011.
3.3 Discussion
As expected, context mattered with respect to the eeriness of humanoid robots. In the
unspecified context, the pattern of results found in our first set of experiments was replicated.
Robots were uncanniest when their mind comprised experience, least uncanny when they had no
mind at all, with agency falling in-between, eliciting more eeriness than robots in the control
condition, but less eeriness than robots with experience. When the robot’s work context was
specified to be nursing, the effect of the robot mind manipulation was reduced. In particular,
eeriness in the nursing context was reduced for robots who can feel. These findings corroborate
the notion that robots with the same attributes can elicit more or less eeriness in different contexts
(Tay et al., 2014). A substantial amount of the research on the uncanny valley is based on robotic
stimuli that are placed in an unspecified environment (cf. Wang et al., 2015). The lack of context
appears to be a remarkable caveat to this line of work.
MIND AND MACHINE 23
In addition to our results on eeriness, we found that robots in a nursing context were
ascribed more femininity and less masculinity than robots in an unspecified context. This result is
in line with work on user’s application of stereotypes regarding humans to robots (e.g., Bartneck
et al., 2018; Eyssel & Hegel, 2012). Given the link between experience and female gender (and
agency and male gender), we examined how a match/mismatch may affect users’ eeriness in
Experiment 3.
4. Experiment 3
Experience is the aspect of mind that is associated with the female gender role (e.g., communion,
Bakan, 1966; femininity, Spence, Helmreich & Stapp, 1975; cf. Abele & Wojciszke, 2014),
whereas agency is the aspect of mind that is associated with the male gender role (e.g., agency,
Bakan, 1966; masculinity, Spence et al., 1975; cf. Abele & Wojciszke, 2014). In Experiment 1A,
a robot with experience was rated to be most feminine. Given that the uncanny valley is often
explained as a conflict between existing schemas and expectations, we tested the assumption that
a robot with experience and female gender would be perceived to be less eerie than a robot with
experience and male gender, whereas the effect of ascribed gender on eeriness should be reversed
for robots with agency.
4.1 Method
4.1.1 Participants. Experiment 3 was again conducted online and US members of the
MTurk participant pool were recruited. Of the 561 participants who finished the study, ten
participants needed less than 50 seconds to complete the questionnaire, which indicated that these
participants did not work on the survey thoroughly. An additional two participants did not answer
the questions on several scales and were also excluded. Moreover, 45 participants did not
remember the gender of the robot at the end of the survey, indicating low data quality (cf. Meade
MIND AND MACHINE 24
& Craig, 2012). The remaining sample consisted of 504 US residents (237 female) with an
average age of 34.33 years (SD = 10.51, range between 18 and 73 years).
4.1.2 Stimuli. Participants received a description of a new generation of robots like in the
previous studies. In addition to the three different human likeness conditions, we manipulated the
robot’s supposed gender by changing its name. The participants were assigned to one out of three
robot gender conditions (neutral vs. female vs. male). In the neutral condition, the robot’s name
was ‘Ellix’, while for the female condition one of five female names (‘Emily’, ‘Madison’,
‘Emma’, ‘Olivia’, ‘Hannah’) and for the male condition one of five male names (’Jacob’,
’Michael’, ’Joshua’, ’Matthew’, ‘Daniel’) was provided. The specific names were chosen because
they were the five most popular female or male given names for newborns in the 2000s, based on
US social security card application data (Social Security Administration, 2017). Thus, robot
descriptions were identical to the descriptions of Experiment 1A except for the fact that for one
third of the participants the robot’s name was Ellix, for one third the robot had a female name,
and for one third the robot had a male name.
4.1.3 Measures. The same scales and measures as in Experiment 1A and 2 were
employed. The eeriness scale showed good reliability (α = .95). The scales on perceived agency
(α = .80) and experience (α = .93), which served as a manipulation check, were also internally
consistent. Furthermore, we asked for the robot’s femininity and masculinity with the same two
items as before. Again, all items went with a five-point scale ranging from not at all (1) to
extremely (5).
4.2 Results
4.2.1 Manipulation check. The results of the manipulation check indicated that the
introductions were successful and elicited the intended representations of the robot’s agency and
experience in our participants. Significant differences between the tool, agent, and experiencer
MIND AND MACHINE 25
conditions were found for agency, F(2,501) = 157.09, p < .001, ηp2 = .38, as well as for
experience, F(2,501) = 171.54, p < .001, ηp2 = .41. Robots in the agent condition (M = 3.83, SD =
0.90) were perceived to possess more agency than the robots in the experiencer condition (n =
169, M = 2.95, SD = 1.02), and those in the tool condition (M = 1.94, SD = 1.00), all ps < .001.
Robots in the experiencer condition had higher experience ratings (M = 3.16, SD = 1.24)
compared to robots in the agent (M = 2.00, SD = 1.10) and in the tool condition (M = 1.15, SD =
0.45), all ps < .001. The robot’s gender (neutral vs. female vs. male name) neither affected
agency, F(2,501) = 1.58, p = .21, ηp2 = .006, nor experience, F(2,501) = 2.41, p = .09, ηp2 = .009.
Robot gender had the intended effect on perceived femininity and masculinity, as the
results showed significant differences for femininity, F(2,501) = 79.31, p < .001, ηp2 = .24, and
masculinity, F(2,501) = 69.36, p < .001, ηp2 = .21, between the neutral, male, and female
conditions. Robots with a female name were rated higher in femininity (M = 3.46, SD = 1.21)
than robots with a male name (M = 2.01, SD = .88) or with a neutral name (M = 2.49 SD = 1.10),
all ps < .001. Moreover, robots with a male name (M = 3.22, SD = 1.17) were perceived as more
masculine than robots with a female name (M = 1.88, SD = .87) or a neutral name (M = 2.71 SD
= 1.13), all ps < .001. No differences were found between the mind perception conditions for
femininity, F(2,501) = 0.27, p = 0.77 , ηp2 = .001, and for masculinity, F(2,501) = 0.12 , p = 0.88 ,
ηp2 < .001. Descriptive statistics for all conditions are displayed in Table 2.
MIND AND MACHINE 26
Table 2
Effects of robot mind and robot gender on perceived eeriness, agency, experience, masculinity,
and femininity ratings: Descriptive statistics (Experiment 3)
n
Eeriness
M (SD)
Agency
M (SD)
Experience
M (SD)
Masculinity
M (SD)
Femininity
M (SD)
Neutral
Robot
Tool
56
1.50 (0.63)
1.93 (0.99)
1.11 (0.37)
3.09 (1.07)
2.41
(0.95)
Agent
53
2.22 (0.86)
3.70 (0.97)
1.71 (0.96)
2.41 (1.10)
2.60
(1.23)
Experiencer
62
2.25 (1.02)
2.96 (0.94)
3.20 (1.27)
2.61 (1.14)
2.45
(1.13)
Female
Robot
Tool
61
1.73 (0.73)
1.98 (1.08)
1.17 (0.52)
1.82 (0.90)
3.36
(1.20)
Agent
50
1.99 (0.91)
3.77 (0.89)
1.85 (0.98)
1.88 (0.96)
3.48
(1.33)
Experiencer
57
2.33 (0.95)
2.94 (0.98)
2.99 (1.20)
1.93 (0.75)
3.56
(1.12)
Male
Robot
Tool
48
1.74 (0.77)
1.88 (0.92)
1.16 (0.44)
3.15 (1.27)
1.85
(0.77)
Agent
67
2.24 (0.91)
3.97 (0.84)
2.34 (1.21)
3.25 (1.06)
2.13
(0.97)
Experiencer
50
2.26 (1.03)
2.95 (1.17)
3.29 (1.26)
3.26 (1.21)
2.00
(0.86)
4.2.2 Effect of mind and ascribed gender on eeriness. Like in Experiments 1 and 2, our
main focus was on eeriness. The mind of a robot yielded a significant effect, F(2,495) = 24.12, p
< .001, ηp2 = .088, whereas the gender of the robot did not, F(2,495) = 0.51, p = .60, ηp2 = .002.
The two-way interaction between both factors was not significant, F(4,495) = 1.25, p = .33, ηp2 =
.009. Robot mind influenced the eeriness of a female robot, F(2,165) = 7.24, p < .001, ηp2 = .081,
as well as the eeriness of a male robot, F(2,162) = 5.37, p < .01, ηp2 = .062. Follow-up
comparisons for the mind factor showed that the robot without mind (tool, M = 1.65, SD = 0.72)
was perceived to be significantly less eerie than the robot with experience (M = 2.28, SD = 0.99)
MIND AND MACHINE 27
and the robot with agency (M = 2.16, SD = 0.89), ps < .001. The difference between the
experiencer and the agent conditions was not significant, p = .13.
4.3 Discussion
Gender is one of the defining characteristics in the perception of other humans. When the
distinction of social content in two dimensions is applied (cf. Abele & Wojciszke, 2014), women
are associated with communion or experience, whereas men are associated with agency (Bakan,
1966; Spence et al., 1975). Given that social categories for humans tend to be applied to
computers and robots as well (Nass et al., 1994) – including category of gender (e.g., Eyssel &
Hegel, 2012) – the eeriness of a robot’s mind was assumed to vary with the robot’s gender. In
contrast to our expectations, however, a feeling robot named Emily (or Hannah or Olivia) was not
less eerie than a feeling robot named Jacob (or Michael or Daniel). Likewise, the male agentic
robot provoked neither lower nor higher eeriness than the female agentic robot. Thus, simply
giving a robot a female name does not reduce its eeriness, even if experience is its prominent
characteristic.
5. Agents, Experiencers, and Participants’ Age
Several authors have suggested that the uncanny valley could be a function of the
participants’ age (e.g., Ishiguro, 2007; Kuo et al., 2009; Stein & Ohler, 2017), but the nature and
direction of this effect is unclear. Therefore, we analyzed the data from Experiments 2 and 3 to
test whether age moderates the effect of a robot’s mind (tool vs. agent vs. experiencer) on
eeriness. We refrained from building arbitrary age groups and examined both linear and nonlinear
(quadratic) interaction effects (we did not test this moderator hypothesis on data from
Experiments 1A and 1B, because the sample size was limited and much smaller than in
Experiments 2 and 3).
5.1 Method of Data Analysis
MIND AND MACHINE 28
The data from Experiments 2 and 3 was analyzed separately to be able to investigate
whether a potential moderator effect of age replicates across studies. The data of two participants
from Experiment 2 were excluded from the analyses because they did not indicate their age.
Thus, the analyses were based on a sample of 384 participants in Experiment 2 who were
between 19 and 74 years old (M = 33.37; SD = 10.42), and on 504 participants in Experiment 3
whose age ranged between 18 and 73 years (M = 34.33; SD = 10.51).
To test whether age moderates the effect of robot mind on eeriness in a linear or quadratic
fashion, we conducted moderated regression analyses. We used two dummy-coded indicator
variables for the robot mind factor (reference category: experiencer) and we centered the
continuous variable age before calculating quadratic terms for age and interaction terms with the
dummy variables. We specified a linear model (i.e., a model including a linear effect of age on
eeriness as well as interactions between the linear age term and the dummy variables):
!"#$"%& '()* '+,-../%)* '0,12"34%) '5,12"%)* '6,
7
-../%,12"%
8
*) *'9,
7
12"34%,12"%
8
)* :%
(1)
and a quadratic model (i.e., a model including both a linear and a quadratic term for age as well
as interactions between the linear and quadratic age terms and the dummy variables):
!"#$"%& '()* '+,-../%)* '0,12"34%) '5,12"%)* '6,12"0
%)* '9,
7
-../%,12"%
8
)* ';,
7
12"34%,12"%
8
)* '<,
7
-../%,12"=%
8
)* '>,
7
12"34%,12"=%
8
)* :%
(2)
MIND AND MACHINE 29
For both datasets, we tested whether there was a moderating effect of age and whether the
quadratic model (2) fitted the data better than the linear model (1).
5.2 Results and Discussion
The results of the moderator analyses are presented in Table 3.
4
In Experiment 2, there
was no significant difference between the linear and the quadratic model, F(3, 375) = .72, p =
.54. In the linear model, we found no age effects on eeriness in the experiencer condition (main
effect of age) and no interaction effects between age and robot mind.
5
In Experiment 3, the
quadratic model showed a significantly better fit to the data than the linear model, F(3, 495) =
2.89, p = .034. In Experiment 3, there was a significant quadratic relationship between age and
eeriness in the experiencer condition (main effect of the quadratic term of age), b = .001, t(495) =
2.34, p = .020. Moreover, the interaction between the quadratic term of age and the tool dummy
variable (representing the difference between the tool and the experiencer condition) was
significant, b = -.002, t(495) = -2.94, p = .003. To gain more insight into the form this interaction
effect took, we plotted the predicted curves for each experimental condition. As shown in Figure
2, the difference in eeriness between robots without minds (robots-as-tools) and robots with
experience were more pronounced for young as well as old participants than for middle-aged
participants.
4
For Experiment 2, we also tested a model which additionally included the experimental nursing context
factor. There was no interaction between the linear or quadratic terms of age and nursing context, and no
three-way interactions between age, nursing context, and robot mind. Therefore, we did not include this
model in the results section.
5
There was a significant interaction effect between age and the tool indicator variable in the quadratic
model. However, given that the quadratic model did not fit the data better than the linear model and given
that the interaction between age and the tool indicator variable was not significant in the linear model, we
refrained from interpreting this age*tool interaction effect that turned significant in the quadratic model
(possibly due to suppressor effects).
MIND AND MACHINE 30
Figure 2. Eeriness in response to robots in the tool, agent, and experiencer condition. Quadratic
interaction with age (Experiment 3)
The latter findings need to be interpreted with caution, as the moderator effect of age was
only found in one of the two datasets we analyzed. We consistently found no support for a linear
effect of age on eeriness. These findings are in contrast to the assumption that the relatively
higher eeriness of experiencer robots decreases with participants’ age. As a caveat of our
findings, it needs to be noted that our sample was recruited via MTurk, and therefore, did not
comprise the complete age range. It mainly included individuals in the young old-age segment
(65-74 years of age), but few fell into the old segment (75-84 years) and no one belonged to the
oldest-old age segment (85 years and above). Thus, these results hold only for a younger segment
of the older adult population. Moreover, further research is needed to draw a distinction between
age and cohort effects in an ever-changing technological environment.
MIND AND MACHINE 31
Table 3a
Results of the age effect analysis in Experiment 2 for the linear and quadratic model
Note. Experiment 2: N = 384. Experiencer was used as reference category. b = unstandardized regression weight; β = standardized
regression weight.. 95% CI = 95% confidence interval.
* p < .05; ** p < .01
Linear Model
0.282 .0793**
Intercept 2.584** 0.102 [2.384 , 2.784]
Tool -0.737** 0.136 [-1.004, -0.470] -0.321 [-0.437, -0.204]
Agent -0.354** 0.139 [-0.627, -0.081] -0.151 [-0.267, -0.034]
Age 0.014 0.009 [-0.005, 0.033] 0.131 [-0.043, 0.305]
Tool x Age -0.018 0.013 [-0.044, 0.007] -0.101 [-0.243, 0.041]
Agent x Age -0.013 0.013 [-0.039, 0.013] -0.072 [-0.212, 0.069]
Quadratic Model
Intercept 2.669** 0.121 [2.431 , 2.908]
0.291 .0845**
Tool -0.877** 0.171 [-1.213, -0.540] -0.381 [-0.528, -0.235]
Agent -0.449** 0.171 [-0.785, -0.113] -0.191 [-0.334, -0.048]
Age 0.028 0.014 [0.000 , 0.056] 0.262 [-0.003, 0.526]
Ag -0.001 0.001 [-0.002, 0.000] -0.146 [-0.367, 0.076]
Tool x Age -0.038* 0.019 [-0.075, -0.001] -0.210 [-0.415, -0.004]
Tool x Ag 0.001 0.001 [-0.001, 0.003] 0.138 [-0.060, 0.336]
Agent x Age -0.028 0.019 [-0.065, 0.009] -0.153 [-0.353, 0.047]
Agent x Age² 0.001 0.001 [-0.001, 0.003] 0.093 [-0.098, 0.283]
Predictors
b
SE b
b
95 % CI
[LL,UL]
β
β
95 % CI
[LL,UL]
R
MIND AND MACHINE 32
Table 3b
Results of the age effect analysis in Experiment 3 for the linear and quadratic model
Note. Experiment 3: N = 504. Experiencer was used as reference category, b represents unstandardized regression weights; β
indicates the standardized regression weights; LL and UL indicate the lower and upper limits of a confidence interval, respectively. *
p < .05; ** p < .01
Linear Model
.307 .0943
Intercept 2.295** 0.068 [2.162 , 2.429 ]
Tool -0.636** 0.097 [-0.826, -0.446]
-0.326 [-0.424, -0.228]
Agent -0.133 0.096 [-0.321, 0.056 ]
-0.068 [-0.166, 0.029 ]
Age 0.010 0.008 [-0.005, 0.025 ]
0.112 [-0.058, 0.282 ]
Tool x Age -0.014 0.010 [-0.033, 0.005 ]
-0.100 [-0.235, 0.036 ]
Agent x Age -0.003 0.010 [-0.022, 0.017 ]
-0.018 [-0.151, 0.114 ]
Quadratic Model
Intercept 2.165** 0.087 [1.994, 2.336 ]
.330 .109
Tool -0.420** 0.121 [-0.658, -0.181] -0.215 [-0.337, -0.093]
Agent -0.009 0.124 [-0.253, 0.235 ] -0.005 [-0.130, 0.121 ]
Age -0.003 0.009 [-0.020, 0.015 ] -0.029 [-0.234, 0.176 ]
Ag 0.001* 0.001 [0.000, 0.003 ] 0.298 [0.052, 0.543 ]
Tool x Age 0.011 0.013 [-0.015, 0.037 ] 0.078 [-0.106, 0.262 ]
Tool x Ag -0.002** 0.001 [-0.004, -0.001] -0.351 [-0.586, -0.117]
Agent x Age 0.009 0.013 [-0.016, 0.034 ] 0.062 [-0.110, 0.233 ]
Agent x Age² -0.001 0.001 [-0.003, 0.000 ] -0.189 [-0.399, 0.020 ]
R
b
SE b
b
95 % CI
[LL,UL]
β
β
95 % CI
[LL,UL]
Predictors
MIND AND MACHINE 33
6. General Discussion
In many societies around the world, intelligent personal assistants (such as Siri and
Alexa), autonomous vehicles, and smart homes have become part of people’s everyday lives or
are projected to be mass phenomena in the very near future. Humans have a tendency to ascribe
mind to non-human entities (Gray et al., 2007) and they sometimes respond to computer
technologies in ways similar as to real social beings (Nass et al., 1994). With the proliferation of
these smart systems and increasing computational power, it is relevant to understand users’
responses to these humanlike but non-human systems. Responses to these technologies may
provide insight to basic human experience and behavior and they are relevant for creators and
marketers.
The technological innovation in the field that is arguably most challenging for theory and
research in the social sciences (as well as for legislators, e.g., EU Directorate-General for Internal
Policies, 2016) are humanoid service robots, meant to provide sexual pleasure, to accomplish
tasks in military operations, or to assist in hospitals or nursing homes. A dominant framework to
predict user responses to humanoid robots is the uncanny valley hypothesis, positing that
humanlike but not perfectly human robots elicit feelings of eeriness among (future) users (Mori,
1970; Wang et al., 2015). According to a recent perspective at understanding the eeriness of
humanoid robots, this negative response is a function of perceiving mind in a machine (Gray &
Wegner, 2012; Stein & Ohler, 2017; Wegner & Gray, 2016). Based on a two-dimensional
approach to mind perception (Gray et al., 2007; Tanibe et al., 2017), it has been assumed more
specifically that humanoid robots with experience elicit eeriness because experience or emotions
are exclusively associated with the concept of humans. Agency, the second mind perception
dimension, however, was considered to be unrelated to eeriness (Gray & Wegner, 2012; Wegner
& Gray, 2016). User responses to a future supercomputer provided initial evidence for this
MIND AND MACHINE 34
assumption (Gray & Wegner, 2012), but findings in an applied context were somewhat
contradictory (Stafford et al., 2014). If experience but not agency would be responsible for the
uncanny valley phenomenon, this would not only inform theory, but it could also have important
implications for the design and the marketing of future robots.
A series of four experiments was conducted, to test the impact of a robot’s agency and
experience on eeriness, including moderation effects of the context in which the robot is set, its
gender, and users’ age. In all of our experiments, we found that a robot with experience elicited
stronger feelings of eeriness than a robot without mind. In our initial two experiments
(Experiments 1A and 1B), we further found that a robot with agency elicited stronger feelings of
eeriness than a robot without mind. Whereas the results on robots who can feel corroborate
previous research (Gray & Wegner, 2012), the latter result is novel in demonstrating that both
dimensions of mind perception can elicit eeriness and both may be responsible for the eeriness of
humanoid robots examined in uncanny valley research. The subsequent experiment (Experiment
2) showed that the impact of a robot’s mind on eeriness is reduced when the robot is set in a
nursing context, particularly, a robot with experience is perceived to be less eerie when its task is
to feed or to clean in a nursing context than to operate without the given context. Recognizing the
role of context could be important for other research questions on the uncanny valley hypothesis
as well. For roboticists and HRI developers, our results suggest that it might not be feasible to
work towards general design implications in the field of social robotics; instead, we consider it
more reasonable to identify different „sets“ of robot characteristics that work more or less in
different contexts.
Despite the strong link between experience and agency attributions and gender (e.g.,
Bakan, 1966; Fiske et al., 2007; Spence et al., 1975), ascribing female or male gender to a robot
had no attenuating effect on the eeriness of robotic minds (Experiment 3). Today, the distinction
MIND AND MACHINE 35
between male and female is still one of the most influential categories in human-human
perception (Ellemers, 2018). Prior research shows that a robot’s gender affects human responses
to the robot as well (e.g., Eyssel & Hegel, 2012). Our findings indicate, however, that
stereotypical expectations about a robot’s gender cannot override the general influence of the
robot mind dimensions. In our final analyses, the moderating influence of participants’ age was
examined. Findings were mixed, as the influence of robot mind on eeriness was unrelated to
respondents’ age in Experiment 2. The same analysis conducted for the data of Experiment 3,
however, yielded a significant quadratic interaction, suggesting that the difference between robots
with experience and with no mind at all is smallest between the ages of 30 and 50 with larger
differences at an earlier or older age. This finding is in contrast to hopes that robots that elicit
eeriness for many elicit rather low eeriness among individuals of young old age and older.
The design of our series of experiments was focused on testing the influence of
experience and agency on users’ eeriness in a most internally valid manner. To this end, we used
the attributes that defined both dimensions (Gray et al., 2007) in our description of a future
generation of robots. Our research was not aimed at identifying factors that lead users to ascribe
experience and agency in the first place. Wang and Krumhuber (2018), for example, showed that
the proposed function of a robot (social vs. sales) affects the ascribed experience of the robot.
The embodiment of an artificial character is another factor that likely influences ascriptions of
mind. Hoffmann and colleagues (2018) proposed that embodiment predicts the perceived
attributes of nonverbal expressiveness, (shared) perceptions, mobility, tactile interaction, and
corporeality (physical existence) which in turn determine user responses. Research is encouraged
to enrich our understanding as to how embodiment factors change ascribed agency and
experience.
MIND AND MACHINE 36
Our series of studies provided novel insights, but limitations need to be noted. First, our
studies were based on vignettes, that is, descriptions of a new generation of robots, which were
varied to manipulate the mind of the robot (along with its occupational role, and its gender in
Experiments 2 and 3). This methodological approach had been used in closely related work
before (Gray & Wegner, 2012, Study 2) and we believed that a textual description of a robot’s
mind provides the most exact operationalization of both dimension of mind that drove our
research questions (Gray et al., 2007). Much of the research on the uncanny valley is based on
images of robots and or morphs between humans and robots (Wang et al., 2015). Such visual
stimuli are arguably able to elicit perceptions of mind, but a differentiation between agency and
experience that adheres to theoretical distinctions is difficult to fathom. We need to acknowledge
that the description of the robot in the tool condition was shorter than the description of the robot
in both other conditions. We have no indication that the differences between the conditions were
a function of the mere amount of attributes of the robot. That said, future research should strive
for constant length or complexity. Moreover, our experimental design did not include a condition
in which robots are ascribed both agency and experience. Stein and Ohler (2017) compared
agents with both, agency and experience, to agents who were programmed by humans (similar to
the tool condition) showing that the former elicited more eeriness. Users’ responses to robots
with agency and experience – as compared to robots with only one component – are still to be
examined.
Second, our approach to mind perception closely followed earlier theory. Our description
of the robot with agency included the characteristics of self-control, morality, memory, emotion
recognition, and planning, whereas the robot with experience was characterized by hunger, fear,
pain, pleasure, consciousness, personality, and other emotions. These descriptors were almost
identical to the attributes that founded each of the mind perception factors in the principal
MIND AND MACHINE 37
component analysis presented by Gray and colleagues (2007). We believe that our distinction
between the agency and experience dimensions of mind perception and the operationalization of
robots that are closely aligned to the descriptions of the mind perception dimensions makes our
work readily accessible to the research community. Our findings, however, do not rule out the
feasibility of alternative conceptions of mind (e.g., potential one-, three-, or more-dimensional
solutions which would be based on different markers of mind). On a related note, our findings do
not rule out the possibility that some characteristics within a mind dimension (e.g., morality)
could elicit more eeriness than others (e.g., memory; morality and memory are both constituents
of agency).
Last, all experiments were conducted online, using participant pools such as MTurk and
Clickworker. People who are actively involved in an online participant pool might show higher
levels of openness towards innovation or technology. Further, all participants were recruited in
Western cultures (i.e., the US and Germany). In addition to the limited age distribution discussed
above, these aspects limit the generalizability of our findings.
7. Conclusion
Connecting research on mind perception and research on the uncanny valley hypothesis, we
showed that a humanoid robot who can feel (experience) as well as a humanoid robot who can
think and plan ahead (agency) elicit more eeriness than a robot without mind (robot-as-tool), with
experience yielding highest eeriness. This supports previous findings on the eeriness of robots
with feelings, but theory and practice should acknowledge that robots with agency do elicit
eeriness as well. The effects of a robot’s mind are attenuated when a nursing context is
introduced. A robot’s gender had no influence on the eeriness of robots with minds. Non-linear
effects of participants’ age on the robot mind-eeriness link were found, but need corroboration.
MIND AND MACHINE 38
References
Abele, A.E. & Wojciszke, B. (2014). Communal and agentic content in social cognition: A dual
perspective model. In M. P. Zanna & J. M. Olson (Eds.), Advances In Experimental
Social Psychology, Vol. 50 (pp. 195-255), Burlington CT: Academic Press.
Archibald, M. M., & Barnard, A. (2018). Futurism in nursing: Technology, robotics and the
fundamentals of care. Journal of Clinical Nursing, 27, 2473-2480. doi:
10.1111/jocn.14081
Bakan, D. (1966). The duality of human existence: An essay on psychology and religion.
Chicago, IL: Rand McNally.
Bartneck, C., Yogeeswaran, K., Ser, Q. M., Woodward, G., Sparrow, R., Wang, S., & Eyssel, F.
(2018, February). Robots and racism. In Proceedings of the 2018 ACM/IEEE
International Conference on Human-Robot Interaction (pp. 196-204). ACM. doi:
10.1145/3171221.3171260
Bartneck, C., Kanda, T., Ishiguro, H., & Hagita, N. (2009). My robotic doppelgänger - A critical
look at the Uncanny Valley. Proceedings - IEEE International Workshop on Robot and
Human Interactive Communication, 31, 269–276. doi: 10.1109/ROMAN.2009.5326351
Broadbent, E. (2017). Interactions with robots: The truths we reveal about ourselves. Annual
Review of Psychology, 68, 627-652. doi: 10.1146/annurev-psych-010416-043958
Broadbent, E., Kuo, I. H., Lee, Y. I., Rabindran, J., Kerse, N., Stafford, R., & MacDonald, B.
(2010). Attitudes and reactions to a healthcare robot. Telemedicine Journal and E-Health,
16, 608–613. doi: 10.1089/tmj.2009.0171
Brown, M. B., & Forsythe, A. B. (1974). 372: the ANOVA and multiple comparisons for data
with heterogeneous variances. Biometrics, 30, 719-724.
MIND AND MACHINE 39
Burleigh, T. J., Schoenherr, J. R., & Lacroix, G. L. (2013). Does the uncanny valley exist? An
empirical test of the relationship between eeriness and the human likeness of digitally
created faces. Computers in Human Behavior, 29, 759–771.
doi:10.1016/j.chb.2012.11.021.
Cadman, C., & Brewer, J. (2001). Emotional intelligence: a vital prerequisite for recruitment in
nursing. Journal of Nursing Management, 9, 321-324. doi: 10.1046/j.0966-
0429.2001.00261.x
Čapek, K. (1920/2001). R.U.R. (Rossum's Universal Robots). New York: Dover.
Carstensen, L. L., & Mikels, J. A. (2005). At the intersection of emotion and cognition. Current
Directions in Psychological Science, 14, 117–121. doi: 10.1111/j.0963-
7214.2005.00348.x
Carstensen, L. L., Turan, B., Scheibe, S., Ram, N., Ersner-Hershfield, H., Samanez-Larkin, G. R.,
… Nesselroade, J. R. (2011). Emotional experience improves with age: Evidence based
on over 10 years of experience sampling. Psychology of Aging, 26, 21–33. doi:
10.1037/a0021285
Charles, S. T., Mather, M., & Carstensen, L. L. (2003). Aging and emotional memory: The
forgettable nature of negative images for older adults. Journal of Experimental
Psychology: General, 85, 163-178. doi: 10.1037/0096-3445.132.2.310
Cheetham, M., Suter, P., & Jäncke, L. (2011). The human likeness dimension of the “uncanny
valley hypothesis”: behavioral and functional MRI findings. Frontiers in Human
Neuroscience, 5(November), 1–14. doi: 10.3389/fnhum.2011.00126
Creed, C., Beale, R., & Cowan, B. (2014). The impact of an embodied agent’s emotional
expressions over multiple interactions. Interacting with Computers, 27, 172–188. doi:
10.1093/iwc/iwt064
MIND AND MACHINE 40
EU Directorate-General for Internal Policies (2016). European civil law rules in robotics.
Brussels: European Parliament Policy Department C.
Ellemers, N. (2018). Gender stereotypes. Annual Review of Psychology, 69, 275-298.
10.1146/annurev-psych-122216-011719
Eyssel, F., & Hegel, F. (2012). (S) he's got the look: Gender stereotyping of robots. Journal of
Applied Social Psychology, 42, 2213-2230. doi: 10.1111/j.1559-1816.2012.00937.x
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using
G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods,
41, 1149-1160.
Ferrari, F., Paladino, M. P., & Jetten, J. (2016). Blurring human–machine distinctions:
anthropomorphic appearance in social robots as a threat to human distinctiveness.
International Journal of Social Robotics, 8, 287-302. doi:10.1007/s12369-016-0338-y
Fiske, S. T., Cuddy, A. J. C., & Glick, P. (2007). Universal dimensions of social cognition:
Warmth and competence. Trends in Cognitive Sciences, 11, 77–83. doi:
10.1016/j.tics.2006.11.005
Freud, S. (1919/2003). Das Unheimliche [The uncanny]. London, UK: Penguin.
Gilly, M. C., & Zeithaml, V. A. (1985). The elderly consumer and adoption of technologies.
Journal of Consumer Research, 12, 353-357.
Gnambs, T., & Appel, M. (2019). Are robots becoming unpopular? Changes in attitudes towards
autonomous robotic systems in Europe. Computers in Human Behavior, 93, 53-61.
doi:10.1016/j.chb.2018.11.045
Gray, H. M., Gray, K., & Wegner, D. M. (2007). Dimensions of mind perception. Science, 315,
619. doi: 10.1126/science.1134475
MIND AND MACHINE 41
Gray, K., & Wegner, D. M. (2012). Feeling robots and human zombies: Mind perception and the
uncanny valley. Cognition, 125, 125–130. doi: 10.1016/j.cognition.2012.06.007
Hegel, F., Krach, S., Kircher, T., Wrede, B., & Sagerer, G. (2008, March). Theory of Mind
(ToM) on robots: a functional neuroimaging study. In Proceedings of the 3rd ACM/IEEE
International Conference on Human-Robot Interaction (pp. 335-342). ACM. doi:
10.1145/1349822.1349866
Ho, C. & MacDorman, K. F. (2010). Revisiting the uncanny valley theory: Developing and
validating an alternative to the Godspeed indices. Computers in Human Behavior, 26,
1508–1518. doi: 10.1016/j.chb.2010.05.015
Ho, C., & MacDorman, K. F. (2017). Measuring the uncanny valley effect. International Journal
of Social Robotics, 9, 129-139. doi:10.1007/s12369-016-0380-9
Hoffmann, L., Bock, N., & Rosenthal-von der Pütten, A. M. (2018, February). The peculiarities
of robot embodiment (EmCorp-Scale): development, validation and initial test of the
embodiment and corporeality of artificial agents scale. In Proceedings of the 2018
ACM/IEEE International Conference on Human-Robot Interaction (pp. 370-378). ACM.
doi: 10.1145/3171221.3171242
International Federation of Robotics (2018). World robotics survey: Service robots are
conquering the world. Retrieved from https://ifr.org/news/world-robotics-survey-service-
robots-are-conquering-the-world-/
Ishiguro H. (2007) Android science. In S. Thrun, R. Brooks, & H. Durrant-Whyte (Eds.),
Robotics Research. Springer Tracts in Advanced Robotics (28). Berlin: Springer. doi:
10.1007/978-3-540-48113-3_11
Jentsch, E. (1906). Zur Psychologie des Unheimlichen. Psychiatrisch-neurologische
Wochenschrift, 8, 195–198, 203–205. [English version: Jentsch, E. (1997). On the
MIND AND MACHINE 42
psychology of the uncanny. Angelaki: Journal of the Theoretical Humanities, 2, 7–16.]
doi: 10.1080/09697259708571910
Kachouie, R., Sedighadeli, S., Khosla, R., & Chu, M. T. (2014). Socially assistive robots in
elderly care: a mixed-method systematic literature review. International Journal of
Human-Computer Interaction, 30, 369-393.
Kätsyri, J., Förger, K., Mäkäräinen, M., & Takala, T. (2015). A review of empirical evidence on
different uncanny valley hypotheses: support for perceptual mismatch as one road to the
valley of eeriness. Frontiers in Psychology, 6. doi: 10.3389/fpsyg.2015.00390
Kuo, I. H., Rabindran, J. M., Broadbent, E., Lee, Y. I., Kerse, N., Stafford, R. M. Q., &
MacDonald, B. A. (2009, September). Age and gender factors in user acceptance of
healthcare robots. In Robot and Human Interactive Communication, 2009. RO-MAN
2009. The 18th IEEE International Symposium on (pp. 214-219).
Liang, Y., & Lee, S. A. (2017). Fear of Autonomous Robots and Artificial Intelligence: Evidence
from National Representative Data with Probability Sampling. International Journal of
Social Robotics, 9, 379–384. doi:10.1007/s12369-017-0401-3
Lischetzke, T., Izydorczyk, D., Hüller, C., & Appel, M. (2017). The topography of the uncanny
valley and individuals’ need for structure: A nonlinear mixed effects analysis. Journal of
Research in Personality, 68, 96–113. doi:10.1016/j.jrp.2017.02.001
MacDorman, K. F. (2006, July). Subjective ratings of robot video clips for human likeness,
familiarity, and eeriness: An exploration of the uncanny valley. In ICCS/CogSci-2006
long symposium: Toward social mechanisms of android science (pp. 26-29).
MacDorman, K. F., & Chattopadhyay, D. (2016). Reducing consistency in human realism
increases the uncanny valley effect; increasing category uncertainty does
not. Cognition, 146, 190–205. doi: 10.1016/j.cognition.2015.09.019
MIND AND MACHINE 43
MacDorman, K. F., & Entezari, S. (2015). Individual differences predict sensitivity to the
uncanny valley. Interaction Studies, 16, 141–172. doi:10.1075/is.16.2.01mac
MacDorman, K. F., Green, R. D., Ho, C. C., & Koch, C. T. (2009). Too real for comfort?
Uncanny responses to computer generated faces. Computers in Human Behavior, 25,
695–710. doi:10.1016/j.chb.2008.12.026
MacDorman, K. F., & Ishiguro, H. (2006). The uncanny advantage of using androids in cognitive
and social science research. Interaction Studies, 7, 297–337. doi:10.1075/is.7.3.03mac
Mara, M., & Appel, M. (2015). Science fiction reduces the eeriness of android robots: A field
experiment. Computers in Human Behavior, 48, 156–162. doi:10.1016/j.chb.2015.01.007
Mather, M., Canli, T., English, T., Whitfield, S., Wais, P., Gabrieli, J. D. E., … Ochsner, K.
(2004). Amygdala responses to emotionally valenced stimuli in older and younger adults.
Psychological Science, 15, 259–263.
Mather, M., & Carstensen, L. L. (2005). Aging and motivated cognition: The positivity effect in
attention and memory. Trends in Cognitive Sciences, 9, 496–502.
doi:10.1016/j.tics.2005.08.005
Meade, A. W., & Craig, S. B. (2012). Identifying careless responses in survey data.
Psychological Methods, 17, 437–455. doi: 10.1037/a002808
Mori, M. (1970). Bukimi no tani (the uncanny valley). Energy, 7, 33–35. Translated in: Mori, M.,
MacDorman, K. F., & Kageki, N. (2012). The uncanny valley [from the field]. Robotics &
Automation Magazine, IEEE, 19(2), 98–100. doi: 10.1109/mra.2012.2192811
Mostaghel, R. (2016). Innovation and technology for the elderly: Systematic literature review.
Journal of Business Research, 69, 4896-4900.
Nass, C., Steuer, J., & Tauber, E. R. (1994). Computers are social actors. In: Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems: Celebrating
MIND AND MACHINE 44
Interdependence (72-78). Boston, MA: ACM.
Nomura, T., & Nakao, A. (2010). Comparison on identification of affective body motions by
robots between elder people and university students: A case study in Japan. International
Journal of Social Robotics, 2, 147–157. doi:10.1007/s12369-010-0050-2
Nomura, T. T., Syrdal, D. S., & Dautenhahn, K. (2015). Differences on social acceptance of
humanoid robots between Japan and the UK. in M. Salem, A. Weiss, P. Baxter & K.
Dautenhahn (Eds), Proceedings 4th Int Symposium on New Frontiers in Human-Robot
Interaction (pp. 115-120). Canterbury , United Kingdom: The Society for the Study of
Artificial Intelligence and the Simulation of Behaviour (AISB)
Piwek, L., McKay, L. S., & Pollick, F. E. (2014). Empirical evaluation of the uncanny valley
hypothesis fails to confirm the predicted effect of motion. Cognition, 130, 271–277.
doi:10.1016/j.cognition.2013.11.001
Reich-Stiebert, N., & Eyssel, F. (2017). (Ir) relevance of gender?: On the influence of gender
stereotypes on learning with a robot. In Proceedings of the 2017 ACM/IEEE International
Conference on Human-Robot Interaction (pp. 166-176). ACM. doi:
10.1145/2909824.3020242
Rosenthal-von der Pütten, A. M., & Krämer, N. C. (2014). How design characteristics of robots
determine evaluation and uncanny valley related responses. Computers in Human
Behavior, 36, 422-439. doi: 10.1016/j.chb.2014.03.066
Rosenthal-von der Pütten, A. M., & Krämer, N. C. (2015). Individuals’ evaluations of and
attitudes towards potentially uncanny robots. International Journal of Social Robotics, 7,
799-824. doi: 10.1007/s12369-015-0321-z
Rosenthal-von der Pütten, A. M., & Weiss, A. (2015). The uncanny valley phenomenon: Does it
affect all of us? Interaction Studies, 16, 206-214. doi: 10.1075/is.16.2.07ros
MIND AND MACHINE 45
Shibata, T., & Wada, K. (2011). Robot therapy: A new approach for mental healthcare of the
elderly–a mini-review. Gerontology, 57, 378–386.
Spence, J. T., Helmreich, R., & Stapp, J. (1975). Ratings of self and peers on sex role attributes
and their relation to self-esteem and conceptions of masculinity and femininity. Journal of
Personality and Social Psychology, 32, 29-39. doi: 10.1037/h0076857
Stafford, R. Q., MacDonald, B. A., Jayawardena, C., Wegner, D. M., & Broadbent E. (2014).
Does the robot have a mind? Mind perception and attitudes towards robots predict use of
an eldercare robot. International Journal of Social Robotics, 6, 17-32. doi:
10.1007/s12369-013-0186-y
Stein, J. P., & Ohler, P. (2017). Venturing into the uncanny valley of mind—The influence of
mind attribution on the acceptance of human-like characters in a virtual reality setting.
Cognition, 160, 43-50. doi: 10.1016/j.cognition.2016.12.010
Social Security Administration (2017). Top names of the 2000s. Retrieved from
https://www.ssa.gov/OACT/babynames/decades/names2000s.html
Tanibe, T., Hashimoto, T., & Karasawa, K. (2017). We perceive a mind in a robot when we help
it. PloS One, 12(7), e0180952. doi: 10.1371/journal.pone.0180952
Tay, B., Jung, Y., & Park, T. (2014). When stereotypes meet robots: the double-edge sword of
robot gender and personality in human–robot interaction. Computers in Human Behavior,
38, 75-84. doi: 10.1016/j.chb.2014.05.014
Van de Vijver, F. J., & Leung, K. (1997). Methods and data analysis for cross-cultural research.
Thousand Oaks, CA: Sage.
Wang, S., Lilienfeld, S. O., & Rochat, P. (2015). The uncanny valley: Existence and
explanations. Review of General Psychology, 19, 393-407. doi:10.1037/gpr0000056
MIND AND MACHINE 46
Wang, X., & Krumhuber, E. G. (2018). Mind perception of robots varies with their economic
versus social function. Frontiers in Psychology, 9: 1230. doi: 10.3389/fpsyg.2018.01230
Waytz, A., Gray, K., Epley, N., & Wegner, D. M. (2010). Causes and consequences of mind
perception. Trends in Cognitive Sciences, 14, 383-388. doi: 10.1016/j.tics.2010.05.006
Wegner, D. M., & Gray, K. (2016). The mind club: Who thinks, what feels, and why it matters.
New York: Viking.
Wiggins, J. S., & Broughton, R. (1991). A geometric taxonomy of personality scales. European
Journal of Personality, 5, 343–365. doi: 10.1002/per.2410050503
Wong, B., Cronin-Golomb, A., & Neargarder, S. (2005). Patterns of visual scanning as predictors
of emotion identification in normal aging. Neuropsychology, 19, 739–749.
doi:10.1037/0894-4105.19.6.739
Złotowski, J., Proudfoot, D., Yogeeswaran, K., & Bartneck, C. (2015). Anthropomorphism:
opportunities and challenges in human–robot interaction. International Journal of Social
Robotics, 7, 347-360. doi: 10.1007/s12369-014-0267-6
MIND AND MACHINE 47
Appendix
Descriptions of Robots with Mind and Context manipulated (Experiment 2)
Note: Participants were randomly assigned to read one out of the following descriptions and to
indicate how they feel when thinking about the robot. Note that the “Carry and Grasp” and the
“Cleaning” variants were developed to increase the generalizability of the results. They did not
show any different effects. Thus, the results for these conditions were collapsed.
Tool / Unspecified Context / Carry and Grasp
Ellix is a robot with arms and hands to carry and to grasp things. Ellix assists people in their
everyday chores.
Ellix was developed with the ability to act on orders of an individual. The user can command the
robot to execute actions.
Tool / Unspecified Context / Cleaning
Ellix is a robot with arms and hands to clean things. Ellix assists people to keep their
surroundings tidy.
Ellix was developed with the ability to act on orders of an individual. The user can command the
robot to execute actions.
Tool / Nursing / Carry and Grasp
Ellix is a nursing robot with arms and hands to carry and to grasp things. Ellix assists elderly
people who are in need of support. One of its main tasks is to feed people.
Ellix was developed with the ability to act on orders of an individual. The user can command the
robot to execute actions.
Tool / Nursing / Clean
Ellix is a nursing robot with arms and hands to clean things. Ellix assists elderly people who are
in need of support. One of its main tasks is to keep the elderly and everything around them tidy.
MIND AND MACHINE 48
Ellix was developed with the ability to act on orders of an individual. The user can command the
robot to execute actions.
Agent / Unspecified Context / Carry and Grasp
Ellix is a robot with arms and hands to carry and to grasp things. Ellix assists people in their
everyday chores.
Ellix was developed with the ability of self-control, morality, memory, and emotion recognition.
Ellix has the capacity to plan ahead and to independently execute actions.
Agent / Unspecified Context / Cleaning
Ellix is a robot with arms and hands to clean things. Ellix assists people to keep their
surroundings tidy.
Ellix was developed with the ability of self-control, morality, memory, and emotion recognition.
Ellix has the capacity to plan ahead and to independently execute actions.
Agent / Nursing / Carry and Grasp
Ellix is a nursing robot with arms and hands to carry and to grasp things. Ellix assists elderly
people who are in need of support. One of its main tasks is to feed people.
Ellix was developed with the ability of self-control, morality, memory, and emotion recognition.
Ellix has the capacity to plan ahead and to independently execute actions.
Agent / Nursing / Cleaning
Ellix is a nursing robot with arms and hands to clean things. Ellix assists elderly people who are
in need of support. One of its main tasks is to keep the elderly and everything around them tidy.
Ellix was developed with the ability of self-control, morality, memory, and emotion recognition.
Ellix has the capacity to plan ahead and to independently execute actions.
MIND AND MACHINE 49
Experiencer / Unspecified Context / Carry and Grasp
Ellix is a robot with arms and hands to carry and to grasp things. Ellix assists people in their
everyday chores.
Ellix was developed with the ability to feel some form of hunger, fear, pain, pleasure, and other
emotions. Ellix is characterized by consciousness and personality.
Experiencer / Unspecified Context / Cleaning
Ellix is a robot with arms and hands to clean things. Ellix assists people to keep their
surroundings tidy.
Ellix was developed with the ability to feel some form of hunger, fear, pain, pleasure, and other
emotions. Ellix is characterized by consciousness and personality.
Experiencer / Nursing / Carry and Grasp
Ellix is a nursing robot with arms and hands to carry and to grasp things. Ellix assists elderly
people who are in need of support. One of its main tasks is to feed people.
Ellix was developed with the ability to feel some form of hunger, fear, pain, pleasure, and other
emotions. Ellix is characterized by consciousness and personality.
Experiencer / Nursing / Cleaning
Ellix is a nursing robot with arms and hands to clean things. Ellix assists elderly people who are
in need of support. One of its main tasks is to keep the elderly and everything around them tidy.
Ellix was developed with the ability to feel some form of hunger, fear, pain, pleasure, and other
emotions. Ellix is characterized by consciousness and personality.
... Mate selection theories propose that uncanniness of deviating features is caused by an evolutionary mechanism to avoid potential mates with bad fitness, and threat avoidance theory proposes that the uncanny valley is part of a mechanism to detect and avoid indicators of contagious disease [8]. Other theories of the uncanny valley focus on the perception of mind or animacy,or lack thereof [34][35][36], the detection of possibly dangerous (malevolent) intention in another human actor [9], or dehumanization of near humanlike entities [10,37]. Thus, several theories presuppose human (or animal) specificity of the uncanny valley. ...
... Some theories on the uncanny valley explain uncanniness through changes in animacy perception [10,[34][35][36][37]: stimuli may be uncanny because they straddle boundaries of animacy perception, or because they are "dehumanized" through a subtraction of animacy perception. Past research associated greeble expertise training with animacy [72], and it has been argued that greebles already look animate [73]. ...
Article
Full-text available
Humanlike entities deviating from the norm of human appearance are perceived as strange or uncanny. Explanations for the eeriness of deviating humanlike entities include ideas specific to human or animal stimuli like mate selection, avoidance of threat or disease, or dehumanization; however, deviation from highly familiar categories may provide a better explanation. Here it is tested whether experts and novices in a novel (greeble) category show different patterns of abnormality, attractiveness, and uncanniness responses to distorted and averaged greebles. Greeble-trained participants assessed the abnormality, attractiveness, uncanniness of normal, averaged, and distorted greebles and their responses were compared to participants who had not previously seen greebles. The data show that distorted greebles were more uncanny than normal greebles only in the training condition, and distorted greebles were more uncanny in the training compared to the control condition. In addition, averaged greebles were not more attractive than normal greebles regardless of condition. The results suggest uncanniness is elicited by deviations from stimulus categories of expertise rather than being a purely biological human- or animal-specific response.
... This paper also found differences between the Chinese and American participants in their perception of service robots. These results complement cross-cultural studies on human attitudes toward robots [66,67]. ...
Article
Full-text available
Previous studies on the human likeness of service robots have focused mainly on their human-like appearance and used psychological constructs to measure the outcomes of human likeness. Unlike previous studies, this study focused on the human-like behavior of the service robot and used a sociological construct, social distance, to measure the outcome of human likeness. We constructed a conceptual model, with perceived competence and warmth as mediators, based on social-identity theory. The hypotheses were tested through online experiments with 219 participants from China and 180 participants from the US. Similar results emerged for Chinese and American participants in that the high (vs. low) human-like behavior of the service robot caused the participants to have stronger perceptions of competence and warmth, both of which contributed to a smaller social distance between humans and service robots. Perceptions of competence and warmth completely mediated the positive effect of the human-like behavior of the service robot on social distance. Furthermore, Chinese participants showed higher anthropomorphism (perceived human-like behavior) and a stronger perception of warmth and smaller social distance. The perception of competence did not differ across cultures. This study provides suggestions for the human-likeness design of service robots to promote natural interaction between humans and service robots and increase human acceptance of service robots.
... Even shortterm greetings have been reported to induce positive emotional response in observers during human-robot interaction in a recent study by Fischer et al. (2019). Design and the level of human likeness of the robot faces also changes emotional response in observers (Appel et al., 2020). ...
Article
Full-text available
Functional near infrared spectroscopy (fNIRS) has been gaining increasing interest as a practical mobile functional brain imaging technology for understanding the neural correlates of social cognition and emotional processing in the human prefrontal cortex (PFC). Considering the cognitive complexity of human-robot interactions, the aim of this study was to explore the neural correlates of emotional processing of congruent and incongruent pairs of human and robot audio-visual stimuli in the human PFC with fNIRS methodology. Hemodynamic responses from the PFC region of 29 subjects were recorded with fNIRS during an experimental paradigm which consisted of auditory and visual presentation of human and robot stimuli. Distinct neural responses to human and robot stimuli were detected at the dorsolateral prefrontal cortex (DLPFC) and orbitofrontal cortex (OFC) regions. Presentation of robot voice elicited significantly less hemodynamic response than presentation of human voice in a left OFC channel. Meanwhile, processing of human faces elicited significantly higher hemodynamic activity when compared to processing of robot faces in two left DLPFC channels and a left OFC channel. Significant correlation between the hemodynamic and behavioral responses for the face-voice mismatch effect was found in the left OFC. Our results highlight the potential of fNIRS for unraveling the neural processing of human and robot audio-visual stimuli, which might enable optimization of social robot designs and contribute to elucidation of the neural processing of human and robot stimuli in the PFC in naturalistic conditions.
... Much research has endeavored to understand and improve humanrobot interactions (HRI) and human-computer interactions (HCI), often focusing on the interpersonal interactions between humans and AIs (e.g., Abubshait & Wiese, 2017;Admoni & Scassellati, 2017;Belpaeme, Kennedy, Ramachandran, Scassellati, & Tanaka, 2018;Leichtmann & Nitsch, 2020;Nomura et al., 2008;Ren & Bao, 2020). Other research has focused on humans' reactions to AIs, machine behavior, and the moral influence of AIs on humans (e.g., Appel, Izydorczyk, Weber, Mara, & Lischetzke, 2020;Köbis, Bonnefon, & Rahwan, 2021;Köbis & Mossink, 2021;Laakasuo, Palomäki, & Köbis, 2021;Naneva et al., 2020;Rahwan et al., 2019;Renier et al., 2021;Shank & DeSanti, 2018). The Computers As Social Actors (CASA) paradigm established that humans use similar social principles to interact with machines and humans (Nass et al., 1994;Reeves & Nass, 1996). ...
Article
Understanding the moral consideration of AIs as moral patients is increasingly critical given their rapid integration into daily life and the projected proliferation of advanced AIs. We present the results from a preregistered online survey with 300 U.S. Americans on the psychological predictors of the moral consideration of AIs to develop psychological theory surrounding this phenomenon. We tested an array of psychological predictors inspired by the literature on human-human and human-animal relations: perspective (future orientation, con-strual level), relational (social dominance orientation, sci-fi fan identity), expansive (human-centric norms, anthropomorphism, global citizenship, openness to experience, techno-animism), technological (affinity for technology, substratism, human-AI overlap, realistic threat, identity threat), and affective (emotions felt towards AIs). The strongest predictors were substratism, sci-fi fan identity, techno-animism, and positive emotions. We also identified three conceptual dimensions of moral consideration with an exploratory factor analysis of eight moral consideration indices drawn from prior literature: mind perception, psychological expansion, and practical consideration. Additionally, the temporal existence of AIs impacted moral consideration: AIs existing in the future were attributed more emotional capacity and more value as feeling entities than were current AIs. These results illustrate nuances in the moral consideration of AIs and lay the foundation for future research.
... For Study 2, small modifications were made to both lists to enhance plausibility ('performer' now 'musician') and confusion between personality and functionality ('helpful' now 'warm'). Additionally, participants chose their preferred name for the robot from a list of 28 names that are highly and distinctly prevalent in Black (e.g., Lakisha and Jabari) and White communities (e.g., Allison and Connor; [65]), maleand female-cued robot names from science-fiction media as a potential indicator of sex perception but not humanizing (e.g., Ava from Ex Machina and Primus from R.U.R.) as well as sex-neutral fictional robot names (e.g., Ellix; see [66]). Again, stereotypes were clustered for analysis based on social and ontological categories and valences. ...
Article
Full-text available
Increasingly humanlike social robots are praised and critiqued for implementing human social-group cues to facilitate adoption, but potentials for robot race and sex cues to provoke application of human stereotypes are yet unclear. This investigation explored discrete and intersectional stereotyping effects of robot race and sex cues in (1) live encounters and (2) mediated exposures. Participants (Study 1: N = 93; Study 2: N = 351) considered one of four humanoid social robots (2 × 2: black-/white-cued × male-/female-cued) and ascribed stereotype-indicative social roles, traits, and identifiers. Results indicated scant influence of visual/verbal cues on stereotyping, suggesting those cues do not prompt people to categorize robots as they do humans. Instead, consistently ascribing helpfulness, thinker, and servant attributes suggests stereotyping robots as robots rather than according to human categories—a pattern more pronounced when people were primed to think about human stereotypes. We infer that humanoid robots are seen as a distinct kind, apart from humans—so distinctly apart that exploratory analyses demonstrated that even self-similar robots are considered more different than are dissimilar humans.
... In contrast, AI systems with low levels of agency require the constant input of human commands or guidance. As studies suggest, agency can influence the perception of a non-human agent (Appel et al., 2020;Brink and Wellman, 2020;Zafari and Koeszegi, 2020). In Human-Robot Interaction research, agency has been linked to increased anthropomorphism (the tendency to infer human-like traits to non-human entities; Nowak and Biocca, 2003;Epley et al., 2008;Crowell et al., 2019). ...
Article
Full-text available
Artificial Intelligence (AI) is supposed to perform tasks autonomously, make competent decisions, and interact socially with people. From a psychological perspective, AI can thus be expected to impact users’ three Basic Psychological Needs (BPNs), namely (i) autonomy, (ii) competence, and (iii) relatedness to others. While research highlights the fulfillment of these needs as central to human motivation and well-being, their role in the acceptance of AI applications has hitherto received little consideration. Addressing this research gap, our study examined the influence of BPN Satisfaction on Intention to Use (ITU) an AI assistant for personal banking. In a 2×2 factorial online experiment, 282 participants (154 males, 126 females, two non-binary participants) watched a video of an AI finance coach with a female or male synthetic voice that exhibited either high or low agency (i.e., capacity for self-control). In combination, these factors resulted either in AI assistants conforming to traditional gender stereotypes (e.g., low-agency female) or in non-conforming conditions (e.g., high-agency female). Although the experimental manipulations had no significant influence on participants’ relatedness and competence satisfaction, a strong effect on autonomy satisfaction was found. As further analyses revealed, this effect was attributable only to male participants, who felt their autonomy need significantly more satisfied by the low-agency female assistant, consistent with stereotypical images of women, than by the high-agency female assistant. A significant indirect effects model showed that the greater autonomy satisfaction that men, unlike women, experienced from the low-agency female assistant led to higher ITU. The findings are discussed in terms of their practical relevance and the risk of reproducing traditional gender stereotypes through technology design.
Chapter
Socially active humanoid robots (SAHRs) are designed to communicate and interact with humans in humancentric environment using speech, movements, gestures, or facial expressions to communicate with their users following some set of social behavior while providing their assistance. Just like humans interact in an adaptive manner with others by changing their speech, tone, and body language intuitively, such type of adaptive behavior can be developed in SAHRs to get a human-like rich interaction capabilities. Therefore, a lot of research work and studies are going on to replicate various behavioral aspects of humans into SAHRs, so that human-robot interaction can be improved further. Besides interacting with humans, humanoid robot should be able to perform the assigned tasks remotely and also in real time with better accuracy. Thus, these social robots designed can be used in a diversified field of applications like education, healthcare, entertainment, communication, constructions, medical, collaborations, hazard management systems, etc.
Chapter
We analyze socio-psychological process wins and losses in relation to Human-AI collaborative performance. The reported heterogeneous effects of the body of literature on group performance in relation to structural and processual determinants are briefly summarized. Based on this, two of the most relevant socio-psychological aspects of Human-AI collaborative performance are highlighted: Accuracy of the shared mental model and fulfillment of basic human needs. The paper concludes by proposing an empirical and experimental research program that addresses under-researched socio-psychological phenomena in Human-AI collaboration that can be held accountable for process wins and losses.KeywordsHuman-AI interactionSocio-psychologyCollaborative performance
Article
Full-text available
Many societies are on the brink of a robotic era. In the near future, various autonomous computer systems are expected to be part of many people’s daily lives. Because attitudes influence the adoption of new technologies, we studied the attitudes towards robots in the European Union between 2012 and 2017. Using representative samples from 27 countries (three waves, total N = 80,396), these analyses showed that, within five years, public opinions regarding robots exhibited a marked negative trend. Respondents became more cautious towards the use of robots. This tendency was particularly strong for robots at the workplace, which are, despite the drop, still more positively evaluated than robots performing surgeries or autonomous cars. Attitudes were more positive among men and people in white-collar jobs. Moreover, countries with a larger share of older citizens evaluated robotic assistance more favorably. In general, these results highlight increasing reservations towards autonomous robotic systems in Europe.
Article
Full-text available
While robots were traditionally built to achieve economic efficiency and financial profits, their roles are likely to change in the future with the aim to provide social support and companionship. In this research, we examined whether the robot's proposed function (social vs. economic) impacts judgments of mind and moral treatment. Studies 1a and 1b demonstrated that robots with social function were perceived to possess greater ability for emotional experience, but not cognition, compared to those with economic function and whose function was not mentioned explicitly. Study 2 replicated this finding and further showed that low economic value reduced ascriptions of cognitive capacity, whereas high social value resulted in increased emotion perception. In Study 3, robots with high social value were more likely to be afforded protection from harm, and such effect was related to levels of ascribed emotional experience. Together, the findings demonstrate a dissociation between function type (social vs. economic) and ascribed mind (emotion vs. cognition). In addition, the two types of functions exert asymmetric influences on the moral treatment of robots. Theoretical and practical implications for the field of social psychology and human-computer interaction are discussed.
Conference Paper
Full-text available
Most robots currently being sold or developed are either stylized with white material or have a metallic appearance. In this research we used the shooter bias paradigm and several questionnaires to investigate if people automatically identify robots as being racialized, such that we might say that some robots are 'White' while others are 'Asian', or 'Black'. To do so, we conducted an extended replication of the classic social psychological shooter bias paradigm using robot stimuli to explore whether effects known from human-human intergroup experiments would generalize to robots that were racialized as Black and White. Reaction-time based measures revealed that participants demonstrated 'shooter-bias' toward both Black people and robot racialized as Black. Participants were also willing to attribute a race to the robots depending on their racialization and demonstrated a high degree of inter-subject agreement when it came to these attributions.
Article
Full-text available
People sometimes perceive a mind in inorganic entities like robots. Psychological research has shown that mind perception correlates with moral judgments and that immoral behaviors (i.e., intentional harm) facilitate mind perception toward otherwise mindless victims. We conducted a vignette experiment (N = 129; Mage = 21.8 ± 6.0 years) concerning human-robot interactions and extended previous research’s results in two ways. First, mind perception toward the robot was facilitated when it received a benevolent behavior, although only when participants took the perspective of an actor. Second, imagining a benevolent interaction led to more positive attitudes toward the robot, and this effect was mediated by mind perception. These results help predict what people’s reactions in future human-robot interactions would be like, and have implications for how to design future social rules about the treatment of robots. © 2017 Tanibe et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Article
Full-text available
People vary in the extent to which they report fear toward robots, especially when they perceive that the robot is autonomous or has artificial intelligence. This research examines a specific form of sociological fear, which we name as fear of autonomous robots and artificial intelligence (FARAI). This fear may serve to affect how people will respond to and interact with robots. Applying data from a nationally representative dataset with probability sampling (N = 1541), research questions examine (1) the extent and frequency of FARAI, (2) demographic and media exposure predictors, and (3) correlates with other types of fear (i.e., loneliness, drones, and unemployment). A latent class analysis reveals that approximately 26% of participants reported experiencing a heightened level of FARAI. Demographic analyses show that FARAI is connected to participant sex, age, education, and household income; albeit these effects were small. Media exposure to science fiction predicts FARAI above and beyond the demographic variables. Correlational results indicate that FARAI is associated with other types of fear, including loneliness, becoming unemployed, and drone use. In sum, these findings render a much needed glimpse and update regarding how much individuals fear robots and artificial intelligence.
Article
We analyze for the first time the economic contributions of modern industrial robots, which are flexible, versatile, and autonomous machines. We use novel panel data on robot adoption within industries in 17 countries from 1993-2007, and new instrumental variables that rely on robots’ comparative advantage in specific tasks. Our findings suggest that increased robot use contributed approximately 0.37 percentage points to annual labor productivity growth, while at the same time raising total factor productivity and lowering output prices. Our estimates also suggest that robots did not significantly reduce total employment, although they did reduce low-skilled workers’ employment share
Conference Paper
We propose a new theoretical framework assuming that embodiment effects in HAI and HRI are mediated by users' perceptions of an artificial entity's body-related capabilities. To enable the application of our framework to foster more theoretical-driven research, we developed a new self-report measurement that assesses bodilyrelated perceptions of the embodiment and corporeality - which we reveal as not being a binary characteristic of artificial entities. For the development and validation of the new scale we conducted two surveys and one video-based experiment. Exploratory factor analysis reveal a four-factorial solution with good reliability (Study 2, n = 442), which was confirmed via confirmatory factor analysis (Study 3, n = 260). In addition, we present first insights into the explanatory power of the scale: We reveal that humans? perceptions of an artificial entity's capabilities vary between virtual and physical embodiments, and that the evaluation of the artificial counterpart can be explained through the perceived capabilities. Practical applications and future research lines are discussed.
Article
There are many differences between men and women. To some extent, these are captured in the stereotypical images of these groups. Stereotypes about the way men and women think and behave are widely shared, suggesting a kernel of truth. However, stereotypical expectations not only reflect existing differences, but also impact the way men and women define themselves and are treated by others. This article reviews evidence on the nature and content of gender stereotypes and considers how these relate to gender differences in important life outcomes. Empirical studies show that gender stereotypes affect the way people attend to, interpret, and remember information about themselves and others. Considering the cognitive and motivational functions of gender stereotypes helps us understand their impact on implicit beliefs and communications about men and women. Knowledge of the literature on this subject can benefit the fair judgment of individuals in situations where gender stereotypes are likely to play a role. Expected final online publication date for the Annual Review of Psychology Volume 69 is January 4, 2018. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.
Article
Aims: To explore the concept of futurism and the emergence of robotics in relation to the fundamentals of care, highlighting how nurses need a more anticipatory and contemporary position towards technology to maintain relevance in the future. Background: The future of nursing in Western countries will soon be linked with the emergence of robotics for efficient and cost effective provision of fundamental care. Their emergence and roles with care of the body and more broadly assisting people with their daily living activities has enormous implications for the profession and health care. Despite this importance, how nursing understands and will respond to technological trends and developments is insufficiently reflected in the professions discourse. Design: A discursive article METHODS: Literature from nursing fundamentals of care / fundamental care, information science, technology, humanities and philosophy informed the arguments in this paper. Conclusions: This paper examines the intersection of futurism and the fundamentals of care and how adopting an anticipatory and post-human perspective towards technological-care integration is necessary amidst a robot revolution in the techno-era. Relevance to clinical practice: Nurses are currently challenged to understand, prioritize and deliver fundamental care. Health systems are challenged by a lack of care predicated by shortfalls in skilled staff and deficiencies in their mobilization. Both challenges can be compounded or alleviated by further integration of technology, but to maximize benefit requires forethought and understanding. This article can help open needed dialogue around planning for the future and is a call to action for the nursing profession to conceptualize their position on exponential technological growth and fundamental care provision. This article is protected by copyright. All rights reserved.