Acceptance of Healthcare Robots: the Integrated Model

(1)

Acceptance of Healthcare Robots:

the Integrated Model

Bachelor Thesis

Information Science - University of Amsterdam

Edward Gubler

10536892

Supervisor:

Mw. Dr. J.A.C. (Jacobijn) Sandberg

J.A.C.Sandberg@uva.nl

May 19, 2018

Abstract

This research provides an Integrated Model regarding the factors that have influence on the acceptance of healthcare robots. These factors con-sist of factors found in existing models; the UTAUT(2) by Venkatesh et al., the Almere Model by Heerink et al. and the TAM by Davis et al. and two new factors based on recent studies on the acceptance of healthcare robots. The factors were operationalized the same way as in the mod-els stated above and data was collected by using convenience sampling. the Integrated Model (research model) was tested by performing path analysis on the model. Based on this analysis most of the factors were confirmed to have an (in)direct effect on the Intention to Use health-care robots. After the analysis a Confirmed Integrated Model remains. These confirmations contribute to the current insight in factors that have influence on the acceptance of healthcare robots.

(2)

1 Introduction

In the last decades the life expectancy of citizens in Europe has increased signif-icantly. This results in an increasing percentage of the population being older than 60 years. In 2017 the percentage of the population being over 60 years of age in Europe was 25%, where in 2050 it’s expected to reach 35%. The ex-pectancy is that the percentage of the population that is over 60 years of age will keep increasing in the following decades(United Nations, 2017).

Besides the increasing population over the age of 60, The number of births in Europe has also decreased. The number of births in all European countries is currently below the required level for replacement of the population in the long run (around 2.1 births per woman, on average) and, for most European countries, has been below the required replacement level for multiple decades already(United Nations, 2017). Together this leads to an increasing general population ageing in Europe.

Because of the increasing population ageing there is an increasing number of elderly in need of caretakers to cater for them, but there are not enough care-takers available for the total elderly population and this problem is growing. The use of healthcare robots may be a solution for the lack of human caretak-ers, but the elderly population has to accept these robots instead of real-life human caretakers first(Forlizzi, 2005).

Nowadays, healthcare robots are already being used in the care sector. For instance it is possible for healthcare robots to assist elderly with reminding them to take their medicines, lift the senior citizens out of bed or being a social companion for them when there is no one else around to have a form of social interaction with(Broadbent et al., 2009). In Japan the use of robots is already widely accepted, but in the Western world the acceptance and usage of robots has not reached the same level yet(Kaplan, 2004). Of course there are cultural differences that could play a part in the difference of implementation between these two parts of the world. Some of these cultural factors are that overall Japanese citizens are technology fans, where Westerners consider technology as less important. In Japan the difference between natural and artificial entities is less essential than in the Western world. The building of all sorts of machines is in Japan seen as a way to bend the natural laws in a positive way. In the Western world citizens seem to be more afraid of the impact machines can have on human race(Kaplan, 2004).

Besides cultural factors, there could be other factors that have a significant pos-itive or negative effect on the acceptance of healthcare robots in the Western world.The research presented in this paper investigates if there are factors that have a significant positive or negative influence on the acceptance of healthcare robots. The main goal of this research is to give scientific insight into the fac-tors that are currently of importance and create an Integrated Model of the

(4)

factors that are found to be important by other research. Integrated Model will be tested to see which factors can be confirmed. Insight into these factors could help the developers of healthcare robots by showing producers of health-care robots what are important factors for people to accept healthhealth-care robots.

2 Theoretical Background

2.1 The concept of healthcare robots

A healthcare robot can mean different things. First a robot can be defined as “a very powerful computer with equally powerful software housed in a mobile body and able to act rationally on the world around it.”(Broadbent et al., 2009, p. 319). Besides defining a general robot, a healthcare robot can be defined as follows: “a healthcare robot is primarily intended to improve or protect the health and lifestyle of the human user.”(Broadbent et al., 2009, p. 319). There are multiple types of healthcare robots that focus on different aspects of pro-viding healthcare. For instance there are surgical robots which can assist a surgeon with performing surgery. A popular surgical robot is daVinci. Surgi-cal robots can differ in size and are not always humanoid, also surgiSurgi-cal robots do not have to be social robots(Hockstein et al., 2005). There are robots that perform other tasks, lifting patients out of a bed for instance, like RIBA. These kind of healthcare robots are not always humanoid and do not have to be social robots(Mukai et al., 2010). Besides robots to increase physical health, there are robots to increase emotional health of people as well. These kind of robots are social robots. A good example of such a robot is PARO, is a robotic seal doll that is being used to act as a social companion for elderly people that are lonely(Sabanovic et al., 2013). Combining both physical and emotional health increasing features could possibly create a healthcare robot that can be of seri-ous use in, for instance, homes for the elderly. For the remainder of this paper a healthcare robot will be defined as a social humanoid robot that can provide medical care and medical advice.

2.2 Earlier research on the acceptance of technology in

general and of healthcare robots in particular

Research in the area of the acceptance of healthcare robots takes place, within the context of technology acceptance in general. The latter has a long history. One of the first theoretical models of technology and an important part of the research to technology acceptance is the Technology Acceptance Model (TAM) by Davis et al. (1989). This model was built to give insight into the acceptance of people of technology to be able to improve the technology to increase accep-tance rate. The model was operationalized by creating variables for the factors in the model and creating a questionnaire of these variables. This method for

(5)

“user acceptance testing” could be used for other systems as well to test the acceptance rate of new systems. Important factors in the model are “Perceived Usefulness” of the system, “Perceived Ease of Use” of the system and the “At-titude Toward Using” the system. Higher scores on these factors should result in a higher “Actual System Use”. For every factor Likert scale statements were created to measure these factors for a system by using the questionnaire stated above (Davis et al., 1989).

Figure 1: TAM by Davis et al. (1989, p. 985)

Venkatesh et al. created a new model on the acceptance of technology called the Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh et al., 2003). The TAM by Davis et al. was a main source for this new model. The UTAUT model was created because there were different theories and mod-els on technology acceptance including the TAM model, but Venkatesh et al. wanted to create a unified model out of all these theories and models and vali-date the newly created model, the UTAUT model. The validation method used was comparable to the method used for the TAM model. A questionnaire with statements that were created for every factor was used here as well. The fac-tors “Performance Expectancy” and “Effort Expectancy” in the UTAUT are factors that are comparable with “Perceived Usefulness” and “Perceived Ease of Use” from the TAM. “Social Influence” and “Facilitating Conditions” were added to the UTAUT model, as well as the following four key moderating fac-tors “Gender”, “Age”, “Experience” and “Voluntariness of Use”, because these were expected to be of significant influence based on other studies and were indeed confirmed to be of significant influence by Venkatesh et al. Venkatesh et al. found that “Performance Expectancy”, “Effort Expectancy” and “Social Influence” had a significant direct effect on “Intention to Use”. “Facilitating Conditions” and “Intention to Use” had a significant direct effect on “Actual Use”. Later the UTAUT was extended to create a model that fits consumer acceptance better, called the UTAUT2 (Venkatesh et al., 2012). “Hedonic Mo-tivation”, which is the enjoyment that users of the technology experience, was

(6)

added to the model. “Habit” was added to see if using the technology had become a routine for the users. “Price Value”, which could be important for consumers, was added to the model as well(Venkatesh et al., 2012). These three new factors were all confirmed to have significant influence on “Intention to Use”. For “Habit” a direct significant effect on “Use” was also found. In the UTAUT2 Venkatesh et al. also found “Facilitating Conditions” to have a signifi-cant effect on “Intention to Use” where in the original UTAUT only a signifisignifi-cant effect on “Use” was found for “Facilitating Conditions”. A significant influence of “Age” and of “Gender” on “Social Influence” and on “Price Value” was found as well. “Age”, “Gender” and “Experience” were also found to have influence on “Hedonic Motivation”. “Experience” was found to have significant effect on “Behavioral Intention”. All the definitions of these factors can be found in appendix I.

Figure 2: UTAUT2 by Venkatesh et al. 2012, (p. 160)

Instead of performing research regarding the acceptance of technology in gen-eral, Heerink et al. performed research focused on the acceptance of healthcare robots. Heerink et al. found that a robot with more social abilities has a higher score on measuring ‘’Social Presence”(Heerink et al., 2008). Social presence means when using the robot, it can be seen as a social entity by the user. which

(7)

leads to a higher ‘’Perceived Enjoyment” level for the user. The higher level of “Perceived Enjoyment” has a positive influence on the “Intention to Use” a robot. From this research can be concluded that in the end a healthcare robot with more social skills will more likely be accepted by (potential) users.

In later research they confirmed with a case study that the “Ease of Use” of robots and the “Attitude” of people towards robots plays a significant part in the intention to use robots and in the end to the actual use of healthcare robots(Heerink et al., 2009). Heerink et. al created a model that could be used to test the acceptance of technology as well. This model is based on the general ideas of the UTAUT. Heerink et al. found a significant influence of “Trust” on “Perceived Sociability” and on “Intention to Use”, a significant influence of “Perceived Sociability” on “Social Presence” and a significant influence of “So-cial Presence” on “Perceived Enjoyment”. Also “Perceived Adaptability” was found to have significant influence on “Perceived Usefulness”. “Perceived Ease of Use” was found to have a significant influence on “Perceived Usefulness” and on “Intention to Use”. “Perceived Usefulness” was found to have a significant influence on “Intention to Use”. “Anxiety” was found to have a significant in-fluence on both “Perceived Usefulness” and on “Perceived Ease of Use”. Finally “Attitude” was confirmed to have a significant influence on “Trust”. In total “Trust”, “Perceived Sociability”, “Social Presence”, “Adaptability” and “Anx-iety” were added to the model by Heerink et al. because they were confirmed to be of influence on the acceptance of healthcare robots, but not yet in any model on this subject. The final model was called the Almere Model(Heerink et al., 2010)

In a study on the acceptance of healthcare robots for the older population, Broadbent et al. found, besides factors already known, that the individual medical needs of a person and the education level of a person could play a serious role in the acceptance of healthcare robots (2009). They stated that a single perfect design for a healthcare robot is unlikely, assessing individual needs and preferences and matching these may enable greater acceptance(Broadbent et al., 2009). Broadbent et al. based these findings on a literature review. For some factors the results differ between studies. Kuo et al. found that age isn’t an important factor, but gender is. Men were found to accept healthcare robots more easily than women. Besides that they also point out the importance of the social skills of the robot. The voice and the interactiveness of the robot could be improved according to the participants of the study(Kuo et al., 2009). Kuo et al. used a method similar to the method of Davis et al., Heerink et al. and Venkatesh et al. A few years later, Flandorfer found that age is in fact an important factor, besides confirmed factors gender and education level. Flandorfer also pointed out the importance of adaptability to individual needs (2012). Flandorfer used a literature review as well. Broadbent and Flandorfer found different results for the importance of age(Flandorfer, 2012). The ques-tion remains if age is indeed an important factor.

(8)

Figure 3: The Almere Model by Heerink et al.(redrawn),2010 (p. 372)

3 The Integrated Model

This section introduces an Integrated Model, based on previously established relationships and some new factors arising from recent literature. The purpose was to create and subsequently validate a model that accurately reflects the current status of our knowledge on the factors influencing healthcare robot ac-ceptance. Based on recent research on the acceptance of healthcare robots and on the acceptance of technology in general, two new factors were found that are not in the UTAUT(2) or in the Almere model. The first new factor found was “Personal Needs”. An important part of healthcare nowadays is “patient-centered care” (Epstein et al., 2010). Every patient is different and has different needs, which can play a part in the acceptance of healthcare robots as well, as pointed out by Broadbent et al. (2009) and by Flandorfer (2012). That’s why in the model “Personal Needs” was added, which was likely to have influence on “Adaptability” and on “Performance Expectancy”.

The second new factor is a moderating factor; “Education Level”. The key mod-erating factors in this Integrated Model consist of “Age”, “Gender”, “Com-puter Experience” and “Education Level”. For these key moderating factors

(9)

the possible influence on all other variables were tested. “Education Level” has also not been a part of earlier models of Heerink et al. or Venkatesh yet, but is suggested to be relevant by Flandorfer (2012) and by Broadbent et al. (2009). The factor “Culture” was removed from these general individual factors because the participants that took part in the survey had quite the same cultural back-ground.

Some factors in the UTAUT2 and the Almere Model by Heerink et al. have a different name, but mean the same. In this case the names of the UTAUT2 were used. On the left side of the created model, the Almere Model by Heerink et al. was reconstructed and connected to the UTAUT2 on the right side. Between both models “Personal Needs” was implemented. Because of the lack of data on “Use” all relations going to “Use” in the earlier models were linked to “Intention to Use”.

• H1: ‘’Trust” has significant influence on ‘’Perceived Sociability”.

• H2: ‘’Perceived Sociability” has a significant influence on “Social Pres-ence”.

• H3: “Social Presence” has a significant influence on “Hedonic Motivation”. • H4: “Anxiety” has a significant influence on “Performance Expectancy”

and on “Effort Expectancy”.

• H5: “Attitude” has a significant influence on “Intention to Use”.

• H6: “Performance Expectancy” has a significant influence on “Intention to Use”.

• H7: “Effort Expectancy” has significant influence on “Intention to Use” and on “Performance Expectancy”.

• H8: “Hedonic Motivation” has a significant influence on “Intention to Use”.

• H9: “Social Influence” has a significant influence on “Intention to Use”. • H10: “Facilitating Conditions” has a significant influence on “Intention to

Use”.

• H11: “Habit” has a significant influence on “Intention to Use”.

• H12: “Personal Needs” has a significant influence on “Performance Ex-pectancy” and on “Adaptability”.

• H13: “Adaptability” has a significant influence on “Intention to Use”. • H14: “Gender” has a significant influence on one or more of the latent

(10)

• H15: “Age” has a significant influence on one or more of the latent vari-ables.

• H16: “Computer a Experience” has significant influence on one or more of the latent variables.

• H17: “Education Level” has a significant influence on one or more of the latent variables.

These hypotheses were derived from the Integrated Model. The goal was to finish with a model with all the confirmed factors and relations in there and to throw away any factors and relations that were not confirmed by this research. After this process the difference between the confirmed factors and relations in the Almere Model by Heerink et al., the UTAUT2 by Venkatesh et al. and the TAM by Davis et al. was compared with the confirmed factors and relations found in this research.

(11)

(12)

4 Method

4.1 Participants

The participants for this study were collected by convenience sampling. All of the participants are Dutch citizens. In total 112 persons participated in this study. 74 of the participants were male and 38 were female. The educa-tion level of the participants was for 83 participants WO/VWO, for 26 persons HBO/HAVO and for 3 persons MBO/VMBO. Details on Age and Computer Experience can be found in Appendix III.

4.2 Design

Correlational non-experimental research was performed because the predictor variables were not controlled, manipulated or altered. Also random assignment of participants to specific conditions was not performed.

4.3 Materials

Measurement: A survey was created to investigate the factors depicted in the Integrated Model and a video was selected to give the participants a good impression of a healthcare robot. The factors measured are Trust, Perceived Sociability, Social Presence, Anxiety, Attitude, Performance Expectancy, Ef-fort Expectancy, Hedonic Motivation, Social Influence, Facilitating Conditions, Habit, Personal Needs, Adaptability, Intention to Use, Use, Age, Computer Ex-perience, Gender and Education Level. All factors in the Integrated Model were connected to about five variables each, which in total was 69 variables. A variable is a statement where the participant can agree with on a Likert scale from 1 to 5. 1 would mean “Strongly Disagree” and 5 would mean “Strongly Agree”. For instance this is a variable for “Habit”: Using the robot could be-come natural to me. These variables were partly already there in the UTAUT2 and have been edited to work for healthcare robots as well. In table 1 the sources per factor for the statements can be found. Besides editing the exist-ing variables, some variables were added so every factor would have about the same number of variables. For every factor five variables were created, except for “Facilitating Conditions” where 4 variables were created. In most cases the variables were formulated positive (60/69), but sometimes there was a negative statement (9/69) in there as well to make sure the participant was not filling in random values in the survey. To collect data on this subject a survey was created in Google Forms. The complete survey can be found in appendix IV.

(13)

Table 1

Number of variables per factor Factor Number of vari-ables in the Almere Model Number of vari-ables in the UTAUT2 Number of variables in the Integrated Model

Trust 2 X 5 (2 existing + 3 new) Perceived Sociability 4 X 5 (4 existing + 1 new) Social Presence 5 X 5 existing

Anxiety 4 X 5 (4 existing + 1 new) Attitude 3 X 5 (3 existing + 2 new) Performance Expectancy 3 4 (4 existing + 1 new) Effort Expectancy 5 3 5 existing

Hedonic Motivation 5 3 5 existing

Social Influence 2 3 5 (3 existing + 2 new) Facilitating Conditions 2 4 4 existing

Habit X 4 5 (4 existing + 1 new) Adaptability 3 X 5 (3 existing + 2 new) Personal Needs X X 5 new

Intention to Use 3 3 5 (3 existing + 2 new)

Healthcare robot video: To familiarize the participant with the concept of a healthcare robot,the following freely available video (link) was selected and edited a bit so the duration was just 2:27. This a video of Dr. Pepper, a social healthcare robot. Dr. Pepper is a model of the Pepper robot developed by Softbank, a Japanese based telecom company(Dignan, 2014). The social robot can be used for different functions, but in the video it is being used as a healthcare advisor. First of all dr. Pepper welcomes the patient. After welcoming the patient, the robot does some general analyses on the patients health including stress level and sleep pattern. Also the blood pressure is being measured. The robot gives some tips to improve these health indicators for the patient. This was the best video available because it can be seen as a social humanoid robot that can provide medical care and medical advice as defined earlier in the Theoretical Background section. Other options found missed one or more of the necessary features according to the definition used in this research.

4.4 Procedure

The participants received a link so they could view the video and fill in the survey on their own computer or mobile device. First of all participants were informed that their data was kept anonymous. After this, some general information was

(14)

collected; the age (digit), Gender (Man/Woman), Computer Experience (Likert scale 1-5) rated by the participant for himself or herself and the Education Level following the three general Dutch education levels (1 = MAVO/VMBO, 2 = HAVO/HBO, 3 = VWO/WO on a scale level). Next the participant was offered the following introductory text: Probably you are currently not in need of medical care whatsoever. If not, please imagine you would be in need of medical care living as an elderly person in an elderly home. At this point you would be in need of medical care and wouldn’t be able to manage taking full care of yourself.

Next, the participant received a message that it was very important to watch the video that was shown, which was the video of Dr. Pepper as described above. After the video, the introductory text was presented once again to stress the research context once more.

After viewing the video, the participant had to give his or her answer to the 69 statements. When the questionnaire was done, the participant was thanked for his or her time.

4.5 Analysis

For the analysis first of all SPSS (version 23.0) was used to convert all the data to the right formats. After this the data was loaded in SmartPLS (version 3.2.7) to perform Structural Equation Modelling to test the Integrated Model. In SmartPLS a path analysis was performed with the latent variables to see if there were any significant influences between the variables. This way the hypotheses derived from the Integrated Model section were tested.

5 Results

5.1 Reliability of Constructs

First performing a Reliability of Constructs test had to be done to see if all the factors explained their variables (indicators). First of all the Cronbach’s alpha test was performed. The alpha has a threshold minimum value of 0.7. Facilitating Conditions (0.44), Personal Needs (0.40), Social Influence (0.66) and Trust (0.68) had a overall Cronbach’s alpha below 0.7. By removing the variables that were lower than 0.7 the quality of the model was increased. After the removal of variables Personal Needs was removed from the model completely because there was only 1 variable left for this factor. Except for Facilitating Conditions (0.69) and Social Influence (0.67), all the factors and remaining variables met the Cronbach’s alpha requirements. Facilitating Conditions and Social Influence were kept in the model because these factors were very close to the threshold and the Cronbach’s Alpha is a severe test. Removing the variables with a Cronbach’s alpha below 0.7 also resulted in

(15)

meeting the Average Variance value above 0.5 for all factors. The remaining model was now ready to perform Path Analysis on the model. The complete removals of the variables can be found in Appendix II.

Table 2

Cronbach’s alpha and Average Variance before and after the removal of variables Factor Cronbach’s Alpha be-fore Cronbach’s Alpha af-ter Average Variance before Average Variance after Adaptability 0.777 0.768 0.532 0.591 Anxiety 0.793 0.744 0.545 0.660 Attitude 0.809 0.856 0.577 0.777 Effort Expectancy 0.716 0.786 0.484(X) 0.697 Facilitating Conditions 0.436(X) 0.692(X) 0.398(X) 0.763 Habit 0.733 0.748 0.493(X) 0.666 Hedonic Motivation 0.784 0.764 0.538 0.588 Intention to Use 0.813 0.790 0.573 0.618 Performance Expectancy 0.777 0.762 0.529 0.582 Sociability 0.804 0.709 0.562 0.613 Social Influence 0.656(X) 0.672(X) 0.417(X) 0.601 Social Presence 0.701 0.751 0.463(X) 0.666 Trust 0.678(X) 0.777 0.457(X) 0.690

(16)

(17)

5.2 Path Analysis

A PLS (Partial Least Squares) Algorithm analysis was performed on the Integrated Model in SmartPLS(Schubring et al., 2016). This Algorithm is a sequence of regression analyses with weighted vectors. Next, bootstrapping was performed to test the statistical significance of the results found in the PLS Algorithm. A set of 5000 sub samples was used to ensure stability of results, these sub samples were generated from the original data. A threshold for statistical significance of 0.05 was chosen for the p-value because a two-tailed test was used. The path coefficients derived from the PLS Algorithm analysis and the P-values derived from the bootstrapping can be found for the relations in the model in table 4. For the variables “Age” and “Computer Experience” only the significant relations were included because a lot of relations were not statistically significant. “Education Level” and “Gender” were removed from the model because a multi-Group analysis would be necessary to obtain trustworthy results, but the group sizes were too small. For a Multi-Group analysis the data would have been divided in groups to measure the difference between the different groups divided on “Gender” or “Education Level” , but the groups would be too small to obtain trustworthy results. A small effect size is a path coefficient between -0.2 and 0.2, a medium effect size is a path coefficient above 0.2 and below 0.8, or below -0.2 and above -0.8, a big effect size is a path coefficient above 0.8 or below -0.8. The effect sizes for all relations can be found in Table 3. The relations in H1, H2, H3, H4, H5, H6, H8, H11 and H15 were confirmed by the algorithm and the bootstrapping. For H7, only one of the two statements was confirmed. H12, H14 and H17 were not tested because the data was not sufficient. For H12 (Personal Needs) the factor was not operationalized well as described in the Reliability of Constructs section. H14(Gender) and H17(Education level) were also not tested because it was not possible to perform a multi-group analysis on a sample of this size. H9, H10 and H13 were tested, but not confirmed to be statistically significant.

The complete analysis for the Confirmed Integrated Model results in an R2_{(explained variance) of 0.79 for ”Intention to Use”, which is exceptionally}

high. This can be found in Appendix 5 in the Smart PLS model. The Factors which had the most influence on Intention to Use were ”Habit”, ”Attitude” and ”Performance Expectancy” which can be found in Table 3.

(18)

Table 3

Path analysis results for the hypothesis (”*” = p-value below 0.05) Hypothesis Independent Factor Dependent Factor Path Coefficient

+ Effect Size

p-value H1 Trust Perceived Sociability 0.563(medium) 0.000* H2 Perceived Sociability Social Presence 0.73(medium) 0.000* H3 Social Presence Hedonic Motivation 0.539(medium) 0.000* H4.1 Anxiety Performance Expectancy -0.368(medium) 0.000* H4.2 Anxiety Effort Expectancy -0.452(medium) 0.000* H5 Attitude Intention to Use 0.350(medium) 0.000* H6 Performance Expectancy Intention to Use 0.285(medium) 0.000* H7.1 Effort Expectancy Intention to Use -0.058(small) 0.464 H7.2 Effort Expectancy Performance Expectancy 0.286(medium) 0.021* H8 Hedonic Motivation Intention to Use 0.172(small) 0.014* H9 Social Influence Intention to Use 0.067(small) 0.481 H10 Facilitating Conditions Intention to Use 0.131(small) 0.075 H11 Habit Intention to Use 0.226(medium) 0.001* H12.1 Personal Needs Performance Expectancy X X H12.2 Personal Needs Adaptability X X H13 Adaptability Intention to Use 0.153(small) 0.069

H14 Gender X X X

H15.1 Age Anxiety -0.243(medium) 0.037* H15.2 Age Attitude 0.254(medium) 0.015* H15.3 Age Perceived Sociability -0.154(small) 0.045* H16 Computer Experience Effort Expectancy 0.258(medium) 0.011*

H17 Education Level X X X

6 Discussion

6.1 Theoretical Contributions

This research makes a contribution to the research regarding the acceptance of healthcare robots. Several factors presented in earlier research by others on this subject have been confirmed again in this research. This strengthens the theoretical framework already present in the healthcare robot acceptance field. For some factors the influence on “Intention to Use” were already confirmed by Davis et al., Venkatesh et al. and Heerink et al. and are confirmed by this research as well. These factors are “Performance Expectancy” and “Ef-fort Expectancy”. For some factors, only Heerink et al. used and confirmed them, but these were confirmed by this research as well. These factors are “Trust”, “Perceived Sociability”, “Social Presence” and “Anxiety”. Only the factor “Adaptability” was confirmed solely by Heerink et al, but not by this

(19)

research. overall, an R2(explained variance) of 0.79 is exceptionally high, so the model says a lot about the factors that have influence on Intention to Use for healthcare robots. By chance this is exactly the same R2 value as the value found by Heerink et al. in the Almere Model. The two factors with the highest path coefficients on ”Intention to Use” were ”Performance Expectancy” and ”Attitude”. For the Almere Model by Heerink et al. these two factors were also the ones with the highest path coefficients on ”Intention to Use”.

Some factors were not depicted in the TAM by Davis et al., but were found and confirmed later by Venkatesh et al. and Heerink et al. One of these factors was also confirmed by this research, this was “Hedonic Motivation”. “Social Influ-ence” was not confirmed by this research as well as “Facilitating Conditions”. “Attitude” was earlier confirmed by Davis et al. and by Heerink et al., and also by this research. “Habit” was earlier solely confirmed by Venkatesh et al., but also by this research now. The moderating factors “Age” and “Experi-ence” were confirmed by both Venkatesh and this research, but because this research couldn’t perform a Multi-Group analysis, “Gender” is only confirmed by Venkatesh et al.

The newly added factors “Education Level” and “Personal Needs” unfortunately couldn’t be tested because of insufficient data.

In summary this research confirms 11 of the 15 factors found and confirmed in earlier models consisting of the TAM by Davis et al., the UATAUT(2) by Venkatesh (et al.) and the Almere Model by Heerink et. al. as can be found in Table 4.

(20)

Table 4

Confirmed (in)direct effect on Intention to Use per factor per research Factor Heerink et al. Venkatesh et al. Davis et al. Gubler Trust + x x + Perceived Sociability + x x + Social Presence + x x + Anxiety + x x + Attitude + x + + Performance Expectancy + + + + Effort Expectancy + + + + Hedonic Motivation + + x + Social Influence + + x -Facilitating Conditions - + x -Habit x + x + Personal Needs x x x -Adaptability + x x -Age x + x + Experience x + x + Education Level x x x x Gender x + x x

”+” = Confirmed, ”-” = not Confirmed, ”x” = no analysis performed.

6.2 Limitations and Future Research

As stated earlier, the analysis of “Personal Needs”, “Gender” and “Education Level” was not possible because of insufficient data. First of all, the variables created for “Personal Needs” were probably not clear enough and possibly be-cause of this, not correlating with each other. This resulted in removing all the variables for this factor, except for one variable. One variable would probably not be enough to be able to trustworthy confirm or not confirm this factors relevance. In the future the statements in the questionnaire could be improved to operationalize this factor better.

The moderating factors “Gender” and “Education Level” needed a Multi-Group analysis, which was not possible because of the sample size which wasn’t big enough. This could be solved in future research by increasing the sample size. Another problem for “Education Level” would have been that 83 of the 112 participants were on a “WO/VWO” level and only 3 participants were on a “MBO/MAVO” level. The participants in a future study should be from more differing education levels to be able to perform an analysis on this factor. Another possible improvement for this research in the future could be to use a real-life healthcare robot, instead of a video of a healthcare robot. This could improve the experience for the participants and give them a better impression of a healthcare robot.

(21)

7 Conclusion

This study confirms that the factors “Trust”, “Perceived Sociability”, “So-cial Presence”, “Anxiety”, “Attitude”, “Performance Expectancy”, “Effort Ex-pectancy”, “Hedonic Motivation”, “Habit”, “Age” and “Experience” have a significant (in)direct influence on the Intention to Use healthcare robots. These confirmations are in line with earlier models on this subject found in past re-search. Improving these factors when designing healthcare robots could possibly lead to a higher acceptance rate in the long run.

(22)

References

Broadbent, E., Stafford, R., and MacDonald, B. (2009). Acceptance of health-care robots for the older population: Review and future directions. Inter-national Journal of Social Robotics, 01(01):319–330.

Davis, F., Bagozzi, R., and Warshaw, P. (1989). User acceptance of computer technology: A comparison of two theoretical models. Management Science, 35(08):982–1003.

Dignan, L. (2014). Softbank, aldebaran launch pepper, an emotional robot. https://www.zdnet.com/article/ softbank-aldebaran-launch-pepper-an-emotional-robot/.

Epstein, R., Fiscella, K., Lesser, C., and Stange, K. (2010). Why the nation needs a policy push on patient-centered health care. HEALTH AFFAIRS, 29(08).

Flandorfer, P. (2012). Population ageing and socially assistive robots for elderly persons: The importance of sociodemographic factors for user acceptance. International Journal of Population Research, 2012.

Forlizzi, J. (2005). Robotic products to assist the aging population. Interactions, 12(02):16–18.

Heerink, M., Krose, B., and Evers, V. a. W. B. (2008). The influence of social presence on acceptance of a companion robot by older people. JOURNAL OF PHYSICAL AGENTS, 02(02).

Heerink, M., Krose, B., and Evers, V. a. W. B. (2009). Measuring acceptance of an assistive social robot: a suggested toolkit. The 18th IEEE International Symposium on Robot and Human Interactive Communication.

Heerink, M., Krose, B., and Evers, V. a. W. B. (2010). Assessing acceptance of assistive social agent technology by older adults: the almere model. Inter-national journal of social robotics, 02(04):361–375.

Hockstein, N., Nolan, J., O’Malley, B., and Woo, J. (2005). Robotic microlaryn-geal surgery: A technical feasibility study using the davinci surgicalrobot and an airway mannequin. The Laryngoscope, (115):780–785.

Kaplan, F. (2004). Who is afraid of the humanoid? investigating cultural differences in the acceptance of robots. International Journal of Humanoid Robotics, 01(03):465.

Kuo, I., Rabindran, J., Broadbent, E., Lee, Y., Kerse, N., Stafford, R., and Mac-Donald, B. (2009). Age and gender factors in user acceptance of healthcare robots. The 18th IEEE International Symposium onRobot and Human In-teractive Communication.

(23)

Mukai, T., Hirano, S., Nakashima, H., Kato, Y., Sakaida, Y., Guo, S., and Hosoe, S. (2010). Development of a nursing-care assistant robot riba that can lift a human in its arms. The 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems.

Sabanovic, S., Bennett, C., Chang, W., and Huber, L. (2013). Paro robot affects diverse interaction modalities in group sensory therapy for older adults with dementia. 2013 IEEE International Conference on Rehabilitation Robotics, (115):780–785.

Schubring, S., Lorscheid, I., Meyer, M., and Ringle, C. (2016). The pls agent: Predictive modeling with pls-sem and agent-based simulation. Journal of Business Research, 69(10):4604–4612.

United Nations, T. (2017). World population prospects 2017 key find-ings. https://esa.un.org/unpd/wpp/Publications/Files/WPP2017_ KeyFindings.pdf.

Venkatesh, V., Morris, M., Davis, G., and Davis, F. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(03):425– 478.

Venkatesh, V., Thong, J., and Xu, X. (2012). Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly, 36(01):157–178.

(24)

8 Appendix

8.1 Appendix I: Factor, Source and Definition

Factor Name, Original Source and used Definition Factor Original Source Definition

Trust Heerink et al. The belief that the system performs with personal integrity and reliability. Perceived Sociability Heerink et al. The perceived ability of the system to

perform sociable behaviour.

Social Presence Heerink et al. The experience of sensing a social entity when interacting with the system. Anxiety Heerink et al. Evoking anxious or emotional reactions

when using the system.

Attitude Davis et al. Positive or negative feeling about the appliance of the technology.

Performance Expectancy Davis et al. The degree to which a person believes that using the system would enhance his or her daily activities.

Effort Expectancy Davis et al. The degree to which the user believes that using the system would be free of effort.

Hedonic Motivation Venkatesh et al. Feelings of joy or pleasure associated by the user with the use of the system. Social Influence Venkatesh et al. The user’s perception of how people

who are important to him/her think about him/her using the system. Facilitating Conditions Venkatesh et al. Objective factors in the environment

that facilitate using the system. Habit Venkatesh et al. The extent to which people tend to

per-form behaviors automatically because of using the system.

Personal Needs Gubler The extent to which people find patient-centered care important. Adaptability Heerink et al. The perceived ability of the system to

be adaptive to the changing needs of the user.

(25)

8.2 Appendix II: Removed Variables

Factor Name, Removed Variable Name and Cronbach’s Alpha below 0.7

Factor Variable Cronbach’s

Alpha (be-low 0.7) Trust I would trust a real life person more than the robot. 0.591 Trust I would only trust the robot if an expert would tell

me I could.

0.200

Perceived Sociability I think the robot would understand me. 0.697 Social Presence It really looks like the robot makes eye contact. 0.577 Social Presence It’s obvious for me the robot is not like a real person. 0.322 Anxiety If I should use the robot I would be afraid to make

mistakes with it.

0.634

Anxiety I would be afraid the robot would hurt me 0.677 Attitude I think robots could support humans. 0.572 Attitude The robot would make my life more interesting. 0.606 Performance Expectancy The robot wouldn’t really be able to help me. 0.658 Effort Expectancy I would be able to use the robot when I would have

a good manual.

-0.457

Effort Expectancy I would only use the robot when there’s someone around to help me with it.

0.472

Hedonic Motivation I would enjoy the robot talking to me. 0.673 Social Influence I would lose contact with people using the robot if I

wouldn’t use it.

0.025

Social Influence Me using the robot would impress other people. 0.553 Facilitating Conditions I think I could get help from others when using the

robot.

0.253

Facilitating Conditions Using the robot would be comparable to other tech-nology I use.

0.551

Habit I could get addicted to using the robot. 0.594 Habit If I had such a robot I would use it regularly. 0.489 Personal Needs I don’t think specific personal needs would matter

for the robot to function.

0.229

Personal Needs I think my needs would be different than the needs of other persons.

-0.238

Personal Needs I would only use the robot if it would be able to help me with my personal needs.

0.220

Personal Needs Personal needs would have priority for me over gen-eral needs for the robot to cater for.

0.519

Adaptability I think the robot would just do what I need at a particular moment.

0.626

(26)

8.3 Appendix III: Descriptive Statistics on Participants

Descriptive Statistics

N Minimum Maximum Mean Mean Std. error Std. Devia-tion Age 112 12 62 25.91 1.312 18.886 Computer Experience 112 1 5 3.54 0.090 0.948 Valid N (list-wise) 112

(27)

8.4 Appendix IV: Statements per factor used in

question-naire

Trust

1. I would trust the robot if it gave me advice.

2. I would follow the advice the robot gives me.

3. I would trust a real-life person more than the robot.

4. I would only trust the robot if an expert would tell me I could.

5. I would trust the robot with my life.

Perceived Sociability

1. I would consider the robot a pleasant conversation partner.

2. I would find the robot pleasant to interact with.

3. I think the robot would understand me.

4. I think the robot is nice.

5. I think I could become friends with the robot.

Social Presence

1. I have the feeling the robot interacts like a real person.

2. It really looks like the robot makes eye contact with the person.

3. can imagine the robot to be a living creature.

4. I realise the robot is not a real person.

5. The robot seems to have real feelings.

Anxiety

1. If I should use the robot, I would be afraid to make mistakes with it.

2. If I should use the robot, I would be afraid to damage something.

3. I find the robot scary.

4. I find the robot intimidating .

5. I would be afraid the robot would hurt me.

Attitude

(28)

2. The robot would make my life more interesting.

3. It’s good to make use of robots.

4. I think more people should use robots.

5. I think robots could support humans.

Performance Expectancy

1. The robot could be useful to me.

2. It would be convenient for me to have the robot.

3. The robot could help me with many things.

4. The robot could perform the tasks I have for him/her.

5. The robot wouldn’t really be able to help me.

Effort Expectancy

1. I would quickly know how to use the robot.

2. The robot would be easy to use.

3. I would be able to use the robot without any help.

4. I would only be able to use the robot when there’s someone around to help me.

5. I would be able to use the robot when I had a good manual.

Hedonic Motivation

1. I would enjoy the robot talking to me.

2. I would enjoy doing things with the robot.

3. I would find the robot enjoyable.

4. I find the robot fascinating.

5. I find the robot boring.

Social Influence

1. Me using the robot would impress other people.

2. The people around me would like it if I used the robot.

3. People that I don’t know would like me when using the robot.

4. If the people around me would use the robot, I would like to use it as well.

(29)

Facilitating Conditions

1. I have the knowledge necessary to use the robot.

2. Using the robot would be comparable to other technology I use.

3. I think I could get help from others when using the robot.

4. I would easily find out how to use the robot.

Habit

1. When using this robot, it could become a habit for me.

2. I could get addicted to using the robot.

3. Using the robot could become natural to me.

4. If I had such a robot I would use it every day.

5. If I had such a robot I would not use it regularly.

Adaptability

1. I think the robot would be able to adapt to my needs.

2. I think the robot would just do what I need at a particular moment.

3. I think the robot would help me when I would consider it necessary.

4. I think the robot would know when I don’t need any help.

5. When my needs change I think the robot will be able to recognise this.

Personal needs

1. I would only use the robot when it’s able to help me with my personal needs.

2. I think my needs would be different than the needs of other persons.

3. I think my special needs would make it more difficult for the robot to assist me.

4. I don’t think specific personal needs would matter for the robots function-ing.

5. Personal needs would have priority for me over general needs the robot is able to cater for.

Intention to Use

1. I would intent to use the robot in the future.

(30)

3. I would be using the robot frequently if I could.

4. I am not intending to ever use a robot in the future.

5. I have no interest in using a robot whatsoever.

Use

(31)

Acceptance of Healthcare Robots: the Integrated Model