//Logo Image
「世大智科/天才家居」-我們創業囉
PDF Version
AuthorsJun-Ming Lu and Yeh-Liang Hsu (2015-09-01)Recommend: Yeh-Liang Hsu (2015-10-21).
Note: This paper is published on Journal of Industrial and Production Engineering, vol. 32, no.4, pp.449-456, September, 2015.

The use of psychophysiological methods in the evaluation of mental commitment robots for elderly care

Abstract

This study aims to conduct objective user evaluation of mental commitment robots for elderly care using psychophysiological methods, including heart rate, skin temperature, skin conductance, and eye gaze data. In stage 1, four males and five females aged from 57 to 67 years old were recruited to interact with eight robot prototypes. The purpose was to validate whether the measures are capable of differentiating the semi-elderly and elderly users’ feelings about the robots, as well as capturing a rough preference of the robot’s anthropomorphism, size, weight, voice pitch, and covering material. In stage 2, two teleoperated robots and two autonomous robots were adopted in real human–robot interaction (HRI) with six males and 10 females aged from 52 to 80 years old. The results reflect the semi-elderly and elderly users’ preference in real HRIs, which may facilitate the developers to respond to the market more closely and effectively.

Keywords: human–robot interaction (HRI), eye tracking, heart rate, skin temperature, skin conductance

1.     Introduction

In the early stages of robot development for elderly care, the emphasis was mostly on the maintenance or improvement of physical health in terms of more independent livings and higher quality of care. However, the older adults themselves may expect the enhancement of mental health as well, in which the needs were usually underestimated. For example, the robot playmate Primo Puel became quite popular among females in their 50s or above, which was not expected before being released [7]. For these users who usually stayed at home all alone in the daytime, this robot perfectly served as their companion and brought back the feeling of “being needed.” Besides, based on the concept of animal-assisted therapy, the therapeutic robot, Paro, clinically satisfied the needs for mental health care of the demented elderly [17]. Following the success, it is now widely adopted by more institutions for dementia care in Japan, United States, and some European countries.

In addition to the autonomous robots mentioned earlier, teleoperated robots also came to the market to serve as the agent of remote controllers. For instance, enabled by telepresence technologies, Telenoid developed by Ogawa et al. [12] allowed remote controllers to interact with the local elderly users through physical contacts and vocal communication. TRiCmini is another example that integrated facial expression, expressive body movements, and video conferencing to provide the elderly and their family members with an alternative of “staying with each other remotely,” so that mental commitment can be realized [5].

In response to these potential needs, some researchers had been devoting to the understanding of elderly users’ views about social robots. However, they mostly focused on the elderly users’ expectation toward the robot’s functions or capabilities [3, 4, 14]. Considering user acceptance and the willingness to use, some other human factors need to be addressed as well. For example, Broadbent et al. [2] reported that the robot’s function influenced the user’s expectation toward its physical appearance. Anthropomorphism is usually one of the key criteria characterizing a robot’s appearance. As indicated by Riek et al. [15], a higher level of anthropomorphism encouraged the user to engage more in the human–robot interaction (HRI). In other words, users were more likely to treat a more anthropomorphic robot as real humans, and hence resulted in a higher level of acceptance and willingness to use. However, according to Mori [10], if a human-like robot performs unnatural body movements that are not consistent with its anthropomorphism in the appearance, it will make the user fall into the “uncanny valley,” which leads to the dramatically reduced satisfaction.

Regarding the issue of human likeness, Austermann et al. [1] invited 16 engineers ranging from 22 to 52 years old to interact with a humanoid robot (ASIMO) and a zoomorphic robot (AIBO). They found that the users tended to make more complicated verbal (such as talking to it) and less frequent nonverbal (such as touching it) communication with the humanoid robot than the zoomorphic robot. A possible reason was that the users treated the humanoid robots as real humans, so that they had a stronger motivation to have a conversation but was reluctant to make body contacts with such a stranger. In addition, as the user became confident with the robot’s capability to react in real time, an increasing speed in the speech was observed. However, the changes in the tone, pitch, and volume of speech were not investigated.

Haring et al. [6] further compared the users’ impression about a highly anthropomorphic robot, a moderately anthropomorphic robot, and a zoomorphic robot. Sixty-eight males and 33 females (21.1 years old in average) were recruited to join the questionnaire survey. The results demonstrated that the anthropomorphism of a robot affects not only user acceptance but also the expectation of its body movements and functions. In the case of humanoid robots, a higher level of anthropomorphism seemed to make the users expect the robot to perform high-level anthropomorphic body movements as well. However, it gave the users the impression of being less clever than a less human-like humanoid robot. In addition, similar as indicated in [1], the users tended to touch the zoomorphic robots more frequently than the humanoid robots.

From another point of view, Li et al. [8] considered the effect of cultural differences on user acceptance of robots. One hundred and eight participants from China, Korea, and Germany (23.7 years old in average) participated in the questionnaire survey. The findings showed that highly anthropomorphic robots generally attracted the users more, but there was a cultural difference. Chinese and Korean participants were used to be influenced by others more, so they engaged in the HRI more. On the other hand, German participants preferred the larger-sized robots that moved faster. As for the overall preference, there was no significant difference between humanoid and zoomorphic robots, and no cultural difference was found. Another study [13] on 40 participants (23.8 years old in average) declared that a robot’s appearance affected the user’s expectation toward its capability. More specifically, compared against the appearance, the anthropomorphism of a robot’s body movements was in fact more important than its appearance. In other words, it would be recommended to give a higher priority to the development of human-like movements rather than its appearance.

In addition to visual and haptic concerns, the auditory features of a robot also play an important role in user acceptance and preference. For example, Niculescu et al. [11] made a questionnaire survey with 28 young adults regarding their preference of a robot’s voice pitch. They reported that robots with high-pitched voice were accepted more. Besides, Siegel et al. [16] presented a robot with neutral appearance (neither masculine nor feminine) to 76 male and 58 female participants (35.6 years old in average). In one condition, the robot was with a male voice (low-pitched), while others presented a female voice (high-pitched). According to their findings, male participants generally preferred the one with a female voice, no matter whether they came alone or with others. As for females, they also preferred the female voice when coming with others. However, in the case of coming alone, their preference changed to the robot with a male voice. Similar as reported in [16], users tended to have different preference of the voice pitch of a robot.

As been seen so far, most of the studies were conducted with young adults and hence limited the applications in the mental commitment robots for elderly care. One of the very few studies was conducted by Wu et al. They invited 15 participants ranging from 66 to 89years old to conduct a questionnaire survey regarding the preference of different types of robots [18]. The more popular robots were found to be common in the characteristics of “small,” “cute,” and “consistency between the robot’s appearance and functions.” Following the findings, the elderly users seemed to exhibit similar expectation toward the robot’s appearance as young users did, and they had a clear preference of small-sized or light-weight robots.

In summary, the anthropomorphism of a robot plays a key role of the user’s first impression, and it affects the willingness of physical contacts with the robot as well. Besides, the preference of a robot’s size and weight may depend on the user’s cultural background, age, experience, and some other factors. A study focusing on the target users will hence be necessary. Further, the robot’s voice pitch is another interesting topic. Although few studies showed that high-pitched voice is preferred, the individual differences and user scenario should be considered as well. Moreover, in conventional studies of users’ perception of robots, subjective responses were often adopted. Nevertheless, sometimes the participant may give untruthful answers or fail to accurately describe the feelings.

Therefore, this study aims to conduct objective user evaluation of mental commitment robots for elderly care through the use of psychophysiological methods, including heart rate, skin temperature, skin conductance, and eye gaze data. Aiming at the continuing development within the next 15years, not only the elderly (65years old and more) but also the semi-elderly (50–64years old) were considered. The semi-elderly and elderly users’ preference of the robot’s appearance and functions will be studied to achieve a better understanding of the relationship between these design elements and the performance of mental commitment.

2.     Stage 1: simulated HRI

Prior to the evaluation of mental commitment robots using psychophysiological methods, it is necessary to validate whether these measures are capable of identifying the emotional changes of semi-elderly and elderly users while interacting with robots. Thus, in the first stage, some robot prototypes were adopted to conduct a simulated HRI.

2.1      Experimental design and methods

In order to observe how psychophysiological measures reflect the semi-elderly and elderly users’ feeling about mental commitment robots with different visual, auditory, and haptic features, a variety of robot prototypes need to be prepared. As shown in Table1, there were two levels for each of the five features including the robot’s anthropomorphism (humanoid or non-humanoid), size (small or medium), weight (light or medium), voice pitch (low-pitched or high-pitched), and covering material (soft or hard). Theoretically, this should result in a total number of 32 prototypes. Considering time and cost efficiency, a 25−2 factorial design was employed to determine the main factor effects, where only eight prototypes are required.

Table 1. The characteristics of the eight robot prototypes adopted in stage 1

In this study, the ProComp Infiniti biofeedback system was used to measure the participant’s heart rate, skin conductance, and skin temperature. Assuming that a mental commitment robot will well demonstrate its role as it makes the user feel more relaxed or less anxious, it will come along with the decreased heart rate, skin conductance, and skin temperature. In case that the individual differences may lead to the misinterpretation of psychophysiological measures, normalization was performed by dividing the original measurement by that obtained under the baseline condition (when each participant stayed calm and relaxed).

Moreover, the Viewpoint eye-tracking system was employed to obtain the participant’s eye gaze data. If the user prefers a specific type of robot, he/she is likely to spend more time looking at it during the interaction. To interpret such behaviors, the percentage of time spent looking at the robot’s head, trunk, or limbs were analyzed, respectively. Besides, combined with a digital camera, the percentage of time that the participant touched the robots was analyzed as well. As Figure 1 illustrates, the participant kept wearing the biofeedback sensors and the eye-tracking system to facilitate data collection during the HRI sessions.

In addition to the objective measures, it is also important to know how the participant perceives his/her own emotions. Thus, subjective responses were collected using the six-item short-form State-Trait Anxiety Inventory-State (STAI-S) [9], as presented in Table 2. The total STAI-S score ranges from 6 to 24, where a higher score implies a higher level of anxiety. In other words, if a specific condition produces a lower STAI-S score, it generally makes the better effect of mental commitment. Further, the results were compared against the psychophysiological measures and eye gaze data to determine where there’s a consistency.

http://www.tandfonline.com/na101/home/literatum/publisher/tandf/journals/content/tjci21/2015/tjci21.v032.i07/21681015.2015.1078420/20150923/images/medium/tjci_a_1078420_f0001_oc.gif

Figure 1. A participant wearing the sensors (left) and interacting with a robot prototype (right)

Table 2. The six-item short-form STAI-S

2.2      Participants and procedures

Nine older adults aged from 57 to 67 years old (with the average of 61.8 years old) were recruited from a few neighboring communities in Taiwan, including four males (62.0 years old in average) and five females (61.5 years old in average).

As the participant arrived, he/she was given the instructions of the experiment and then signed the inform consent. After wearing the sensors and the eye-tracking system with the staff’s assistance, the participant was asked to sit with ease under the baseline condition, in which a two-minute relaxing music was played. The psychophysiological measures were recorded for reference in the meantime. Subsequently, one of the eight robot prototypes was presented (in a random order) to the participant for a two-minute interaction. The participant could freely decide what to do with the robot prototype. As he/she touched it, a recorded voice will be played through the wireless speaker installed inside the robot. If there were no actions from the participant’s side, a “hello” word was soon played through the wireless speaker by the staff to encourage the participant to start the interaction more actively. As an HRI session ended, the participant took a rest for one minute. The staff then confirmed whether the measures returned to the level as collected in the baseline condition. If the measures indicated that the participant was still in the state of excitement or anxiety, the resting time was then extended for one more minute. Only if the requirement was met, the participant was then directed to the next HRI session. Eventually, the experiment was finished as all the eight HRI sessions were completed, which took around 60 min for a participant.

2.3      Results and discussion

Following the significant differences between levels determined by ANOVA (Table 3), a robot prototype in general performed better (directing the participant to a more relaxed state) as it was in the non-humanoid form and with small size, light weight, and a low-pitched voice. One possible attribution is that the user can easily hold such a robot in the arms or on the laps, which is similar to the experience of staying with a pet. Following the theory of “uncanny valley,” humanoid robots might be more acceptable as their expressions and movements can perform as naturally as real humans. As for the voice pitch, there seems to be no specific reason why low-pitched ones were preferred.

The only inconsistency of factor effect among the STAI-S score and the psychophysiological measures was found in the robot’s covering material. The skin conductance suggested that the soft covering material made the participant feel more relaxed, whereas the skin temperature implied that the better performance can be found with hard ones.

In addition, the eye gaze data revealed that the participant tended to spend a significantly longer time looking at the humanoid robots’ limbs than the non-humanoid ones, whereas no differences were found in the percentage of gaze on the robot’s head or trunk. On the other hand, according to the percentage of time that the participant spent touching the robot, a more frequent contact was found during the interaction with non-humanoid robots. The possible reason is that the participant tended to treat the humanoid ones as real humans, in which body contact is usually avoided against strangers. As for the non-humanoid robots, they were treated just like the participant’s pets. Hence, it was more natural to exhibit more body contact during the HRI.

In summary, it seems that the semi-elderly and elderly users’ preference of mental commitment robots in terms of visual, auditory, and haptic factors can be assessed through the use of psychophysiological measures. Thus, a more detailed investigation can be followed up to get a clearer view.

Table 3. Summary of factor main effects on psychophysiological measures and eye gaze data in stage 1

3.     Stage 2: real HRI

In stage 1, the pilot investigation was conducted with robot prototypes to exclude the possible effects of their functions. Definitely this needs to be considered in the real HRI. Thus, in stage 2, a few robots with complete functions were adopted to help further understand how semi-elderly and elderly users may react to the different designs.

3.1      Experimental design and methods

Four robots covering two ways of operations and different levels of anthropomorphism were selected to interact with the semi-elderly and elderly participants. As shown in Figure 2, they are TRiCmini+ (teleoperated and humanoid) and Wobot (teleoperated and less anthropomorphically humanoid) developed by GRC, Taiwan, and Paro (one in white and one in yellow; autonomous and non-humanoid) developed by AIST, Japan.

http://www.tandfonline.com/na101/home/literatum/publisher/tandf/journals/content/tjci21/2015/tjci21.v032.i07/21681015.2015.1078420/20150923/images/medium/tjci_a_1078420_f0002_oc.gif

Figure 2. The four robots considered in stage 2 (left to right: TRiCmini+, Wobot, white Paro, and yellow Paro)

While interacting with a teleoperated robot (TRiCmini+ or Wobot), the staff remotely controlled the robot’s facial expressions and body movements, as well as initiating a video chat with the participant. As for the interaction with an autonomous robot (white Paro or yellow Paro), the robot just reacted to the participant’s actions based on the pre-programmed functions. For each of the four robots, the participant could freely decide whether to make physical interaction (such as hugging or touching it) or not.

Moreover, the interpersonal communication with the staff (who made remote control of the teleoperated robots) was also conducted. During a two-minute session, the participant was asked to talk about a happy topic or an unhappy topic (in a randomized order) that he/she had experienced within the past one month. The purpose was to make comparisons against the HRI in terms of the users’ behaviors.

3.2      Participants and procedures

Sixteen older adults aged from 52 to 80 years old (63.1 years old in average) were recruited from a few neighboring communities in Taiwan. There were six males (66.3 years old in average) and 12 females (61.1 years old in average). Six of the 16 participants had never seen a robot in front of them, and the other 10 were with such experiences for at least one time.

After registration and signing the informed consent, the participant started the preparation session. Here, a set of four video clips were presented to the participant, including ASIMO (autonomous robot developed by Honda Motor Co., Ltd.), KABO-chan (autonomous robot developed by PIP Co., Ltd.), Actroid-F (teleoperated robot developed by Osaka University and manufactured by Kokoro Co., Ltd.), and Telenoid (teleoperated robot developed by Osaka University and ATR). The purpose was to make a better balance of the familiarity with robots, so as to reduce the effect of the individual differences.

Subsequently, the staff helped the participant wear the sensors as well as the eye-tracking system, and then directed him/her to move on to the interpersonal and HRI sessions. In the first session, the participant began with the baseline condition for two minutes, which is the same as been described in Section 2.2. Next, the anxious condition was created by asking the participant to play cognitive games using a tablet PC for two minutes. After that, six sessions (two interpersonal and four HRI) were performed in a randomized order. Between any two sessions, a one-minute rest was given to let the participant return to the baseline level. As each session ended, the six-item short-form STAI-S was presented to the participant for filling it out. The experiment finished as all the eight sessions were completed, and it took around 90 min for one participant.

3.3      Results and discussion

The results of one-way ANOVA demonstrated that the condition (anxious condition and four HRIs) had a significant effect on the normalized STAI-S score, heart rate, and skin conductance. Further, as shown in Table 4, by performing Duncan’s multiple range test, a set of groups with significantly differences were identified. For example, considering the normalized STAI-S score, the anxious condition led to the highest level of anxiety, followed by the HRIs with Wobot and white Paro, finally yellow Paro and TRiCmini+. In other words, interacting with TRiCmini+ and yellow Paro resulted in a more relaxed state than the other two robots, as well as the anxious state. In the case of normalized heart rate, TRiCmini+ and Wobot had the better performance of mental commitment than the other two robots, and the level of relaxation was beyond that of the anxious condition. As for the normalized skin conductance, Wobot outperformed the other three robots and helped the participant get out of the anxious state.

Table 4. Comparison of psychophysiological measures among different conditions in stage 2

Generally speaking, according to the participants’ subjective responses, interacting with the robots did help calm them from being anxious. However, there’s no sufficient evidence whether humanoid or non-humanoid robots generate the stronger effect. Unlike the interactions with robot prototypes, the users might engage more in the HRI due to the more comprehensive functions. In such cases, they probably felt as if interacting with a doll or a pet. Further, the humanoid robots considered in this stage are less human-like, which might help reduce the negative impact of “uncanny valley.” To be noted, it is surprised that the performance of the white Paro differed from that of the yellow Paro. This implies that the color of a robot may influence its effect of mental commitment.

Moreover, the eye gaze data were also analyzed to evaluate the effects of mental commitment delivered by robots. Here, two kinds of dialogists were discussed. One is the robot, and the other is the real human who had a conversion with the participant on a happy or unhappy topic. First, the percentage of time focusing on the dialogist’s head was taken into account. The assumption is that a higher percentage indicates more interest associated with the interaction. As shown in Table 5, the humanoid robots tended to attract the participant to focus more on the head during the HRI, which may be due to the attractiveness of the changing facial expressions. Further, in such cases, the participant became even more immersed than in interpersonal communication. This should be an interesting topic worth further investigation. On the other hand, the percentage of time focusing on the dialogist’s body was compared as well, but there was no significant difference found. Finally, according to the percentage of gaze out of the dialogist, the results are quite consistent with that of gaze on the dialogist’s body. In other words, as the participant spent more time looking at the dialogist’s body, the less time the gaze was found out of the dialogist.

Table 5. Comparison of eye gaze data among different conditions in stage 2

4.     Conclusions

This study demonstrated the feasibility of using psychophysiological measures to evaluate mental commitment robots for elderly care. Results showed that the heart rate, skin conductance, and eye gaze data seem to be feasible solutions for assessing HRI. However, due to the low consistency among psychophysiological measures or between objective and subjective measures, further studies will be needed to obtain a better understanding. In addition, the semi-elderly and elderly users’ preferences were determined. Considering the robot’s appearance in terms of visual, auditory, and haptic features, the non-humanoid robots with small size, light weight, and a low-pitched voice tend to exhibit a better performance of relaxation. As the functions are included, there is no evidence whether humanoid or non-humanoid robots are preferred. However, if physical contact is required, non-humanoid robots would provide a stronger motivation. Besides, user preferences were found to be similar no matter whether the robot is autonomous or tele-operated. Thus, developers may choose either one to meet the technology limitations or other concerns. Moreover, the semi-elderly and elderly users usually make more frequent eye contacts with mental commitment robots than real humans, but it is not always consistent with a favorable impression. Further investigation is encouraged to understand what lies behind, which may contribute to a more engaging HRI.

Notes on contributors

Jun-Ming Lu currently serves as a contract assistant professor in the Department of Industrial Engineering and Engineering Management at National Tsing Hua University, Taiwan. He received his PhD degree in Industrial Engineering from National Tsing Hua University, Taiwan, in 2009. His research interests include Ergonomics, Digital Human Modeling, Man-Machine System, and Gerontechnology.

Yeh-Liang Hsu currently serves as a professor in the Department of Mechanical Engineering and the director of Gerontechnology Research Center at Yuan Ze University, Taiwan. He received his PhD degree in Mechanical Engineering from Stanford University, United States, in 1992. His present research interests include Mechanical Design and Optimization, Gerontechnology, and Home Telehealth System.

Acknowledgments

The authors would like to express their gratitude to Miss Hsiao-Man Liu, Mr Hui-Bang Yu, Mr Chuan-Hua Chen, Mr Chih-Yin Tai, and Miss Jin-Ni Lee for the assistance in data collection.

Notes

An asterisk (*) indicates that the p-value is smaller than 0.05, where a significant difference was found between the two levels of the factor. The term inside a parenthesis denotes the level in which a better effect of relaxation was produced. In the column of Duncan’s MRT (multiple range test), the means in the same group are separated by commas, while the significantly different groups are separated by the “>” mark. For normalized STAI-S score, hear rate, skin conductance, and skin temperature, a smaller value indicates a more relaxed state or a lower level of anxiety.

In the column of Duncan’s MRT (multiple range test), the means in the same group are separated by commas, while the significantly different groups are separated by the “>” mark.

References

1.          Austermann, A., S. Yamada, K. Funakoshi and M. Nakano, “Similarities and differences in users' interaction with a humanoid and a pet robot,” Proceedings of Human Robot Interaction, Osaka, 73–74 (2010).

2.          Broadbent, E., Y. I. Lee, R. Q. Stafford, I. H. Kuo and B. A. MacDonald, “Mental schemas of robots as more human-like are associated with higher blood pressure and negative emotions in a human-robot interaction,” International Journal of Social Robotics, 3, 291–297 (2011).10.1007/s12369-011-0096-9 [CrossRef]

3.          Broadbent, E., R. Stafford and B. MacDonald, “Acceptance of healthcare robots for the older population: Review and future directions,” International Journal of Social Robotics, 1, 319–330 (2009).10.1007/s12369-009-0030-6 [CrossRef]

4.          Broekens, J., M. Heerink and H. Rosendal, “Assistive social robots in elderly care: A review,” Gerontechnology, 8, 94–103 (2009). [CrossRef]

5.          Chen, Y.-S., Y.-L. Hsu and J.-M. Lu, “TRiCmini – telepresence robot for interpersonal communication for older adults,” Journal of the Chinese Society of Mechanical Engineers, 34, 121–127 (2013).10.3901/JME.2013.02.121 [CrossRef], [Web of Science ®]

6.          Haring, K. S., K. Watanabe and C. Mougenot, “The influence of robot appearance on assessment,” Proceedings of HRI 2013, 131–132 (2013).

7.          Japan’s hi-tech carers [Internet]. BBC News; c2015 [cited Jun 7, 2015]. Available online at:http://news.bbc.co.uk/2/hi/programmes/this_world/golden_years/4436633.stm

8.          Li, D., P. L. Rau and Y. Li, “A cross-cultural study: Effect of robot appearance and task,” International Journal of Social Robotics, 2, 175–186 (2010).10.1007/s12369-010-0056-9 [CrossRef]

9.          Marteau, T. M. and H. Bekker, “The development of a six-item short-form of the state scale of the Spielberger State-Trait Anxiety Inventory (STAI),” British Journal of Clinical Psychology, 31, 301–306 (1992).10.1111/bjc.1992.31.issue-3 [CrossRef],[PubMed], [Web of Science ®]

10.      Mori, M., “Bukimi no tani (The uncanny valley),” Energy, 7, 33–35 (1970), in Japanese.

11.      Niculescu, A., B. van Dijk, A. Nijholt and S. S. Lan, “The influence of voice pitch on the evaluation of a social robot receptionist,” Proceedings of the 2011 International Conference on User Science and Engineering (i-USEr), 18–23 (2011).10.1109/iUSEr.2011.6150529

12.      Ogawa, K., S. Nishio, K. Koda, K. Taura, T. Minato, C. T. Ishii and H. Ishiguro, “Telenoid: Tele-presence android for communication,” Proceedings of ACM SIGGRAPH 2011 Emerging Technologies, 15 (2011).

13.      Park, E., H. Kong, H.-T. Lim, J. Lee, S. You and A. P. Del Pobil, “The effect of robot's behavior vs. Appearance on communication with humans,” Proceedings of the 6th ACM/IEEE International Conference on Human-Robot Interaction, 219–220 (2011).

14.      Pigini, L., D. Facal, L. Blasi and R. Andrich, “Service robots in elderly care at home: Users' needs and perceptions as a basis for concept development,” Technology and Disability, 24, 303–311 (2012).

15.      Riek, L. D., T.-C. Rabinowitch, B. Chakrabarti and P. Robinson, “How anthropomorphism affects empathy toward robots,” Proceedings of ACM/IEEE International Conference on Human-Robot Interaction, 223–224 (2009).

16.      Siegel, M., C. Breazeal and M. I.NORTONhttp://cdncache-a.akamaihd.net/items/it/img/arrow-10x10.png, “Persuasive robotics – The influence of robot gender on human behavior,” Proceedings of the 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2563–2568 (2009).10.1109/IROS.2009.5354116

17.      Wada, K., T. Shibata, T. Saito, K. Sakamoto and K. Tanie, “Psychological and social effects of one year robot assisted activity on elderly people at a health service facility for the aged,” Proceedings of the 2005 IEEE International Conference on Robotics and Automation (ICRA 2005), 2785–2790 (2005).10.1109/ROBOT.2005.1570535

18.      Wu, Y.-H., C. Fassert and A.-S. Rigaud, “Designing robots for the elderly: Appearance issue and beyond,” Archives of Gerontology and Geriatrics, 54, 121–126 (2012).10.1016/j.archger.2011.02.003 [CrossRef], [PubMed], [Web of Science ®]