Designing an Accepted Look for an Assistive Robot Results of a mixed methods study with people of different age groups

Abstract

In this contribution, the participatory design process for a service robot to assist people in old age or with disabilities is presented. In order to provide the platform with an attractive and accepted embodiment, a three-stage approach with potential users is envisaged. This article reports the results of phase 1. Three models with different head and body shapes were evaluated by a younger and an older age group. They were exposed to 3D models they saw in the room through augmented reality glasses. A mixed methods design was chosen: Respondents were asked about each of the models by using open questions in a questionnaire. Additionally, they had to rate the robot in a semantic profile. The potential users preferred a design with a rounded body and an implied face. Tendencies could be seen towards different results in the age groups concerning colour and human likeness. Most people in the younger age group voted for a model with a less humanlike face, while in the older age group a high proportion also chose a model with a human face on a monitor. Changes such as a friendlier expression, a lower height of the robot and a rounder shape of the body seem to be necessary. Involving potential user groups in the process proves useful, as it provides deeper insight into their needs. Augmented reality evaluation promises to be a time and material saving method, but further research is needed to validate procedures and results.

Full text

Designing an Accepted Look for an
Assistive Robot
Results of a mixed methods study with people
of different age groups
Jessica Sehrt*1, Marina Ringwald1, Barbara Klein1, Héctor Solis1, Julian Umansky1
and Peter Nauth1
1Frankfurt University of Applied Sciences, Frankfurt, Germany
Abstract. In this contribution, the participatory design process for a service robot
to assist people in old age or with disabilities is presented. In order to provide the
platform with an attractive and accepted embodiment, a three-stage approach
with potential users is envisaged. This article reports the results of phase 1. Three
models with different head and body shapes were evaluated by a younger and an
older age group. They were exposed to 3D models they saw in the room through
augmented reality glasses. A mixed methods design was chosen: Respondents
were asked about each of the models by using open questions in a questionnaire.
Additionally, they had to rate the robot in a semantic profile. The potential users
preferred a design with a rounded body and an implied face. Tendencies could be
seen towards different results in the age groups concerning colour and human
likeness. Most people in the younger age group voted for a model with a less
humanlike face, while in the older age group a high proportion also chose a model
with a human face on a monitor. Changes such as a friendlier expression, a lower
height of the robot and a rounder shape of the body seem to be necessary. Involv-
ing potential user groups in the process proves useful, as it provides deeper in-
sight into their needs. Augmented reality evaluation promises to be a time and
material saving method, but further research is needed to validate procedures and
results.
Keywords: Assistive Robot, Robot Design, Embodiment, Participatory Ap-
proach, Acceptance
1 Introduction
People in older age or with disabilities can benefit from digital technologies and robotic
systems to stay self-determined and live more independently. The acceptance of a robot
depends on its functionality, but also on its appearance as it evokes emotions in users
(Hwang et al., 2013; Otterbacher & Talias, 2017) and reflects roles (De Grad &
Allouch, 2015; Esposito et al., 2019).
In the following, the participatory design process for a service robot developed at
Frankfurt University of Applied Sciences in Germany (Nauth et al., 2016) is presented.
The platform ROSWITHA (Robot System WITH Autonomy) aims to navigate in peo-
ple’s flat and to fetch and bring objects, such as a glass of water. In order to provide it
with an attractive and accepted embodiment, a three-stage participatory approach with

2
potential users is envisaged. In phase 1 and 2, based on a literature research, differently
developed versions are realised as 3D models in augmented reality, which are evaluated
by respondents from a younger and an older group. Finally, a physical prototype will
be created (see Fig. 1). This article reports the results of phase 1.
2 Methods
2.1 Design process
Three designs were created based on a systematic literature review of people's prefer-
ences relating to the appearance of robots, the dimensions and technical requirements
of the platform and the desired design elements. The objective of the exterior shape was
that the robot should make an attentive, friendly, competent, but also reserved impres-
sion, so that it would not be perceived as intrusive in people’s homes.
In phase 1, different head and body shapes were evaluated. For the head design, a
neutral humanoid head, a playful cartoon-like head, and a head with a human-like face
in a monitor were chosen. The bodies varied from angular to conical to round (see Fig.
2).
Fig. 1. Design and evaluation process: In the first and second evaluation phase, initially three and
then two models are discussed after the potential user groups have seen them in augmented reality.
In phase 3, they will evaluate a physical prototype (M1-5 = Model 1-5; picture ROSWITHA: J.
Umansky, graphic: J. Sehrt).

3
2.2 Potential users
A younger and an older age group were selected as there were indications of possible
differences regarding their preferred robot appearances in the literature (Prakash &
Rogers, 2015; Tu et al., 2020). The younger group consisted of 16 students (8 female,
8 male) between 23 and 52 years (M = 29.19; SD = 7.20) from subjects with a health
or social, architectural or IT background. The older group was recruited through senior
citizens' advisory boards and offers in city districts. It consisted of 14 persons (8 female,
6 male) aged 69-87 years (M = 77.36; SD =4.98).
2.3 Procedure
The potential users were exposed to the 3D models they saw in the room through aug-
mented reality glasses during appointments at the university, urban premises or in their
home environment. A mixed method design was chosen: Respondents were asked about
each of the models by means of open questions in a questionnaire, which were qualita-
tively evaluated. In the first and second question they were asked about their first im-
pression and what feelings the robot evoked in them. After that, they had to rate the
robot in a semantic profile on 7 characteristics which were chosen by the research team
because they were intended to reflect the design characteristics of the robot in a range
between -3 and 3. The items were: clumsy – elegant, incompetent – competent, striking
– discreet, scattered – attentive, intrusive – withdrawn, unfriendly – friendly, unreliable
– reliable. The last question for each model was about the changes the respondents
would make.
In a more general part, the potential users should first rate to what extent they could
imagine using such a robot in their home on a 4-point Likert-scale from very hard to
imagine to very well to imagine and to name potential obstacles. Then they were asked
Fig. 2. The three options for the first evaluation phase

4
to specify the tasks for which they would like to use such a robot. Finally, they chose
their preferred model of the three drafts and decided a colour out of green, blue and
orange.
The answers to the open questions were categorized according to main topics, those
from selection options were analysed quantitatively. Results are reported descriptively
below. Due to the small number of persons, the median was calculated for the semantic
profiles, since the mean value would have been stronger influenced by individual re-
sults.
3 Results
A strong majority of the younger group (12 out of 16) voted for the second model (see
Fig. 2), as did a much narrower majority (6 out of 14) of the older group. This was
followed by model 3 (two persons in the younger group and four persons in the older
group) and finally 1 (one person among the younger, two persons among the older).
Two persons in the older group and one person in the younger group indicated "none".
Among the younger group of potential users, a clear majority (11/16) chose the colour
green, among the older group orange (9/14).
The change requests of the younger group of potential users were mainly related to
the face and body. For model 1, one person stated that the face and two persons that the
body should be more human. For model 1 and 2, one respondent each wished for a less
human face, and for model 2, two people wished for a less childlike face. For model 3,
five people said they favoured a more abstract or robotic face. A friendlier facial ex-
pression was desired for all three designs. More rounded shapes were favoured for the
corpus as well as a change of wheels. The round shelf on the front of model 2 was rather
rejected, as was the flat colouring in model 3.
In addition to the face and body, the older group also had many comments about the
size and controls. The size of approx. 1.50 m was rated as too high several times, and a
screen as a control element at the front as well as more interaction options were desired.
Four persons of the older group stated that model 1 should be more human (three of
them the face) and one for model 3. Two persons noted this for model 2, one of them
specifically for the face. A full face with a mouth was also desired by three persons and
a friendlier look by two. However, two favoured a non-human appearance for model 3.
Members of the group also preferred a rounder body and hiding the undercarriage.
The semantic profiles of the younger group showed medians in the neutral to slightly
positive range (0 to 1 out of 3) for all traits recorded. Model 2 was rated median 2 as
friendlier than the others, and model 3 was rated median -1 as more conspicuous than
the others. The older group rated model 1 and 2 as clumsier (-0.5) and more conspicu-
ous (-1,5) than the average and model 3 as more conspicuous (-1), the other values were
in the neutral to slightly positive range.
When asked about the intended use for such a robot, a majority in both groups an-
swered for household tasks and for fetching and bringing objects. The younger group
could imagine the use of such a robot better (see Fig. 3) than the older group (see Fig.

5
4). Ten persons stated that they could imagine using it very well or well, and four poorly
or very poorly. Two persons were undecided and voted between well and poorly.
Fig. 3. Response to the question of how well younger people could imagine using the robot
Half of the older group could well imagine the use of such a robot, the other half poorly
or very poorly.
Fig. 4. Response to the question of how well older people could imagine using the robot
2
8
222
very well well undecided
between well
and poorly
poorly very poorly
NUMBER OF PERSONS
RESPONSE
Younger age group (N = 16)
0
7
0
34
very well well undecided
between well
and poorly
poorly very poorly
NUMBER OF PERSONS
RESPONSE
Older age group (N = 14)

6
4 Discussion
As the number of potential users was not representative, the results cannot be general-
ised, nevertheless often-mentioned features to improve face and body can be carried
over to the next design phase, in which specific parts of the robot shall be further de-
veloped. The results indicate to continue the work with models 2 and 3 which includes
one design with a physical face on the head and one with a face on a monitor. Reducing
the height of the robot and providing a rounder shape for the corpus also seems neces-
sary to be more responsive to the needs of potential user groups.
As in Prakash & Rogers (2015) and Tu et al. (2020) there could also be seen tenden-
cies towards different results in the age groups concerning human likeness. Most people
in the younger age group voted for model 2 which has a less humanlike face than model
3 for which also a high proportion in the older age group voted. This was also the case
when asked about future design changes with persons in the older age group preferring
more humanlike designs in all conditions and a more diverse picture in the younger age
group. Especially, model 3 with the humanlike face was rejected by more younger than
older persons. This aspect will have to be considered in further development stages as
well as the characteristics identified by the potential users in the semantic profile. How-
ever, since the scales were difficult to complete for some people in the older group, a
more narrative approach should be adopted in future surveys to identify people’s per-
ceptions.
The use of augmented reality allows to see the models in real size in the existing
room. It is seen as an innovative and flexible tool for product design (Sahin & Togay,
2016) and for imaging the equipment of rooms (Joshi et al., 2020). It also can be con-
sidered ecologically sustainable, since several prototypes do not have to be physically
produced in each evaluation round. Despite the potential, use is currently not wide-
spread (Bottani & Vignali, 2019). However, experience with the older group of poten-
tial users in particular showed that good personal support is necessary so that the test
persons can see the model in an appropriate way and also use the three-dimensionality
for their evaluation, for example by moving around in the room.
5 Conclusion
The phase 1 evaluation of the initial options for a possible embodiment of the assistive
robot ROSWITHA revealed useful aspects for the further design process. Involving
potential users in the process proves to help designing products that are closer to peo-
ple's lives and ultimately lead to greater acceptance. Therefore, the design process will
be continued in a participatory way.
Evaluation in augmented reality promises to be a time- and material-saving method,
which offers the possibility to get a realistic image of the model e.g. at home. However,
further research is needed, especially with older target groups, in order to validate pro-
cedures and to verify the results.

7
Acknowledgments. The research leading to these results received funding from the
Commerzbank Foundation, Frankfurt am Main, Germany.
References
1. Bottani, E. & Vignali, G.: Augmented reality technology in the manufacturing industry: A
review of the last decade. IISE Transactions 51(3), pp. 284-310 (2019).
https://doi.org/10.1080/24725854.2018.1493244.
2. Esposito, A., Amorese, T., Cuciniello, M. et al.: Elders prefer female robots with a high
degree of human likeness. In: 2019 IEEE 23rd International Symposium on Consumer Tech-
nologies (ISCT). IEEE, pp. 243–246. https://doi.org/10.1109/ISCE.2019.8900983.
3. Graaf, M. M. A. de, Ben Allouch, S.: The evaluation of different roles for domestic social
robots. In: 2015 24th IEEE International Symposium on Robot and Human Interactive Com-
munication (RO-MAN). IEEE, pp. 676–681.
https://doi.org/10.1109/ROMAN.2015.7333594.
4. Hwang, J., Park, T., Hwang, W.: The effects of overall robot shape on the emotions invoked
in users and the perceived personalities of robot. Applied Ergonomics 44(3), pp. 459-471
(2013). https://doi.org/10.1016/j.apergo.2012.10.010.
5. Joshi, S., Walavalkar, P., Shetkar, P., Joshi, S., Mahale, S., Bhogan, S., Sawant, K.: Product
Visualization Using Augmented Reality. International Research Journal of Engineering and
Technology 7(5), pp. 6924-6929 (2020).
6. Nauth, P., Sharan, S., Michalik, R., Umansky, J., Schafner, R.: Development of a robot as-
sisting elderly people. Healthy Aging in Digitized Societies (HeADS), pp. 156 - 162 (2016).
7. Otterbacher, J., Talias, M.: S/he’s too Warm/Agentic! In: Mutlu, B., Tscheligi, M., Weiss,
A. et al. (eds) Proceedings of the 2017 ACM/IEEE International Conference on Human-
Robot Interaction. ACM, New York, NY, USA, pp. 214–223.
https://doi.org/10.1145/2909824.3020220.
8. Prakash, A., Rogers, W. A.: Why Some Humanoid Faces Are Perceived More Positively
Than Others: Effects of Human-Likeness and Task. Int J Soc Robot 7, pp. 309–331 (2015).
https://doi.org/10.1007/s12369-014-0269-4.
9. Şahin, D. & Togay, A.: Augmented reality applications in product design process. Global
Journal on Humanites & Social Sciences 3, pp. 115-125 (2016).
10. Tu, Y.-C., Chien, S.-E., Yeh, S.-L.: Age-Related Differences in the Uncanny Valley Effect.
Gerontology 66, pp. 382–392 (2020). https://doi.org/10.1159/000507812.