Living with a Vacuum Cleaning Robot

Abstract

Little is known about the usage, adoption process and long-term effects of domestic service robots in people’s homes. We investigated the usage, acceptance and process of adoption of a vacuum cleaning robot in nine households by means of a six month ethnographic study. Our major goals were to explore how the robot was used and integrated into daily practices, whether it was adopted in a durable way, and how it impacted its environment. We studied people’s perception of the robot and how it evolved over time, kept track of daily routines, the usage patterns of cleaning tools, and social activities related to the robot. We integrated our results in an existing framework for domestic robot adoption and outlined similarities and differences to it. Finally, we identified several factors that promote or hinder the process of adopting a domestic service robot and make suggestions to further improve human-robot interactions and the design of functional home robots toward long-term acceptance.

Full text

Int J Soc Robot (2013) 5:389–408
DOI 10.1007/s12369-013-0190-2
Living with a Vacuum Cleaning Robot
A 6-month Ethnographic Study
Julia Fink ·Valérie Bauwens ·Frédéric Kaplan ·
Pierre Dillenbourg
Accepted: 2 May 2013 / Published online: 13 June 2013
©SpringerScience+BusinessMediaDordrecht2013
Abstract Little is known about the usage, adoption process
and long-term effects of domestic service robots in people’s
homes. We investigated the usage, acceptance and process of
adoption of a vacuum cleaning robot in nine households by
means of a six month ethnographic study. Our major goals
were to explore how the robot was used and integrated into
daily practices, whether it was adopted in a durable way,
and how it impacted its environment. We studied people’s
perception of the robot and how it evolved over time, kept
track of daily routines, the usage patterns of cleaning tools,
and social activities related to the robot. We integrated our
results in an existing framework for domestic robot adoption
and outlined similarities and differences to it. Finally, we
identified several factors that promote or hinder the process
of adopting a domestic service robot and make suggestions
to further improve human-robot interactions and the design
of functional home robots toward long-term acceptance.
Keywords Domestic robots ·Ethnographic study ·
Human-Robot interaction ·Social issues in robotics ·
Long-term interaction ·Adoption of technology
J. Fink (!)·V. B a u we n s ·F. K a plan ·P. Dillenbourg
Computer-Human Interaction in Learning and Instruction
(CHILI), Ecole Polytechnique Fédérale de Lausanne (EPFL),
Rolex Learning Center, Station 20, 1015 Lausanne, Switzerland
e-mail: julia.fink@epfl.ch
V. Ba u w en s
e-mail: valerie.bauwens@epfl.ch
F. Ka p lan
e-mail: frederic.kaplan@epfl.ch
P. Dillenbourg
e-mail: pierre.dillenbourg@epfl.ch
1 Introduction
With domestic service robots entering people’s homes, it
becomes increasingly important to understand the process
of long-term adoption of these devices, taking into account
user needs, characteristics of the home, and the (social) im-
pact that these devices can have on the home ecosystem [11].
Still little is know about these aspects, despite the fact that
several million units of domestic service robots have already
been sold (mostly vacuum cleaning robots). To date, only
few studies investigated Human-Robot Interaction (HRI) in
homes over a longer period of time. However, these kind of
real-world studies can provide interesting insights into us-
age and experience from a user point of view, and help to
identify factors promoting and hindering adoption. By un-
derstanding better how people use and adopt domestic ser-
vice robots, the functionality, user interaction and general
design could be improved toward the goal of making domes-
tic service robots useful, usable, and acceptable everyday
tools. Studies investigating HRI during a longer period of
time are also crucial, as it has been recognized that novelty
plays a role and that interactions with technologies change
over time and with growing experience [1,8,11,18,30,33].
It is our aim to study the adoption of domestic robots
and explore niches for this technology, addressing practical
problems in daily lives of humans. As a starting point, we
took the possibility of using a commercially available vac-
uum cleaning robot (iRobot’s Roomba) for a 6-month ethno-
graphic study with nine households using the robot. To the
best of our knowledge, no such study has been conducted in
Europe so far. Our study contributes to a better understand-
ing of long-term implications in HRI in two ways.
First, findings contribute as itself to the limited number
of long-term studies on HRI. We provide detailed insights
into what happens when a functional robot like the Roomba
is deployed in a home, with respect to:
Author's personal copy

390 Int J Soc Robot (2013) 5:389–408
•people’s cleaning routine and how it was affected by the
introduction of the vacuuming robot (Sect. 4.1)
•how the physical and social context of the home impacts
the usage of and interaction with the domestic service
robot (Sect. 4.2)
•people’s understanding of robots in general, their expec-
tations and perception of a domestic service robot, and
social activities related to the robot (Sect. 4.3)
•initial reactions to the robot and its adoption, along with
factors related to this process (Sect. 5)
•implications for the acceptance of robots in domestic en-
vironments (Sect. 6)
Second, our work adds in a meaningful way to an ini-
tial framework of long-term acceptance of robots in homes,
namely the ‘Domestic Robot Ecology’ (DRE) [33], along
which we analyze and integrate our observations. The DRE
is based a long-term field study with 30 households in the
US. We carried out a similar study, with similar methodol-
ogy, but in a culturally different context using different mod-
els of the same domestic service robot. By this, we provide
new data to the framework and are able to show major simi-
larities and interesting qualitative differences. Consequently,
our work contributes also to making the DRE framework
more robust and holistic.
Overall, the novelty of our work lies not in the experi-
mental setup but in the way the data is analyzed and inte-
grated in an existing framework of domestic robot adoption.
2 Related Work
Roomba is one of the most widely deployed vacuum clean-
ing robots and has been used in a series of studies focusing
on user needs [31], design and personalization of home tech-
nologies [32], demographic profiles of robot owners [29],
social implications of domestic robots [10,11,28], and
long-term adoption [30,33]. In the following we first give
an overview of previous work that explored usage, user pro-
files, and the adoption of domestic service robots, and then
describe how our study can contribute to this.
People’s Expectations of Domestic Robots Sung et al. and
Forlizzi et al. have conducted interviews [31], surveys [29],
and ethnographic research [11] with Roomba owners. They
found that within their concept of robots, people distin-
guished between ‘Roomba’ and the great mass of ‘other
robots’ [11]. While people’s expectations of robots in gen-
eral were high, they were rather low for the Roomba and
its practical functionality [11]. Nevertheless, participants ex-
pected a domestic robot to be intelligent and able to learn.
The important point for our work is that, people’s specific
expectations of Roomba have been shown to shape their ini-
tial experience with it which is in turn important for adop-
tion. This makes it essential to assess participants’ attitudes
first in order to better understand their reaction.
User Profiles and Usage of Domestic Service Robots Sung
et al. [29] assessed demographic profiles and usage pat-
terns in a survey among more than 350 Roomba owners.
Data revealed that Roomba users were equally likely men or
women, and tended to be younger with higher levels of ed-
ucation or technical backgrounds [29]. Half of the Roomba
owners who took part in the survey lived with one or more
pets and those with children at home expressed greater sat-
isfaction with the robot’s performance. This suggests that
the composition of a household influences how people use
and experience a domestic robot [10]. Concerning the usage
of vacuum cleaning robots, Kim et al. discovered a discrep-
ancy between the cleaning path participants in Korea used
when manually vacuuming, and the paths chosen by vac-
uum cleaning robots [19]. Consequently, the authors sug-
gest to adapt the robot’s path planning to the specific area
of a home and the according user needs, as the user tends to
use area specific methods for vacuuming.
Domestic Robots can Impact Routines and Create Social
Dynamics In terms of the impact on housekeeping, it has
been shown that in comparison to a traditional vacuum
cleaner (VC), the Roomba changed people’s cleaning activ-
ities and how they used other tools [10]. The robot affected
cleaning routines, by increasing cleaning activities, encour-
aging multi-tasking, collaboration, and making cleaning a
concern for everyone in the home [11,33]. On one hand,
the Roomba was described as “a tool to improve the clean-
liness of the home” [11,33], though the majority of Roomba
owners reported doing extra cleaning with the manual vac-
uum cleaner [29]. On the other hand, the robot enhanced
(social) activities other than cleaning: It led people to adjust
the physical environment or to create and remove obstacles.
In both ethnographic studies with the Roomba, people en-
gaged socially with the robot, such as by giving it a name,
talking directly to it, or ascribing intentions and personality
traits [11,28,33].1
Sung et al. further observed that people customized their
Roomba (e.g. with drawing or putting stickers on it) to ex-
press its ‘identity’ (or gender) and show its value to the
household [32]. Intimacy and positive emotional attach-
ment, such as assigning an identity to the robot, led to
greater acceptance of the product in general and the per-
ceived usability [37]. These social aspects and their positive
1Social implications of home technologies and domestic robots, such
as intimacy, affective quality, and emotional attachment have been
studied in more detail [10,22,28,37].
Author's personal copy

Int J Soc Robot (2013) 5:389–408 391
effect have also been mentioned by others and in respect to
different kinds of robots and technologies [5,8,28]. It is
suggested that social factors in the usage of household tech-
nology are crucial to understand usage patterns and long-
term acceptance [36,40].
Novelty Effects with Robots and Long-Term Usage Nov-
elty effects have been described with interactive technolo-
gies [22]andalsoinseverallong-termstudiesonHRIin
non-domestic environments, such as workplaces [15,20],
schools [17,18]oreldercarecenters[23]. However, also
with domestic robots, long-term studies found that peo-
ple’s interest, engagement and fascination with the robot de-
creased over time [8,30]. Concerning long-term usage and
the adoption of robots in homes, two ethnographic studies
with the Roomba have been conducted in the US, involving
14 and 30 households respectively [10,11,33]. These stud-
ies led to the development of an initial framework, the ‘Do-
mestic Robot Ecology’ (DRE), which is proposed by Sung
et al. [33]. The DRE applies a holistic view to the relation-
ships that robots shape in the home and takes into account
long-term effects. DRE is a first step towards a comprehen-
sive understanding of HRI in a domestic environment.
Based on the interesting findings provided by the two ma-
jor ethnographic studies with the Roomba in households in
the US, we decided to add to this branch of research a sim-
ilar study. We deployed recent models of Roomba in nine
households in Switzerland.By this we take into account the
specific context for which we aim to develop new robotic
solutions in the future. A replication of Sung et al.’s long-
term study with Roombas makes sense for the goal of ex-
ploring similarities and differences in the process of adop-
tion in two culturally different regions. Since cultural differ-
ences seem to play a role in the perception of robots and
HRI [1], and further Swiss homes might look and be or-
ganized/maintained differently than US homes [3], we also
wanted to explore how far our findings can be integrated and
analyzed in relation to the data from Sung et al./Forlizzi et
al., and the DRE initial framework of long-term acceptance
of robots in homes, in particular. Consequently, the original-
ity of our work mainly lies in the way our data is analyzed,
namely taking the DRE framework into account and show-
ing that it can be used as a tool to structure and understand
the process of adoption of a domestic robot, also in a differ-
ent context.
3 Study Design
We set up an ethnographic study with nine households to
which we gave a Roomba vacuum cleaning robot (two dif-
ferent models, depending on whether the household owned
apet,seeTable1). It was our aim to investigate in a qualita-
tive and explorative way how usage and experience evolved
over time and study the impact on cleaning patterns. This
requires a holistic understanding of the household’s every-
day routines, their cleaning activities, as well as their expec-
tations of domestic service robots. Ethnography is a qual-
itative research method that attempts to generate a holistic
account of cultures or groups of people, such as a house-
hold. Ethnography has already been applied to study HRI
and long-term adoption of robots in homes [8,10,30]. In
its origin, ethnography is a form of anthropological practice
and both a methodology and perspective [3]. According to
Bell, ethnography is considered a type of field work where
the researcher spends time in and with the culture or peo-
ples studied, participating in everyday life and attempting to
make sense of the patterns of that culture [3]. We understand
ethnography not in its traditional anthropological sense of
‘going native’ but as a research technique to obtain qualita-
tive insights into what happens in the real world in contrast
to a well defined but limited laboratory setting. Roboticists,
engineers, and interaction designers acknowledge the rich
insights qualitative research in real world environments can
provide to inform future developments and refine design [8,
24,33]. However, longitudinal ethnographic research brings
along several challenges in terms of data acquisition, data
structuring, and interpretation. These and other difficulties
have also been outlined by Fernaeus et al. [8]andSunget
al. [30].
3.1 Methodology
The methodology applied to study long-term acceptance of
a domestic robot needs to match the described difficulties
and constraints created by the private nature of the home.
Potentially relevant events can take place not only in a pre-
defined space but within the whole social living space, as
well as at times when the researcher is not present. How-
ever, using logging mechanisms or extensive video record-
ing would create an unnatural setting which would not meet
the goals of our study. Another challenge is to capture rou-
tines of cleaning activities. A routine commonly happens
less consciously; it is consequently hard for participants to
talk about it explicitly and thus it is difficult to investigate.
We addressed these difficulties by using a similar study
design and methodology to that proposed by Sung et al. [30]
and also inspired by how Forlizzi et al. set up their studies
about cleaning in the home with and without the Roomba
[10,11]. More concretely, we also used ethnographic meth-
ods, including home visits, a ‘home tour’, qualitative in-
terviews, and cleaning diaries. Our study took place from
March to October 2011 and was composed of five visits at
each of the participating households:
1. Approximately one week prior to handing out the Roomba,
we had an introductory visit and home tour at each home
and got to know the household, people’s routines and
their attitudes towards robots.
Author's personal copy

392 Int J Soc Robot (2013) 5:389–408
2. At our second visit, about one week after the first one, we
brought the Roomba with us, interviewed people about
their expectations and observed their initial reactions to
the robot.
3. Approximately two weeks later, we came back to see
how the household experienced Roomba and what it was
used for. Prior to this visit, participants were asked to re-
port their usage of the robot in a diary.
4. About two months after the Roomba was deployed, the
households were visited again. We wanted to examine
how usage and perception of the robot changed over time
and how far the robot was integrated into everyday life.
For ten days prior to this visit, householders completed a
cleaning-diary.
5. A concluding visit was conducted about six months after
we brought the Roomba to the household. This was to
investigate changes in the usage patterns, to assess social
dynamics, and whether the experience with the domestic
robot changed people’s perception of robots in general.
Participants were again asked to fill out a cleaning-diary
during ten days prior to the visit.
In terms of the methods used, we applied several quali-
tative as well as quantitative techniques. Data was collected
at all households throughout the 6-month period, regardless
whether they stopped using the robot for cleaning. At each
visit, semi-structured qualitative interviews were conducted
(44 interviews, each lasting 1–1.5 hours), audio-recorded
and qualitatively re-transcribed. At each interview, partici-
pants were interviewed collectively and asked to describe
how they used the robot (for cleaning and other activities),
their satisfaction with it, and perceived benefits and con-
straints. The first visit also included a home tour where the
household’s main contact person showed us their technolog-
ical equipment, cleaning tools, and spots in the home that
they considered challenging to clean. By this we also learned
about their conceptions of cleanliness and order [21]. At
each visit, we collected field notes and photos or videos from
the on-site observations. Prior to each visit (except the first
one), each household filled out a daily diary to capture clean-
ing activities and Roomba usage. Self-reported data such as
diaries have been used by others, such as for ‘cultural prob-
ing’ [12], in sociological research [27] as well as in pre-
vious studies with the Roomba [10]. We verified people’s
self-reported data through the interviews and observations.
People’s perception of the robot was quantified by a series
of seven point Likert scales integrated into a small question-
naire to be filled out at each visit. Participants rated the robot
in terms of intelligence, usefulness, ease of use, its impact
on the household, fun, and emotional attachment, as well
as overall impression. We adapted this questionnaire from
Bartneck et al. [2] and Sung et al. [30]. The assessed aspects
are relevant for describing people’s perception of a robot and
crucial for the adoption of technology [38]. To make it eas-
ier for participants to talk about their cleaning routines and
their experience with the robot, we enriched the interviews
with projective techniques such as ‘Bubble Talk’ [30], ‘Day
Reconstruction Method’ [16], and drawing or tinkering ac-
tivities. Participants generally appreciated these activities.
The ethnographic approach of going to the households
combined with the described qualitative and quantitative re-
search techniques worked well. It allowed us to gain de-
tailed insights into how people were living with their robotic
vacuum cleaner and helped us to understand the changes it
evoked in some of the households. The before-mentioned
activities during the interviews made the discussions more
enjoyable and helped people to be natural and open with
us. Nevertheless, it has to be mentioned that the prepara-
tion of these research tools required a lot of effort and we
were happy to have the participation of a professional ethno-
graphic researcher, who is one of the co-authors.
The gathered data was mostly analyzed in a qualitative
way where two researchers studied the collected material
related to the research topics, cleaning activities, and ex-
perience with the Roomba. We extracted relevant observa-
tions from field notes in several spreadsheets and organized
them in chronological time-series as well as in topic-specific
categories. This process was motivated by Grounded The-
ory [4]andamethodforanalyzinguserbehaviorinhome
environment [13]. Later, data was integrated and analyzed
according to the DRE framework of adoption [33]. Data
from the cleaning diaries was translated in English (if nec-
essary) and typed in a spreadsheet for a descriptive analysis.
The paper-and-pencil diaries captured the following aspects
of a cleaning/household task: day and date, time, relation
to previous/following activity (if any; e.g. before guests ar-
rived), duration, who carried out the task, which rooms/parts
of the home were cleaned, what was cleaned (e.g. floors,
windows, toilet, etc.) using which tools (e.g. towel, vac-
uum cleaner, brush, dish-washer, Roomba, etc.). These were
all self-reported data that may be incomplete or biased by
the nature of the study (i.e. people actually carry out more
cleaning activities than usual because they consciously think
about cleanliness as it is part of the study). However, we dis-
cussed cleaning tasks with the participants during the inter-
views to confirm and clarify some of the entries. Quantita-
tive data from the Likert ratings was subjected to a statistical
analysis. Preliminary results have been published [9].
3.2 Participants/Sample
As the goal of this study was to qualitatively explore how
people live with a vacuum cleaning robot, we decided to
select a diverse sample of households rather than focusing
on one specific type. Households were recruited through
word of mouth, a request distributed via e-mail, and an an-
nouncement in a school’s newsletter. The call invited house-
holds with and without children and pets to participate in
Author's personal copy

Int J Soc Robot (2013) 5:389–408 393
Tab l e 1 Participants of
6-month ethnographic Roomba
study; the cleaning strategy was
determined from self-reported
data in diaries and on-site
observations; the adoption group
refers to the degree of household
use or adoption of the robot
# type People, age Pets House
type
Cleaning
service
Cleaning
strategy
Roomba
model
Adoption
group
H1 single m, 40 no flat no spartan 520 adopter
H2 single f, 71 1 dog
2 turtles
house yes minimalistic 563 PET user
H3 couple f, 62
m, 58
2 cats flat no minimalistic 563 PET user
H4 family mother, 43
father, 47
2 girls, 10, 8
1 cat house yes carer 563 PET user
H5 family mother, 45
father, 45
3girls,18,12,12
boy, 15
1dog house no carer 563PET adopter
H6 family mother, 28
father, 28
girl, 5
2boys,4,1/2
1 cat house yes manic 563 PET adopter
H7 family mother, 29
father, 34
boy, 1 1/2
no flat no manic 520 rejecter
H8 family father, 41
3boys,11,10,8
no flat no spartan 520 rejecter
H9 family mother, 41
father, 45
2boys,8,7
no flat yes carer 520 user
a study about how people live and interact with a domestic
robot. The text briefly described that households were go-
ing to receive a Roomba vacuum cleaning robot over a pe-
riod of several months, and that the study would take place
in their home with the researcher visiting each household
several times. We wanted to avoid giving a Roomba to peo-
ple who were not at all interested in deploying a robot in
their home. This fact has to be kept in mind when analyz-
ing and interpreting the data and makes results not gener-
alizable. The sample consisted of nine households with a
total of 30 participants: 16 men and 14 women, 15 children
(six months to 18 years old), and 15 adults ranging from
26 to 71 years old (mean age 43.6 years). More specifically,
the study included three single-headed (one single parent)
and six double-headed households (one couple without chil-
dren). Each household (H) profile is briefly described in the
following:
H1 consists of a 40 year old single male participant without
pets. Originally from France, he lives in an apartment in
the periphery of town and works in the IT domain. He
also programs in his leisure time.
H2 consists of a 71 year old single woman who is origi-
nally from Austria. Now retired, she used to work in
the health care sector and has been living for more than
20 years in a small house with garden in the suburbs.
She has a small dog and two tortoises. She barely uses
her laptop and internet but is open though skeptical of
new technology. When she has difficulties in using her
computer or cell phone, she appreciates help from her
younger friends.
H3 is a couple consisting of a 62 year old British woman
and her 58 year old Swiss common-law spouse. They
share a flat in the center of the city with two cats. Both
describe themselves as not being much into technology
but open to new developments as long as they are useful.
While he is working in the media sector, she works in-
dependently for a company that sells organic products.
H4 consists of parents (40ies), two girls aged 10 and 8, and
a cat. This German family lives in a house with garden
in the periphery of town close to the lake. They describe
themselves as open to new technologies but the mother
is skeptical. She used to work in the health care sector
but now stays at home. The father has a background in
engineering but now does freelance management.
H5 consists of parents (40ies), 4 children (18 year old
daughter, 15 year old son, 12 year old twin girls), and a
dog. This Danish family lives in the periphery of town
and is quite open to new technology. The father works
in an international company while the mother stays at
home.
H6 consists of the parents (late 20ies), the 5 year old son,
4 year old daughter, a 6 month old baby boy, and a cat.
This Swiss family lives in the countryside, maintains an
Author's personal copy

394 Int J Soc Robot (2013) 5:389–408
Fig. 1 Adapted and simplified version of the Domestic Robot Ecol-
ogy framework, Sung et al. [33]. The overview of interactions between
avacuumingrobot(inthemiddle), and the key attributes: the related
tasks, the environmental context, and the social actors; which form to-
gether the ecology of the home
organic lifestyle, and uses new technology as long as
it is practical (e.g. they do not have a TV). The father
works as dentist and the mother manages her own tea-
shop part-time in the town nearby.
H7 consists of parents (early 30ies), and a 1 1/2 year old
son. This Swiss-French family lives in the center of the
city and we would describe them as an ‘early adopter’
household, since they have all latest technologies in
their home. They emphasize design and live an organic,
healthy lifestyle. The father (Swiss) works as a free-
lance graphic designer, partly from his home office.
The mother (French) works part-time in an advertising
agency.
H8 consists of a single father (41 years old), and three boys
(8, 10, and 11 years old) who also live at their Japanese
mother’s place. This Swiss-Japanese family lives in a
flat in the center of the city and we would describe them
as an ‘early adopter’ household, based on the latest tech-
nology devices they are using. The father (Swiss) works
as a teacher and has a background in robotics.
H9 consists of parents (40ies), and two boys (7 and 8
years). This Swedish-American family lives in an ex-
clusive apartment in the suburbs with view on the lake.
They have latest technologies in their home. The father
(Swedish) is working in an international company and
the mother (French-American) gives courses in an inter-
national business school. The family left the study after
5 months because they moved unexpectedly to another
country. The data for the concluding 5th visit is thus
missing.
4 Findings
We integrated and structured results along the ‘Domestic
Robot Ecology’ framework (DRE) [33]. We did this because
first, the DRE provides a concrete structure to organize our
observations, and second because it makes findings of two
similar studies easy to compare. Through outlining similar-
ities and differences we aim to extend previous work and
contribute to a robust holistic framework of acceptance of
domestic robots. Initially we wanted to develop our own
model of domestic robot adoption. During data analysis and
synthesis we saw that our results mainly support the DRE
framework. This is why in the end we did not build another
model but added data to an existing one. Figure 1shows an
adapted and simplified visualization of the DRE. It gives an
overview of interactions between a vacuuming robot (pic-
tured in the middle), the related tasks (such as cleaning),
the environmental context of the home (physical and social
space), and the social actors in the household. We use this
simplified version to highlight at which stages of adoption
process the robot impacts which key attributes. The reader
can refer to more details in the original work [33].
The DRE gives a holistic view of the relationships that
robots shape in the home and how these change over time.
On one hand, the framework identifies four temporal stages
of how households accept robots in homes: (1) pre-adoption,
(2) adoption,(3)adaptation,and(4)use/retention (see
Sect. 5). It also describes three key attributes that influence
the interaction experiences across these temporal stages:
physical and social space,social actors,andintended tasks
[33]. Interactions between the robot, the task characteristics
(vacuuming) (Sect. 4.1), environmental context (Sect. 4.2),
and social actors (Sect. 4.3)aredescribedinfiverelation-
ships: the robot formed relationships as (1) a tool to perform
tasks, (2) an agent that impacts the environment,(3)ame-
diating factor that motivates people to make changes in the
environment, (4) a mediator that enhances social relation-
ships within the household, and (5) an agent that engages
with people in social events [33].
4.1 Task Characteristics: Cleaning Routine
According to Sung et al.[33], tasks refer to the activity that
the robot is designed to perform (vacuuming and more gen-
erally cleaning). Domestic tasks are inter-related and there-
fore automating one task with a robot may bring changes
to the connected tasks [33]. In the following, we describe
and compare people’s cleaning routine without and with the
robotic vacuum cleaner (robotic VC). This data was col-
lected through the paper-and-pencil diaries and rounded it
out with interviews and day-reconstruction technique [16].
One of our aims was to understand each household’s atti-
tude toward cleaning and cleanliness. We expected the us-
age and adoption of the robotic VC would be impacted by
Author's personal copy

Int J Soc Robot (2013) 5:389–408 395
the respective cleaning strategy. Therefore, we tried to cate-
gorize households according to the effort spent for cleaning
the house. We are aware that this is a risky step given the
small number of participating households.
Cleaning Strategies Without the Robot Based on people’s
remarks and our observations, we classified households ac-
cording to their cleaning strategy. We derived the following
four types, based on the motivation that a household has to
keep the home clean, the efforts made and amount of time
spent cleaning:
Spartan cleaners barely notice dirt and do very little about
it. We would describe them as lacking motivation to
clean, and hardly use the few cleaning tools they have
(e.g. vacuuming once in two months). However, they
feel comfortable. Cleaning is not important to them.
(H1 and H8)
Minimalistic cleaners notice dirt around the house which
makes them feel a little uncomfortable (which creates
some intrinsic motivation to clean). They do what is nec-
essary but not more. Vacuuming is done only when they
have time to do so, e.g. once a week or once in two
weeks. Cleanliness is not a priority. (H2 and H3)
Caring cleaners really care to have a clean and nice look-
ing home to show visitors that they have a well working
‘home ecosystem’ (which creates some extrinsic moti-
vation). They like to keep the home clean by vacuum-
ing several times per week and they enjoy the resulting
cleanliness. (H4, H5, and H9)
Manic cleaners almost obsessively clean. They are very
picky, notice every little piece of dirt, and probably con-
stantly feel some pressure to clean or tidy up. Thus, they
constantly engaged in cleaning and these tasks are a pri-
ority for them. Vacuuming happens more or less on a
daily basis. (H6 and H7)
This classification of cleaning strategies is not based on
demographic factors but on our observations, interviews,
and the analysis of each households’ cleaning diaries. For
instance, we discussed with people when we found dirty ar-
eas in their home and asked them how they felt about it. In-
terview questions asked whether one constantly felt the need
to clean and tidy up or would be uncomfortable with leaving
unwashed dishes when departing the house, for instance.
Comparison of Cleaning with and without the Robot Vac -
uuming was generally considered as an “annoying”,“bor-
ing”,“time consuming” repetitive task, that needs to be
done over and over again. However, everyone in our sample
described the vacuum cleaner (VC) as an important, indis-
pensable cleaning tool. People didn’t like its noise, its large
size, heaviness, that it looked ugly, and needed to be plugged
in to an electrical socket when used, which is not convenient
for a device that needs to be moved around manually. Pet-
owning households reported problems with the efficiency of
the VC and some others with storing it. H1 and H4 needed
to attach the VC inside the closet so that it did not fall out.
This made it annoying to get the VC out of the closet and
to put it back in which made it a tool that was not ‘ready
for use’. The 71 year old woman in H2 stored her VC in the
basement, since it was mainly the cleaning lady who used
it. However, when the woman needed it, she had to carry it
upstairs, which was a heavy task for her and thus she tried
to avoid it.
Future developments need to take aspects of how a de-
vice is used into account and develop appropriate design so-
lutions. Domestic appliances that are in regular use, such as
the VC, need to be easy to store and handle. Future domestic
robots need to enable an overall positive user experience to
become useful and valuable tools.
How did people in our study keep their homes clean, how
did they use the VC and the robotic VC? In the following,
we describe and compare people’s cleaning routine in terms
of who is cleaning, what, when, where, how, and why [13].
Overall, 634 distinct cleaning tasks were documented in the
cleaning diaries, 193 activities of what involved the robotic
VC, and 65 used the VC.
Who: In most households, there was a clear division of labor
between household members in matters of responsibilities of
specific household tasks. Generally, our findings support the
traditional (gender) roles of housekeeping [25,27]. Women
were the main housekeepers, independent from whether they
were housewives, working part- or full-time. Women in our
sample carried out 69 % of the reported cleaning activities
and 75 % when only vacuuming is considered. Of the 75 %
of vacuuming sessions, however, 20 % were done by the
cleaning services, and another 10 % by girls and female
teens in the study. Surprisingly, women also turned out to
be the Roomba’s main users (67 % of the n=178 valid
cases) which was different in results from an online survey
among Roomba users [29]. Asked, why she was Roomba’s
main user the mother in H7 explained: “Roomba is a vac-
uum cleaner, and since my husband never vacuums it’s only
me using it.” Still, in comparison to the usage of the VC,
men’s proportion for Roomba usage was higher: men carried
out 14 % of all vacuuming with the traditional VC but 23 %
of all Roomba usages. What is notable, is the high propor-
tion of vacuuming that was carried out by children and teens
(21 %) in comparison to their share of all cleaning activi-
ties (1 %). In respect to teamwork, generally, when people
cleaned, they did it on their own. Only about 10 % of clean-
ing happened in collaboration of several people at the same
time. Contrary to Sung et al. [33], in our study, the robotic
VC did not enhance persistent collaborative cleaning. Initial
effects of collaboration with the robot are addressed later
(Sect. 5.2 ‘adoption phase’).
Author's personal copy

396 Int J Soc Robot (2013) 5:389–408
Fig. 2 An elderly participant reading the robot’s manual (left); a family unpacking their vacuuming robot (center), and an example for adjustments
of the home due to the robot (right)
In respect to people’s age,twoelderly people in our study
(62 and 71 years) actively tried to avoid using the VC, due
to its heaviness and their own physical restrictions. They
both appreciated that the Roomba was much smaller and
lighter than their traditional VC. We did not find differences
in the frequency and amount of Roomba usage between el-
der and younger adults. However, how participants experi-
enced the robot seemed to be affected by age, this was espe-
cially linked to usability. Age differences also played a role
in how the robot was approached and used the first time.
Children showed a great fascination for the robot and their
intuitiveness in operating it was surprising. Even infants in-
tuitively managed to switch on the robot and felt rewarded
by the sound the robot makes when starting. Younger chil-
dren tended to prefer the robot to the VC but teens were not
interested in it and stopped using it after the first try.
Regarding the type of household, we cannot make gen-
eral statements with a sample size of 9 households. How-
ever, analysis of the diary reports suggest that families used
the Roomba more often for cleaning (the robot represented
35 % of the various cleaning tools) than households without
children (19 %). This is in accordance with Sung et al. [29]
who found that households with children expressed signif-
icantly greater satisfaction with the Roomba’s performance
than households without children. We did not find differ-
ences in the amount and frequency of the robotic VC use in
respect to whether the household benefited from a cleaning
service or had a pet. However, these aspects shaped people’s
expectations of and hopes for the robot.
What: Half of the reported tasks were ‘small things’, such
as cleaning the dishes (or loading the dishwasher) (19 %),
wiping the table (18 %), tiding up something (14 %), or
cleaning up crumbs from the floor (11 %). In addition to that,
23 % of the household tasks concerned the laundry, folding
clothes, or ironing. Surprisingly, overall only about 10 % of
all cleaning tasks involved the VC. During the course of the
study, the proportion of cleaning activities that involved the
VC decreased over time from 17 % before the Roomba was
deployed, to 15 % at two months after the introduction of
the robot, to 8 % after six months. We think that the study it-
self might have triggered additional vacuuming activities in
the beginning of the study. Similar to the VC usage, the pro-
portion of Roomba usage (among the other cleaning tools
in the household) decreased over time; the robot was used
much more frequently during the first phase.
When: There was a tendency for people to clean more often
on Sundays (18 % of n=634 cleaning activities). This was
especially true for the usage of the VC (26 % on Sundays).
However, Roomba usage was distributed more equally over
the week with a peak on Wednesday (20 %). This confirms
Sung et al.’s finding that the robotic VC helps people clean
more regularly [33]. Concerning the time of day, participants
carried out most cleaning tasks during the morning (46 %
from 5–11 am), in comparison to 24 % in the afternoon
(1–5:30 pm), and finally 27 % in the evening (6–10:30 pm).
A similar distribution was found for vacuum cleaning as
well as for Roomba usage. Whereas some of the households
had a clear schedule (e.g. Sunday morning after breakfast),
others did cleaning in a more opportunistic fashion when it
was needed or they felt uncomfortable [11]. Most cleaning
activities were related to meals.
On average, people cleaned 26 minutes, however 53 % of
all cleaning activities took in between 1–15 minutes. People
spend a bit longer for vacuuming: usually between 20–30
minutes but still 27 % of the reported vacuuming sessions
took 60 minutes or longer. The amount of time and effort
spent on cleaning differed significantly between the house-
holds but was surprisingly unrelated to the amount of dust
that we noticed in each home. In contrast to vacuum clean-
ing, it was surprising that one third of the n=141 valid
Roomba usages did not take longer than 15 minutes. This
reflects the fact that people tended to stop the robot manu-
ally after a short period of time instead of letting it finish and
return to its charging station itself.
Where: More often than anywhere else in participants’
homes, cleaning took place in the kitchen (58 %), followed
by 11 % in the bathroom, and 10 % in the living room (see
Author's personal copy

Int J Soc Robot (2013) 5:389–408 397
Tab l e 2 A comparison of where people clean, with the vacuum
cleaner (VC), and with the Roomba; percentages in each column do
not sum to 100 % as not all possible locations are listed
Where General
cleaning
VC
usage
Roomba
usage
Kitchen 58 % 17 % 35 %
Bathroom 11 % 35 % 2 %
Living room 10 % 10 % 28 %
Children room 6 % 17 % 17 %
Bedroom 4 % 14 % 10 %
Table 2). However, only about 17 % of the vacuuming hap-
pened in the kitchen, but 35 % in the bathroom, and another
17 % in the children’s room. Further, the Roomba was used
in the living room (28 %), and the bedroom (10 %). People
avoided using the robotic VC in the bathroom (2 %) which
was the favorite room for the traditional VC (35 %). People
explained to us that they were afraid that there could be wet
spots in the bathroom and the robot could get damaged.
How: In most cases, people vacuumed ‘room-by-room’.
Most participants wanted to use the Roomba similarly to
how they used their VC, which has also been found in a
study on matching a vacuum robot’s path planning to user
expectations [19]. People imagined using the robot in one
room, passing only once over every spot and they expected
the robot to work better than their VC in less time. However,
afatherwhohadabackgroundinrobotics(H8)hadmore
realistic expectations about the robot’s capabilities than the
other participants. Overall, there was no common strategy
for using the robot. As we did not give instructions, we let
people experiment themselves and observed the first usage.
At the later visits, we asked how and why they used the robot
as they did. In H1, H6, and H7 usage strategies changed over
time. It seemed that these households had found their opti-
mal way of using the robot after several attempts whereas
the other households were not willing to spend the effort in
learning how to use the robot in an optimal way and gave up
after some disappointment (see Sect. 5.3).
The four defined categories of cleaning strategies (spar-
tan, minimalistic, caring, manic), served well to explain how
a household cleaned but were not transferable to Roomba
usage. With traditional vacuuming, spartan and minimalis-
tic cleaners planned their vacuuming sessions in advance,
e.g. H1 had a monthly alert on his smart phone to remind
him of vacuuming. In contrast, caring and manic cleaners
vacuumed regularly, more or less on a daily basis. We also
expected that the frequency of Roomba usage would impact
a household’s cleaning strategy. However, this was not the
case. H1, who vacuumed once per month, turned out to use
the robotic VC on a daily basis and put great effort in making
it work well, whereas H7 who vacuumed every day, tried out
the robot a couple of times but then refused to use it. These
and other observations highlight that the robot was used as
an independent tool and differently from a classic VC.
How people cleaned did not seem to be related to whether
they had children or were pet owners. Each household com-
plained about specific problems related to cleaning and used
different strategies to keep efforts as low as possible. For ex-
ample, H4 lived in a house with garden. To keep the ground
floor clean, they always entered through the garage. Several
households (H2, H4, H6, H7, H9) had a broom or dustpan in
the kitchen to be able to quickly clean up the kitchen floor
without having to get out the VC.
Why: We asked participants why they cleaned in the man-
ner they cleaned, for example they tended to store their VC
in a ‘remote place’ such as the furthest corner of the closet,
despite the fact that it was used regularly. Participants were
not very clear in their answers but mentioned the VC was an
ugly and unpleasant tool that they didn’t like at all and thus
tried to “keep it out of sight”. One participant said: “I don’t
want to see the VC every time I open my closet and be re-
minded of the need to vacuum.” Another one said: “I need
the VC every single day to clean up around the table. Sure,
it would be easier to have it just right next to the table. But it
wouldn’t look nice. I don’t want my VC or any other clean-
ing tool to be that present.” We conclude that traditional VCs
would need a lot of improvements to be considered as a more
pleasant tool.
Intermediate Essential: A comparison of cleaning activ-
ities shows that people use a robotic VC in a different
way than a classical VC. Both tools are preferred according
to their own characteristics (e.g. suction power, immediate
availability) and the characteristics of the area which needs
to be cleaned (e.g. surface, obstacles). The difference in us-
age might suggest that a robotic vacuum cleaner might not
aprioribe able to replace a traditional vacuum cleaner.
4.2 Environmental Context: Physical and Social Space
In Sung et al.’s DRE, the environmental context of the home,
is understood as the physical, social, as well as technical
space of a home [33]. This comprises the indoor physical
environment (floors, rooms, furniture), the social lifestyle of
the household, as well as the configuration and organization
of technologies in the home. With mobile and ubiquitous
technology in the home, the location of a technology has a
new meaning. The ubiquitous spatial presence [33]ofau-
tonomously moving robots impacts upon the broader phys-
ical and social space. Previous work suggests that the envi-
ronmental context of the home is likely to impact on how a
domestic robot is used and experienced [11] as well as vice
versa and our observations generally confirm this.
Author's personal copy

398 Int J Soc Robot (2013) 5:389–408
Physical Space: Participants expected the robot to work
equally well in the entire house and intended to use it every-
where. However, the physical layout of the home impacted
the robot’s performance. Two households (H3, H7) had door
sills that the robot was not able to pass over. Whereas H7 felt
Roomba’s inability to overcome this obstacle made the robot
less usable, H3 acknowledged that this meant she could eas-
ily keep it in the kitchen. Households with stairs in their
home didn’t feel this constrained how they used the vacu-
uming robot, however, usage data showed that Roomba was
used slightly more often in homes without stairs. Another
aspect that constrained where and how the robot was used
was the location of electrical sockets, as the robot’s charging
station needs to be plugged in so that the robot could auto-
matically go back to it. In three homes there was a lack of
electrical sockets in the living room and in two of them, the
height of the sockets meant that the charging station could
not be placed on the ground because the cable was too short.
This was clearly impractical and it made it difficult to use
the robot in its intended way because it constrained its au-
tonomy. In the end, this negative experience hindered people
in integrating the robot in their cleaning routine.
Regarding different kinds of floors, homes in our sam-
ple included parquet floor, tiles, as well as various types
of carpets. The robotic VC worked equally well on each
of the surfaces, however people reported difficulties when
Roomba was moving from tiled or parquet floor to carpet
(the robot dropped dust or even got stuck). This shaped the
experience with the robot and was a crucial factor of accep-
tance/adoption (see Sect. 6). Not all participants were will-
ing to (re-)move the ‘difficult’ carpets. For instance, an el-
derly woman (H2) who had Persian carpets was afraid the
robot might leave marks on them and only used the robot in
the kitchen where there was no Persian carpet.
In general, participants found it easier to use the Roomba
in a home characterized by large open spaces and few sep-
arating walls. Additionally, participants were positively sur-
prised to see the robot cleaning underneath furniture where
they could not reach with the traditional VC and placed the
Roomba intentionally right in front of the bed or closet. This
shows that the physical layout of the home can also trigger
specific use cases with a robot.
Social Space: How people were using the robot was further
impacted by the social space of the home. Some homes were
characterized by a harmonious arrangement of furniture and
decor objects. The specific lifestyle of H3 made it impossi-
ble for the woman to use “something as artificial as a robot”
in her painting room. She explained the robot would not fit
the decor of the room and when she once tried out using it
there she felt that it was disturbing the mood. The robot’s de-
sign interfered with the social space of several other homes.
Characteristics of the robotic VC design seemed to be more
important than the classical VC. This could be due to the
fact that in contrast to a usual VC, Roomba was not hidden
in a closet when not in use but always visible and present.
Further it moved autonomously through the home and some
people found this intruded their privacy. Consequently, they
didn’t like to use it when they were not at home and able
to “observe it”. This however, was not a good solution be-
cause the robot’s vacuuming noise was annoying for most
of the people. Only one participant (H2) described she liked
the robot’s noise because it gave the impression it was “ac-
tually working” and “doing its job”:“It’s nice when the
Roomba moves around the house making little noises and
talking to me occasionally.” The other participants decided
to use the robot while they were not at home. However, not
everybody felt comfortable about this. Some people had dif-
ficulties with having the robot’s charging station visible in
a prominent open space, such as the living room. The liv-
ing room is a highly social space, where people relax, host
guests and have conversations, it serves to present a specific
lifestyle. People did not want to have either the robot or its
charging station visible in the living room. As the mother in
H9 explained “Still, this is a vacuum cleaner and you don’t
want to have your vacuum cleaner next to your dining table
when you are having dinner or friends are coming over. It
looks too much like work!”
4.3 Social actors: The Robot within the Household
According to the ‘Domestic Robot Ecology’, social actors
are the living members in the home, such as householders,
guests, and pets. Sung et al. [33] divide social actors into
‘users’ who interact with robots on a regular basis to com-
plete a task, and ‘non-users’ who do not regularly use but
engage in social activities with this technology. Pets can
be regarded as part of the group of social actors. Social
responses to robots (e.g. anthropomorphism) are, amongst
other things, based on how people perceive and relate to the
robot [7]. To capture people’s perception of the Roomba, we
let them evaluate it in respect to various topics. We were
interested in whether and how people’s perception changed
over time and how far it would be related to individual and
external factors.
People’s Perception of the Robot over Time Participant’s
perception of the robot was assessed using a recurring ques-
tionnaire that was integrated in the semi-structured inter-
views at each household visit. All present family members
were asked to rank the robot on a 7-point Likert scale (1–7)
according to the aspects intelligence, usefulness, ease of use,
fun, attachment, impact and overall impression. The method
was adapted from Sung et al. [30] and Scopelliti et al. used
similar scales [26]. To enhance discussion and keep the situ-
ation as natural as possible, participants of the same house-
hold were interviewed collectively, thus children were not
Author's personal copy

Int J Soc Robot (2013) 5:389–408 399
Tab l e 3 Means (M) and standard deviations (SD) for each topic over
five times (T1–T5) on a 7-point Likert scale (1–7). Repeated measures
ANOVA with time as within subjects factor, n=15
Top i c M T1T2T3T4T5F
(4, 56) p
SD
Intelligence 3.9 4.1 4.2 3.8 3.3 2.2 0.09
1.6 1.6 1.6 1.7 1.3
Usefulness 4.9 5.2 5.1 4.5 4.2 2.6 0.05
1.4 1.0 1.2 1.9 1.9
Ease of use 5.6 6.1 6.3 6.3 6.1 2.2 0.08
1.2 0.6 0.7 0.8 1.0
Fun 4.0 4.1 4.6 4.1 4.3 0.7 0.59
1.6 1.6 1.7 1.6 1.4
Attachment no valid data due to wrong measurement
Impact 3.5 4.2 4.0 3.2 2.8 3.4 0.02
1.6 1.4 1.1 1.3 1.5
Overall
impression
5.2 5.2 5.0 4.9 4.5 1.3 0.30
1.2 1.1 1.2 1.3 1.5
separated from their parents, and participants evaluated the
robot in front of each other. This could have caused some
biases, which should be kept in mind when interpreting the
data. Participants filled out the questionnaire at five time
points: (T1) before they had seen the robot, (T2) right after
they had unpacked and tried out the robot for the first time,
(T3) two weeks after they had received the robot, (T4) two
months afterward, and (T5) after six months living with the
robot. Participants’ ratings were compiled in a spreadsheet,
to one decimal place (two coders agreed on the interpreta-
tion of each mark, e.g. if the mark was understood as 5.4
or 5.5). Preliminary results after T3 have been published al-
ready [9]. We used repeated measures ANOVA tests with
time (5 time points) as within-subjects factor. Further, we
defined gender,family code,andadopter group as between-
subjects factors. Gender had two categories (female/male),
family code had nine categories (H1–H9), and adopter group
had three levels (rejecter/user/adopter) according to how far
at the end of the study a household was still using the robot
for cleaning or not. Rejecters didn’t use the robot any more,
users sporadically used it but would not buy it for them-
selves, whereas adopters integrated the robot in their routine
and could not imagine living without it. In general, the re-
sults of these statistical analysis are not strong, since only a
small number of valid ratings (n=15) were considered (Ta-
ble 3). For this reason, we could also not statistically analyze
age differences.
Overall, there was a ‘positive’ peak for most of the ratings
at T2, right after participants used the robotic VC for the
first time. This suggests that the sample had a rather positive
experience when trying out the robot and most of the expec-
tations (T1) were initially met (or even exceeded). However,
most ratings became less positive after six months with the
robot, with slight variations during the study. In more detail:
Perceived Intelligence: Overall on a scale from 1–7, par-
ticipants rated the robot as ‘somewhat intelligent’ (values
from 3.3–4.2) but were not really convinced about it being
smart. The rating did not change significantly over time (see
Table 3). Interestingly, at all five time points, male partici-
pants rated the robot as more intelligent (average 4.5) than
females (average 3.4), however, this difference was not sig-
nificant. When asked why they thought Roomba possessed
intelligence, men more often referred to its abilities to detect
obstacles and stairs, and being able to recharge itself. Con-
trary, women referred to its rather limited vacuuming func-
tion and found it not as intelligent because “It doesn’t see
where the dirt is, it sometimes leaves it, it can’t be smart!”.
There was no significant difference between adopter groups
(F(2,12)=0.2, p=.79), however, adopters rated the robot
more intelligent at each time point.
Perceived Usefulness: Participants rated the functional
robot as quite useful (values from 4.2–5.2), however, values
decreased significantly over time. Women rated the robot
less useful (average 4.4) than men (average 5.2) (differ-
ence not significant.) This again was due to the fact that
women (as the main users) evaluated the robot’s vacuum-
ing power whereas men regarded more generally that the
robot could help cleaning and would thus be a useful device.
A significant difference was found for the adopter groups
(F(2,12)=6.7, p=.01): adopters rated Roomba signifi-
cantly more useful (average 6.1) than users (average 4.6),
and rejecters (average 3.9). This observed difference is in
accordance with models of the adoption of technology [22,
37] and acceptance of robots [14].
Perceived Ease of Use: People found the robot very easy
to use (values from 5.6–6.3) and the only thing they would
change about it, was how one had to empty the dust bin. Still
some participants had difficulties in using the infrared walls,
that come with some Roomba models, and especially the el-
der participants did not know how to assemble the charg-
ing station. There was no significant gender difference in
the perceived ease of use, however, data suggests that with
growing experience, especially women rated the robot as
even easier to use (contrary, men rated the robot less usable
in the end of the study, probably because they had not gained
much experience with it). Accordingly, adopters rated the
robot significantly easier to use (average 6.7) than rejecters
(average 5.9) (F(2,12)=4.6, p=.03). This confirms again
traditional models of technology adoption and acceptance of
robots [22,37] and acceptance of robots [14].
Experienced Fun: Participants rated the robot as ‘some-
what fun’ (values from 4.0–4.6). Again, there was a qual-
Author's personal copy

400 Int J Soc Robot (2013) 5:389–408
itative gender difference (not significant), in the sense that
the experienced fun remained quite stable for male partici-
pants, whereas it tended to decrease for female participants.
Males described the robot more as a ‘fun gadget’ than fe-
males, who referred to it as “In the end, Roomba is a clean-
ing tool and it is not really fun to use it.”. In respect of the
three adopter groups, there was no significant difference for
the ratings (F(2,12)=0.3, p=.73), however, at each time
asked, adopters rated having experienced more fun (average
4.5) than rejecters (average 3.9).
Emotional Attachment: The ratings for people’s emotional
attachment are unfortunately not consistent because we dis-
covered they were strongly related with the respective in-
structor who conducted the interview process. The two in-
structors described ‘attachment’ in an inconsistent way dur-
ing the study and at different households. Because of this we
cannot use the data for attachment for this study.
Perceived Impact on the Household: Participants esti-
mated the robot’s impact on the household significantly dif-
ferent at different points of time. While in the beginning at
T2(rightafterfirstusage)andT3(aftertwoweeks),peo-
ple believed Roomba could maybe change something in the
household, after two and six months (T4andT5) they more
or less lost their initial hope the robot could evoke some
changes. There was a significant difference for perceived
impact of the robot between adopter groups (F(2,12)=5.4,
p=.02). Adopters rated the robot’s impact higher (average
4.6) than users (average 3.7), and rejecters (average 2.9).
Overall Impression: People’s expectations before they
used the robot (T1) were quite high (5.2, on a scale from
1–7 with 7 as the most positive) and though not significant,
men expected even more (5.7) than women (4.8). However,
with time, participant’s overall impression constantly got
worse, even though with an average rating of 4.5 in the end it
was still rather positive. The three adopter groups rated their
overall impression significantly different (F(2,12)=23.3,
p≤.05): adopters (average 6.2), users (average 4.5), re-
jecters (average 3.8). Interestingly, before they had seen the
robot, the later adopters had the lowest expectations (4.5),
whereas rejecters had the highest (5.7). We could interpret,
that those with the highest expectations had a rather negative
first experience since the robot probably did not meet their
expectations, while those who expected less were maybe
positively surprised.
Social Activities with Roomba The robot, even though
functional, created some few social dynamics that had im-
plications also on the participant’s social life in general. It
has already been described that domestic robots can serve
as socially interactive agents [10,28]. In our study, people
also engaged in social activities with the robot, however,
we would like to stress, that these social dynamics were
quite rare and not persistant. For example, previous work
described people gave a name to their vacuuming robot and
liked to customize it [28,32]. However, we did not observe
any activity of customizing the robot and only one woman
(H3) gave a name to her robot (“Elvis”) which seemed to
be more for fun than actually reflecting a kind of social at-
tachment to the robot. We think that no participant perceived
the robot as a social agent, as it is suggested in the DRE
framework [33].
A similar rather weak social interaction that only oc-
curred during the stage of initial adoption was collaborative
cleaning. This finding is in contrast to Sung et al.’s conclu-
sion that the robot would have the social impact of “making
cleaning a concern for everyone in the household” [11,33].
As described earlier, the robot was not able to encourage in
alongtermwayhouseholdmemberstocleanwhodidnot
clean before. Only in the very beginning did people assist
the robot in cleaning (notably children). Children collected
crumbs and put them right in front of the robot, they put up
the bathroom carpet or chairs that hindered the robot from
passing through (Fig. 7center part). This immediate activ-
ity was observed in all families with younger children but
also in the elderly single woman initially started collecting
dust while the robot was vacuuming. She explained that she
didn’t want to see the robot working on its own and would
feel bad when she didn’t at least help him a little. However,
she stopped using the robot completely after about 3 months.
Another immediate activity that the robot encouraged
particularly children to do was to build walls out of ob-
stacles and furniture to prevent the robot from going some-
where where they didn’t want it to go or to keep it some-
where. However, people also stopped this activity soon.
Children up to the age of about six years kept on putting
toys on top of the robot or offered their stuffed animal a ride.
The positive impact of children and pets on the experiences
with Roomba has already been described by [29]. Younger
children interacted with the robot physically (Fig. 7left part)
and emotionally by playing and experimenting with it. For
example, they let it bump into parts of their body, carried it
around, and arranged objects as a circuit or labyrinth for the
robot. Some children (mostly boys) made the robot part of
their games and plays, e.g. they let it make a footrace with
the baby that just learned to crawl or played hide and seek
with the robot. Some also found creative ways of integrating
it in their games with and without other toys: Roomba as a
shark and the sofa as the boat where one is safe; Roomba
as a spaceship with lasers. The pets in our study reacted
in different ways to the robot. Though the pets were curious
(Fig. 7right part), two cats and one dog were afraid, one dog
tried to bite the robot and only H3 derived fun from watch-
ing their two cats’ playing with it. Dogs barked at Roomba
(H5) when it was vacuuming or anxiously hid behind the
owner’s feet (H2).
Author's personal copy

Int J Soc Robot (2013) 5:389–408 401
Nevertheless, the robot also served as a mediator for so-
cial interactions with persons who did not belong to the
household, such as work colleagues or guests. During the
first weeks, when participants hosted guests, they showed
the robot and gave demonstrations. This enhanced discus-
sion or served as entertainment. Especially male participants
described they liked to show their friends the “fancy new
robot” they had in their home. Female participants rather
liked to receive feedback from their friends which is de-
scribed a usual activity in the process of adopting inno-
vative technology [22]. One participant (H1) reported the
robot would have helped him to socialize at his workplace
as colleagues were asking him for news about the robot and
his experiences. Also within a household, Roomba and the
fact of taking part in the study triggered discussions about
future technologies, robotics, cleanliness, and the cleaning
routine. It seemed to create an awareness of the otherwise
rather unnoticed routine of cleaning of things lying around.
Most people reported Roomba made them more aware of
their clutter and encouraged them to tidy up.
We found the most interesting social relationship to the
robot with a 71 years old single woman who lived with a
dog and two tortoises. She was socially active and regularly
met friends. Her dog seemed to be one important social in-
teraction partner and she talked directly to it as if it were
a human and treated it in quite human-like ways (anthro-
pomorphism). When she saw and tried out the robotic VC,
she compared it to her dog and told us (after hesitation) she
would feel emotionally attached to the robot. Though she
did not give it a name, she talked directly to it and cared for
it more than one would have to ‘care’ for an object that can
be switched off. For instance, she phoned us because she felt
her Roomba could lack attention during the time she wanted
to go on holidays, and asked us whether she should give it
to her neighbor so that it would still be in use. For her, the
robot seemed to be a companion and she confirmed this sev-
eral times.
5 Adoption Process
Concerning temporal stages of adoption, our study generally
supports Rogers’ ‘Diffusion of Innovation’ [22]aswellas
Sung et al.’s ‘Domestic Robot Ecology’ (DRE) framework
[33]. In the following, we integrate the qualitative observa-
tions from our 6-month study into the latter initial frame-
work of domestic robot adoption. This step of analysis is a
second main contribution of our work. We are able to out-
line major similarities and some qualitative differences in
how households in Sung et al.’s study and our study adopted
(or rejected) their vacuum cleaning robot. This is an impor-
tant finding, as the two studies have been conducted with
different models of the robot and in a different cultural re-
gion. Consequently, it shows that the DRE, though still an
Fig. 3 Pre-adoption stage: main interaction between the robot (in the
middle) and the key attributes are highlighted with a bold arrow;people
expect the robot to work as a tool that performs vacuuming (robot im-
pacts related task); people expect the robot to improve the cleanliness
of the home (robot impacts environment); inspired from [33]
initial framework, can in a sufficiently robust way describe
the process of adopting a domestic service robot.
According to the four temporal steps described in Sung
et al.’s DRE, our findings are structured as follows:
1. Pre-adoption: Forming expectations
2. (Initial) Adoption: Getting the first impressions
3. Learn/Adaptation: Learning affordances and limitations
4. Use and Retention: Routine practice and maintenance
We observed different durations for these four steps for
different households. It seemed that some households al-
ready formed a strong initial decision to adopt or reject the
robot right after the first tryout while others seemed unde-
cided for about two months or even longer. In general, most
things happened during the first two weeks and two months
but only few dynamics have been observed during the last
four months. In the following, we describe how each of the
adoption phases was characterized in our study.
5.1 Pre-adoption
Pre-adoption (see Fig. 3) refers to a phase that is tempo-
rally before the purchase of the system (in our study, this
is the period from when we confirmed to a household the
participation in the study until the second visit, when we
brought the robot with us). Rogers describes this phase in-
volving two steps: knowledge and persuasion [22]. In this
phase, people learn that there exist domestic vacuum clean-
ing robots and gain some understanding and first knowledge
of it. People form expectations and attitudes towards vacu-
uming robots, e.g. they expect the robot to work as a tool
Author's personal copy

402 Int J Soc Robot (2013) 5:389–408
that performs vacuuming (see Fig. 3bold arrow, robot im-
pacts related task), and how it might improve the cleanliness
of the home, thus the robot impacts the environment (see
Fig. 3bold arrow). These are the main interactions occur-
ring in the pre-adoption stage. People’s expectations formed
during this stage can have a strong impact on their initial
reaction to it and powerfully shape the experience [11]. Our
participants had very different expectations about robots and
the Roomba, in particular. The assumptions that they made
about possible robot functions were based on previous ex-
perience with robots (when available), on articles they read
about robots in journals or newspaper (mostly about indus-
trial robots, robot toys or technical aspects), or on media
(science-fiction, novels). Participants with a lower affinity
for technology had more difficulties in describing their ex-
pectations about the robot. However, generally, people’s un-
certainty about what domestic robots are capable of and the
lack of experience with robots, created skepticism. People
were not at all sure what they should expect of a vacuum
cleaning robot. Participants were curious to try out ‘a robot’
in their home, also the elder participants. Some participants
already thought about trying out a robotic VC before be-
cause they had heard about it but the financial barrier was
too high to just purchase one without really knowing what
it would be able to do. Non of the households had tried
out a Roomba before but two families had had experiences
with other robots (LEGO mindstorms (H9), lawn-mowing
robot(H6)). When using Roomba the first time, children
in H6 thought then it was going to cut the carpet’s ‘hair’.
How powerfully expectations can shape the initial experi-
ence with a device can also be illustrated with the reaction
of a 8 years old girl in H4. After having unpacked, set up and
switched on the robot, she folded her arms, looked down and
went upstairs to her room. We later asked her what had hap-
pened and she explained she was sad because she expected
the robot to have arms, and a head and be more human-like.
Most participants expected the robot to be bigger but less
noisy as it actually was. All hoped Roomba would make
their home cleaner and take over some part of the vacuuming
(or replace the VC). Participants wished vacuuming would
become less cumbersome, they could save time and do other
things while the robot was vacuuming (multitasking). Espe-
cially pet owners wanted to decrease the amount of pet hair.
Households with children had concerns whether the robot
could work properly due to the children toys on the floor and
the expected fascination the robot would have for the chil-
dren. Parents also hoped the robot could motivate the family
members to tidy up. Two parents wanted to use Roomba to
introduce robotics to their children because they believed
robots would be part of the future and the children should
get in touch with this. Mothers of younger children were
concerned about the robot might attract too much the child’s
attention and it could be a harm for the infant or contrary the
Fig. 4 In the initial adoption stage: people watch and observe the robot
(social actors relate to the robot); the robot makes people talking about
it and evokes an impression (robot impacts social actors); the robot en-
courages people to make changes to the environment (dashed arrow);
inspired from [33]
robot might get damaged when the child would play with it.
Elder participants were much more concerned about the us-
ability of the robot and were more hesitant when using it the
first time. Expected challenges were mostly that the robot
would not clean well, e.g. in corners or around the table and
some were also afraid the robot could fall down stairs, leave
marks on the floor or bump heavily into furniture.
5.2 Initial Adoption
The initial adoption phase describes people’s first impres-
sions and initial reactions to the robot and includes the first
interactions with Roomba (see Fig. 4). In this stage, main ef-
fects included people watching the robot (see Fig. 4,boldar-
row) and also, the robot evoking an impression in the social
actors (see Fig. 4, bold arrow). Additionally, we observed
participants making changes to the environment (see Fig. 4,
dashed bold arrow) encouraged by the robot.
How did participants approach the robot and interact with
it during the first usage? We structured each household’s
initial use of the robot in several steps. A comparison re-
vealed common stops and differences in how the robot was
approached. Two families opened the packaging and assem-
bled the parts collectively (H5, H6). In contrast to this in
other families there was one person who gave instructions
from the manual while the others were sticking parts to-
gether (H4, H7). In two households the parents were pas-
sively watching how the children unpacked and set up the
robot (H8, H9). Not all households had a look at the man-
ual: some immediately pushed the start button after having
unpacked the robot, some read all or some parts of the man-
ual before switching it on (Fig. 2left and center part), and
Author's personal copy

Int J Soc Robot (2013) 5:389–408 403
some asked questions to the passively observing researcher.
What all households did (most before switching the robot
on) was that they checked its parts and tried to find out how
it actually worked: people turned the robot around, touched
the brushes and wheels and discovered the sensors and front
bumper. Some made a couple of adjustments to the space, to
ensure the robot would not damage something or get dam-
aged, e.g. they put away cables or a sheet of paper that was
lying on the floor, or pushed chairs aside. After this phase
of quick preparation which took between 1–20 minutes (but
was skipped by some households) participants pressed the
‘clean button’ to switch on the robot (we charged its bat-
teries first). When the robot started to move, most partici-
pants stepped back and some children started screaming and
jumping around. In all households the robot was carefully
observed and verbal comments were made on its abilities
(“Oh! It slows down in front of an obstacle!”)orinhope
to guide it around (“But go over there then, there are still
some crumbs!”). While the robot was running, some of the
households had to solve further issues (calm down the dog,
push away delicate objects) or were looking for more in-
formation (in the manual or online). In all households, par-
ticipants discovered the robot, interacted with it and tested
how it would react to some disturbance, for instance. The
first time the robot was touched while it was moving, peo-
ple let it bump into their foot and just after they were sure
that nothing bad would happen, they also kneed and touched
it with their hands. People picked up the robot while it was
moving to see what would happen, and some even put it
on a table to see whether it would fall down when it ap-
proached the rim. All these little experiments helped people
to gain knowledge about how the robot was working and
to form an initial decision. Overall, these first interactions
with the robot revealed a lot about social roles, responsibili-
ties, the household’s process of introducing new products in
the home, and their attitude towards the Roomba. Neverthe-
less, people’s initial reaction to the robot remained difficult
to capture. Though it was already the second visit at each
household, we had the impression that some participants felt
a bit uncomfortable and might have behavedunnaturally due
to the observer. Above all, children’s naturalness in discov-
ering the robot was notable. Most reacted with great enthu-
siasm for the moving thing but little infants were also afraid
and started crying or wanted to be back on the mother’s arm.
However, it has already been described that the first phase
with a new product is shaped by novelty effects [8,18,22,
30]. Novelty effects are the first responses to an artifact and
characterized through high levels of attention and interest,
and increased usage. In respect to domestic robots, Sung
suggests that strong effects can be observed within the first
two weeks with a robot and that after about two months us-
age beyond novelty effects can be observed [30]. Our study
confirms this suggestion. How people used and described
Fig. 5 In the learn and adaptation stage: people’s attitude towards the
robot impacts how much they try it out (social actors relate to the
robot); the robot entertains and makes people giving demonstrations
for guests (robot impacts social actors); the robot encourages people
to make lasting changes to the environment (dashed arrow); the en-
vironmental context impacts how the robot is tried out (environment
impacts robot); people need to maintain the robot which impacts its
performance (related tasks impact robot); inspired from [33]
the robot during the first two weeks was shaped by its nov-
elty. It was seen as a ‘fancy new device’, shown to other
people, and tried out in a variety of ways. When visiting the
households after they lived with the robot for two weeks,
most of them still described a fascination for it, used it fre-
quently and assessed its limitations. This changed after two
months. The majority of the households decreased their fre-
quency of Roomba usage, and described it as “just another
cleaning tool”.Aftersixmonths,twohouseholds(H7,H8)
had completely stopped using the robot (‘rejecters’), four
households (H2, H3, H4, H9) used it sporadically but could
not really take advantage of it and would not buy it (‘users’),
and three households (H1, H5, H6) had integrated the robot
in their cleaning routine (‘adopters’).
5.3 Learn and Adaptation
According to Sung et al. during the learn and adaptation
period (see Fig. 5), people try to learn more about the ar-
tifact by experimenting with complexity in use and com-
patibility with the current environment, and make necessary
changes to better incorporate the device [33]. Through this
stage, which Rongers [22] labels ‘implementation’, people
determine whether they will accept or reject further use. The
main interaction effects during this stage are highlighted in
Fig. 5. In addition to the mutual interactions between the so-
cial actors and the robot (which had already occurred in the
previous stage), now as experimenting how the robot works
begins, ‘cleaning’ (as ‘related task’) and the physical and
Author's personal copy

404 Int J Soc Robot (2013) 5:389–408
social context of the home (‘environmental context’) have a
direct impact on the robot (bold arrows in Fig. 5.
Interestingly, as already described in [29,32,33], the
robot encouraged people to make a variety of adjustments to
their home (mostly during the first two weeks of the study).
For example, people put away obstacles (wires, cables, toys,
decoration, etc.), moved furniture (e.g. so that the robot was
able to pass in between two shelves), and tidied up. One
family (H5) also bought covers for the cables of the TV and
play console and took quite some time to assemble them in
the living room. The mother said the cables had annoyed
her for a long time already, especially when vacuuming but
that the robot motivated her to actually make this adjust-
ment. A similar statement came from the woman in H2, who
kept stacks of magazines and newspapers for more than two
years in the living room: “I felt so sorry for the robot when
it bumped into the stack again and again that I just had
to put it away. Actually, I was thinking of tidying them up
before but never found it was urgent. Now I am happy it’s
out of my view. I have more space now.” The robotic VC
had a strong impact on the physical environment of people’s
homes (‘mediating factor’), more than a usual VC and trig-
gered activities to adjust that environment. Another solution
to make the apartment ‘Roomba safe’ was to adjust things
every time right before the robot was used. The whole six
months during the study, when H1 was using the robot, he
placed the coffee table on the sofa and put away the chairs
(Fig. 2right part). During the first days it happened quite
often that when he came back home after work he found
the robot stuck somewhere. He learned what had caused
the error and fixed the specific aspect. After one week he
had made his flat ‘Roomba-safe’ and was happy that he did
not have to vacuum manually anymore. H7 wanted to let
Roomba vacuum the ground floor during night, when they
were sleeping upstairs. This would have solved the problem
of having the children interacting to much with robot while
it was “doing its job”.However,therobotwastoonoisy
and woke up the parents. The family changed their strategy
to switching on the robot during the day before leaving to
go for a walk or play in the garden. However, some house-
holds were not willing to engage in the effort of learning
how to optimally use the robot and their process of adoption
stopped in this stage. Consequently, they decided to not use
the robot any longer.
5.4 Use and Retention
This step indicates the period when people begin to show
aroutinewiththerobotandseektoconfirmits‘exis-
tence’ in the household (see Fig. 6). In other words, the
user “seeks reinforcement of an innovation-decision that
has already been made, but the individual may reverse this
previous decision if exposed to conflicting messages about
Fig. 6 In the use and retention stage: the robot is used as a tool to
vacuum (robot impacts related tasks); people need to maintain the
robot (related tasks impact robot); the robot vacuums regularly and
thus changes the environment (robot impacts environment); however:
people stopped making changes to the environment and as it has be-
come optimal, it does not impact anymore how the robot is used; peo-
ple stopped giving robot demonstrations and do not relate socially to it
anymore but view it as an everyday tool; inspired from [33]
the innovation” [22]. The main interaction that occurs dur-
ing this stage is related to how far the robot improves the
state of cleanliness in the home (see Fig. 6, bold arrow
to environmental context). As the robot is now regularly
used as a tool to clean, it has a strong impact on the re-
lated task.Contrary, how well people maintain the robot im-
pacts how well it can work (see Fig. 6, bold arrows related
tasks).
For the three households that adopted the robot, this
phase began after about two months. Different use cases had
already been tried out by then and the optimal way of using
the robot was found. Participants then stuck to their strategy
and it became more or less a routine to run Roomba every
morning when leaving the home (H1, H5), and/or addition-
ally after meals, when leaving with the children or before
going to sleep (H5, H6). In H6 it had become a routine that
the father emptied Roomba’s dust bin in the evening so that
the mother could use it again the next day. During this stage,
Roomba was not perceived as a ‘fancy new robot’ anymore
but as ‘just another cleaning tool’. Depending on the adopter
group (adopter, user, or rejecter) in the use and retention
stage, households were either able to confirm the positive
impact that the robot had as a cleaning tool (H1, H5, H6), or
were still not sure about its value as an additional vacuum
cleaner but used it occasionally (H2, H3, H4, H9), or had
not even tried to use it again after the initial disappointment
(H7, H8). It was interesting that four of the households had
not formed a clear decision during the six months. When
asked whether these sporadic users would buy a Roomba,
Author's personal copy

Int J Soc Robot (2013) 5:389–408 405
Fig. 7 Children touched the moving robot first with their feet and then with hands (left); an example of initial assistance with cleaning, a boy
picking up the carpet to allow the robot to go everywhere (center); a cat exploring the opened robot (right)
they all answered “no”,however,allofthemwantedtokeep
the robot though.
6 Implications
In this section, we suggest seven implications that we de-
rived from our observations and participant’s feedback.
These seven implications are adapted to functional domestic
service robots but also match most of the factors that are de-
scribed in traditional theories for the adoption of technology
[6,22,37]. We hope these suggestions can help developers
to go one step toward acceptable, long-term HRI and can
provide them useful insights into human factors and user
needs for a vacuum cleaning robot in people’s homes.
Practical Utility/Usefulness: One of the probably most im-
portant aspect is the practical utility (usefulness, Rogers: rel-
ative advantage) of a domestic robot. Perceived usefulness is
defined as the degree to which a person believes that using
a particular technology will enhance the respective task (or
daily life in general).This includes decreasing the time for
doing the task, more efficiency, and accuracy. The majority
of households (6 out of 9) did not perceive the robot as use-
ful, as they didn’t feel their home became cleaner or they
could save some time (Rogers: observable results).
Physical Space/Environmental Context: As already de-
scribed, some physical layouts constrain the way a vacuum
cleaning robot can work and navigate around. Tiny rooms
with a lot of corners and homes with door sills are not suit-
able for Roomba, as it decreases its efficiency or can make
it impossible for it to move around autonomously. Besides
this, households generally didn’t like to make adjustments
to their home to enable the robot to work. For people liv-
ing alone it seemed easier to adjust the space but as soon as
several people share one place changes are not made easily
and it would cause some effort to get everybody’s agreement
(e.g. to put the problematic carpet away).
Ease of Use/Learn how to use Device: Perceived ease of
use refers to the degree to which a person believes that using
aparticulartechnologywillbefreeofeffort.Itincludesthe
expected effort required to utilize the system and the effort
of using it (Rogers: Simplicity). It has been described that a
huge performance benefit can outweigh the effort that users
expect to spend [6]. Though the robot was fairly ease to use,
it required to learn how to optimally use it, namely by just
letting it do its job without big interventions from the user
side. The difficulty for most participants was to overcome
their expectation that the robot would vacuum more efficient
than their traditional vacuum cleaner in even less time. The
robot needs to be used in a different way than a usual vac-
uum cleaner but not everybody was willing to learn how to
use it.
Compatibility with Habits/Routine: Another huge barrier
when it comes to integrating a robot into a daily routine is
the aspect of its compatibility with a person’s beliefs and
attitudes in respect to the context of the system. Cleaning
routines are based on personal beliefs and attitudes towards
cleaning and cleanliness. These are subjective judgments
and certainly not easy to consider in the design process
of domestic service robots. However, how a person under-
stands herself and her environment impacts how a robot is
used and accepted or rejected. One participant believed no
one could do the cleaning as efficient and properly as herself
and thus she did not accept any help with it, neither from her
husband, nor from a cleaning service, nor from a robot. The
robot was simply not compatible with her personal beliefs.
(Rogers: compatibility with existing values and practices).
Personal Value/Attachment/Reliability: Some devices in
a home might not be very practical or even hard to use, how-
ever, they find their place as they hold a great personal value
(e.g. evoke attachment). The elderly woman in H2 could not
derive much practical utility from the robot and she also was
afraid of doing something wrong with it. However, she liked
to keep it because she simply “liked” it, when it was mov-
ing and required her ‘help’ when it was tangled up in the
Author's personal copy

406 Int J Soc Robot (2013) 5:389–408
carpet. Closely related to this kind of attachment is relia-
bility which plays an important role in HRI. When people
did not trust/rely on the robot, they didn’t want to leave the
room/home when the robot was switched on. However, this
does not meet the robot’s intended way of use.
Social Compatibility/Subjective Norm: According to [39]
the subjective norm is a person’s belief that most of her im-
portant others think she should (or should not) perform the
behavior in question (e.g. using a vacuuming robot). H1’s
work colleagues usually did not talk much to him but sud-
denly showed interest in his new robot, which triggered con-
versations and even led H1 to invite some other colleagues
to his place to give him/her a robot demo. This encouraged
H1 to keep on using the robotic VC, as he experienced a
positive social impact.
Economical Compatibility/Financial Benefits: Last but
not least, a domestic appliance needs to be affordable. Not
only its shelf price but also the financial costs for mainte-
nance are taken into account. Roomba requires new brushes
fairly often when used daily. As we gave the robot for free
to our participants they did not have this financial barrier but
all of them measured the gained benefits to the expenses it
consumed. An important aspect also in respect to the envi-
ronmental responsibility concerns the energy consumption
of the robot. People do not easily accept to leave a device
plugged in all the time as costs for electricity constantly in-
crease. A study on electricity consumption and cleaning ef-
ficiency can be found in [34].
Another important aspect in making the decision upon
adopting a domestic service robot is the possibility of try-
ing it out without having significant negative implications
(financial costs, e.g.). Rogers refers to this as ‘trialability’
[22] and we think that especially with innovative robotic
technologies for the home, this is a crucial factor, as it can
decrease people’s uncertainty with (domestic) robots.
7 Conclusions
We described in detail mainly qualitative observations from
a 6-month long ethnographic study with nine households in
which we deployed a vacuum cleaning robot. The results
along with the implication that we can draw from our study
are, however, limited. This is mainly due to the very small
sample size. Further, the households were using two differ-
ent models of Roomba, which might impact have an im-
pact. Despite these limitations, our qualitative descriptions
can contribute to an understanding of Human-Robot Inter-
action in the real world. Our study also adds new and rich
examples from households around Lausanne in Switzerland
to Sung et al.’s ‘Domestic Robot Ecology’, which is based
on data from the US. This addition helps it become a more
robust framework in the domain and a useful tool for others.
To conclude, we would like to highlight some key as-
pects about living with a functional robot at home. We were
surprised to find that only three out of the nine households
adopted the robot in a durable way though we recruited only
households that were generally interested in a vacuum clean-
ing robot. At this moment, we believe that, as soon as robots
and humans are sharing the same space, they need to adapt
to each other to be a good match: People need to learn how
to use a robot in an effective way, by building trust in it and
by letting it do its intended task. This of course requires that
the robot is an acceptable agent that meets moral and ethical
standards, is safe and trust-able. Several key aspects for do-
mestic service robots and the acceptance of domestic robots,
that we found through our study can be summarized as fol-
lows:
People are curious to try out a robot in their home.
People of all ages are generally open to trying out a func-
tional robot in their home, as long as they don’t have
to expect a risk. Their understanding of what domestic
functional robots are able to do and how they can be used
is not very clear and people’s expectations are still based
on media. The lack of previous experience with a robotic
system creates uncertainty but also curiosity.
Cleaning strategies are deeply rooted habits.
How people clean is something very private and seems
to be based on personal convictions. Though differ-
ent strategies of cleaning and attitudes toward cleanli-
ness did not determine whether the cleaning robot was
adopted, some personal convictions are not compatible
with having a robot doing part of the domestic labor.
Arobotdoesnotprofoundlychangecleaningroles.
A robotic VC does not change well defined cleaning
roles in a household, thus, when the mother has been
the main responsible for vacuuming, the father will not
suddenly engage lastingly in vacuuming with the robot.
Men liked the robot more for its symbolic value while
women appreciated its function to work as a vacuum
cleaner.
Roomba is not a replacement of the vacuum cleaner.
A vacuum cleaning robot does not (yet) replace the tra-
ditional VC. However, it can reduce the amount of vac-
uuming sessions and helps people to keep their homes
clean. [35]
Beliefs are more important than the environment.
The environmental context of the home certainly plays
a role in how a robot is and can be used and further im-
pacts the process of adoption of a domestic robot. How-
ever, personal attitudes and underlying household rules
are even more important and first of all a robot needs to
be practical in the eye of the user.
Roomba is not perceived as a robot but as a tool.
People’s general attitude towards robots and robotics
does not seem to change when using a simple functional
Author's personal copy

Int J Soc Robot (2013) 5:389–408 407
robot such as Roomba. However, the experience broad-
ened most people’s mind and more importantly, how
the robot appeared changed qualitatively over time from
‘fancy new robot’ to ‘just another cleaning tool’. This
confirms the existence of novelty effects and their strong
effects on human-robot interaction and perception which
both are likely to change over time and with growing
experience with the system. After novelty effects have
worn off, Roomba was not perceived as a robot but as a
tool.
The social impact of functional robots is overestimated.
People tend to basically anthropomorphize even a sim-
ple functional robot, by talking to it directly or by using
communication traits that are comparable when relat-
ing to a pet or other human. However, the phenomenon
wears off when people become familiar with the robot.
The social impact of a vacuum cleaning robot seems to
be overestimated. Apart from children and one single el-
derly woman, people did not view or treat the robot as a
social agent.
We hope that more long-term studies on HRI and the
adoption of robots in domestic spaces will be carried out,
so that in the future meaningful robotic solutions will be ac-
ceptable as everyday objects.
Acknowledgements We th ank all o ur part i cipa n ts for th eir eng a ge-
ment in the study as well as iRobot and iRobotics GmbH Switzerland
for their support. Thanks also to the reviewers and to JaYoung Sung.
This research was supported by the Swiss National Science Foundation
through the National Centre of Competence in Research Robotics.
References
1. Bartneck C, Suzuki T, Kanda T, Nomura T (2006) The influence of
people’s culture and prior experiences with aibo on their attitude
towards robots. AI Soc 21(1–2):217–230
2. Bartneck C, Kuli´
cD,CroftE,ZoghbiS(2008)Measurement
instruments for the anthropomorphism, animacy, likeability, per-
ceived intelligence, and perceived safety of robots. Int J Soc Robot
1(1):71–81
3. Bell G (2001) Looking across the atlantic: Using ethnographic
methods to make sense of Europe. Intel Technol J Q3:1–10
4. Corbin JM, Strauss A (1990) Grounded theory research: proce-
dures, canons, and evaluative criteria. Qual Sociol 13(1):3–21
5. Dautenhahn K, Woods S, Kaouri C, Walters ML, Koay KL,
Wer r y I (200 5 ) What i s a r obot co mpani o n—fr i end, a s sista nt or
butler? In: IEEE Proceedings IROS’05, pp 1192–1197
6. Davis FD (1989) Perceived usefulness, perceived ease of use, and
user acceptance of information technology. MIS Q 13(3):319–340
7. Epley N, Waytz A, Cacioppo JT (2007) On seeing human: a three-
factor theory of anthropomorphism. Psychol Rev 114:864–886
8. Fernaeus Y, Haakansson M, Jacobsson M, Ljungblad S (2010)
How do you play with a robotic toy animal?: A long-term study of
pleo. In: Proceedings IDC’10. ACM, New York, pp 39–48
9. Fink J, Bauwens V, Mubin O, Kaplan F, Dillenbourg P (2011) Peo-
ple’s perception of domestic service robots: same household, same
opinion? In: Mutlu B, Bartneck C, Ham J, Evers V, Kanda T (eds)
Social robotics, vol 7072. Springer, Berlin, pp 204–213
10. Forlizzi J (2007) How robotic products become social products:
an ethnographic study of cleaning in the home. In: Proceedings
HRI’07. ACM, New York, pp 129–136
11. Forlizzi J, DiSalvo C (2006) Service robots in the domestic envi-
ronment: a study of the roomba vacuum in the home. In: Proceed-
ings HRI’06. ACM, New York, pp 258–265
12. Gaver B, Dunne T, Pacenti E (1999) Design: cultural probes. In-
teractions 6(1):21–29
13. Ha TS, Jung JH, Oh SY (2005) Method to analyze user behavior
in home environment. Pers Ubiquitous Comput 10(2–3):110–121
14. Heerink M, Kröse B, Evers V, Wielinga B (2009) Measuring ac-
ceptance of an assistive social robot: a suggested toolkit. In: IEEE
Proceedings RO-MAN’09, pp 528–533
15. Hüttenrauch H, Severinson-Eklundh K (2003) Fetch-and-carry
with CERO: observations from a long-term user study with a ser-
vice robot. Tech Rep IPLab-213, Royal Institute of Technology
(KTH), Stockholm
16. Kahneman D, Krueger AB, Schkade DA, Schwarz N, Stone AA
(2004) A survey method for characterizing daily life experience:
the day reconstruction method. Science 306(5702):1776–1780
17. Kanda T, Ishiguro H (2005) Communication robots for elementary
schools. Tech rep, CiteSeerX. http://citeseerx.ist.psu.edu/viewdoc/
summary?doi=10.1.1.152.3557
18. Kanda T, Hirano T, Eaton D, Ishiguro H (2004) Interactive robots
as social partners and peer tutors for children: a field trial. Hum-
Comput Interact 19(1):61–84
19. Kim H, Lee H, Chung S, Kim C (2007) User-centered approach
to path planning of cleaning robots: analyzing user’s cleaning be-
havior. In: Proceedings HRI’07. ACM, New York, pp 373–380
20. Mutlu B, Forlizzi J (2008) Robots in organizations: the role of
workflow, social, and environmental factors in human-robot inter-
action. In: Proceedings HRI’08. ACM, New York, pp 287–294
21. Pantofaru C, Takayama L, Foote T, Soto B (2012) Exploring
the role of robots in home organization. In: Proceedings HRI’12.
ACM, New York, pp 327–334
22. Rogers EM (1995) Diffusion of innovations. Simon and Schuster,
London
23. Sabelli AM, Kanda T, Hagita N (2011) A conversational robot
in an elderly care center: an ethnographic study. In: 2011 6th
ACM/IEEE international conference on human-robot interaction
(HRI), pp 37–44
24. Salvini P, Laschi C, Dario P (2010) Design for acceptability: im-
proving robots’ coexistence in human society. Int J Soc Robot
2(4):451–460
25. Schön-Bühlmann J, Freymond C, Koch D, Renfer JP (2006) Le
ménage pour lieu de travail: le temps consacré au travail domes-
tique et familial et son estimation monétaire. Actualités OFS 779-
0600, Office Fédéral de la Statistique (OFS), Neuchatel
26. Scopelliti M, Giuliani MV, Fornara F (2005) Robots in a do-
mestic setting: a psychological approach. Univers Access Inf Soc
4(2):146–155
27. Sullivan O (2000) The division of domestic labour: twenty years
of change? Sociology 34(3):437–456
28. Sung J, Guo L, Grinter RE, Christensen HI (2007) “My roomba
is rambo”: intimate home appliances. In: Krumm J, Abowd GD,
Seneviratne A, Strang T (eds) Proceedings UbiComp’07, vol
4717. Springer, Berlin, pp 145–162
29. Sung J, Grinter RE, Christensen HI, Guo L (2008) Housewives
or technophiles?: Understanding domestic robot owners. In: Pro-
ceedings HRI’08. ACM, New York, pp 129–136
30. Sung J, Christensen HI, Grinter RE (2009) Robots in the wild:
understanding long-term use. In: Proceedings HRI’09. ACM, New
York, pp 4 5 – 5 2
31. Sung J, Christensen HI, Grinter RE (2009) Sketching the future:
assessing user needs for domestic robots. In: IEEE Proceedings
RO-MAN’09, pp 153–158
Author's personal copy

408 Int J Soc Robot (2013) 5:389–408
32. Sung J, Grinter RE, Christensen HI (2009) “Pimp my roomba”:
designing for personalization. In: Proceedings CHI’09. ACM,
New York, pp 193–196
33. Sung J, Grinter RE, Christensen HI (2010) Domestic robot ecol-
ogy. Int J Soc Robot 2(4):417–429
34. Vaussard F, Bonani M, Rétornaz P, Martinoli A, Mondada F
(2011) Towards autonomous energy-wise RObjects. In: Hutchi-
son D et al (eds) Towards autonomous robotic systems, vol 6856.
Springer, Berlin, pp 311–322
35. Vaussard F, Fink J, Bauwens V, Rétornaz P, Hamel D, Dillenbourg
P, Mondada F (2013) Lessons learned from robotic vacuum clean-
ers entering in the home ecosystem. Robotics and Autonomous
Systems. (Submitted)
36. Venkatesh A (1996) Computers and other interactive technologies
for the home. Commun ACM 39(12):47–54
37. Venkatesh V (2000) Determinants of perceived ease of use: inte-
grating control, intrinsic motivation, and emotion into the technol-
ogy acceptance model. Inf Syst Res 11(4):342–365
38. Venkatesh V, Bala H (2008) Technology acceptance model 3 and
a research agenda on interventions. Decis Sci 39(2):273–315
39. Venkatesh V, Davis FD (2000) A theoretical extension of the tech-
nology acceptance model: Four longitudinal field studies. Manag
Sci 46(2):186–204
40. Young JE, Hawkins R, Sharlin E, Igarashi T (2008) Toward ac-
ceptable domestic robots: applying insights from social psychol-
ogy. Int J Soc Robot 1(1):95–108
Julia Fink is a PhD student in Human-Robot Interaction at EPFL. She
has a multi-disciplinary background in Media, Technology, Communi-
cation, and Interaction Studies. Her research interests are social impli-
cations and acceptance of robots in daily life.
Valérie Bauwens is an ethnographer, studying the process of adoption
of technology in homes and work spaces. She has a background in
business administration and psychology and has been working for 10
years in finance and business development before she founded Human
Centricity in 2010. She applied her expertise in ethnographic research
in several projects with EPFL.
Frédéric Kaplan worked ten years for Sony inventing new artificial
intelligence algorithm for the AIBO robots. He joined EPFL in 2006 to
design new kinds of interactive furniture and robotic objects. In 2008
he founded OZWE a company selling robotic computers.
Pierre Dillenbourg is professor of learning technologies at EPFL. For-
mer teacher in elementary school, he graduated in educational science
(University of Mons, Belgium). He obtained a PhD in computer sci-
ence from the University of Lancaster (UK), in the field of educa-
tional applications of artificial intelligence. His work covers various
domains of computer-supported collaborative learning (CSCL), rang-
ing from novel interfaces for face-to-face collaboration (interactive fur-
niture, tangibles, paper computing) to more cognitive projects on dual
eye tracking.
Author's personal copy