Content uploaded by Julie Carpenter
Author content
All content in this area was uploaded by Julie Carpenter on Jan 27, 2017
Content may be subject to copyright.
Int J Soc Robot (2009) 1: 261–265
DOI 10.1007/s12369-009-0016-4
ORIGINAL PAPER
Gender Representation and Humanoid Robots Designed
for Domestic Use
Julie Carpenter ·Joan M. Davis ·Norah Erwin-Stewart ·Tiffany R. Lee ·
John D. Bransford ·Nancy Vye
Accepted: 15 February 2009 / Published online: 19 March 2009
© Springer Science & Business Media BV 2009
Abstract Humanoid robots’ appearance and behavior pro-
vide social cues about their purpose and abilities. However,
little is known about how a robot’s gender representation
will affect users in everyday home use scenarios. This paper
presents the results of a study exploring people’s expecta-
tions of humanoid robots, or androids, designed for home
use. Results of this study demonstrated participants’ will-
ingness to attribute human roles and tasks to an android, al-
though they did not indicate an overall preference for the
robot as a social actor. In addition, following the viewing
of video stimulus featuring human-robot interactions, robot
gender issues surfaced during open-ended interviews.
Keywords Android ·Design ·Expectations ·Gender ·
Human ·Humanoid ·Human-robot interaction ·
Interaction ·Robot ·Social ·Stereotypes
J. Carpenter (!)·J.M. Davis ·N. Erwin-Stewart ·T.R. L ee ·
J.D. Bransford ·N. Vye
College of Education/Educational Psychology,
University of Washington, Seattle, WA, USA
e-mail: julie4@u.washington.edu
J.M. Davis
e-mail: joand@u.washington.edu
N. Erwin-Stewart
e-mail: nes2@u.washington.edu
T.R . Lee
e-mail: tlee13@u.washington.edu
J.D. Bransford
e-mail: bransj@u.washington.edu
N. Vye
e-mail: nancyvye@u.washington.edu
1 Introduction
We are entering an age where living and working with ro-
bots imbued with functionality and socialness is a reality.
Designing humanoid robots for home use requires under-
standing how people currently feel about robots. One chal-
lenge to be considered when designing robots for the home
is the representation—intentional or accidental—of the ro-
bot’s gender. For the purpose of this paper, the discussion of
robots centers on androids, or robots with an aspect of hu-
manlikeness portrayed through appearance, behavior, con-
text of use, speech/voice, or a combination of these charac-
teristics.
Gender presentations affect interactions among humans.
Central questions for robotics development and design are:
How will gendered robots influence everyday use? Con-
versely, how will a user’s gender affect his or her inter-
actions with a gendered robot? Humans often assign gen-
der to inanimate objects; for example, cars and boats are
commonly female in Western culture, and dolls are typi-
cally designed to be clearly male or female. Some humanoid
robots are designed with obvious gender orientation, such
as Repliee (Osaka University and Kokoro, Ltd.) while oth-
ers are less distinct, such as Robovie (ATR Laboratories).
Morphology, overall appearance, behavior, voice, gaze, ges-
ture, functionality, context and cultural expectations all im-
pact the user’s concept of gendered robots. While it is likely
that robots with distinct gender appearance and actions will
make human-robot communication more effective in some
scenarios, the absence of gender in other robot designs may
work better by causing less distraction for the user [8].
Potential research questions include:
•What traits and characteristics do users identify as male,
female or neutral in robots?
262 Int J Soc Robot (2009) 1: 261–265
•What level of gender identification is necessary in overall
robot design for the user’s ease of communication?
•What is the effect of the user’s gender and their willing-
ness to engage with the robot (gendered or neutral)?
•What context and use scenarios require robot gender as-
signment for effective communication?
•What cultural stereotypes do robots mimic or create by
being gendered?
•How is robot gender identified through design?
This paper is a theoretical inquiry about the concepts of gen-
der and humanoid robots based in human-robot interaction,
gender studies, cognitive science and design theories, and
will present the findings of an exploratory study examining
user expectations of social robots developed for the home.
We use the findings of our recent study about user expec-
tations of humanoid domestic robots as a springboard for
discussion.
Robots are machines without an organic gender or sex-
uality. However, robots are also rooted in our cultural ex-
pectations as servant, enemy, friend, pet, slave, toy, com-
panion, and other roles presented in popular mythology
[3,4]. These roles are loaded with user stereotypes and
related expected signifiers. Humanoid robots offer another
unique set of issues for the user, who must recognize that a
mobile thing with some humanlike morphology, behaviors,
and of varying intelligence and autonomy is not natural, but
human-made. Haraway [6] refers to this perceptual dilemma
as the “distinction between animal-human (organism) and
machine” (p. 152). Highly humanlike robot development is
on the cusp of becoming invisible machinery; mechanical,
yet through humanlike appearance or behavior, triggering a
sense of perceived humanness in the user to the point where
they respond to the thing as something alive and natural.
Reeves and Nass [9]“computersassocialactors”
(CASA) theory explains how humans unconsciously ascribe
agency, personality, and intentionality to computer-mediated
technologies. In humanoid robots, the combination of hu-
man appearance and user-projected human intentionality
creates a complex mixture of attachment-related responses
for users of android products in which the drive to respond to
the android as if it was human is at odds with the realization
that the android is a machine.
This study incorporated a questionnaire to examine peo-
ple’s perceptions of a robot’s overall appearance and behav-
iors after viewing videos of each robot. Previous studies [4]
have used questionnaires as a tool to investigate user atti-
tudes towards domestic robots.
Through a semi-structured interview, participants were
also de-briefed about their opinions of the robots and in-
teractions, and their responses were recorded through obser-
vation and interview notes. Field notes from the interview
portions of these sessions were analyzed using qualitative
data analysis techniques [1,10].
2 Method
Nineteen participants, all students at the University of Wash-
ington, completed the study. The mean participant age was
22.8 years (SD =4.9); 53% were women (n=10) and
47% were men (n=9). Two video clips of separate ro-
bots, Repliee and Robovie, were the stimulus for discussion.
Each robot was contextually introduced as “designed to be a
friend or member of the family, offering assistance through
social interaction.” Participants first viewed the Repliee and
the Robovie videos, shown in random order. After watch-
ing the videos, participants completed a Likert-scale survey
with three questions about their perception of each robot as
appearing machine- to human-like, friendly/unfriendly and
the comfort level of having that robot in their homes. Par-
ticipants completed the same questionnaire for each video.
They next completed a short demographics survey.
A semi-structured interview protocol was used for the
post-study interviews, and two people manually coded the
entire data set using Thematic Analysis and an inductive
approach [2]. Here, a theme captures something important
about the data and represents some level of patterned re-
sponse or meaning within the data set. The key themes de-
scribed here (a) demonstrate something important related to
gender representation in robots and (b) characterize a preva-
lent response from participants. Prevalence was counted in
terms of the number of different speakers who articulated
the theme across the entire data set.
3 Materials
The questionnaire asked about humanlikeness vs. machine-
likeness, friendliness and their comfort level of having the
robot in the home.
Two video clips were used as stimuli for discussion. Each
of the two videos fit the following criteria:
•Time was approximately three minutes long.
•Presented only one robot per video.
•Featured robot was humanoid.
•Demonstrated a human-robot language interaction (in
English).
•Showed the robot’s entire body at least once in the video.
The robots used in the videos were, as shown in Fig. 1
(Repliee Q2, Osaka University and Kokoro Co., Ltd.)
and Fig. 2(Robovie, ATR) below.
The content of each video varied. Robovie interacted
with one primary human actor in an extended scene, con-
versing about the weather, asking to play games and en-
gaging in simple conversation with the actor. In the video,
Robovie moved on its wheeled base and demonstrated its
arm and head movement clearly. In the other video, Repliee
Int J Soc Robot (2009) 1: 261–265 263
Fig. 1 Repliee
Fig. 2 Robovie
acted as a receptionist in a brief opening scene, offering di-
rections to an actor; subsequent scenes showed the robot act-
ing as a journalist, interviewing two different actors sepa-
rately. Repliee appeared stationary in all scenes, moving its
head, arms and torso during interactions.
4Findings
Aone-wayrepeatedmeasuresGeneralLinearModelanaly-
sis of variance (ANOVA) test was conducted to determine
the differences between participant (n=19) responses to
three questions in the post-video questionnaire. Each ques-
tion was analyzed as a within-subjects factor (response to
Robovie vs. response to Repliee). A significant difference
was found between participant responses to the human-
and machinelikeness question F(1,18)=110.250, p<.05.
No significant difference was found for questions regarding
friendliness or comfort level of the robot in the home.
However, the comfort level approached significance with
gender as the between-subjects variable (p=.052). Women
felt less comfortable having a robot in the home than did
men. For Robovie, the mean rating from women was 2.10
(SD =.738), while the mean rating from men was 2.56
(SD =1.424). For Repliee, the mean rating from women
was 1.90 (SD =.994), while the mean rating from men was
2.67 (SD =.707). Although these differences are not sig-
nificant, an increase in sample size could possibly yield a
significant result.
Emergent themes from the interviews demonstrated hu-
manoid robots in the home:
•Trigger parallel human gender associations, especially
about appearance and socialness.
•Elicit associations of robot use with menial tasks.
•Challenge existing concepts about robots and social inter-
action.
We will present our findings about user expectations and
gender issues surrounding domestic robots with represen-
tative comments from cases to illustrate major themes and
areas of interest.
Pretty and Comforting: Human Cues Map to Robots
The majority of comments related to robot gender were pre-
dictably about Repliee, who is purposely very humanlike in
appearance and is modeled after an adult woman. As ex-
pected, humanlike indicators in the design of the robot such
as voice, clothing and morphology were the key triggers for
discussing gender-related issues, and participants revealed
preferences for a female robot for in-home use.
“She’s pretty because she looks really human and she’s
slim and she has hair. She just has the human girl features
about her.” (Repliee)
“That one looks better, a lot better. It looks like a woman.
That helps me feel better.” (Repliee)
“Well, it’s female, so that’s a positive. I’d have to say that
it’s just because of the whole idea of the woman being at-
tractive. Well, attractiveness...I probably shouldn’t say that
[blushes]. This is where safety comes into play. The femi-
nine form is typified as being weak or fragile in some form,
but really inviting and warm and more interactive. Whereas
if it were a male robot and masculine design, then there’s a
safety issue of, ‘OK. I gotta protect myself possibly’... you
don’t know [laughs]. No matter how simplistic or mechan-
ical it is, a male robot just seems more... Idon’twantto
say competitive, but... I’m thinking of the word... Amale
robot would be not competitive, but it would be a challenge
in some way.” (Repliee)
“I guess for me, when I’m around girls, I feel like it’s
special. When you have a girl at home, it’s someone you
love. Guys can love each other, I guess.” (Repliee)
“It’s like a child robot. It is quite fun to play with. It
would be fun for children to play with because she—I as-
sume it’s a lady robot—she likes to say some childish things.
Quite superficial and she likes to play, ‘touch me,’ ‘hug me,’
‘shake my hand.”’ (Robovie)
Servant or Assistant: Expected Robot Roles and Functions
Both robots elicited similar themes in response to the in-
terview question “What would you like this robot to do
264 Int J Soc Robot (2009) 1: 261–265
for you in your home?” Most participants identified me-
nial tasks as their expectation of robot duties. Chores men-
tioned most often included such tasks as washing dishes, do-
ing laundry, general cleaning, ironing clothes and heavy lift-
ing of objects. In addition, many participants named specific
roles they envisioned the robots fulfilling, such as recep-
tionist, librarian or doctor’s assistant. These responses are
especially noteworthy since the robots’ use was framed as
“in the home,” yet participants most often named job roles
that are not normally done in a domestic setting. Partici-
pants were conflicted about roles and functions for the ro-
bots that included more intimate or prolonged social con-
tact, such as childcare and companionship. Frequently, par-
ticipants fluctuated in their opinion, often within sentences,
about whether they would want a robot to take care of their
child or act in a purely social way.
“I wouldn’t want her to answer the phone or interact with
children. Anything other than that, cleaning or getting the
mail would be fine. Anything that doesn’t involve interac-
tion.” (Repliee)
“I guess... have it do chores like vacuuming, doing
dishes, and cleaning the bathroom. I wouldn’t have it take
care of kids. I wouldn’t have it answer the phone. I wouldn’t
have it interact with other people, other than adults that knew
they were interacting with a robot.” (Robovie)
Don’t Touch Me: Robots and Social Interaction
As stated previously, the majority of participants indicated
they would not like to see either robot in their home with a
social purpose. Identified concerns included potential con-
fusion about whether a robot was a machine or organic (es-
pecially with children), the robot touching the human in a
social way, and the false pretense of a machine expressing
humanlike emotions.
“I don’t want any kids that are in my home—whether or
not they’re mine—to really be interacting with someone like
that because they might mistake an actual human for a robot
like that. They’ll be interacting with a piece of machinery
with the idea that it’s someone who’s real.” (Repliee)
“I didn’t like how it wanted physical interaction...Idon’t
want to see it interact with me personally like that. If I was
married, I wouldn’t expect that kind of interaction, whereas
it’s just kind of like, ‘Caress or touch me.’ If it’s a handshake,
that’s fine.” (Robovie)
“Some of the stuff was creepy—like ‘touch me’ and ‘hug
me.’ That struck me as odd. I can’t imagine that the act of
touching a robot would have some sort of inherent or not
pre-programmed response for the robot. Someone’s asking
me to do something that’s not for them, but it’s for me, and
they’re not getting anything out of it... that makes me un-
easy.” (Robovie)
5 Discussion
Before discussing the implications of these findings, it is
important to note several aspects of this study. This re-
search was based on participants’ impressions after watch-
ing videos rather than from more prolonged interactions,
which might result in participants reframing their initial re-
sponses. And, although the videos used in this study fit the
criteria listed under Materials,theywerenotnecessarily
crafted to demonstrate the robot in either a social or service
role. In addition, Repliee was more clearly intended to be a
female robot while Robovie’s gender cues were vague.
Although most participants said in interviews they did
not want a robot for social purposes, conflicting opinions
emerged from individuals in some cases, bearing further
investigation. Some aspects of socialness, such as speech,
were preferred for ease of use, while specific behaviors, such
as the emotion demonstrated by Robovie, upset people.
People’s natural tendencies to categorize others [5]—
including humanoid robots—are demonstrated in the find-
ings of this research. Without specific prompting, partici-
pants frequently commented on each robot’s perceived gen-
der, race or nationality, and social standing within the house-
hold. Unlike disembodied computers, robots will be more
difficult to claim as simply “feminine, masculine or beyond
gender” [11], p. 10. The very nature of a robot with a hu-
manoid form mixed with implied (or real) functionality, the
social characteristics of the robot and the context of use are
combined with the individual user’s cultural expectations
and will encourage a set of interaction norms. Von Zoo-
nen concluded [11] that as technologies are integrated into
everyday domestic settings, a seamless blending of social
structures, symbolic representations of gender and the users’
individual identity hails the positioning of men and women.
Levy [7] states that “humans attribute others with having
minds” (p. 53), and in fact, over-attribute feelings and in-
tentions, such as in the case of the anthropomorphism of
pets. Our participant examples shown in the themes above
demonstrate the tendencies of both (a) relying on human-
human norms to explain humanoid robots and (b) facile
anthropomorphism among our study participants when dis-
cussing both robots. In addition, the robots are clearly an
“other,” even if that other is ill-defined in their own minds
without real experience with a humanoid robot. It is crucial,
therefore, that all of these elements are taken into account
when designing and developing humanoid robots because
they will directly affect efficiency, ease of use and pleasure
in any human-robot interaction. In addition, the gendering
of humanoid robots, whether with intentional design cues or
not, will likely perpetuate aspects of certain human-human
roles and the ideologies that go with them.
Acknowledgements This study was funded by the LIFE Center in
the College of Education, University of Washington. The authors of
Int J Soc Robot (2009) 1: 261–265 265
this paper wish to thank Drs. Hiroshi Ishiguro and Takashi Minato
from Osaka University (Repliee) and Dr. Takayuki Kanda from ATR
(Robovie), who generously contributed video used for this study. Fi-
nally, special thanks to Elizabeth A. Sanders, University of Washing-
ton, for her help and support.
References
1. Boyatzis R (1998) Transforming qualitative information: The-
matic analysis and code development. Sage, Thousand Oaks
2. Braun V, Clarke V (2006) Using thematic analysis in psychology.
Qual Res Psychol 3:77–101
3. Carpenter J, Eliot M, Schultheis D (2006) Machine or friend: Un-
derstanding users’ preferences for and expectations of a humanoid
robot companion. In: Proceedings of 2006 design emotion confer-
ence, Gothenburg, Sweden, 27–29 September 2006
4. Dautenhahn K, Woods S, Kaouri C, Walters M, Koay KL, Werry I
(2005) What is a robot companion, friend, assistant or butler? In:
Proceedings of IEEE IROS, Edmonton, Canada, pp 1488–1493
5. Goffman E (1959) The presentation of self in everyday life. Dou-
bleday, New York
6. Haraway DJ (1991) Simians, cyborgs, and women: The reinven-
tion of nature. Routledge, New York
7. Levy D (2007) Intimate relationships with artificial partners. Un-
published Doctoral Dissertation, Universiteit Maastricht, Maas-
tricht, The Netherlands
8. Powers A, Kramer ADI, Lim S, Kuo J, Lee S, Kiesler S (2005)
Eliciting information from people with a gendered humanoid ro-
bot. In: Proceedings of 2005 IEEE international workshop on ro-
bot and human interactive communication (RO-MAN 2005)
9. Reeves B, Nass C (1996) The media equation: How people treat
computers, television, and new media like real people and places.
Cambridge University Press, Cambridge
10. Ryan GW, Bernard HR (2003) Techniques to identify themes.
Field Methods 15(1):85–109
11. Van Zoonen L (2002) Gendering the Internet: claims, controver-
sies, and culture. Eur J Commun 17:5–23
