Effect on User Impression of Robot’s Task Dependent Uniform

Kazunari Yoshiwara

and Kazuki Kobayashi

Graduate School of Science and Technology, Shinshu University, Nagano 380-8553, Japan

Academic Assembly, Shinshu University, Nagano 380-8553, Japan

Keywords:

Uniform Changing Robot, Robot Clothing, Housekeeping Robot.

Abstract:

In this paper, we investigated the impressions of users when a humanoid robot is dressed in uniforms that

match tasks. Increasingly more robots, such as humanoid robots, can perform multiple tasks on a single

machine. In such a case, it is important to design the appearance of robots and their uniforms according to

the content of the task because these factors are known to express their level of expertise and technical ability.

In our proposed method, a robot changes its appearance dynamically by replacing its uniform with one that

matches a task before performing it. The results of our simulator-based experiments indicated that, compared

with robots that do not wear a uniform, robots that changed their uniform to match a task were interpreted as

being more appropriate, even after making mistakes.

1 INTRODUCTION

In recent years, the range of applications for robots is

rapidly expanding because of advances in not only the

industrial ﬁeld, but also ﬁelds such as medical care

(Heerink et al., 2016), nursing (Huisman and Kort,

2019), agriculture (Hejazipoor et al., 2021), forestry

(Parker et al., 2016) and ﬁsheries (Takagi et al., 2016),

and education (Belpaeme et al., 2018). When robots

become more widespread than they are today, their

appearance will be one of the most important factors

because it is an element that can provide clear infor-

mation about a robot and its intended function.

Many researchers have been studying the appear-

ance of robots. Złotowski et al. (Złotowski et al.,

2020) reported that the appearance of a robot is im-

portant for people in regard to their perceptions of its

capabilities and preferences for it to perform a spe-

ciﬁc task. Komatsu et al. (Komatsu and Yamada,

2008) reported that the same information is inter-

preted differently by users depending on the appear-

ance of a robot. In a prisoner’s dilemma game on a

computer, Kiesler et al. (Kiesler et al., 1996) found

that users who had owned a dog were more likely to

cooperate with a dog-shaped robot than were those

who had not. Komatsu et al. (Komatsu and Yamada,

2011) reported that an adaptation gap, that is, a dif-

ference between the functions users expect based on

a robot’s appearance and the functions users perceive

through actual interactions, signiﬁcantly affects robot

evaluations. Song et al. (Song and Yamada, 2019)

investigated users interpretation for LED expression

by a cleaning robot. Although these ﬁndings indicate

that the appearance of a robot leads to signiﬁcant bi-

ases among users, few studies have investigated robot

clothing or software agents.

Here, we focus on clothing in terms of a robot’s

appearance and investigate the effects of changing

clothing to match a task on the impressions of users

and their evaluations of its work performance. The

public has shown interest in robotic clothing, and

an event on robot fashion has already been held

From a functional perspective, clothing provides in-

sulation against cold and protection against ultravi-

olet rays, wind, rain, and other environmental haz-

ards (Havenith, 2002). Also, as social effects, uni-

forms can convey expertise, position, qualiﬁcations,

and membership in a group (Joseph and Alex, 1972).

Singer et al. (Singer and Singer, 1985) investigated

the effects of police uniforms and found that police

ofﬁcers who were photographed in their uniforms

were perceived as more competent, reliable, intelli-

gent, and helpful compared with a face-only photo-

graph or a full-length photograph dressed in civilian

clothing. Shao et al. (Shao et al., 2004) investi-

gated the relationship between consumers’ motivation

to buy and employees’ clothing and found that more

appropriate employee clothing improved customers’

motivation to buy.

https://www.nippon.com/en/views/b00911/

Yoshiwara, K. and Kobayashi, K.

Effect on User Impression of Robot’s Task Dependent Uniform.

DOI: 10.5220/0010684500003060

In Proceedings of the 5th International Conference on Computer-Human Interaction Research and Applications (CHIRA 2021), pages 90-97

ISBN: 978-989-758-538-8; ISSN: 2184-3244

Although there are not many examples of research

that allow robots to wear clothes, Ledge et al. (Legde

and Cunningham, 2019) have reported that the virtual

agents’ personality characteristics that users interpret

are inﬂuenced by the agents’ clothing. However,

their investigation did not cover dynamically chang-

ing clothing. By adapting the effects of a uniform to

robots, functionally, the robot will be prevented from

being soiled by substances such as oil, and socially,

the robot will express its specialty as well as a hu-

man would. In addition, since autonomous robots can

carry out tasks without clear instructions, the expres-

sion of clothing may become a natural way to transmit

such information to the user, in addition to the actual

function. For these reasons, uniforms may be beneﬁ-

cial as an appearance effect for robots. Therefore, this

simulation-based study aimed to investigate the pos-

sible impression effects on users of a robot changing

its uniform to match a task.

2 SIMULATOR-BASED UNIFORM

CHANGING ROBOT

We developed a uniform changing robot and an origi-

nal housekeeping simulator on a Web-based platform.

2.1 Uniform Changing Robot

Figure 1 shows the robots we designed in the house-

keeping simulator, one wearing no uniform and the

other wearing a task-speciﬁc uniform. The robot’s

appearance is inspired by Pepper, the widely known

semi-humanoid robot manufactured by SoftBank, be-

cause Pepper has a commonly known design that was

developed considering the uncanny valley (Pandey

and Gelin, 2018). The uncanny valley is a concept

in which an artifact resembles an existing organism

so closely that it causes people to have an emotional

response (Mathur and Reichling, 2016; Ho and Mac-

Dorman, 2010). This effect is one of the important

design guidelines for the robot’s appearance.

The uniforms were designed based on the robot’s

three housekeeping tasks: cleaning, washing, and

cooking. In the cleaning task, the robot wears a blue

cloth and a hat like a building cleaner. In the washing

task, the robot wears an apron and a bandana. In the

cooking task, the robot wears a chef’s cap and coat.

2.2 Housekeeping Task Simulator

Bugmann et al. reported that housework was the most

important job for domestic robots, as many people

want robots to perform the cleaning, washing, and

No Uniform Cleaning

Uniform

Washing

Uniform

Cooking

Uniform

Figure 1: Appearance of robot in each task.

cooking (Bugmann and Copleston, 2011). Moreover,

Kang et al. conducted a study based on the three

scenarios of cleaning, washing, and cooking in re-

search to investigate the necessary functions for do-

mestic robots (Kang et al., 2020). Based on those pre-

vious studies, the present study designed household

cleaning, washing, and cooking tasks for the robot to

perform. Figure 2 shows the developed housekeeping

simulator consisting of a room in an ordinary home.

The robot in the simulator can move up, down, left,

and right within the room to perform the three tasks of

cleaning, washing, and cooking. Objects in the simu-

lator such as the refrigerator and kitchen are animated

when the robot manipulates them. Figure 3 shows the

ways the kitchen is animated before, during, and after

cooking.

2.3 User Interface

The right side of Figure 2shows a remote control and

messages for users. Users request the robot to per-

form a task indicated by a button on the remote con-

trol. The tasks to be performed are automatically dis-

played on the screen in the message area, and the

users can then manipulate the remote control.

2.4 Robot Behavior

Figure 4 shows the robot behavior for each task. The

robot changes its uniform whenever it begins to per-

form a task at the initial position near the door in the

room. The tasks to be performed by the robot were

set as cleaning, washing, and cooking. In the cleaning

task, the robot and moves slowly across the ﬂoor and

vacuums. As the robot cleans, the scattered debris on

the ﬂoor gradually fades in color until an animation

plays in which the ﬂoor is shiny. In the washing task,

the robot moves to a clothesline and collects washed

clothes before moving to place them on a table. In the

cooking task, the robot moves in front of the refrig-

erator and removes food before using the frying pan

and bringing the prepared food to a table.

Effect on User Impression of Robot’s Task Dependent Uniform

Refrigerator

Kitchen

Door

Table

Washed clothes

Messages

for user

Robot

Remote control

Please send a command

of "cleaning"

Garbage

Figure 2: Housekeeping task simulator.

Figure 3: Animations in kitchen.

3 EXPERIMENT

To investigate user impressions of the robot chang-

ing its uniform according to the requested task, we

performed experiments with 20 male university stu-

dents (mean age = 22.1, standard deviation = 0.72).

The participants were requested to send commands to

the robot in the simulator according to the instructions

displayed on the screen.

3.1 Task Incompleteness

In the experiment, the robot was set to not accomplish

each task completely to increase the degree of free-

dom of the subjective evaluations of its behavior and

clarify the effects of uniform changing. In the clean-

ing task, a few pieces of debris were left on the ﬂoor.

In the washing task, a few hangers were dropped and

left on the ﬂoor. In the cooking task, the pans and cut-

ting boards used for cooking were left in the sink, and

some small pieces of food were left around the range.

3.2 Task Evaluation by the Participants

At the end of each task, to conﬁrm the users’ indi-

vidual interpretations of and satisfaction with the per-

formed task, the participants were asked to input their

judgment of the quality of each task performed by the

robot along with their subjective satisfaction. Figure

5 shows the evaluation window and score sliders for

each task. The score ranged from 0 (unsatisﬁed or

low quality) to 9 (satisﬁed or high quality). As soon

as a task was ﬁnished, the evaluation window was dis-

played at the top of the simulator screen. When a

participant ﬁnished inputting scores, messages with

further instructions were displayed at the top of the

remote control.

3.3 Experimental Procedure

The participants were instructed about the experiment

as follows.

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

①

②

③

④

⑤

⑥

Cleaning Task

①

②

③

④

⑤

⑥

Cooking Task

*After moving to the table

①

②

③

④

⑤

⑥

Washing Task

*After moving to the table

Figure 4: Robot behavior in each task.

• It will take about 15 minutes to complete this ex-

periment.

• Manipulate the remote control on the screen to re-

quest that the robot perform a task.

• Observe the behavior of the robot and check that

the task was performed appropriately.

• Whenever the task is completed, input your sub-

jective satisfaction and quality scores.

We did not provide any information about the

robot’s uniform or task incompleteness. During the

experiment, the participants sat in a chair in front of a

24-inch monitor on a desk.

3.4 Experimental Conditions

The experiments were performed under two condi-

tions: a uniform changing condition and a no-uniform

condition. We used a between-participants experi-

mental design. In the uniform changing condition,

the robot changes its uniform to match the task before

starting it. After ﬁnishing the task, the robot moves

back to the initial position near the door in the sim-

ulation room and removes the uniform. In the no-

uniform condition, the robot performs a task while

wearing no uniform. The order of the robot’s tasks

was randomized under each condition.

Effect on User Impression of Robot’s Task Dependent Uniform

Figure 5: Evaluation window and score sliders for each

task.

3.5 Evaluation Indices

The participants were asked to reply to a question-

naire survey after the robot had ﬁnished all its tasks.

The questionnaire was composed of seven Likert-type

scaled questions and a free writing area. The ques-

tionnaire items were based on part of Heerink’s model

(Heerink et al., 2010), which includes Davis’s Tech-

nology Acceptance Model (TAM) (Davis, 1989) and

items related to perceived enjoyment and trust. The

items related to trust, perceived usefulness, intention

to use, and social presence were used in the present

questionnaire. In addition, the items described in

section 3.2 related to the clothing change, subjective

quality of the task, and subjective satisfaction of the

task were used. The independent variable was uni-

form changing or not, and the dependent variables

were the satisfaction score, the quality score, and the

questionnaire items.

4 RESULTS

The Mann–Whitney U test was used to conduct the

statistical analysis for the task evaluation (satisfaction

and quality) and questionnaire items after the experi-

ment (Mann and Whitney, 1947). This is because the

number of experimenters was small (N = 10 for each

condition), so it was difﬁcult to judge whether to fol-

low a normal distribution (Nachar, 2008).

Table 1 shows the satisfaction scores for the robot

tasks and the results of the statistical analysis. Al-

though no signiﬁcant difference was seen in the sat-

isfaction scores between conditions, all mean values

for the uniform changing condition were higher than

those for the no-uniform condition.

Table 2 shows the quality scores for the robot tasks

and the result of the statistical analysis. We conducted

the Mann–Whitney U test for both scores and found

no signiﬁcant difference between conditions. How-

ever, all mean values for the uniform changing con-

dition were the same or higher than those for the no-

uniform condition.

Table 3 shows the questionnaire scores and the

results of the statistical analysis using the Mann–

Whitney U test. A signiﬁcant difference was found

between conditions for Q4 (U = 18.0, d f = 18.0, p =

0.015).

Table 4 shows the frequency and order of occur-

rence of each task throughout the experiment. Al-

though the order of the tasks were randomized, no

extreme bias was observed.

5 DISCUSSIONS

5.1 Effects of Changing Uniforms by

Robots

No signiﬁcant differences in satisfaction and quality

scores were found between conditions. However, all

mean values for the uniform changing condition were

the same or higher than those for the no-uniform con-

dition. These results suggest that users of the robot

under the uniform changing condition perceived the

same or better satisfaction and quality compared with

the robot under the no-uniform condition, and that no

users experienced any negative effects of the uniform

change.

A signiﬁcant difference was found between con-

ditions for Q4. This result shows that the participants

interpreted the behavior of the robot under the uni-

form changing condition as being more appropriate,

even if the task was incomplete, compared with that

under the no-uniform condition. One participant’s

free comment (”I felt like the robot was choosing the

right clothes for the job”) also supports the effect of

the uniform changing condition. Although no sig-

niﬁcant differences were seen in the other items be-

tween conditions, Q5 and Q9 had higher values under

the uniform changing condition. These results sug-

gests that the participants noticed the incompleteness

of the tasks more clearly under the uniform chang-

ing condition than under the no-uniform condition;

however, the participants preferred the robot in the

uniform changing condition over the robot in the no-

uniform condition. Regarding whether the uniform

was appropriate, the results of Q1 indicated that more

participants recognized the change of clothing. How-

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

Table 1: Satisfaction scores for robot tasks.

Uniform changing condition No-uniform condition

Satisfaction Mean SD Mean SD U value df p value

Cleaning 4.90 1.85 4.20 2.20 42.5 18.0 0.591

Washing 6.80 1.55 5.40 2.76 35.0 18.0 0.266

Cooking 5.90 2.38 5.70 2.83 32.5 18.0 0.193

Table 2: Quality scores for robot tasks.

Uniform changing condition No-uniform condition

Achievement Mean SD Mean SD U value df p value

Cleaning 4.80 1.81 3.50 1.96 42.0 18.0 0.558

Washing 7.30 1.57 6.90 1.66 49.0 18.0 0.969

Cooking 7.40 1.84 7.40 1.51 47.5 18.0 0.869

Table 3: Questionnaire scores.

Uniform changing No-uniform

condition condition

No. Question Mean SD Mean SD U value df p value

Q1 How many kinds of clothes did you

think there were?

2.89 1.01 1.44 0.78 – – –

Q2 Do you think the robot completed

the task quickly?

5.60 0.70 5.20 1.87 50.0 18.0 1.000

Q3 Do you think the robot is reliable? 4.30 1.16 4.40 1.78 44.0 18.0 0.670

Q4 Do you think the robot’s actions in

the task were appropriate?

6.20 1.03 4.30 1.77 18.0 18.0 0.015

Q5 Do you think the robot made many

mistakes during the task?

5.30 1.34 4.40 2.01 38.0 18.0 0.369

Q6 Do you think the robot has feel-

ings?

2.70 2.16 2.10 1.66 41.5 18.0 0.522

Q7 Was it easy to check the movement

of the robot on the screen?

6.50 0.53 6.60 0.70 42.5 18.0 0.539

Q8 Did you want to ask the robot to

perform other tasks?

4.60 1.90 4.70 2.16 47.0 18.0 0.847

Q9 Would you want to use the robot

daily?

4.20 1.75 3.40 1.78 36.5 18.0 0.313

Q10 Did you command the robot cor-

rectly?

5.80 1.87 6.10 1.29 47.0 18.0 0.835

Table 4: Frequency and order of occurrence of each task.

Task 1st 2nd 3rd

Cleaning 8 6 6

Washing 6 7 7

Cooking 6 7 7

ever, no comments indicating that the participants felt

uncomfortable about the robot’s uniform were seen

in the questionnaires or interviews after the experi-

ment. On the other hand, one participant noted that

”I thought it was strange to wear a hat while cook-

ing when the robot didn’t have hair.” We designed the

robot in order to protect it from effects such as foul-

ing, but we found that there was still a certain sense

of incongruity regarding its clothing. Therefore, it is

considered necessary to redesign the robot clothing to

avoid this incongruity.

5.2 Limitations

This study did have some limitations. First, the partic-

ipants were all young males (average age: 23.1 years)

and the small number of participants. Since this study

targets robotic housework, it is necessary to increase

the number of participants of a wider age range and

female participants to verify the effectiveness of the

proposed method.

Second, the experiment used a virtual robot shown

in a Web browser, not a real robot. Therefore, the

feasibility of the task, the presence or absence of me-

Effect on User Impression of Robot’s Task Dependent Uniform

chanical noise from the robot, and the speed of move-

ment are factors that differ from the case of using an

actual robot.

Finally, in the experiment, we used an animation-

based virtual space instead of a real space. There-

fore, participants may react differently when using an

actual robot in a real space. In addition, the tasks

performed by the robot were limited to housekeep-

ing tasks. In housekeeping tasks, a user and a robot

share a relatively small space, and there are many sit-

uations in which their activity ranges overlap. It is

necessary to investigate the applicability of the pro-

posed method not only to housekeeping tasks but also

to tasks in public spaces such as commercial facilities.

6 CONCLUSION

This paper proposed a uniform changing robot and

conducted experiments involving a housekeeping task

simulator. In the experiments, a robot performed

cleaning, washing, and cooking tasks according to the

participants’ commands. We investigated user im-

pressions of robots that either wore no uniform or

changed its uniform to match the task. The exper-

imental results suggested that the participants inter-

preted the behavior of the robot under the uniform

changing condition as being more appropriate, even

if the task was incomplete, compared with that under

the no uniform condition. Although the appearance

of a robot can lead to signiﬁcant biases among users,

there have been few studies on the effects of clothing.

In the present study, we investigated how the uniforms

worn by robots can affect such biases. This effect can

inform users that a robot has a specialty and can per-

form multiple tasks well, and may be effective for ex-

pressing the internal state of an autonomous robot. At

present, having a robot actually change its clothes by

itself remains a technical problem; however, a ﬂexi-

ble display on the body of the robot could provide the

same function as uniform changing. We plan to ex-

pand the scope of this research with participants and

conduct experiments from different perspectives, be-

cause a uniform that conveys expertise, position, qual-

iﬁcations, and group membership could be effective

for the design of robots.

REFERENCES

Belpaeme, T., Kennedy, J., Ramachandran, A., Scassellati,

B., and Tanaka, F. (2018). Social robots for education:

A review. Science Robotics, 3:eaat5954.

Bugmann, G. and Copleston, S. N. (2011). What can a per-

sonal robot do for you? In Proceedings of the 12th

Annual Conference on Towards Autonomous Robotic

Systems, TAROS’11, page 360–371, Berlin, Heidel-

berg. Springer-Verlag.

Davis, F. D. (1989). Perceived Usefulness, Perceived Ease

of Use, and User Acceptance of Information Technol-

ogy. MIS Quarterly, 13(3):319–340.

Havenith, G. (2002). Interaction of Clothing and Ther-

moregulation. Exogenous Dermatology, 1(5):221–

230.

Heerink, M., Kr

ose, B., Evers, V., and Wielinga, B. (2010).

Assessing Acceptance of Assistive Social Agent Tech-

nology by Older Adults: the Almere Model. Interna-

tional Journal of Social Robotics, 2(4):361–375.

Heerink, M., Vanderborght, B., Broekens, J., and Alb

Canals, J. (2016). New Friends: Social Robots in

Therapy and Education. International Journal of So-

cial Robotics, 8(4):443–444.

Hejazipoor, H., Massah, J., Soryani, M., Asefpour Vakilian,

K., and Chegini, G. (2021). An intelligent spraying

robot based on plant bulk volume. Computers and

Electronics in Agriculture, 180:105859.

Ho, C.-C. and MacDorman, K. F. (2010). Revisiting the

uncanny valley theory: Developing and validating an

alternative to the Godspeed indices. Computers in Hu-

man Behavior, 26(6):1508–1518.

Huisman, C. and Kort, H. (2019). Two-Year Use of Care

Robot Zora in Dutch Nursing Homes: An Evaluation

Study. Healthcare, 7(1):31.

Joseph, N. and Alex, N. (1972). The Uniform: A Socio-

logical Perspective. American Journal of Sociology,

77(4):719–730.

Kang, D., Kwak, S. S., Lee, H., and Choi, J. (2020).

First Things First: A Survey Exploring Key Ser-

vices and Functions of a Robot. In Companion

of the 2020 ACM/IEEE International Conference on

Human-Robot Interaction, HRI ’20, pages 278–280,

New York, NY, USA. Association for Computing Ma-

chinery.

Kiesler, S., Sproull, L., and Waters, K. (1996). A prisoner’s

dilemma experiment on cooperation with people and

human-like computers.

Komatsu, T. and Yamada, S. (2008). Effect of Agent Ap-

pearance on People’s Interpretation of Agent’s Atti-

tude. In CHI ’08 Extended Abstracts on Human Fac-

tors in Computing Systems, CHI EA ’08, pages 2919–

2924, New York, NY, USA. Association for Comput-

ing Machinery.

Komatsu, T. and Yamada, S. (2011). Adaptation gap hy-

pothesis: How differences between users’ expected

and perceived agent functions affect their subjective

impression. Journal Systemics, Cybernetics and In-

formatics, 9(1):67–74.

Legde, K. and Cunningham, D. W. (2019). Evaluating the

Effect of Clothing and Environment on the Perceived

Personality of Virtual Avatars. In Proceedings of the

19th ACM International Conference on Intelligent Vir-

tual Agents, IVA ’19, pages 206–208, New York, NY,

USA. Association for Computing Machinery.

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

Mann, H. B. and Whitney, D. R. (1947). On a Test of

Whether one of Two Random Variables is Stochasti-

cally Larger than the Other. The Annals of Mathemat-

ical Statistics, 18(1):50–60.

Mathur, M. B. and Reichling, D. B. (2016). Navigating a

social world with robot partners: A quantitative car-

tography of the Uncanny Valley. Cognition, 146:22–

32.

Nachar, N. (2008). The Mann-Whitney U: A Test for

Assessing Whether Two Independent Samples Come

from the Same Distribution. Tutorials in Quantitative

Methods for Psychology, 4(1):13–20.

Pandey, A. K. and Gelin, R. (2018). A Mass-Produced So-

ciable Humanoid Robot: Pepper: The First Machine

of Its Kind. IEEE Robotics Automation Magazine,

25(3):40–48.

Parker, R., Bayne, K., and Clinton, P. (2016). Robotics in

forestry. New Zealand Journal of Forestry, 60:8–14.

Shao, C. Y., Baker, J. A., and Wagner, J. (2004). The ef-

fects of appropriateness of service contact personnel

dress on customer expectations of service quality and

purchase intention: The moderating inﬂuences of in-

volvement and gender. Journal of Business Research,

57(10):1164–1176.

Singer, M. S. and Singer, A. E. (1985). The effect of po-

lice uniform on interpersonal perception. The Jour-

nal of Psychology: Interdisciplinary and Applied,

119(2):157–161.

Song, S. and Yamada, S. (2019). Designing LED lights

for a robot to communicate gaze. Advanced Robotics,

33:1–9.

Takagi, M., Mori, H., Yimit, A., Hagihara, Y., and Miyoshi,

T. (2016). Development of a Small Size Underwater

Robot for Observing Fisheries Resources – Underwa-

ter Robot for Assisting Abalone Fishing –. Journal of

Robotics and Mechatronics, 28(3):397–403.

Złotowski, J., Khalil, A., and Abdallah, S. (2020). One

robot doesn’t ﬁt all: aligning social robot appearance

and job suitability from a Middle Eastern perspective.

AI & SOCIETY, 35(2):485–500.

Effect on User Impression of Robot’s Task Dependent Uniform