Content uploaded by Raden Agoeng Bhimasta
Author content
All content in this area was uploaded by Raden Agoeng Bhimasta on Oct 15, 2019
Content may be subject to copyright.
What causes the adoption failure of service robots? A Case of
Henn-na Hotel in Japan
Raden Agoeng Bhimasta
Institute of Service Science
National Tsing-Hua University
Hsinchu, Taiwan, 30013
agoeng.bhimasta@iss.nthu.edu.tw
Pei-Yi Kuo
Institute of Service Science
National Tsing-Hua University
Hsinchu, Taiwan, 30013
pykuo@iss.nthu.edu.tw
ABSTRACT
With the emergence of AI-powered products and services, the
hospitality industry has started to adopt service robots to transform
the guest experience. Despite this growing interest, Henn-na Hotel,
the world’s first robot hotel, recently announced to abandon half of
its robots. This study aims to unveil factors leading to the adoption
failure of service robots in the hospitality context using Henn-na
Hotel as the case study. Through mining online guest reviews from
four different leading online booking sites, we conducted thematic
content analysis on a total of 250 negative online reviews. A total of
six themes emerged from our data (e.g., human intervention,
usefulness, embodiment), illustrating various factors resulting in the
adoption failure. Based on this, we come up with six design
implications for future researchers and designers to re-think about
the interaction process between human and robots, as well as how
service robots could be better designed and used in hospitality
settings to fulfill guest needs.
CCS CONCEPTS
• Human-centered computing →
Empirical studies in HCI
;
Ubiquitous and mobile computing design and evaluation methods;
KEYWORDS
Human-robot interaction, intelligence personal assistance, service
robot, online review, thematic analysis, Henn-na Hotel
1 Introduction
For the past years, we witnessed a growing interest in transforming
hotel services by using service robots in hotel industries. Experts
predict that the growth will continue, and service robots will claim
around twenty-five percent of the workforce in the hospitality
industry by 2030 [1]. It started from Henn-na Hotel, the world first
robot hotel established in 2015, invested in various service robots
such as check-in robot, porter robot, cloakroom robot, facial
recognition, and robot in-room assistant [2]. More recently, more
and more hotels, such as Yotel Hotel [3], Flyzoo Hotel [4], Chase
Walker Hostel [5], also follow the trend. Those hotels mentioned
above invested in robot delivery service to provide room services to
its guests. The adoption of service robots can reduce the labor cost
while also bringing a new experience to the guest.
Interestingly, recent news showed that Henn-na hotel decided to
stop half of its robot service due to many complaints from both of
its staff and guests regarding the robots’ poor performance [6]. F or
instance, the check-in dinosaur robot has consistently failed on
tasks such as photocopying guests’ passport. Also, their in-room
assistant robots make guests feel annoyed by mistaking their
snoring sounds as task commands rather than providing helpful
assistance. The management concludes that these robots do not
satisfy customer needs, and instead, they annoy guests. However,
we believe that this occurred due to the gap between the customer
expectation and Henn-na Hotel’s original motives [7] of adopting
the service robots.
This paper aims to identify this gap, with the ultimate goal to
understand the factors leading to the adoption failure of service
robots through mining and analyzing online reviews from four
popular booking sites (Booking.com, TripAdvisor.com, Agoda.com,
and Expedia.com). In particular, we are interested in identifying the
salient factors leading to un-satisfaction that drives guest
disappointment toward the service robots. In our study, we define
service robots as system-based autonomous and adaptable
interfaces that interact, communicate and deliver service to an
organization’s customers, capable of accomplishing complex series
of actions and making autonomous decisions [8].
We chose Henn-na Hotel as the case study because it is widely
known as a robot hotel, and it has invested a diverse range of service
robots compared to other hotels. In this paper, we focus only on
negative reviews related to service robots. We see online reviews as
an essential information source that reveal the interactions between
guests and service robots.
The paper is organized as follows: we start by summarizing what
prior research using Henn-na Hotel as the care study target found
as well as existing work on human-robot interaction framework.
Next, we describe the study methods, foll owed by findings and
discussion. We then provide six design implications based on our
findings and conclude with stud y limitation and future work.
Permission t o make digital or ha rd copies of part or all of this work for personal o r
classroom use is granted w ithout fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for component s of this work owned by others than ACM
must be honored. Abs tracting with credit i s permitted. To copy othe rwise, or republish,
to post on servers or to redistribute to lists, requires prior specific permission and/or a
fee. Request permissions from Permissions@ac m.org.
Ubicomp/IS WC’19 Adjunct,
September 9-13, 2019, London, United Kingdom
© 2019 Association for Computing Ma chinery.
ACM ISBN 978-1-45 03-6869-8/19/09…$15.00
https://doi.org/10.1145/3341162.3350843
1107
Ubicomp/ISWC’19 Adjunct, September 9-13, 2019, London, UK
RA Bhimasta and PY Kuo
2 Related Work
2.1 The Prior Studies of Henn-na Hotel
Despite Henn-na Hotel’s popularity as one of the world’s first robot
hotels, interestingly, we found little prior work that chose Henn-na
Hotel as their case study target. Ottawa et al. [7] seek an explanation
on how robots could replace human labor and consider Henn-na
Hotel as an appropriate case study. They conducted a field survey
including an interview with the general manager of the Hotel as
well as Hui Ten Bosch corporate planning officer. They also
surveyed the areas and evaluated the robot directly.
According to the interview, the general manager claimed the
adoption of service robots are not intended to replace human
employee but to replace part of their task. Consequently, the task of
full-time human employee become more general, and all of them
shared the same task. For instance, all employees need to monitor
and maintain the service robot, drive guest, and do accounting task.
This change could reduce human employee’s emotional labor
because of less face-to-face interaction with the guests. It worth
noting that all of these strategies emphasize on the improve
efficiency of the operation and employee. In our study, we
considered the strategies mentioned above to relate with actual
guests’ reaction to the interaction with the service robots and find
potential conflicts between the strategies and the reactions.
Yu [9] aim to discover the public’s general perceptions on the use of
service robots in the Henn-na Hotel by mining 1,569 user comments
from the two most frequent viewed Henn-na Hotel videos on
YouTube. To analyze the data, the author used thematic analysis
and utilized the Godspeed dimension [10] as the pre-defined
themes. While those who commented on YouTube were not actual
guests who directly interacted with the robots, our study analyzes
customer reviews from hotel booking sites to analyze guest
interaction experience with the service robots. Therefore, we expect
that our findings will have contrary findings compare to Yu’s
findings [9].
2.2 Human-Robot Interaction Framework
The study of human-robot interaction has been discussed for more
than a decade. We recognized at least four prior studies [10, 11, 12,
13] that attempt to propose practical frameworks to help researchers
and designers analyze and evaluate human-robot interaction.
Among these four guidelines, Weiss et al. [13] proposed a more
holistic framework called USUS framework to support a user-
centric evaluation in HRI beyond p ure usability study. The
framework considers dimensions from frameworks proposed by
other researchers: usability dimension [12], social acceptance [11],
user experience [10], with the addition of societal impact. Societal
impact refers to all potential consequences from the introduction of
a service robot for the social life of the specific community (also
considering cultural differences) in term of quality of life, working
conditions, employment, and education. Further, they indicate that
researchers need to employ mixed methods to address the
evaluation of human-robot interaction adequately. For instance,
questionnaires and focus group can be used to capture social
acceptance, user experience, and societal impact but cannot
measure usability. In contrast, one may use expert evaluation and
user studies to capture usability (except the underlying utility
dimension).
While these frameworks are useful to help us identify themes
emerged from the online reviews, we did not necessarily confine
ourselves to a particular framework as we focus our perspective on
the guests. In particular, Weiss et al. [13] did not consider online
reviews as a source for evaluating human-robot interaction. As part
of the discussions, we examine how our findings relate to the USUS
framework.
2.3 Design Guidelines for Human-AI
Interaction
Recently the Microsoft research team collaborated with the
University of Washington to develop guidelines to evaluate Human-
AI Interaction by synthesizing 20 years of prior AI-Infused
researches and industry implementations. [14]. They proposed a set
of 18 design guidelines for Human-AI Interactions and organized
into four high-level categories based on when user interaction takes
place:
initially
,
during interaction
,
when wrong
, and
over time
.
Initially
, designers need to make clear what system can do and make
clear how well the system can do what It can do.
During interaction
,
the designer should mitigate social biases, match relevant social
norms, show contextually relevant information, and determine time
service based on context. Then,
when the interaction goes wrong
,
the AI should provide efficient invocation, efficient dismissal, and
efficient correction. It also needs to scope service when it doubts and
make clear why the system did what it did. Finally, the AI should
over time
remember recent interaction, learn from user behavior,
update and adapt cautiously, encourage granular feedback, provide
global control, notify user about changes, and convey the
consequences of user actions.
While Human-Robot Interaction and Human-AI Interaction might
have some differences, we hope to use these guidelines as a starting
point to propose design implications that can better support human-
robot interaction. For instance, when guests interact with robot in-
room assistance, the central brain of the robot is an AI. Hence, we
can also consider this guideline to compare and complement our
findings.
3 Methods
We extracted all online review data of Henn-na Hotel from four
popular online booking sites: TripAdvisor.com, Booking.com,
Expedia.com, and Agoda.com, using webscraper.io in May 2019. For
this study, we focus on analyzing the main content of the reviews.
Therefore, we excluded information such as ratings, review title,
photos, or guest demography. In total, we collected a total of 923
online reviews, written in English, Chinese, Japanese, and Korean.
For non-English reviews, we first translated them using Google
translation, followed by an accuracy check performed by one of the
researchers in our team.
We also manually filter out the online reviews that are irrelevant to
human-robot interaction (i.e.
“Location was very good in Hui Ten
Bosch, and the condition of the room was very satisfactory. The
surroundings and restaurant were clean and breakfast as delicious,
too”-P663
). Besides, we omit all online reviews that describe only
positive human-robot interaction experience (i.e., “
It was
remarkable about the breakthrough project called the robot hotel.
The moment when I checked in with a dinosaur staff will be staying
in my memory for very long time”-P917
). This filtering process has
1108
What causes the adoption failure of service robots?
Ubicomp/ISWC’19 Adjunct, September 9-13, 2019, London, UK
UBICOMP’19, September 2019, London, UK
led our final data set to be a total of 250 online reviews for the
subsequent data analysis.
We conduct thematic content analysis [15] with emphasize on
customer-centric perspective. First of all, we proceeded by reading
and re-reading all of the online reviews to get a sense of the online
reviews and noting down initial codes. Initial codes captured
incidents (i.e., “failure to scan passport”, “failure to process
payment”), function (“weather”, “alarm”), effect (i.e., “waste time”),
emotion (i.e., “frustration”, “disappointment”), and actors (i.e.,
“dinosaur”, “Churi”, “Siri”). Next, we collated each of the online
reviews to initial codes. Then, we further collated codes into
potential themes and re-read all of the online reviews to check if the
themes worked. For instance, the code “failure to scan the passport,”
“failure to process payment,” and “did not work properly” collated
under the theme of errors. Finally, we refine these themes by
generating clear definitions for each of them, leading to a total of
six broad themes to be discussed in the following section.
4 Results
Our data analysis identified the following six themes leading to the
adoption failure of service robots (with numbers of reviews coded
for each theme in parentheses):
human intervention
(n=69),
usefulness
(n=135),
ease of use
(n=74),
efficiency
(n=19),
embodiment
(n=43),
and
expectation gap
(n=24). We describe each
theme in details below, with accompanying quotes.
4.1 Human Intervention
Human intervention
refers to reviews indicating instances in which
guests cannot accomplish a certain goal with the service robots. A
large number of guests reported that they were unable to complete
the check-in process through the robot due to several reasons
(“….
check-in robot did not work, a human had to help us check-in ….”-
P91).
Reasons include failures to (1) identify guest’s reservation
number and name differences between reservation versus passport
names
(“…. The name you registered with them must be the same
as your passport, if not the dinosaur check-in counter would not
recognize you. We had to call for assistance to check-in ….”-P851)
,
(2) scan guest’s passport
(“…. Check in with the dinosaur robots
wasn't smooth. We needed help from a staff member. Didn't
recognize my passport scan ….”-P103)
, and (3) process credit card
payment
(“…. Human and dinosaur robots are at the front to process
my check-in. However, my credit card was not recognized many
times during the payment, and a real person came out after all, and
it took more than 20 minutes at the reception ….”-P834)
. In the
situation mentioned above, a human employee needs to assist the
guests.
Our data suggest that human intervention would negatively impact
guest experience in different ways. For instance, it increased the
time spent to complete the check-in process
(“…. Electronic self-
check-in function did not work well. We need help from a limited
number of human staffs. Waste lots of time on check in.”-P10).
Human intervention
will also be likely to affect guest expectation
and consequently ruin their experience
(“…. The reason I chose the
strange hotel was that I wanted to experience the hotel where robots
handle every job, but I met the staff at check-in and explained. It
was disappointing ….”-P458).
Interestingly, one guest shared their
determination on how they want to complete the check-in without
any help of the human employee
(“We got to the counter with two
dinosaurs and a woman robot, but it is not fully staffed by robots.
As soon as we got in, a member of the staff asked for our passports
and tried to process the check-in. We told her to step off since we
wanted to check in with the robots.”-P764).
4.2 Usefulness
Usefulness
refers to onl ine reviews talking about whether the
service robot can provide useful functions guests need. Compared
to
human intervention
,
usefulness
does not necessarily need human
assistance when the service robots cannot fulfill guest requests
(“….
Churi, the robot in the room, can hardly do anything. Siri on the
iPhone is smarter ….”-P909)
. During the check-in process, guests
were concerned about being unable to ask the service robot some
questions. In this case, the service robot was able to complete the
check-in task yet not useful for answering guests’ questions
(“…. I
thought the “dinosaurs” would be more like Siri or Alexa. They were
not-they just responded to the input from the tablet mounted on the
desk. I had questions that could not be answered by recorded tablet
input responses ….”-P79).
The concept of “usable” versus “useful” is discussed often in reviews
about the in-room assistant robots. At Henn-na, the robot is only
able to provide necessary information such as time and weather or
control in-room facilities such as turning on/off the electricity, or
answering questions that are already provided in the written
instructions. While the robot is usable, guests found them not useful.
For instance, guests suggested the robot should provide more useful
information such as forecasted temperature and chance of rain
(“….
When asked for tomorrow’s weather report, the answer was 'fair.'
It could answer with more useful information such as the expected
temperature and chance of rain ….”-P854).
Another
guest suggested
that the in-room assistant robot should offer more conversational
interactions with guests
(“…. There was a small robot named Churi
in the room, but I was disappointed with its low level of
conversational skills. I did not find it useful to provide information
about time and weather, or to turn the electricity on / off (but not to
operate the TV or air conditioner). I think that it would be more
interesting if it can show concierge skills ….”-P807).
We see that
guests tend to compare the performance of hotel’s in-room robots
with that of existing personal assistant robots such as Siri and Alexa
(“…. Robot in room was not very smart and had minimal options.
Siri, Google, and Alexa are all much smarter and more fun ….”-P89)
.
Lastly, there were many cases where the robot in-room assistant,
Churi, mistakenly interpreted guests’ conversations (
“….
Unfortunately, I feel that the technology is not there yet. The robot
in our room is irritating. It speaks when we are in a conversation,
but it could not help us when we needed it….”-P780.
) and sound as
request commands (“….
There was a robot in our room, Churi. I talk
to it and turn on and off the TV and lights. However, suddenly in
the middle of the night, Churi
said, ' I could not hear you, could you
say it again?' and I was surprised and jumped up. It is a hotel where
guests are awakened many times and lack of sleep for many times.”-
P739).
Ultimately, guests felt frustrated by this mistake.
(“…. The
rotten robot on the nightstand woke us up for every 5 minutes, and
we suggest to unplug the robot ….”-P808)
4.3 Efficiency
Efficiency
refers to whether guests perceive service robots as
efficient enough to accomplish specific tasks (i.e.,
“…. I thought that
it took too much time to check in ….”-P197
). Guest spent a long time
1109
Ubicomp/ISWC’19 Adjunct, September 9-13, 2019, London, UK
RA Bhimasta and PY Kuo
to accomplish a particular goal than expected because there was an
error was occurred
(“…. Checking in was irritating. The computer
system was not working corre ctly, and it took way longer than it
should have ….”-P48).
In addition, the service robot transformed the
simple task such as turning off/on lights to be more complicated.
(“…. to turn the lights on and off in the room took ages. As the robot
that turns it on did not work properly and kept o
verheating ….”-
P88). Furthermore, the longer the time required to accomplish a
certain task, it might subsequently affect guest emotion. (“…. Auto
check-in takes a frustratingly long time, making the novelty robot
aspect of the hotel disappear ….”-P90).
Interestingly, the nature of a certain type of service robot makes
guests perceive they are less efficient. This was the case with the
porter robot which tends to move slowly in nature instead of due to
an error
(“…. I think it is not necessary to use an automatic luggage
carrier that delivers luggage to the room as it is slow ….”-P823).
4.4 Ease of Use
Ease of Use
refers to reviews addressing whethe r the interface
allows users to become familiar with it quickly, and be able to make
good use of its features. In most cases, guests only reported the
difficulty to interact with the service robots without further
explaining the detail (i.e., “….
Various procedures took time to
understand the mechanism ….”-P870).
In some cases, guests gave
further explanations related to the touch screen issue (
“…. The
check-in screen was a bit confusing and hard to type ….”-P418),
and
the keyboard position issue
(“…. Although the touch panel on the
reception desk says, 'Please write freehand,' I was supposed to use
the keyboard for input. Improved keyboard position. I was not able
to see with the keyboard whether it was input properly.”-P191).
Besides, guests felt that it was inconvenient when the system
requires them to input much information through kiosk
(“…. It was
inconvenient that I had to respond to many questions asked by the
kiosk ….”-P283).
Interacting with service robots via voice command was also a
challenge. The service robot has issues such as low accuracy to
recognize user input precisely (
“…. Check-in confirmed the name by
voice input on the terminal, but did not recognize it easily …. ”-
P745),
and language barrier difficulty to recognize foreigners’
Japanese pronunciation
(“…. My kid s have fun with the tablet
provided and talking in Japanese to the doll robot –
Churi
chan, but
do not expect to have any proper outcome coz the robot only
response in Japanese….”-P825)
The accuracy of facial recognition was also problematic (i.e.,
“…. I
tried to use facial recognition to open the room door. However, its
accuracy was low. Therefore, I decided just to use the card key ….”-
P761).
Some guests reported that adults were more likely to be
recognized by the system than children (
“… The face of the child
seems to be hard to recognize, and only one person out of three is a
little disappointed.…”-P833).
More importantly, this could lead to
security and privacy issue when the facial recognition feature
mistakenly grants access to a stranger
(“…. The facial recognition
didn’t guarantee better security. After registering with my face,
when I tried to authenticate with the face of a non-registered friend,
it opened. This is no good ….”-P778).
4.5
Embodiment
Embodiment
refers to morphology [13], encompassing the whole-
body communication with both verbal & non-verbal behaviors. In
Henn-na Hotel, the dinosaur was very popular among guests. The
form of dinosaur raised guest expectation to a higher leve l compared
to the self-check-in experience when using kiosk. For instance,
guests wished to be able to have a more interactive conversation
with the dinosaur robot to satisfy their expectation toward
embodiment
(“…. The puppets that we see at the front desk move
and have a speaker but they are not interactive. They tell us to fill
out a paper form and then step over to a machine similar to a train
ticket to check in and receive our key.”- P854).
4.6 Expectation Gap
Expectation gap
speaks to online reviews addressing discrepancies
between their prior beliefs and perceptions and their real experience
after arriving at the hotel (
“…. The place is not a high tech as
advertised! ….”-P120).
Some guests reported that they were
interested of Henn-na Hotel after seeing a television show
(“….
Nothing like the original Henn-na Hotel in Nagasaki that we saw
on TV, when Joanna Lumley did a program on Japan, ….”-P123),
or
after reading online reviews
(“Read many mixed reviews but due to
the novelty component, I want to try it. Not the easiest place to get
to but ended up catching the train from Fukuoka airport with only
one transfer ….”-P761).
However,
the gap occurred when the reality
did not meet their expectation
(“…. The robots are just a marketing
gag - only able to speak a few sentences and tell us to use the
machine next to them to check in. Promoted as a non-human hotel
but we had a person by our side asking to take passport copy seconds
after we started the check in at the machine. Also, cleaning is not
done by robots, and even the robot lawn mower is followed and
monitored by a person. Very disappointing since we only came to
see the robots and did not care about the amusement park at all. ….”-
P828).
5 Discussion
In general, our findings reveal that salient factors leading to the
adoption failure of service robots include usefulness (N=135), ease
of use (N=74), human intervention (N=69), embodiment (N=43),
expectation gap (N=24), and efficiency (N=19). We find that the
themes emerged from the customer reviews are inter-correlated.
These can be seen in several instances but not limited to: correlation
between “
human intervention”
and “
expectation gap” (“…. The
reason I chose the strange hotel was that I wanted to experience the
hotel where robots handle every job, but I met the staff at check-in
and explained. It was disappointing ….”-p458),
between
“embodiment”, “usefulness”, and “expectation gap”
(“…. I thought
the “dinosaurs” would be more like Siri or Alexa. They were not-
they just responded to the input from the tablet mounted on the
desk. I had questions that couldn’t be answered by recorded tablet
input responses ….”-p79).
In addition to these
correlations
, we also
found that “usefulness” tends to co-occur or be reported together
with “ease of use” (n=27), “human intervention” (n=23), and
“embodiment” (n=21) in the reviews. The point here is not to make
causal relationships
among these themes, but to encourage
researchers and designers to further reflect on these
correlations
and
co-occurrences
for future improvement.
An important question emerged from our findings - should hotels
use service robots to replace all prior service touchpoints taken care
by manpower, or selectively use them for particular tasks? As
human intervention is mentioned by the majority of guests as a
1110
What causes the adoption failure of service robots?
Ubicomp/ISWC’19 Adjunct, September 9-13, 2019, London, UK
UBICOMP’19, September 2019, London, UK
factor leading to service failure during the check-in process, we
believe it needs further investigation whether service robots should
be adopted to complete the check-in tasks. This echoes with a recent
statement by Sawada, a manager at the Henn-na Hotel, “
When you
use robots, you realize that there are places where they are not
needed – or just annoy people”
[17]. It is indeed a big challenge to
design a check-in robot, yet we believe AI algorithms could help
eliminate task failures during the check-in process.
Compared to Yu’s findings [9], our results show some similarities
and differences. One of the themes “embodiment” is similar to
“animacy – mechanical”, which refers to when robots are only used
for decoration. Also, the theme “usefulness” is identical to
“perceived intelligence”. In her study, 73.5% of people who viewed
the video of Henn-na Hotel disliked the robot concept by
commenting using negative words such as “creepy”, “scary”, and
“weird”, which is contrary to our customer reviews. This
contradiction could be due to that [9] focuses on people’s online
perceptions of the hotel (without staying at the hotel) whereas our
study focuses on guests’ actual experience. However, our findings
reveal that this type of perception prior to hotel stay tends to lead
to guests’ expectation gap.
Our data shared similarities with the USUS framework proposed by
[13]. The themes emerged from our data (embodiment, efficiency,
ease of use) resemble the “embodiment”, “efficiency”, “learnability”,
and “emotion” dimensions in the USUS framework. Further, our
findings suggest the importance of considering the
usefulness
of
service robots, in addition to usability. In other words, the current
robots should offer features that better tailor to guest needs while
ensuring the ease of use when interacting with them.
Additionally, our findings suggest that there was a discrepancy
between hotel strategy [7] and guest expectation. At Henn-na Hotel,
the service robots were advertised as its unique selling point,
making guests set their expectation too high. Henn-na Hotel seemed
unaware of this high expectation and the fact that guests could have
varying tolerance levels toward robots’ defects (e.g., low
conversational skill, unable to answer questions other than pre-
programmed content)
Moreover, our results speak to several important constructs from
the human-AI interaction guidelines recently proposed by Microsoft
[14]. For example, one guideline emphasizing
the lack of efficient
invocation
could explain one failure scenario of the service robot
(e.g., the in-room assistant robot “Churi” talked to guests when the y
are not supposed to talk). Also, our themes, “embodiment” and
“expectation gap”, relate to two other guidelines during the “initial
interaction” stage:
make clear what system can do
, and
make clear
how well the system does what it can
. In the following Design
Implications section, we will further discuss how service robots can
be better improved in hotels given these high-level constructs.
6 Design Implications
Our findings aim to help hotel managers, researchers, and designers
improve the design of service robots, particularly in hospitality
settings. While the priority of hotel operation is service efficiency
and novelty, the adoption should not ignore whether robots fulfill
user needs and expectations during this paradigm shift (traditional
workforce service model to the service robot model). Otherwise, it
will be hard to engage guests in the long run.
First, our analysis suggests that there should be a distinction
between self-service technology and service robots. In other words,
service robot is not merely a self-service technology wrapped with
robot look. While the look of robots can attract guests, the
embodiment of service robot will also affect higher expectation on
how guests can interact with the robot with the same goals.
Consequently, during the interaction, guests might prefer to interact
with the robot through verbal, expression, and gesture interaction
as they interact with human employees rather than interact through
screen panel. Therefore, we suggest that more multimodal
interaction features can be integrated into future service robots in
hotels.
Second, designer should carefully identify the user flow. This
identification of user flow will come handy to identify what steps
are needed, what functions are needed, and what are the possible
errors during each step when interacting with a robot. Also,
designer should be able to identify all possible actions that guests
might take. For instance, in the case of check-In, the robot cannot
identify foreign passport. The designer of robot failed to implement
that feature; hence , human need to intervene each time when the
situation occurred. In the case of the in-room robot assistant, the
features offered such as weather and time information are not what
guests need. The designer also failed to design a mechanism to
tackle situations such when guests accidentally initiated the
conversation with the robot (i.e., Churi). In results, the robot Churi
falsely identifies snoring or television sound like a command.
Third, we suggest that designers see service robots in hotels as
personal intelligent assistants. Typically, guests treat human
employees as their personal assistants by asking some questions or
recommendations about attractions, restaurants, souvenirs, or
shops. It can happen anywhere anytime, during check-in with the
receptionist, during the trip to the room with bellboy who delivers
the luggage, or during a room call. While it might not be their task
to answer such questions and might not be able to give good
answers, the guests still have some expectation for some answer.
Designer could develop conversational algorithms that cater to this
need, and provides accurate and personalized information to guests.
In other words, adding “personal touch” to robots by enhancing its
intelligence in answering a broader range of questions from guests.
Existing personal assistant robots such as Siri and Alexa, serve as
good points of reference in this regard.
Fourth, we encourage designers and researchers to investigate
further whether human intervention needs to be integrated into the
interaction process with service robots. The human intervention
corresponds to one of the four aspects of Microsoft’s human-AI
interaction design guidelines, examining what AI-powered systems
should do when things go wrong. It is especially important in the
case of required activities such as the check-in process – a process
during which errors are most likely to occur with service robots yet
one of guests’ most anticipated feature at robot hotels. This
consideration also applies when the failures happened. The robot
should be able to provide clear messages about the errors, and be
able to provide appropriate recovery (e.g., having a reset
button/option on the robot, providing a backup solution in the
system for each task completion without involving humans).
Fifth, designers also need to better manage guest expectations. The
designer (and also marketer) should demonstrate what the robot can
do and cannot do. It is particularly crucial since the embodiment of
the robot could create a misleading expectation. Other than the
1111
Ubicomp/ISWC’19 Adjunct, September 9-13, 2019, London, UK
RA Bhimasta and PY Kuo
morphology of the robot, designers need to consider many other
details, including voice characteristic, word choice, speed of speech,
facial expression, body gesture, and culture. By considering all of
these details, it might be possible to create a sense of hospitality in
the service robot. This is especially interesting, in a country like
Japan, which famous for its
omotenashi
culture.
Last but not least, we challenge researchers and designers to think
further about the future of in-room technology and design.
Compared to reception or porter robots, which are relatively
constrained with its functionalities, the design space of hotel’s in-
room service robots is not fully explored yet. Features such as
enhancing the capability to answer more diverse set of questions
and entertaining guests while they relax in the room (e.g., dance ,
music, signing, and game plays) worth further exploration.
7 Conclusion, Limitations, and Future Work
We identified six inter-correlated factors leading to the adoption
failure of service robots in Henn-na Hotel through mining negative
guest reviews from four different online booking sites. These factors
included human intervention, usefulness, efficiency, ease of use,
embodiment, and expectation gap. We offer six design implications
to help researchers and designers improve the current adoption of
service robots, especially in the hotel sector.
Our study has several limitations. First, we used one case study
(Henn-na Hotel) as the starting point to identify factors leading to
service failure. Further validation of our findings across different
robot hotels is needed to generalize our implications. Second, we
realize that online reviews do not provide detailed explanations of
guests’ negative experience. We believe paring the online review
data together with guest interviews and field observation is an
important next step. Third, the online guest reviews did not
represent all guests who have stayed at Henn-na Hotel. In reality,
the distribution of negative experience could be different. For
instance, the number of guests who reported the misbehaviors of in-
room robot assistant might exceed the number of people who
reported via online reviews. Lastly, as part of the future work, we
would like to further contribute to the areas of human-robot
interaction and ubiquitous personal assistance by further exploring
the design space of in-room service robots in the hospitality context.
ACKNOWLEDGMENTS
We’d like to declare that all authors have contributed equally to the
manuscript. We also thank Jyun-Cheng Wang for his support to this
research project.
REFERENCES
[1]
J
Bowen, & C Morosan. 2018. Beware hospitality industry: the robots are coming.
Worldwide Hospitality and Touris m Themes, 10 (6), 726-733.
https://doi.org/10.1108/whatt-07-2018-00 45.
[2] The Guardian. 201 5. Inside Japan’s first robot-staffed hotel. Retrieved on April 7,
2019 from https://www.theguardian.com/travel/2015/aug/14/japan-henn-na-
hotel-staffed-by-robots.
[3] D Hochma n. 2018. Robot Butlers Are Taking Over Hotels (Good News: No
Tipping Requ ired). Retrieved on April 7, 2019 from https://www.forbes.com/
sites/davidhochman/2018/02/23/robot-butlers-are-taking-over-hotels-good-news
-no-tipping-required/#65d6d99f 7636.
[4] T Brennan. 2019. Introducing Alibaba’s Flyzoo Future Hotel. Retrieved on April
7, 2019 from https://www.alizila.com/introducing-alibabas-flyzoo-future-hotel/.
[5] Y Huang. 2016.Taiwan hotel adopts s elf-check-in, pursues luggage-carrying
robots. Re trieved on April 17, 2019 from https://meet.bnext.com.twintl/
articles/view/38 085.
[6] S Shead. 2019. World's First Robot Hotel Fires Half Of Its Robots. Retrieved on
April 7, 2019 from https://www.forbes.com/sites/samshead/2019/01/16/worlds-
first-robot-hotel-fires-half-of-its-robots/#5 67fb886e1b1.
[7] H Osawa, A Ema, H Hattori, N Akiya, N Kanza ki, A Kubo, T Ko yama, and R
Ichise. 2017. What is Real Risk and Benefit on Work with Robots?: From the
Analysis of a Robot Hotel. In Proceedings of the Companion of the 2017
ACM/IEEE International Conference on H uman-Robot Interaction (HRI '17).
ACM, New Yor k, NY, USA, 2 pages. https://doi.org/10.1145/3029798.3038312.
[8] J Wirtz, PG Patterson, WH Kunz, T Gruber, VN Lu, S Paluch, A Martins, 2018.
Brave new world: service robots in the frontline". Journal of Service Management,
29 (5), 907-931. https://doi.org/10.1108/JOSM-04-2018-0119.
[9] CE Yu. 2019. Human like robots as employees in hotel indu stry: Thematic cont ent
analysis of online reviews. Journal of Hospitality Marketing & Management.
https://doi.org/10.1080/19368623.2019.1592733.
[10] C Bartneck, D Kulić, E Croft, S Zoghbi. 2008. Measurement Instruments for the
Anthropomorphi sm, Animacy, Likeab ility, Perceived Intelligence, and Perceived
Safety of Robots. International Journal of Social Robotics, 1(1), 71-81.
https://doi.org/10.1007/s12369-008-0001-3.
[11] M Heerink, and B Kröse, B Wielinga, and V. Evers. 2006. Studying the acceptance
of a robotic agent by elderly users. International Journal of Assistive Robotics and
Mechatronics, 7 (3), 25–35.
[12] A Steinfeld, T Fong, D Kaber, M Lewis, J Scholtz, A Schultz, and M Goodrich.
2006. Common me trics for human-robot interaction. in HRI ’06: Proceedings of
the 1st ACM SIGCHI/SIGAR T conference on Human-robot interaction, pp. 8
pages. ACM, New York, NY, USA.
[13] A Weiss, R Bernhaupt, M Lankes, and M Tscheligi. 2009. The USUS evaluation
framework for human-rob ot inte raction, AISB200 9: Proceeding s o f the
Symposium on New Frontiers in H uman-Robot Interaction, 4 (1), 11-26.
[14]
S Amershi, D Weld, M Vorvoreanu, A Fourney, B Nushi, P Collisson, J Suh, S
Iqbal, PN. Bennett, K Inkpen, J Teevan, R Kikin-Gil, and E Horvitz. 2019.
Guidelines for Human-AI Interaction. In Proceedings of the 2019 CHI
Conference on Human Factors in Computing Systems (CHI '19). ACM, New
York, NY, USA, 13 pages. https://doi.org/10.1145/3290605.3300233.
[15] V Braun & V Clarke. 2006. Using thematic analysis in psychology. Qualitative
Research in Psychology, 3 (2), 77-101, https:// doi.org /10.11 91/14780
88706qp063oa.
[16] DA Norman. 1998. The phycology of everyday things. Basic Books: New York,
NY, US.
[17] M E hrenkranz. 2019. Robots Ruin Robot Hotel. Retriev ed on Jun 20, 2019 from
https://gizmodo.com/robots-ruin-robot-hotel-1831772555.
1112
