How Anthropomorphism Affects User Acceptance of a Robot Trainer

in Physical Rehabilitation

Baisong Liu, Panos Markopoulos and Daniel Tetteroo

Department of Industrial Design, Eindhoven University of Technology, De Zaale, Eindhoven, The Netherlands

Keywords: Socially Assistive Robot, Anthropomorphism, Rehabilitation, User-centred Design, Acceptance.

Abstract: Developments in social robotics raise the prospect of robots coaching and interacting with patient during

rehabilitation training assuming a role of a trainer. This raises questions regarding the acceptance of robots in

this role and more specifically, to what extent the robot should be anthropomorphic. This paper presents the

results of an online experiment designed to evaluate the user acceptance of Socially Assistive Robots (SARs)

as rehabilitation trainers, and the effect of anthropomorphism on this matter. User attitudes were surveyed

with regards to three variations of a scenario where the robot with varying levels of anthropomorphism acts

as a trainer. The results show that 1) participants are accepting towards SAR-assisted rehabilitation therapies,

2) anthropomorphism influences patient’s perceived self-efficacy and attitude towards the system. A second

survey studied inventoried issues regarding patients’ acceptance of such systems, pertaining to technology

acceptance, patient needs for rehabilitation training and the effect of anthropomorphism. Based on the above

findings we propose user-informed design implications for improving user acceptance is rehabilitation

settings.

1 INTRODUCTION

The application of robotic technology in the domain

of physical rehabilitation is an area of ongoing

research (Laut et al., 2016). Projects have developed

robotic technology to support physically impaired

patients, such as mobility aids for aging and motor

function impaired users, assisting users in loaded

walking (Ding et al., 2017) and supporting

rehabilitation training exercises (Feys et al., 2015;

Popescu et al., 2016). Such projects have shown that

robots can help to improve the quality and quantity of

rehabilitation training. However, the current trend

mostly concerns with physically supporting (parts of)

the body (Cardona et al., 2017; Vitiello et al., 2017).

Socially Assistive Robotics (SAR) provides

assistance to users through social interaction (Feil-

Seifer and Mataric, 2005). The use of SAR in

rehabilitation training has been considered and

demonstrated as promising (Eriksson et al., 2005).

Taxonomies of components concerning socially

interactive robots have been proposed (Fong et al.,

2003; Feil-Seifer and Mataric, 2005), suggesting

human-oriented perception is an important part of

SAR interaction design. Thus far, this aspect is yet to

be explored in the context of rehabilitation training.

We have taken a user-centered approach to this

issue, exploring patients’ acceptance towards the

concept of a robot trainer for rehabilitation, and the

effect anthropomorphism has on its acceptance. This

paper provides insights for further design of utilizing

SAR in the context of rehabilitation training.

2 RELATED WORK

2.1 Social Robot Acceptance

General robot acceptance studies have investigated

the effects on acceptance of specific robotic traits,

such as gender of the voice (Eyssel et al., 2012), facial

expressions (Moosaei et al., 2017) and gestures (Zaga

et al., 2017). Regarding SAR, (Fong et al., 2003) have

identified the following factors to be of influence on

acceptance: 1) the user’s attitude towards the robot,

2) the robot’s field performance, 3) robot-displayed

emotions, 4) appearance and dialog, and 5)

personality. These studies explore robot acceptance

regardless of a specific context, and thus provide

general conclusions and directions for further

research.

Liu, B., Markopoulos, P. and Tetteroo, D.

How Anthropomorphism Affects User Acceptance of a Robot Trainer in Physical Rehabilitation.

DOI: 10.5220/0007343600300040

In Proceedings of the 12th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2019), pages 30-40

ISBN: 978-989-758-353-7

Another branch of social robot acceptance study

focuses on specific contexts and user groups,

typically children, elderly and autism patients. For

example, the Almere Model has been proposed for

testing and predicting elderly users’ acceptance of

assistive social agent technologies, suggesting 12

factors to be of influence (Heerink et al., 2010).

Another study employed a zoomorphic companion

robot (Nabaztag) into an elderly user’s home to gain

insights on social robot acceptance, focusing on users

building a long-term relationship with a social robot

in domestic settings (Klamer and Allouch, 2010). The

acceptance of SAR in the specific context of

rehabilitation has not yet been investigated.

2.2 Anthropomorphism

Anthropomorphism of social robots is a powerful

factor influencing the user’s experience, including

empathy (Moosaei et al., 2017), enjoyment, and other

social emotions (Bartneck et al., 2010).

Anthropomorphism is proposed to be expressed in

appearance and behavior (Choi and Kim, 2008). So

far studies have explored different factors of

anthropomorphic appearance embodied in the design

of robots, for example through facial expressions

(Moosaei et al., 2017), voice (Siegel et al., 2009) and

gesture (Salem et al., 2013). As it has been suggested

that user responses to anthropomorphic robotics are

context based (Epley et al., 2007), it is important to

explore the effects of anthropomorphism for specific

contexts and use cases, such as that of physical

rehabilitation.

2.3 Social Robots for Rehabilitation

SAR has been proposed as an alternative to the

therapist for rehabilitation exercises, due to its

potential benefits of cost reduction, privacy,

improving engagement, and open up possibilities for

home training scenarios (Winkle et al., 2018). A

feasibility study has proven the potential of such

application (Kyong Il Kang et al., 2005). A further

study has suggested that even very simple robot

behavior might benefit compliance in stroke

rehabilitation exercises (Gockley and Mataric, 2006).

A more recent study underlined the link between

personalized robot behavior and user task

performance in rehabilitation training (Tapus and

Mataric, 2008.).

3 RESEARCH QUESTIONS

While the above-mentioned study by (Winkle et al.,

2018) examined the design of SAR from a therapist

perspective, this paper presents a study from the

perspective of the patient. The aim of this paper is to

explore patients’ attitude towards having a social

robot trainer to facilitate their rehabilitation training.

Specifically, we are interested in:

RQ1) What is the patient’ attitude towards having

a robot trainer as facilitator of their rehabilitation

training?

RQ2) How does the level of anthropomorphism in

SAR form-design influence the patients’ acceptance

in the context of rehabilitation training?

RQ3) What are patients’ preferences and

concerns regarding SAR within the context of

rehabilitation training?

4 METHOD

We conducted two studies to investigate our research

questions. In both studies, we used an illustration of a

fictional scenario with a patient performing her

rehabilitation exercise with the help of a robot trainer

(see figure 1), participants in the three conditions of

study one were shown with three different robot

trainer concepts presented in table 1.

In study one, aimed at answering RQ1 and RQ2,

we have used standard questionnaires and open

questions to collect the participants’ general attitude

towards social robotics, as well as their attitude

regarding the application of SAR to the scenario.

Study two was aimed at answering RQ3, and to

collect more in-depth information on some of the

answers obtained by study one. This study was

executed as a survey composed of mostly open

questions regarding RQ3.

4.1 Scenarios

Both studies utilized illustrated scenarios describing

a fictional interaction between a patient and a robot

trainer. To focus on only the social interaction

aspects, the robot in this scenario does not support or

enforce movements with the patient, as one might

expect in rehabilitation robotics, but takes only the

role of a coach who provides the patient with

information on the overall progress of the therapy,

instructions for the exercise and encouragement when

the patient is experiencing physical challenges. The

story of the scenario is based on observations of

How Anthropomorphism Affects User Acceptance of a Robot Trainer in Physical Rehabilitation

clinical treatment and was further improved by

consulting experienced physical therapists.

Figure 1: Scenario Used in the Survey.

The scenarios featured three versions of the robot

trainer, differing in their level of anthropomorphic

appearances. In the studies discussed above, robotic

anthropomorphism is expressed mainly through

facial expressions and gestures. Therefore, the three

versions of the robot trainer included one with

human-like body structure and expressions, one with

only expressions, and one with none of the two

factors, to represent high, medium and low levels of

anthropomorphism in robot form design (See Table

1). A small survey was conducted as a manipulation

check, to verify whether participants experienced the

three representations as differing. in terms of

anthropomorphism.

Table 1: Robot trainer designs.

Human-like

body

structure



Human-like

expression



Level of

anthropomo

rphism

High

Medium

Low

4.2 Study One Set-up

We first conducted a between subject study. Three

groups of participants were presented a survey

HEALTHINF 2019 - 12th International Conference on Health Informatics

containing the scenario on rehabilitation training,

each with a different form design for the robot trainer.

We used the Negative Attitude towards Robot

Scale (NARS) questionnaire (Nomura et al., 2004) at

the beginning of the survey to acquire participants’

general attitude towards social robots and to check

whether these were distributed equally over the three

groups. Then, the participants were presented the

scenario. After having experienced the scenario, we

took the Credibility/ Expectancy questionnaire

(Devilly and Borkovec, 2000) and the Technology

Acceptance Model (TAM) questionnaire to evaluate

the participants’ acceptance of the robot assisted

therapy and the robot trainer. At the end of the survey,

we asked the user to rate the look of the robot trainer

on a 10-point Machine-like to Human-like scale for a

manipulation check. The structure of the study is

presented in Table 2.

Table 2: Structure of Study One.

Screening questions

General attitude towards social robots

NARS questionnaire

Open question regarding attitude towards working

with social robots

Scenario “Lisa’s Rehabilitation Training Session with

Robot Trainer”

Attitude towards robot-assisted Rehabilitation

therapy

Credibility/Expectancy questionnaire

Open question regarding attitude towards

robot-assisted rehabilitation therapy

Attitude towards robot trainer in rehabilitation

therapy

TAM questionnaire

Open question regarding attitude towards robot

trainer in rehabilitation therapy

Manipulation check

Rate Robot Trainer’s look on a scale from Machine-like

to Human-like

Open question regarding attitude towards

robot-assisted rehabilitation therapy

4.3 Study Two Set-up

Study two was conducted with a within-subject

design. It presented the same scenario as study one,

but replaced the robot trainer with a blank box. After

participants read through the scenario, we presented

all three robot trainer concepts as available options

and ask them to pick the one they preferred, assuming

they were the patient in the scenario. We then

provided open questions for the participants to give

feedback on the reason of their choice and possible

improvements for the concept of their choice. Finally,

we asked participants to rate all three concepts of the

robot trainer on a ten-point rating scale from

“Machine-like” to “Human-like”. The detailed

structure is of study two is presented in Table 3.

Table 3: Structure of Study Two.

Screening Questions

Scenario “Lisa’s Rehabilitation Training Session with

Robot Trainer” with the trainer left blank

User preference for the given concept

Choose the desired robot trainer from the three given

concepts

Open questions regarding the choice

Open questions about further improvements for the

therapy and the robot trainer

Manipulation check

Rate three Robot Trainers’ look on a scale from

Machine-like to Human-like respectively

4.4 Participants

Participants were recruited through a crowdsourcing

platform for both studies. Participants were expected

to have been (formerly) involved in physical

rehabilitation training. The study involved four

groups of participants (total pool N = 103 after

screening, average age 37). Detailed numbers of

participants and group allocations are listed in Table

4. Study one and two were published on Amazon

Mechanical Turk on July 12, 2018 and lasted 14 days.

Each participant spent around 10 minutes on the

study. Based on the minimum hourly wage, the

reward was set at 1.5 USD.

To ensure the quality of the answers collected, we

set a three-step screening scheme. Firstly, workers on

Amazon Mechanical Turk had to answer three

questions, proving that they had experience in

physical rehabilitation, to access the survey.

Secondly, four reverse questions (changing positive

statements into negative ones, e.g. “I found the robot

trainer easy to interact with” into “I found the robot

trainer difficult to interact with”) were also planted

How Anthropomorphism Affects User Acceptance of a Robot Trainer in Physical Rehabilitation

Table 4: Study participants.

Study

Conditions

Participants

Study One

HA (High

Anthropomorphism)

28 (HA1 - HA28)

MA (Medium

Anthropomorphism)

24 (MA1 - HA24)

LA (Low

Anthropomorphism)

20 (LA1 - LA20)

Study Two

31 (ST1 - ST31)

within the survey to check for satisficing behavior in

answering the survey. Lastly, we put an eight-digit

password at the end of the survey for claiming the

reward, only visible to participants who finished the

survey. We set the survey to be only available to

workers with approval rate higher than 97% percent

and job experience less than 5000 to further ensure

the quality of the answer.

4.5 Analysis

Quantitative data were analyzed through mean scores

for all conditions as an indicator of general

acceptance. Additionally, through one-way ANOVA

tests, we analyzed the effect of anthropomorphism on

the patient’s acceptance for SAR assisted

rehabilitation therapy and robot trainer technology.

We conducted a closed coding and an open coding

analysis on the qualitative data. The closed coding

scheme was applied to the data from Study One, and

consisted of the factors from the

Credibility/Expectancy and TAM questionnaires,

namely: Credibility, Expectancy, Perceived Ease of

Use (PEU), Perceived Usefulness (PU), Behavioral

Intention (BI), Attitude (A), Self-Efficacy (SE) and

Subjective Norm (SN).

Next, we combined the qualitative data from

Study One and Study Two and conducted an

inductive open coding analysis to generalize more

insights.

5 FINDINGS

We present our findings regarding three topics: 1)

general attitude towards SARs, 2) Acceptance of

SAR assisted rehabilitation therapy and robot trainer,

and 3) effects of anthropomorphism on the

acceptance of SAR assisted rehabilitation therapy and

robot trainer.

5.1 General Attitude towards Socially

Assistive Robots

With the NARS questionnaire, we tested Study One’s

participants’ negative attitude towards situations of

interaction with robots, social influence of robots and

emotional interaction with robots. One-way ANOVA

analysis showed that there are no significant

differences among the three groups for Factor 1

(Negative Attitude towards Situations of Interaction

with Robots) (F (2,68) = 2.83, ns), Factor 2 (Negative

Attitude towards Social Influence of Robots) (F

(2,68) = .69, ns), and Factor 3 (Negative Attitude

towards Emotions in Interaction with Robots) (F

(2,68) = 2.97, ns). This suggests that the three groups

have a similar attitude towards socially assistive

robots. The mean scores for the three groups and for

all three factors are under 3. Since NARS is a negative

attitude questionnaire, these lower scores suggest a

rather positive attitude towards the factors (see Figure

2).

Figure 2: Mean score of the three groups from Study One

for the three NARS factors.

Open coding analysis highlighted that robotic

technology is desired for its precision, being

objective, convenience, efficiency, opening up

possibilities for more privacy and eliminating

negative social encounters. Negative opinions

clustered around technological possibilities, and

cultural and ethical concerns. Below are the two

factors concerning acceptance for SARs.

5.1.1 Technological Status-quo

“I think it has the potential to be very interesting and

constructive, but hasn't been fully developed” (HA7)

Factor 1 Factor 2 Factor 3

HA MA LA

HEALTHINF 2019 - 12th International Conference on Health Informatics

“I might feel alone working with robot as robots don't

have cognitive behavior” (MA12)

7 quotes suggested robots are best suited for the

kind of jobs that are simple, repetitive and do not need

complicated judgment. In the participants’ opinions,

current robotic-related technologies still lack

flexibility to handle emergencies and lack proper

judgement for complex situations. Such technological

reality proposed a limitation for participants in terms

of the tasks assumed possible for robots to take on.

5.1.2 Issues of Technology-related Attitude

and Beliefs

“Not comfortable. Humans are losing the ability to

think and react without assistance.” (HA21)

“Not a fan, I can see the appeal, but I'm sure if

there is a living person who needs the money.” (LA6)

This refers to current cultural opinions about the

relationship between people and technology.

Participants regard certain human characteristics as

irreplaceable (e.g., empathy, social perception, etc.),

and worry that humans will be weakened and end up

in an inferior position compared to technology in the

future. The scenarios evoked a sense of anxiety about

human identity among respondents. A recent study

suggested that the anthropomorphic appearances of a

social robot can pose a threat to human

distinctiveness (Ferrari et al., 2016). In this sense,

anthropomorphism can trigger these negative

emotions, and should thus be considered in design

decisions.

Figure 3: Mean scores of three groups for Credibility and

Expectancy.

Figure 4: Means scores of TAM questionnaire.

Credibility Expectancy

HA MA LA

How Anthropomorphism Affects User Acceptance of a Robot Trainer in Physical Rehabilitation

5.2 Acceptance of SAR in

Rehabilitation Therapy and as

Robot Trainer

The Credibility/Expectancy questionnaire was used

in Study One to evaluate how much patients believe

and how much they feel that the social robot can be

an effective aid during rehabilitation training. This

questionnaire has a two-factored structure, three

items are used for each, two of them are nine-point

Likert scales, and the third is a scale from 0 to 100.

The latter is converted to a nine-point scale by

dividing by 12,5 and adding 1. The subscale items are

added resulting in two aggregate scores ranging

between from 3-27 for credibility and for expectancy

respectively, with 15 as the midpoint of the scale. The

mean scores for the three groups show that

participants for conditions HA and LA have positive

opinions on the credibility (HA: M=18.82, SD=5.18,

LA: M=17.26, SD=7.4) and expectancy (HA:

M=17.95, SD=4.26, LA: M=17.16, SD=6.64) of the

training, and participants for condition MA have

neutral scores for credibility (M=15.30, SD=7.21)

and expectancy (M=14.87, SD=6.18), as visible in

Figure 3.

The TAM questionnaire was used to measure the

acceptance of the robot trainer. The results suggest

that participants perceive the robot trainer as easy-to-

use (PEU & SE), have a neutral attitude towards the

concept (A) and are neutral with regards to its

Table 5: Patient Needs for SAR-Assisted Rehabilitation Therapy.

HEALTHINF 2019 - 12th International Conference on Health Informatics

perceived usefulness (PU). Finally, participants have

a below average mean score for the factors

behavioural intention (BI) and subjective norm (SN)

(see Figure 4).

5.2.1 Patients’ Needs for SAR-assisted

Rehabilitation Training

From analyzing the qualitative data, a section labeled

“Patient Needs” emerged. Due to the limited contents

of the scenario provided, participants were left with

doubts about services they considered important in

their own rehabilitation therapies, but which were not

shown in the scenario. Over 100 quotes were

collected and summarized into 9 categories of needs

presented in Table 5.

5.2.2 Perceived Benefits of SAR-assisted

Rehabilitation Training

“My own rehabilitation therapy was with a human,

but it accomplished the same purpose.” (HA17)

Mean scores of Credibility/Expectancy

questionnaire show that participants have above

average expectations from the rehabilitation therapy

provided by a robot trainer.

Self-directed practice is considered to be an

important component in rehabilitation therapy

(Winkle et al., 2018). Quotes from the questionnaire

confirmed that the process is mostly done by the

patient him- or herself, and demands consistency and

certain quantity. To this end, participants have

confidence in having a robot trainer providing

guidance, feedback and motivational prompts.

Participants also mentioned the following additional

benefits a robot trainer could bring about:

1. Providing Stable and Basic Services and

Eliminating Human Errors and Interference.

“The facility I used insisted their doctor see me. He

(the therapist) ignored my doctor's orders. This

(robot trainer) would prevent that from happening.”

(MA19)

“I wouldn't feel judged while doing it. It seems

positive and reassuring.” (HA15)

Some participants mentioned their rehabilitation

experience, during which, there were moments that

they felt neglected, misdiagnosed or given the wrong

instructions and being interrupted by inappropriate

social interactions. These participants believed a

robot trainer to be more goal-oriented, therefore

enabling them to better focus on the training exercise

and eliminating possible human errors.

2. Financial Benefits and Cost Effectiveness.

“My experience was that they were interested in what

they could get from insurance, my needs were not.”

(HA18)

“Development of at-home exercises that can be

performed without purchasing expensive equipment

needed” (LA9)

Cost efficiency and flexibility brought by a robot

trainer were expected by the participants, suggesting

tele-rehabilitation is one of the most promising usage

cases that can be developed for robot trainer.

5.2.3 Concerns about SAR-assisted

Rehabilitation Therapy

Concerns primarily focused on 1) how the robot will

make up for the loss of human specific values in

rehabilitation therapy e.g. empathy, perceptions on

the patients and the overall rehabilitation training

based on experience and expertise, and capability of

dealing with emergencies, 2) the normally

included/expected therapies that are not shown in the

scenario, e.g. massages, hands-on corrections, real-

time demonstrations and exercise walk-through, and

3) the loss of social and emotional interactions. When

it comes to evaluating patients’ training performance,

providing emotional support and dealing with

emergencies and safety issues, participants remained

concerned. The concerns centered on the participants’

lack of trust in the robot trainer, since it does not have

empathy towards the patient. Furthermore, patients

doubted whether the motivational prompts would be

perceived as sincere, or just pre-programmed

prompts.

5.3 Anthropomorphism and

Acceptance

Participants were asked to rate the look of the robot

on a ten-point rating scale (0 = machine-like, 9 =

human-like), as a manipulation check to ensure the

independent variable — in this case the

anthropomorphism level of the robot trainer — was

perceived as intended. The scores in both studies

turned out to be as intended. For study one, a one-way

ANOVA analysis showed the perceived human-

likeliness of the high anthropomorphic concept

(M=4.28, SD=1.77), medium anthropomorphic

concept (M=2.96, SD=2.33) and low

anthropomorphic concept (M=.70, SD=1.03) were

significantly different, F (2-70) =22.79, p<.001. In

study two the three concepts were all shown to every

participant at the same time. Participants were asked

to rate all three concepts on anthropomorphism. One-

How Anthropomorphism Affects User Acceptance of a Robot Trainer in Physical Rehabilitation

way ANOVA analysis also shows a significant

difference between the ratings of human-likeliness (F

(2,89) = 121.34, p<.001) between the three versions

(HA:M=6.67, SD=1.79, MA:M=3.1, SD=1.4,

LA:M=.48, SD=1.46).

A one-way ANOVA analysis showed no

significant effect of anthropomorphism on the

credibility (F (2,68) = 1.89, ns) and expectancy (F

(2,68) = 2.02, ns) of the SAR-assisted rehabilitation

therapy.

A similar analysis on the outcomes of the TAM

questionnaire found a significant preference for the

low anthropomorphic concept (M = 5.34, SD = 1.2),

over the high anthropomorphic (M = 4.43, SD = 1.71)

and the medium anthropomorphic concept (M = 4.02,

SD = 1.7) on the factor Self-Efficacy (SE), F (2,68) =

3.8, p < .05. No significant effects of

anthropomorphism were found for the factors

Perceived Ease of Use (PEU, F (2,68) = 1.84, ns),

Perceived Usefulness (PU, F (2,68) = .7, ns),

Behavioural Intention (BI, F (2,68) = .27, ns),

Attitude (A, F (2,68) = 3.03, ns) and Subjective Norm

(SN, F (2,68) = 1.15, ns).

In Study Two, participants were asked to choose

their preferred version of the robot trainer.

Participants preferred the high anthropomorphic

concept (17 out of 31), followed by the medium

anthropomorphic concept (9/31) and low

anthropomorphic concept (5/31). 13 quotes indicated

that the human-likeliness was the reason for choosing

the high anthropomorphic concept, 7 quotes because

it looks warm and feels personal, and 6 quotes

because it seems most able to provide a better and

more varied service. Reason for participants who

chose one of the other two concepts were 1)

participants feel safe around them, 2) participants are

unfamiliar with the high anthropomorphic concept,

not knowing what to expect, and 3) a robot trainer

with high anthropomorphic appearance, but only

voice interaction, is considered unintelligent. A low

anthropomorphic robot trainer offering the same

interaction was deemed realistic within the current

situation. In line with this reasoning, a unique group

of quotes appeared for the high anthropomorphism

concept. These 15 quotes inquired whether the robot

would provide hands-on training assistance. This

highlights a potential link between the robot’s

appearance and users’ expectations about its

functionality.

All three concepts got quotes suggesting to make

the robot trainer more human-like rather than

machinelike, which indicates a preference for high

anthropomorphic SARs for rehabilitation training

(see Figure 5). Interestingly, most of these quotes

were proposed in relation to the most

anthropomorphic concept.

Figure 5: Number of Quotes for Desired

Anthropomorphism Level.

6 DESIGN IMPLICATIONS

6.1 Managing Form Factor and

Patient’s Expectations

Patient’s expectations are related to their impressions

of the robot trainer, and were found to affect patients

perceived self-efficacy toward the robot trainer. The

form design of the should fit the task that will be

performed, as the form factor of the trainer is

indicative for the patients to assume the service that

will be offered. Therefore, the form design of the

robot trainer should be referred to the task assigned to

it, and made easy for patients to have a realistic

expectation and a positive opinion.

6.2 Connecting to Wearable

Technologies

One of the participants’ major concerns is the quality

of the feedback given by the robot trainer. Also, trust

is a crucial issue in a medical context. The use of

wearable technologies has the potential for dealing

with both issues. Data like heart rate, perspiration,

body posture, speed, muscle tension and more can

provide a more comprehensive understanding of the

training condition of the patient (Beckerle et al.,

2017), giving a more accrete and timely

understanding of the situation in progress.

Furthermore, being able to see the instrument of

measuring would help in explaining and specifying

the process and content of user data collection,

Human-Like Machine-Like

HA MA LA

HEALTHINF 2019 - 12th International Conference on Health Informatics

therefore facilitating better trust in SAR-assisted

rehabilitation therapy.

6.3 Allowing Emotional Input for

Tailored Experience

A physiotherapist is always able to pick up emotional

conditions of the patient, therefore can select the

suitable motivational prompts accordingly. Being

able to acquire and respond appropriately to a

patient’s emotional condition can better help to tailor

the experience offered, and can help to fast-forward

the adaptation phase, where the robot trainer has to

learn about the patient.

7 LIMITATIONS OF CURRENT

STUDY

The main limitations pertain to the nature of the

materials presented to participants (sketches and

scenarios) and the sampling approach

(crowdsourcing). The attitudes expressed are based

on imagined experiences based on very limited

stimuli. More extensive and realistic exposure to a

robot acting as a coach in rehabilitation therapy

would provide higher confidence in the results found,

and further study will also shift focus to elder groups

which fits better with rehabilitation context. Second,

with crowdsourcing one is constrained to the

crowdsourcing platform as a sampling frame. This

allows surveys that cover different geographical

areas, but the sample may be skewed to people very

familiar with internet technology. Further, there could

be potential for more satisficing behaviors from

crowd workers interested to earn the reward rather

than provide good data. On the upside,

methodological research has shown that

crowdsourcing can be effective and valid approach

for accessing participants and that crowd workers can

actually be more motivated to provide good quality

answers(Stewart et al., 2017). Moreover, suitable

checks were made to check on the quality of the data

in accordance to the screening process and verifying

reversed questions.

8 CONCLUSION

This work explores of patients’ acceptance for

socially assistive robot in rehabilitation settings, and

the use anthropomorphic form factor in robotic design

in this context. We discovered that 1) participants

have a neutral to positive attitude towards SARs and

it’s use in rehabilitation therapy as trainer, 2) the SAR

technology in therapies is regarded easy to use but

participants generally lack intention for using the

system, which is possibly due to unfamiliarity with

SARs and lack of trust for them, 3) the level of

anthropomorphism has an effect on patients’ self-

efficacy and attitude for the robot trainer, a potential

link exists between SAR form factor and user

expectation for its service, 4) high anthropomorphic

concept is generally preferred. As we found out, to

bring SAR into rehabilitation therapies, much work is

yet to be done in solving patient concerns, improving

the quality of the therapy and developing detailed

personalized motivational strategies. Further studies

should replicate and elaborate these results while

exposing patients to more realistic experiences of

social-robot assisted rehabilitation therapy and

include proximal (non-crowdsourced) studies with

actual patients.

REFERENCES

Bartneck, C., Bleeker, T., Bun, J., Fens, P., Riet, L. (2010)

'The influence of robot anthropomorphism on the

feelings of embarrassment when interacting with

robots'. Paladyn, Journal of Behavioral Robotics, 1(2).

Beckerle, P., Salvietti, G., Unal, R., Prattichizzo, D., Rossi,

S., Castellini, C., Bianchi, M. (2017) ' A Human–Robot

Interaction Perspective on Assistive and Rehabilitation

Robotics', Frontiers in Neurorobotics, 11.

Cardona, M., Destarac, M. A., Garcia, C. E. (2017)

'Exoskeleton robots for rehabilitation: State of the art

and future trends In' 2017 IEEE 37th Central America

and Panama Convention (CONCAPAN XXXVII)

Managua: IEEE, pp. 1–6.

Choi, J., Kim, M. (2008) 'The Usage and Evaluation of

Anthropomorphic Form in Robot Design', 15.

Devilly, G. J., Borkovec, T. D. (2000) 'Psychometric

properties of the credibility/ expectancy questionnaire',

14.

Ding, Y., Galiana, I., Asbeck, A. T., De Rossi, S. M. M.,

Bae, J., Santos, T. R. T., … Walsh, C. (2017)

'Biomechanical and Physiological Evaluation of Multi-

Joint Assistance With Soft Exosuits' IEEE Transactions

on Neural Systems and Rehabilitation Engineering,

25(2), 119–130.

Epley, N., Waytz, A., Cacioppo, J. T. (2007) 'On seeing

human: A three-factor theory of anthropomorphism'

Psychological Review, 114(4), 864–886.

Eriksson, J., Mataric, M. J., Winstein, C. J. (2005) 'Hands-

Off Assistive Robotics for Post-Stroke Arm

Rehabilitation' In 9th International Conference on

Rehabilitation Robotics, 2005. ICORR 2005. Chicago,

IL, USA: IEEE, pp. 21–24.

How Anthropomorphism Affects User Acceptance of a Robot Trainer in Physical Rehabilitation

Eyssel, F., Kuchenbrandt, D., Bobinger, S. (2012) ‘If You

Sound Like Me, You Must Be More Human’: On the

Interplay of Robot and User Features on Human- Robot

Acceptance and Anthropomorphism, 2.

Feil-Seifer, D., Mataric, M. J. (2005) 'Socially Assistive

Robotics' In 9th International Conference on

Rehabilitation Robotics, 2005. ICORR 2005. Chicago,

IL, USA: IEEE, pp. 465–468.

Ferrari, F., Paladino, M. P., Jetten, J. (2016) 'Blurring

Human–Machine Distinctions: Anthropomorphic

Appearance in Social Robots as a Threat to Human

Distinctiveness' International Journal of Social

Robotics, 8(2), 287–302.

Feys, P., Coninx, K., Kerkhofs, L., De Weyer, T., Truyens,

V., Maris, A., Lamers, I. (2015) 'Robot-supported upper

limb training in a virtual learning environment: a pilot

randomized controlled trial in persons with MS' Journal

of NeuroEngineering and Rehabilitation, 12(1), 60.

Fong, T., Nourbakhsh, I., Dautenhahn, K. (2003) 'A survey

of socially interactive robots' Robotics and Autonomous

Systems, 42(3–4), 143–166.

Gockley, R., Mataric, M. J. (2006) 'Encouraging physical

therapy compliance with a hands-Off mobile robot' In

Proceeding of the 1st ACM SIGCHI/SIGART

conference on Human-robot interaction - HRI ’06 Salt

Lake City, Utah, USA: ACM Press, pp. 150.

Heerink, M., Kröse, B., Evers, V., Wielinga, B. (2010)

'Assessing Acceptance of Assistive Social Agent

Technology by Older Adults: the Almere Model'

International Journal of Social Robotics, 2(4), 361–

375.

Klamer, T., Allouch, S. B. (2010) 'Acceptance and use of a

social robot by elderly users in a domestic environment'

IEEE.

Kyong Il Kang, Freedman, S., Mataric, M. J., Cunningham,

M. J., Lopez, B. (2005) 'A Hands-Off Physical Therapy

Assistance Robot for Cardiac Patients' In 9th

International Conference on Rehabilitation Robotics,

2005. ICORR 2005. Chicago, IL, USA: IEEE, pp. 337–

340.

Laut, J., Porfiri, M., Raghavan, P. (2016) 'The Present and

Future of Robotic Technology in Rehabilitation'

Current Physical Medicine and Rehabilitation Reports,

4(4), 312–319.

Moosaei, M., Das, S. K., Popa, D. O., Riek, L. D. (2017)

'Using Facially Expressive Robots to Calibrate Clinical

Pain Perception' In Proceedings of the 2017 ACM/IEEE

International Conference on Human-Robot Interaction

- HRI ’17 Vienna, Austria: ACM Press, pp. 32–41.

Nomura, T., Kanda, T., Suzuki, T., Kato, K. (2004)

'Psychology in human-robot communication: an

attempt through investigation of negative attitudes and

anxiety toward robots' IEEE, pp. 35–40.

Popescu, N., Popescu, D., Ivănescu, M. (2016) 'Intelligent

Robotic Approach for After-stroke Hand

Rehabilitation' In Proceedings of the 9th International

Joint Conference on Biomedical Engineering Systems

and Technologies. Rome, Italy: SCITEPRESS -

Science and Technology Publications, pp. 49–57.

Salem, M., Eyssel, F., Rohlfing, K., Kopp, S., Joublin, F.

(2013) 'To Err is Human(-like): Effects of Robot

Gesture on Perceived Anthropomorphism and

Likability' International Journal of Social Robotics,

5(3), 313–323.

Siegel, M., Breazeal, C., Norton, M. I. (2009) 'Persuasive

Robotics: The influence of robot gender on human

behavior' In 2009 IEEE/RSJ International Conference

on Intelligent Robots and Systems. St. Louis, MO,

USA: IEEE, pp. 2563–2568.

Stewart, N., Chandler, J., Paolacci, G. (2017)

'Crowdsourcing Samples in Cognitive Science' Trends

in Cognitive Sciences, 21(10), 736–748.

Tapus, A., Mataric, M. J. (2008) 'Socially Assistive Robots:

The Link between Personality, Empathy, Physiological

Signals, and Task Performance' 9.

Vitiello, N., Mohammed, S., Moreno, J. C. (2017) 'Guest

Editorial Wearable Robotics for Motion Assistance and

Rehabilitation' IEEE Transactions on Neural Systems

and Rehabilitation Engineering, 25(2), 103–106.

Winkle, K., Caleb-Solly, P., Turton, A., Bremner, P. (2018)

'Social Robots for Engagement in Rehabilitative

Therapies: Design Implications from a Study with

Therapists' In Proceedings of the 2018 ACM/IEEE

International Conference on Human-Robot Interaction

- HRI ’18. Chicago, IL, USA: ACM Press, pp. 289–297.

Zaga, C., de Vries, R. A. J., Li, J., Truong, K. P., Evers, V.

(2017) 'A Simple Nod of the Head: The Effect of

Minimal Robot Movements on Children’s Perception

of a Low-Anthropomorphic Robot' In Proceedings of

the 2017 CHI Conference on Human Factors in

Computing Systems - CHI ’17, Denver, Colorado, USA:

ACM Press, pp. 336–341.

HEALTHINF 2019 - 12th International Conference on Health Informatics