MASTERTHESIS:
THE TOUCH OF A ROBOTIC FRIEND:
Can a touch of a robot, when the robot and the person have bonded with each other, calm a person down during a stressful moment?
By Vera van der Hout 16th August 2017
THE TOUCH OF A ROBOTIC FRIEND:
Can a touch of a robot, when the robot and the person have bonded with each other, calm a person down during a stressful moment?
Name: Vera van der Hout
Student number: s1543016
Education: Master Human Factors Engineering Psychology Submission date: 16th of August 2017
Organisation: TNO, Soesterberg Duration internship and
masterthesis: (EC): 5 months (30 ECS) 1st supervisor: J. M. C. Schraagen 2nd supervisor: J. B. F. van Erp External supervisor: C. J. A. M. Willemse
PREFACE
My name is Vera van der Hout and I am a master student of the master Human Factors Engineering Psychology. Before you is the masterthesis “The touch of a robotic friend: can a touch of a robot, when the robot and the person have bonded with each other, calm a person down during a stressful moment?”.
A study about the evaluation of a robot-initiated touch that was conducted by means of a 2x2 factorial design experiment, with having a bond or not and having been touched by the robot or not as independent variables. It has been written to fulfil the graduation requirements of the Human Factors Engineering Psychology master at the University of Twente. I was engaged in researching and writing this masterthesis from September 2016 to August 2017. I am now pleased to complete my masterthesis. The research was undertaken at TNO (an independent research organisation) and was part of a PhD project on “Social Touch” (mediated and robot-initiated) of Christian Willemse, who studies whether and if so, how a social touch can be mediated or simulated by technology in order to induce social responses similar to real human touch. I found writing this masterthesis to be both interesting and instructive. I have learned undertaking and processing a quantitative research.
My research question was formulated together with my external supervisor (Christian Willemse) and we have been working closely together. I would first like to thank my external supervisor, Christian Willemse at TNO for the wonderful collaboration. You supported me greatly and were always willing to help me. The research was not always easy, but Christian was always available and willing to help me out and answer my questions.
Also, I would like to thank my first internal supervisor (Jan Maarten Schraagen) for the meetings about my research. Helping me with the questions I had and giving always quick feedback on my research proved to be a valuable guidance to help me to successfully complete my masterthesis.
Finally, I would like to thank my parents, friends and boyfriend for believing in me and always be there for me during this project. It was not easy all the time and there were some personal matters that influenced the performance of my research, but all my supervisors, family, friends and boyfriend kept me motivated in these moments. Thank you all!
I hope you enjoy your reading.
Vera van der Hout
Zwolle, August 2017
ABSTRACT
An important form of social behaviour in humans, is touch. A touch has many positive effects on a human’s wellbeing and is also important for the daily social interactions between people. An important effect is that a touch may calm people down during stressful moments, and it has beneficial effects for people who are not touched. A touch is very complex and there are several aspects that influence the experience of a touch, one of them is the relation between people. A touch could be experienced as inappropriate when touched by someone you do not know and just appropriate when a person you know is touching you. Nowadays we live in a world where the use of (social) robots in different areas increases. Social robots are suitable for touch, because of their physical embodiment. This research focuses on the experience of a robot's touch, especially stress-reduction, after a bond is built between the person and the robot. An important question that needs to be answered is whether a touch of a robot can have the same effects as a touch between people. The main question is: can a touch of a robot, when the robot and the person have bonded with each other, calm a person down during a stressful moment.
Through a 2 x 2 between subjects factorial design I investigated whether the touch of a robot had a calming effect on a person. The two independent variables were having a bond or not, and to be touched or not. Dependent variables were stress reduction, robot-touch experience, robot perception and the Midas Touch effect. Throughout a thrilling movie, tension was created and the robot tried to soothe the participant verbally and touch, depending on the condition. In total, 67 participants were randomly assigned to one of the four conditions.
The results show that it seems possible to create a superficial bond between a robot and a person in a short period of time. But this study found no scientific evidence for any of the examined hypotheses.
Also, an alternative bonding variable gave no interaction effect of bond and touch. Another interesting finding was that participants in the bond condition (who were touched by the robot) experienced the touch of the robot less pleasant than the participants in the no bond condition. But with the alternative bonding variable this is the other way around: participants in the bond condition experienced the robot touch more pleaseant than the no bond condition.
Potential limitations of this study are small effect sizes for interaction effects of the touch and bond, there were small differences in the data. Also, it is important to take into account that the effect of a robot touch could be different instead of a human touch. Future research should focus on how a robot touch could contribute in a positive way for a human being.
SAMENVATTING
Een belangrijke vorm van sociaal gedrag bij mensen is aanraking. Een aanraking heeft vele positieve effecten op het welzijn van de mens en is belangrijk bij de dagelijkse sociale interactie tussen mensen.
Een belangrijk effect is het kalmeren tijdens stressvolle momenten en het heeft voordelige effecten op mensen die minder vaak worden aangeraakt. Een aanraking is erg complex en er zijn verschillende aspecten die de ervaring van een aanraking kunnen beïnvloeden. Een belangrijk aspect is de relatie die je hebt met de ander. Een aanraking kan als ongepast worden ervaren als je wordt aangeraakt door iemand die je niet kent, en juist gepast als je door iemand wordt aangeraakt die je kent. Vandaag de dag leven we in een wereld waar het gebruik van (sociale) robots in verschillende gebieden toeneemt. Dit onderzoek richt zich op de ervaring van de aanraking van een robot, in het bijzonder stressreductie, nadat er een band is opgebouwd tussen de persoon en de robot. Een belangrijke vraag die beantwoord moet worden is of een aanraking van een robot dezelfde effecten kan hebben als een aanraking tussen mensen. De hoofdvraag is: kan een aanraking van een robot, wanneer een persoon en de robot een band hebben gevormd, de persoon kalmeren tijdens een stressvol moment?
Deze vraag wordt beantwoord door middel van een 2 x 2 between subjects factorial design waarin werd onderzocht of de aanraking van een robot effect heeft op een persoon. De twee onafhankelijke variabelen waren het hebben van een band of niet en het worden aangeraakt of niet.
Afhankelijke variabelen waren stressreductie, ervaring van de robot aanraking, perceptie van de robot en de Midas Touch. Door middel van een spannende film werd er een spanning gecreëerd en de robot probeerde de participant verbaal en met een aanraking, afhankelijk van de conditie, te kalmeren. Het betrof een onderzoek met in totaal 67 personen die willekeurig waren ingedeeld in een van de vier condities.
De resultaten van deze studie wijzen erop dat het mogelijk is, om in een korte tijd, een kunstmatige band te creëren tussen een robot en een persoon. Maar deze studie vindt geen wetenschappelijk bewijs voor één van de onderzochte hypotheses. Eveneens met een alternatieve band variabele worden er geen interactie-effecten gevonden voor band en aanraking. Een andere interesante bevinding was dat participanten in de band conditie (die aangeraakt zijn door de robot) een aanraking als prettiger ervaren dan de participanten in de geen band conditie. Maar als de alternatieve band variabele wordt gebruikt geeft dit het tegenovergestelde effect: particianten in de band conditie ervaren de aanraking van de robot als aangenamer dan de participanten uit de geen band conditie.
Potentiële beperkingen van deze studie is de studie een kleine effectgrootte heeft voor de interactie-effecten van aanraking en band, er zijn kleine verschillen in de data. Eveneens is het van belang in acht te nemen dat een robot aanraking wellicht niet dezelde effecten heeft als een menselijke aanraking. Vervolgonderzoek zou zich moeten richten op hoe een robot aanraking positief kan bijdragen in de toekomst voor mensen.
TABLE OF CONTENTS page.
1. INTRODUCTION 8
1.1. Related work 9
1.1.1. Positive effects touch 9
1.1.2. Robot-initiated touch 10 1.1.3. Creating a bond with a robot 12
1.2. Hypotheses 14
1.3. Introducing research method 14
2. METHOD 16
2.1. Participants 17
2.2. Set up and apparatus 17
2.2.1. Set up 17
2.2.2. Robot 18
2.2.3. Interaction (part 1) 18
2.2.4. Film (part 2) 20
2.2.5. Physiology 20
2.3. Materials and measures 20
2.3.1. Bond with the robot 20
2.3.2. Arousal 21
2.3.3. Perception van de robot 22
2.3.4. Midas touch 24
2.3.5. Covariates 24
2.4. Procedure 24
3. RESULTS 26
3.1. Hypothesis I: Effects of touch on stress reduction 31 3.2. Hypothesis II: Effects of touch on evaluations of the robot 32 3.3. Hypothesis III: Effects of the Midas Touch 40
3.4 Alternative analyses 42
4. DISCUSSION 45
4.1. Conclusion 45
4.2. Theoretical implications 45
4.3. Limitations study 48
4.4. Further research 50
REFERENCES 51
APPENDIX 58
Appendix I: Set up experiment 58
Appendix II: Script interaction 59
Appendix III: Questionnaires
63
1. INTRODUCTION
Humans have different forms of social behaviour, behaviour that persons exhibit in relation to each other, for example talking and listening. An important form of social behaviour between humans is touch (Field, 2010). A touch has many positive effects on a human’s wellbeing and is also important for the daily social interactions between people. Among others, a touch can strengthen relationships between people (Field, 2010; Gallace & Spence, 2010) and a touch is the most widely used method to calm people down in stressful moments (Dolin & Booth-Butterfield, 1993). A touch is very complex and there are several aspects that influence the experience of a touch (Hertenstein, 2002). How is the touch feeling, hard or soft? What is the meaning of touch, wishing someone success or comforting someone? Who is touching you, are you touched by a stranger or your partner? A touch could be experienced as inappropriate when you are touched by someone you do not know and as appropriate when a person you know is touching you (Smith & MacLean, 2007).
Nowadays we live in a world where technology is becoming increasingly important and also the use of (social) robots in different areas increases. According Bartneck and Forlizzi (2004), a social robot is an autonomous or semi-autonomous robot that interacts and communicates with people by complying with the behavioural norms accepted by the person interacting with the robot. The advent of social robots will also lead to more social behaviour between humans and robots. Touch as well plays an important role in the development of robots who want to interact and communicate with people in a natural way (Salter, Dautenhahn & Broekhorst, 2006). Social robots are suitable for touch, because of their physical embodiment (Cramer, Kemper, Amin, Wielinga & Evers, 2009; Chen, King, Kemp & Thomaz, 2014;
Bevan & Fraser, 2015). There are social robots that can detect a person’s touch, robots that can produce a touch and also indicate a social touch (Chen et al., 2014). The more a robot interacts socially with a person, the more people are willing to accept the social robot in their lives (public places, work and home) (Pereira, Leite, Mascarenhas, Martinho & Paiva, 2010).
Much research has been done in the area of social touch between people, but less so in social touch by robots (Chen et al., 2014; van Erp & Toet, 2015). At this moment, touch is an underestimated topic in human-robot interaction research, but certainly not unimportant for the interaction between human and robot. There are situations where people are not touched anymore, when needed (for example, lonely elderly people), research shows that touch can work advantageously in these situations (Phelan, 2009; Gleeson & Timmins, 2005). So, because social touch is important for the interaction between people, it is interesting to investigate the possibilities and limitations for touch interaction between human and robot. The main question is whether a robot, in the future, can contribute to the wellbeing of a person in the above situations. Because a lot of research has focused on the technical side of robot touch and less on the psychological side, it's important to look at the psychological side of a social touch: how do people experience the touch of a robot. The psychological approach to human- robot touch research is still in the exploratory phase (Broadbent, 2017).
In order to gain insight, the relationship that people have with each other is important for an appropriate touch (Coan, Schaefer & Davidson, 2006), this could also be the case in human-robot interaction. This study will focus on a robot touch in combination with the relation between the human and the robot. Is it possible to create a bond between a person and a robot and does the relation between a robot and a person have an influence on the experience of a robot touch? With this study, I hope to contribute to the understanding of the possibilities and limitations for touch interaction between human and robot. In the future, maybe a robot touch can have great benefits in human-robot interaction and for the wellbeing of people.
The main purpose of this study is to investigate whether it is possible to create a bond between a person and a robot and how this bond affects the touch that the robot gives. To do so, I carried out an experiment, consisting of two parts. Part one, an interaction where an attempt was made to establish a bond between the person and the robot. Part two, an interaction where the effect of robot touch was investigated in the context of a person watching a thrilling movie with the robot calming down the person.
The remainder of this thesis is structured as follows. First, the related work section where I go deeper into the different constructs that have been used in the study, after that I formulate the hypotheses.
Second, I present the research method that I have used. Third, I report the results of the experiment.
And, finally I discuss the results of the experiment and give recommendations for further research.
1.1. Related work
1.1.1. Positive effects touch
As mentioned earlier, a touch has many positive effects on human well-being. A touch between people is called interpersonal touch (van Erp & Toet, 2013). Interpersonal touch plays an important role in our emotional well-being (Gallace & Spence, 2010). Montagu (1971) stated that a touch is much stronger than verbal or emotional contact, and it affects everything we do. Also, Field (2001) stated that touch is a very important sense. There are several positive effects of interpersonal touch between humans. First of all, touch is important in greeting, intimate communication, correcting and intimate relationships (Field, 2010). A touch can be used to generate or modulate emotions, but also to enhance a message or emotion (Cabibihan, Zheng & Cher, 2012; van Erp & Toet, 2013). Also, a touch can affect a certain social behaviour in humans, for example, modulating their tendency to meet others’ requests (called the
“Midas Touch effect”) (Crusco & Wetzel, 1984). In addition, touch plays a part in creating a bond between couples or groups and enhancing romantic relationships, possibly without the touch being really remembered (Gallace & Spence, 2010). Besides, a touch has a positive physiological effect on humans, as it can lower the stress level of a person (Heinrichs, Baumgartner, Kirschbaum, & Ehlert, 2003). A touch of someone can lower the blood pressure or heart rate of a person (Field, 2010). A touch is the most widely used method of reassuring people who experience stress or negative excitement (Dolin &
Booth-Butterfield, 1993). Research has shown that there are positive effects of touch in healthcare, such
as therapy, the nursing and the well-being of the person (Caris-Verhallen, Kerkstra & Bensing, 1999;
Phelan, 2009; Gleeson & Timmins, 2005).
In conclusion, as touch has many positive effects on humans, the question arises whether a social robot is also able to fulfil these functions? Below, I will discuss what is already known about robot-touch and creating a bond with a robot.
1.1.2. Robot-initiated touch
Robot-initiated touch is an interaction between the robot and the human, where the robot touches the human (Chen et al., 2014). A human touch is very complex (Hertenstein, 2002) and is therefore almost impossible to fully reproduce by a social robot. Social robots have been created to show and communicate social behaviour (van Erp & Toet, 2013). There are studies that claim that a touch of a robot can give physical, emotional and behavioural responses that are similar to people. There are several studies that investigated robot-initiated touch, often in healthcare settings.
Chen et al. (2014) investigated how people respond to robot-initiated touch. They conducted an experiment in which a robot touched the person’s forearm. They used the independent variables whether or not the robot verbally warned the person before touching, and the robot verbally indicated that touch was intended to clean the person’s skin (instrumental touch) or to provide comfort (affective touch).
They concluded that the participants had a generally positive subjective response of the robot touch, but the participant’s response was significantly more favourable with instrumental touch rather than affective touch. Also, they found that the perceived intent of the robot is important and influenced people’s responses. If the participant believed that the robot intended to clean their arm, the participant tended to respond more favourably than if they believed the robot intended to comfort them. So, an important social factor for touch interaction is the perceived intent of robot-initiated touch. Also, Shrestha et al. (2015) stated that functional touch is more accepted than affective touch, persons did not find robot-initiated contact to be uncomfortable if deemed necessary.
Nakagawa et al. (2011) investigated the effect of robot-initiated touch on people’s motivation, for example performing monotonous tasks as exercising. In the experiment, the robot requested participants to perform a monotonous task with a robot’s touch or no touch. They found that participants performed the task significantly longer and with more activity when the robot touched their hands. And the participants felt that the robot was more friendly when touching them compared with no touch. This is similar to the motivational effect of touch between humans (Fisher, Rytting & Heslin, 1976).
Itoh et al. (2006) investigated objective measures of the effect of touch on humans, in decreasing stress. They found that a robot touch decreased a person’s perception of stress with physiological indices. This study was not specific robot-initiated touch because the participant and the robot were holding hands, but nevertheless the effect was a reduction in stress. Also, Bickmore, Fernando and Schulman (2010) stated that a robot touch could be comforting, but only with persons who were comfortable being touched by other people. The place where the touch takes place is also important, Li,
Ju and Reeves (2016) stated that touching less accessible regions of the robot (e.g., buttocks and genitals) was more physiologically arousing than touching more accessible regions (e.g., hands and feet).
Willemse, Huisman, Jung, van Erp & Heylen (2016) found that pleasantness ratings of observed stroking touch to a human arm, showed a similar response pattern for stroking with an actual human hand, mannequin arm, robot arm, and plastic tube. This would suggest that the social setting of the touch may be more important than the stimulus that applies the touch.
Fukuda, Shiomi, Nakagawa and Ueda (2012) investigated whether touch from a robot could inhibit a negative feeling to the robot. They set up an experiment with two conditions: an economic trust game was played, where the robot made an unfair proposal when the robot touched or did not touched the arm of the participant. Results indicate that sense of unfairness was reduced only in the touch condition, which suggests the possibility that the Midas touch effect occurs even in human-robot communication. Wullenkord, Fraune and Eyssel (2016) investigated the effect of robot touch on the attitudes towards robots. Results showed that robot touch effectively reduced negative emotions.
However, for participants with pre-existing negative emotions toward robots, contact sometimes produced more negative attitudes. It is also important that how closer and more active a robot is, the more important the non-verbal signals (like touch) of the robot are to create a positive effect (Nakata, Sato & Mori, 1998).
So, touch comfort as well as experiences and attitudes towards robots play an important role in the evaluation of touch from a robot. Hoffmann (2017) stated that it is possible that the acceptance of robot touch would further increase if robots became more familiar to humans in the future. Van Erp and Toet (2013) formulated design rules as guidance for the development of social agents that can touch.
An important aspect when a robot touches a human, because it is impossible to fully reproduce a touch by a robot, is that a touch of a robot meets a person's expectations. When a robot touch is not in line with the expectations of a person, this will result in a negative evaluation of the robot (Mori, 1970;
MacDorman, 2006; Yohanan & MacLean, 2011). The meaning of the touch depends on social, cultural and individual differences (van Erp and Toet, 2013).
The abovementioned studies create positive expectations about the effects of a robot touch. But there is also an increasing opposing force of studies that prove the opposite effect of robot touch. For example, the study by Willemse, Toet and van Erp (2017) investigated whether a robot-initiated touch could have a soothing effect on a person, during a stressful moment. The study concluded that there was no calming effect due the touch of a robot. According to the study, a possible limitation was that there was no relation between the person and the robot, this could be important for an appropriate touch (Rantala et al., 2013; Gooch and Watts, 2010). A touch of a stranger can even be experienced as inappropriate or unpleasant (Smith & MacLean, 2007). In recent study by Sefidgar et al. (2016), the calming effect of an animal robot was studied, with the robot on the participant’s lap. The results of the study showed that the participant's heart rate and respiratory rate were significantly lower and that the participant felt calmer and happier than the participants without a robot. There are also studies that use
mediated touch (remote touch by a computer or another device), suggesting that touch can only be appropriate if people have created a bond (Gooch & Watts, 2010). But there is also a study that finds no effect in using mediated touch (Erk, Toet & van Erp, 2015). Is this also the case with a robot-initiated touch? There are some studies that demonstrate that a robot-initiated touch can have stress reducing effects on people (Itoh, et al., 2006; Bickmore et al., 2010). Also, the effect of pet robots (that not active touch the human, just lies on the lap of the participant) has been shown to have a stress reducing effect (Yohanan & MacLean, 2012). People are familiar with pets and stuffed animals but less so with robots.
Perhaps by increasing the familiarity towards a robot, for example by building a bond with the robot, the calming effect of the robot can be enhanced. Hence, there are no unequivocal effects of robot touch.
It might even mean that a robot touch does not have the same effect similar to the touch of a human.
The studies mentioned above have a particular effect on functional touches (e.g. instrumental and motivational touches), affective robot touch has less strong effects.
In conclusion, there are studies claim that a touch of a robot gives similar responses as human- touch, such as motivation, help behaviour and comforting. There is an increasing amount of studies that report null-effects or even opposite effects. Several social factors are important in robot-initiated touch:
perceived intent of the robot touch, context, place of touch, attitude towards robots, expectations of the person and the bond between them. In the next paragraph, I will discuss previous research about bonding with a robot, a possible important aspect in robot-initiated touch.
1.1.3. Creating a bond with a robot
As mentioned above, the relationship that people have with each other is important for an appropriate touch (Coan, Schaefer & Davidson, 2006). For an appropriate touch interaction between human and robot, it could also be the case that there has to be a bond between the person and the robot. Below some studies are discussed that have investigated the bond between human and robot.
There are studies that claim that it is possible to create a bond with a robot within one interaction moment. Pereira, Leite, Mascarenhas, Martinho and Paiva (2010) argue that empathy is an important concept in social behaviour, as empathy is related to helping behaviour and friendship. An empathic robot is seen more like a friend than a non-empathic robot. In this study, the perceived friendship towards the robot was measured, with six dimensions of friendship (stimulating companionship, reliability, intimacy, reliable alliance, self-validation and emotional security). During the experiment five chess games were played, and the subjects never interacted with the robot before. There were two conditions, one in which the robot exhibited empathic behaviour and the other condition in which the robot exhibited neutral behaviour. The results of this study were that the users to whom the robot was empathic perceived the robot more as a friend. This study addressed the role of empathic behaviours in social robots that attempt to establish long-term relationships with humans. Also, later research by Leite et al.
(2013) stated the importance of empathic behaviour in human-robot interaction. Kruijff-Korbayová et al. (2015) showed that it is possible, within one hour, that there are participants who feel a bond with
the robot. They claim that familiarity and self-disclosure are important concepts in forming bonds.
Establishing a positive relationship between human and robot, is important in the use of social robots in the long term (Kruijff-Korbayová et al., 2015). Further, Abe et al. (2012) claimed that two factors are important to obtain a good relationship between human and robot, which they refer to as “the degree of interest” and “the degree of familiarity”. It is important that the robot can act according to the person’s (in this study children’s) inner state, for example the robot offers a new game when the child gets bored.
Abe et al. (2012) stated that there are children who interact with a robot for 30 minutes and experience the robot as a friend, but the robot has to sustain the child’s interest for a good relation. Also, Beran &
Ramirez-Serrano (2011) carried out research in which children created a bond with a robot, resulting in the children being positive about the robot and feeling they have bonded. Melson et al. (2009) stated that children treat technological devices as if they were social beings. Age could play a role in human- robot interaction (Hieida et al., 2014).
There are also studies that state that it is not possible to create a bond in such a short time. De Graaf, Allouch end Klamer (2014) stated that it is not possible to create a bond with a robot within one interaction moment. They argue that it's not realistic to have a personal bond with a robot in one day, but that this takes more time. Also, Baxter et al. (2011) state that an interaction over a longer period of time is needed to create a constructive relationship between human and robot. There is no unambiguous effect on creating a bond between human and robot. Emmeche (2014) argues that it is difficult to define a friendship between a person and a robot because it depends on what the person sees as friendship, and to what extent a robot can meet this, and this differs by person. Luo et al. (2008) argues that a negative attitude towards the robot has an influence on the relationship between a person and a robot, as well as the characteristics of a person. People who are positive towards the robot find a touch more appropriate and also the robot more human than people who have a negative attitude towards the robot (Luo et al., 2008).
Creating a bond and touch are also in relation with each other. Touch plays an important role in developing a bond or enhancing a bond (Yohanan & MacLean, 2011; Tanaka, Cicourel & Movellan, 2007). If a robot can touch a person in a natural way, this helps in creating a bond with the robot. Hieida et al. (2014) study was about creating a bond between robot and child. They stated that a good relationship can be built when the child walks hand in hand with the robot. They also stated that physical communication is age-dependent and that this physical touch may have a positive effect on reducing stress in adults. But if the robot is totally strange then touching (hand holding) might be stressful (Hieida et al., 2014). It could therefore be that having a bond with the robot is very important for a positive effect of a touch. There are recent studies that express a positive effect of touch in human-robot interaction (Miyashita et al., 2007; Argall & Billard, 2010); these studies use small pet-like robots. People are already familiar with pets instead of social robots. They see a pet as a friend, maybe when social robots will be more familiar for people, it is the same for humanoid robots?
In conclusion, creating a bond with a robot prior to engaging in a touch interaction seems to be appropriate. Some factors that are important for a robot-touch is, that a touch is appropriate, the robot shows empathy, and the robot is familiar to the human. A positive relation with the robot is important for a long-term relationship in the future between human and robot. But also, the other way around, touch could strengthen the relation between human and robot, but that is not the main purpose of this thesis.
1.2. Hypotheses
Following the above information, the following hypotheses for this study are formulated.
H1: Creating a bond with a robot will lead to more positive effects of robot touch, have beneficial effects on the persons’ arousal level in stressful circumstances, as compared to not having created a bond with a robot. This can be seen in both the physiological measures and subjective measures (self-reports) of the person.
H2: A robot-initiated touch induces more positive evaluations of the robot (physical appearance, behavioural, attitude towards the robot, trust in robot and relation with the robot), and will decrease one’s negative attitude towards the robot, after creating a bond with the robot.
H3: A touch of the robot can generate social behaviour (Midas Touch effect). The persons who are touched by a robot are more willing to stay longer, donate more money or help the experiment leader, than the person who is not touched.
An experiment has been conducted to test these hypotheses. The section below will give a brief explanation of the constructs that will be used in the experiment.
1.3. Introducing constructs study.
In this study, I will investigate whether a touch of a robot can have a calming effect if a bond has already been created between the person and the robot. Below are the various constructs that I will be used in the experiment.
Bond building
To create a bond with the robot I will set up an interaction moment in which the participant and the robot perform several tasks together. I will apply different constructs which can be important to create a bond between the participant and the robot, these constructs are also important to create relations between humans. There are several constructs that could say something about the bond between a person and a robot.
First, an important construct that affects the creation of a bond between a person and a robot is the in-group effect, which means that a person sees the robot as a partner, for example as a teamplayer.
Research, between humans, has shown that in-group members receive higher rewards than out-group members (Tajfel, Billig, Bundy & Flament, 1971), show more prosocial behaviour towards in-group
(Levine, Prosser, Evans & Reicher, 2005), and more positive affects (Beran & Ramirez-Serrano, 2011).
If the robot is seen as an in-group member then this will result in a more positive evaluation of the robot and people are more willing to interact with the robot in general (Kuchenbrandt, Eyssel, Bobinger &
Neufeld, 2013). This in-group effect will be created through put the participant and the robot in the same team during playing a game.
Secondly, the construct self-disclosure is used to measure the extent of a bond between the participant and the robot. Self-disclosure is the sharing of personal information with another person (Tolstedt & Stokes, 1984). According to Tolstedt and Stokes (1984) self-disclosure is one of the two key concepts that are important for developing a good relationship. Self-disclosure can be divided into two dimensions, breadth and depth of self-disclosure. Breadth of self-disclosure refers to the range of topics about which an individual discloses and the depth of self-disclosure is the extent to which the disclosure is personal or private. With an increase of self-disclosure, the relationship will also become stronger (Tolstedt & Stokes, 1984; Derlega, Wilson & Chaikin, 1976). As mentioned earlier, Kruijff- Korbayová et al. (2015) did a study in human-robot bonding and claim that self-disclosure is an important construct in forming a bond with a robot. This study will also use self-disclosure between the robot and the participant.
Finally, persons who are named consistently by their name find the robot friendlier than persons who were not called by their name (Kim, Kwak & Kim, 2012). Familiarity could play a role here Kruijff- Korbayová et al. (2015) and Abe et al., (2012) claimed that familiarity is important in human-robot relations. By calling the person’s name, they feel familiar to the robot and also the other way around.
Similarly, from the study of Klamer, Allouch and Heylen (2011), there is a positive relationship between name calling and building a relationship. Gonsior, Sosnowski, Buß, Wollherr & Kühnlenz, 2012) stated that it strengthens the bond because the person exhibits helpful behaviour towards the robot (Gonsior, Sosnowski, Buß, Wollherr & Kühnlenz, 2012). In this study, the robot will call the participants by their name during the experiment.
Arousal
The participants, together with the robot, will watch a scary movie in order to induce arousal, equal to the study of Willemse et al. (2017). During the movie, the robot will try to soothe the person verbally, and will or will not touch the person on the shoulder. A touch is the most widely used method to lower the stress level of a person (Heinrichs, Baumgartner, Kirschbaum, & Ehlert, 2003; Field, 2010; Dolin &
Booth-Butterfield, 1993). Self-reports and physiological measures will be used to investigate how the participant experiences the movie and the touch of the robot (see paragraph 2.3 for the specific questionnaires). Physiological measurements measure the stress level of a person, namely heart rate, skin conductance and respiration of the person. The heartbeat of a person is in relation to the perceived stress. When a person experiences more stress, the heart rate increase and hearts rate variability decrease in intensity and upon relaxation, intensity decreases (Mandryk, Inkpen & Calvert, 2006). Skin
conductance can also say something about a person's stress experience. Skin conductance exposes emotional response and cognitive activity, with increased stress resulting in increased skin conductance (Lang, 1995). Respiration, the number of breaths per minute increases with an increase in stress (Stern, Ray & Quigley, 2001).
Perception of the robot
Also, different questionnaires will be used to investigate the perception of the person towards the robot.
This will be measured before and after the experiment, to see if there is a difference. The questionnaires will measure the attitude of the person toward the robot, it is possible that participants who create a bond with the robot have a more positive attitude towards the robot at the end of the experiment. Which human characteristics a person attributes to a robot, to which a non-human thing is seen as something human for example a non-human thing like a robot, computer and animals (Bartneck, Croft & Kulic, 2009).
Which impression a person has of the robot, if this impression is positive, for example, from the robot, this leads to more positive evaluations (Robbins & DeNisi, 1994). Human robots are generally better able to interact with human behaviour than less humanlike robots (MacDorman & Ishiguro, 2006).
Social Presence is to what extent a person sees the robot as a human being (Lee, Jung, Kim & Kim, 2006). Results from a study by Heerink, Kröse, Evers and Wielinga (2008) show that a high social presence in the interaction with a robot leads to more fun and greater acceptance of the robot. The more social possibilities the robot has, the higher the Social Presence (Heerink et al., 2008).
Midas touch
A touch can affect social behaviour in humans (Crusco & Wetzel, 1984), as previously described. In this study, three forms of Midas Touch effect will be applied. First, the participant will be asked if he or she is willing to forego a part of the fee that was received for future robot research. Second, there will be a request from by the robot, asking the participant whether they are willing to stay longer for filling in another questionnaire. Third, general help behaviour of the participant will be measured, consisting of willingness to help the experiment leader, with picking up pens (which were fallen). People think that people feel more empathy for friends than for strangers (Krebs, 1970), which could mean that the participant who has a connection with the robot is more willing to meet these requests.
2. METHOD
In this study, I researched the calming effect of a touch from a robot on a person, during a stressful moment. I also looked at whether having a bond with the robot had an effect on the experience of the robot-touch. To investigate this, a quantitative study was conducted in the form of an experiment. The experiment was a follow-up to a previous study by Willemse, Toet & van Erp (2017) that investigated the soothing effect of a robot-intiated touch. The experiment used a 2 x 2 between subjects factorial design (Dooley, 2009), with the two independent variables: bond and touch. The experiment consisted
of two parts. In the first part, there was an interaction between the robot and the participant, trying to establish a bond. In the second part, stress was induced by a thrilling movie and I looked at the calming effect of a touch of a robot on the participant. The robot was controlled by the experiment leader using a Wizard of Oz technique.
2.1. Participants
In total 67 persons participated in the experiment, these persons were recruited from the participant database of TNO. This is a database where people can subscribe on a voluntary basis. To participate in the experiment, there were some criteria that the participant had to meet. The inclusion criteria were that the participant had to be at least 18 years old, should not suffer from hearing or vision problems and they had functioning arms. Because a thrilling movie was shown during the experiment, it was desirable that the participant had no bad experiences with thrilling movies.
The mean age was 47.93 (SD: 19.96, range: 18-78) and 34 of them (50.7%) were male and 33 (49.3%) were female. The participants were randomly assigned to one of the four conditions, (1) bond and touch condition (8 males/ 7 females, mean age: 50.13, range: 22-70), (2) bond and no touch condition (9 males/ 8 females, mean age: 40.07, range: 19-73), (3) no bond and touch condition (8 males/
10 females, mean age: 57.83, range: 20-77) and (4) no bond and no touch condition (9 males/ 8 female, mean age: 42.82, range: 18-78). Participants were financially compensated for their participation (30 euro plus travel expenses up to 15 euro, in accordance with the directives of TNO). The experiment was approved by the Ethics Committee of the University of Twente and the Ethics Committee of TNO.
2.2. Set up and apparatus 2.2.1. Set up
The experiment was conducted in two rooms directly adjacent to each other (see appendix I for a floor map of the set up). The first room, called the test room, was where the participant was received and the experiment was carried out. The second room was the control room where the experimenter was seated.
The test room had a homely feel and had daylight coming in. The test room was divided in two parts, one area for the interaction between the participant and the robot (part 1) and the other area where the film with the robot took place (part 2). A total of three robots were used during the experiment, one robot in the test room and two robots in the control room (one for the interaction part and the other for the movie part). The robot in the test room had been linked to the robot in the control room, depending on the part of the experiment: the interaction or the movie part). The test room consisted of a table with one chair for the participant. The robot was placed on the table facing the participant. Furthermore, a laptop and a paper board game was used. The board game was used for a memory game and had numbers on it, corresponding with the numbers on the memory cards, so it was possible for the experiment leader to make the good combination through the camera. In the film area was a couch for the participant. On the left side of the couch there was a little chair for the robot. The chair was placed so that the robot was
able to touch the participant on the shoulder during the experiment. It was important that the participant took place on the right side on the couch next to the robot, therefore the left side of the couch was blocked by a little table, where the questionnaire was placed that had to filled in after the movie.
The movie was projected by a projector on a white wall. In the test room two cameras were positioned, enabling the recording of the participant during the interaction part and movie. The cameras had been used for observing the participant and for control the robot in the right way from the control room during the experiment. At the same time intercoms were used, so the experimenter could hear and talk to the participant during the experiment.
2.2.2. Robot
The robot was controlled by the experiment leader used a Wizard of Oz (WoZ) technique. The WoZ is a technique where a person (often the experimental leader) remotely controls a robot, such as movement and speech (Kelley, 1984). Three Nao robots1 (version: Nao v4 and operating system NaoQi 1.12.5), were used in a “master-slave” set up. The movements made by the "master" robot in the control room were reproduced by the "slave" robot in the test room. Two robots were used in the control room, one for the interaction and one for the movie. Because different programs were used for the robot's behaviour during the interaction and the movie, the two robots could be switched quickly during the experiment.
The third robot was used in the test room. It was linked to the interaction robot or the movie robot that was in the control room. There was another spare robot for unforeseen situations, such as the battery failure. For the speech of the robot, a custom made software program (NaoCopier) was used in combination with the Acapela Femke Dutch Female 22 kHz Text-to-Speech converter2. This program was able to pre-program sentences that the robot had to say, and it was also possible to manually enter sentences during the experiment. Per part (interaction and film), pre-sentences were made that the experimental leader could play during the experiment. The voice of the robot was not clearly a male or female voice, this was not to affect the participant in relation to the gender of the robot.
2.2.3. Interaction (part 1)
During the interaction part, four different tasks were performed (see Appendix II for the interaction script / protocol). These tasks are based on principles that are used by people, and apparently also robots, to create and maintain relationships. In the bond condition the robot and the participant performed the tasks together. In the no bond condition, the tasks were performed by the experiment leader and the participant (in the explanation below the word robot is used, but it could also be the experiment leader, depending on the condition). The interaction part of both conditions was identical. The following order of the four tasks are carried out.
1
Informatie over Nao robot: https://www.ald.softbankrobotics.com/en/cool-robots/nao
2
http://www.acapel-group.com
Acquaintance
An introduction where the robot presented itself and shortly asked about the participant’s interests (hobbies and interests in movies). There was a short conversation between the robot and the participant, where the robot and the participant disclosed to each other to strengthen the bond. This was done by pre-programmed sentences, such as "Hello my name is Nao, who are you?" Or "We're going to watch a movie, do you like watching movies (name participant)?". During the conversation, the robot used the name of the participant to make it more personal. See appendix II for the whole conversation.
Memory game
The second task was a memory game. It was a paper board game that was stuck on the table, with the numbers on the board corresponding to the numbers on the cards. So, it was possible for the experimental leader to play the game from the control room through camera. During the game, the robot asked the participant if he or she could help the robot by turning the card. Helping behaviour is important for strengthening the bond between robot and the participant. Before the game started the robot asked if the participant could help the robot turn the cards. Also, the robot complimented the participant during the game while using the name of the participant. The robot said sentences like: "Could you turn number 5 for me?" Or "Very good, (name participant)".
Quiz
The participant and he robot did a quiz together. They were a team and had to solve questions, this was to create an in-group effect and strengthen the bond between them. The quiz was done on a laptop consisting of eight questions that were identical for each participant; the participant, however, was told that the questions were randomized out of a database with more than one hundred questions. The robot and the participant had to help each other to come to the right solution. The robot helped the participant by asking critical questions on what the robot himself thought what the answer was and also encouraged the participant. For example: "I think that answer B is" or "A cheetah is going fast, would it be more than 100 km/h?" or "Well done, (name participant)".
Personal story
In this part of the interaction the robot told the participant a personal story that the robot had been through. This is a process in which personal information is shared by the robot, and the participant is thought to be more willing to disclose something personally. The story that the robot told was a situation that the robot had experienced (for the story see appendix II). After the robot had told the story, the robot asked the participant if he or she was willing to telling something similar that happened last week or earlier. When the participant was willing to do so, the robot told he or she could tell the robot at a later time.
2.2.4. Movie (part 2)
The movies used for the experiment were movies not known to the participants. After to the experiment, participants were asked whether they were familiar with one of two thrilling movies, The Descendent (Anderson & Glickert, 2006) and the Red Balloon (Trounce, Mace, & Wasjbrot, 2010). Before the two thrilling movies, a neutral movie was shown ("Coral Sea Dreaming", Hannan, 2010) and in the two thrilling movies, the tension was slowly built up. The total duration of the movies was 27 minutes and 51 seconds, of which 20 minutes and 4 seconds the exciting sections. The exciting sections of the movie consisted of a build-up of tension during the fragments, there were no extreme (violent) images.
2.2.5. Physiology
During the movie, physiological measurements were carried out by measuring heart rate, skin conductance and respiration. This was to measure how the participant physically responded to the robot's touch. The moments when the robot touched the participant were the same as in the previous study by Willemse, Toet and van Erp (2017). By means of a custom made program, these physiological measurements could be linked to the predetermined moments during the movie, for example, how the participant responded to a thrilling moment in the movie. These were times when it was thrilling and there was not much noise, so the robot was well understood by the participant, while the robot is tried to soothe the participant verbally, such as "Fortunely, it is just a movie". The physiological response of the participant was determined by means of electrodes pre-picked up. The Biosemi ActiveView software (v7.06, recording rate 2048 Hz) equipment and six electrodes were used, the Flat Ag-AgCl electrodes (heart rate), passive Nihon Kohden electrodes (skin guidance) and the SleepSense 1387 kit (respiration).
2.3. Materials and measures
The following questionnaires and measurements were used in the experiment for the different constructs (see an overview of the questionnaires in appendix III).
2.3.1. Bond with the robot
In order to be able to say something about the bond between the participant and the robot, different questionnaires have been used.
Inclusion of Other in the Self scale
The Inclusion of Other in the Self Scale (IOS) (Aron, Aron & Smollan, 1992) was answered by the participant to see how the participant considered his or her relationship with the robot. The question consisted of seven images with circles that proposed the relation between the participant and the robot.
The images are ordered from one to seven, whereby the circles in image one are next to each other and in image seven the circles totally overlap. The participant could choose between seven images and must indicate which one was most appropriate to him or her. The IOS was administered before and after the
interaction and at the end of the experiment, to map both the effects of the interaction and the touch during the movie.
Attachment Scale
The Attachment Scale (Schifferstein & Zwartkruis-Pilgrim, 2008) says something about how the participant was connected to the robot. The scale consisted of five questions that had to be answered by the participant with a 5-point Likert scale. An example question is: "I felt emotionally connected to the robot". This scale was administered after the interaction and at the end of the experiment.
Perceived Friendship Scale
The Perceived Friendship Scale, as used in Willemse, Toet and Van Erp (2017) from Pereira, Leite and Mascarenhas (2011) and previously used by Nie and Park (2012), consisted of eight questions that indicate how the participant experienced his or her friendship with the robot with a 7-point Likert scale.
An example question that was asked was: "I would like to watch another movie together with the robot".
This scale was administered at the end of the experiment.
Self-disclosure
At the end of the interaction part, where the robot told a personal story, the participant was asked to what extent he or she was willing to tell something about different personal topics (varied in intimacy).
A questionnaire used by Pearce and Wiebe (1975) was used, which consists of 21 questions different in degree of intimacy (low, medium, high). In this study, two questions from each level will be used, so a total of six. For example: "Music taste” and "Most sad experience in your life". The participant indicated how willing he or she was to tell about it, on a 7-point Likert scale.
2.3.2. Arousal
To measure the amount of stress, objective (physiological) and subjective dimensions were used. Below are the different used measures that were used to measure the amount of stress.
Physiological measures
To measure the heart rate, two electrodes were placed, one on the right collar and one on the left floating rib. Both, the number of beats per minute and heart rate variability have been measured, and also the variance between the heartrates. To measure the skin conductance, two electrodes were placed on the inside of the left hand. One in the middle of the hand and the other just below the index finger, as in Lykken and Venables (1971). Finally, I measured the respiratory rate of the participant, the number of breaths per minute. This was measured by means of a respiration band that was placed under the breast of the participant.
Average scores were determined for heart rate, skin conductance and respiration for different moments during the film. (1) For the baseline measurement, (2) the entire movie, (3) only the thrilling portion of the film and (4) the average of all eight touch moments (duration touch plus 45 seconds for each moment). By means of the averages, calculations could be made by comparing the averages of the different moments.
Self-report
In addition to this objective measurement of the emotional state of a person, a subjective method was used. A questionnaire that measures the emotional state of a person is the Self-Assessment Manikin (SAM) (Bradley & Lang, 1994) and the Positive and Negative Affect Schedule (PANAS) (Watson, Clark
& Tellegen, 1988, Dutch translation: Peeters, Ponds & Vermeeren, 1996). With the SAM, three different constructs are measured, Valence, Arousal and Dominance. The PANAS measures the degree of positive and negative impact. In both questionnaires, a person indicates how he or she feels at the moment. The SAM uses pictures and a 9-point Likert scale. The participant indicates the extent to which his or her feelings correspond to the different pictures. The PANAS measures the degree of positive and negative impact through 20 words, for example words like "upset", "nervous" and "proud". The participant indicated how he or she felt at the time the questionnaire was taken. These questions were asked before and after the movie.
2.3.3. Perception of the robot
Perception of the robot by the participant was measured in three ways: (1) the attitude towards the robot, (2) the physical appearance of the robot and (3) the behaviour of the robot.
Attitude towards the robot
The attitude of the participant towards the robot was measured by the Negative Attitude to Robot Scale (NARS) questionnaire (Normura et al., 2008). The NARS is a questionnaire that consists of three different sub-scales namely, (1) Interaction with Robots, (2) Social Influence of Robots and (3) Emotional Interactions with Robots. This is a questionnaire with 14 statements, such as: "I would feel uncomfortable when robots really had emotions". The answers are a 5-point Likert scale of strongly disagree (1) to strongly agree (5). The NARS questionnaire was taken at the beginning and the end of the experiment to see how the participant thought about the robot prior to the experiment and whether this had changed by the end of the experiment. The average of the NARS subscales was calculated to be able to perform the analyses.
Physical appearance of the robot
To see to what extent the participant gave human attributes to the robot, parts of the Godspeed questionnaire were used (Bartneck, Croft & Kulic, 2009). The questionnaire consists of five constructs,
two of them will be used in this study, namely Anthropomorphism and Likeability. Anthropomorphism is the extent to which a non-human thing is seen as something human and Likeability is the positive impressions reported as a result of the visual and vocal behaviour shown by the robot. These questionnaires were measured with a 5-point Likert scale. For example, Anthropomorphism (moving rigidly-moving elegantly) and Likeability (dislike-like). The Godspeed was administered before and after the interaction and at the end of the experiment.
The Perceived Human Likeness Scale (MacDorman, 2006) was administered by means of three questions with a 7-point Likert scale. The higher the score, the more human the robot is seen. For example: "Machine-like - Human-like". These questions were asked after the interaction and at the end of the experiment.
Behaviour of the robot
The Social Presence is to what extent the participant viewed the robot as a human being (Lee et al., 2006). Social Presence was measured after the interaction moment, consisting of eight questions with a 10-point Likert scale (questionnaire used from Lee et al., 2006). Among other things, the question was asked to what extent the participant found the robot human for instance, by asking: "To what extent did the robot come across as an intelligent being?". By taking the average of these eight questions, Social Presence was determined.
Behaviour that brings people physically or psychologically closer together can you call immediacy behaviour. Four questions from the Immediacy Scale (proximity, "cold-hot") of the robot were asked, as used in Kidd (2003). And the participant is asked how credible the robot seems to be.
For this, fifteen questions from the Credibility Scale (credibility, "fair-unfair") were asked (Kidd, 2003).
These questions (Immediacy and Credibility scale) were asked at the end of the experiment, to answer with a 7-point Likert scale.
Touch by the robot
At the end of the experiment, all the participants were asked if they had been touched and if so, how the touch was experienced. All participants were asked if they were touched, regardless of whether they were assigned to the ‘touch’ condition or not. This questionnaire consisted of six statements, for example, "The robot's touch was appropriate", and participants responded by means of a 7-point Likert scale (not at all (1) to totally (7)). By taking the average of these six statements, a score was obtained that indicates the extent to which the participant experienced the touch as positive.
Trust in robot scale
First of all, four questions were used of the Affective Trust Scale (adopted from Johnson & Grayson, 2005, as applied by Kim et al., 2012), in which the participant indicated to what extent the robot, for example, was strange or familiar to him or her. Next, six questions from the Perceived Trust Scale (Kidd, 2003 and Rubin et al., 2009) were used, where the participant indicated to what extent he or she agreed
with the different statements, for example: "The robot was sincere". These statements were answered on a 7-point Likert scale.
2.3.4. Midas Touch
As mentioned in the introduction, there were also a number of questions in the experiment that investigated the Midas Touch effect. A combination of elements was measured that focused on a request specifically directed to the robot, so at the end of the last questionnaire the participant was asked whether he or she was willing to forego part of the fee received so that this money could be used for future robot research. If yes, then it was asked how much money the participant would like to donate. Second, there was another request from the robot, on whether the participant was willing to complete another questionnaire for other research after the experiment. If so, how long did the participant have time for it? Finally, to assess the general help behaviour of the participant, the experimental leader dropped a bowl of pens. It was then checked whether the participant was helping to pick up the pens and how many pens were picked up. The experimental leader picked up one pen per second and when all the pens had been picked up, the experimental leader counted how many pens the participant had picked up.
2.3.5. Covariates
Several covariates were included in this study, which could have influenced the outcome of the perception of the robot.
First, the Robot Anxiety Scale (RAS) (Nomura, 2008). The RAS measures the fear that the participant experiences with the robot. It is a questionnaire of 11 statements with a 6-point Likert scale (I do not feel any anxiety at all (1) to I feel anxiety very strongly (6)). An example of a statement is:
"Robots may talk about something irrelevant during conversation" The questionnaire consists of three sub-scales, Communication Capability of Robots, Behavioural Characteristics of Robots and Discourse with Robots. The RAS was only requested prior to the experiment.
Second, Touch Receptivity (Bickmore, Fernando, Ring & Schulman, 2010) has also been included in the study. This questionnaire says something about how people experience a touch and whether they like it when they are touched by someone. This could affect the experience of the robot's touch. The questionnaire asks questions like: "I feel uncomfortable when someone casually touches me"
and "I like people who shake hands with me".
Finally, gender, age and experience with technology and robot were taken as a covariate (Hieida et al., 2014; Schermerhorn, Scheutz & Crowell, 2008). Experience with technology plays an important role in robot acceptance (Heerink et al., 2008).
2.4. Procedure
When the participant arrived, the experiment leader told what the apparent purpose of the experiment was. I did not tell in advance what the real purpose of the experiment was, so as not to influence the
behaviour of the participant. A cover story was told, the purpose of the first part (interaction) was that the robot observed the participant and collected data, such as speech and movement. In part 2 (film), the information collected by the robot in part 1 was used to show appropriate behaviour to the participant during the movie. After explaining the experiment and signing the informed consent, the participant started with the first paper questionnaire. During the experiment three composite questionnaire (with the questionnaires mentioned above in paragraph 2.3) were filled in, before, after the interaction part and at the end of the experiment. See table 1 for a schematic representation of the questionnaires in chronological order and appendix III for the full questionnaires. The first questionnaire consisted of the demographic characteristics, knowledge and experience of technology and robots and perception of the robot. The first questionnaire differed between the touch and no touch condition. In the touch condition, the Touch Receptivity was also measured. After completing the first questionnaire, the interaction moment started with the robot or the experiment leader, depending on the condition. Four tasks were carried out. First, the acquaintance between the robot and the participant was carried out, after that a memory game was played, then a quiz was played and at the end the robot told a personal story to the participant. From the control room, the robot was controlled by the experiment leader. The experiment leader responded from the control room to the participant's behaviour with pre-set sentences and sometimes by entering a text herself. Several sentences were generated to give quick responses. In the control condition where the participant did the task with the experiment leader, the experiment leader tried to show as much as possible the same behaviour as the robot did. In the control condition, the robot sat on the table and moved his head as if the robot observed the participant. After the interaction moment, the second questionnaire was filled in. In the bond condition, the robot asked the participant to complete the second questionnaire, and whether the participant wanted to put the robot on the couch for part two part (film). After questionnaire two was filled in the electrodes were placed by the experiment leader for part two. After the electrodes were placed it was checked they were properly connected. In the touch condition, the robot had to touch the participant, so it was important that the robot was placed in such a way that it could touch the shoulder of the participant with his hand. When everything was ready for part two, the light was dimmed and the experiment leader started the movie from the control room.
During the film, the experimental leader steered the movements (in the touch condition) and speech of the robot. The robot spoke calming texts and at the same time laid his hand on the shoulder of the participant. The eight touch moments and duration were the same as in the study of Willemse, Toet and Van Erp (2017), but the sentences were customised. The sentences were such that participants were able to interpret them in different ways and they were both for participants who found the movie exciting and for participants who did not find it exciting (see appendix II for the sentences that were used). At the end of the movie, the robot asked the participant to fill in the third questionnaire. Finally, the Midas Touch effect was queried. At the end of the experiment, the participants were informed of the actual purpose of the experiment, the payment form was completed and the participant was returned to the reception. The experiment lasted approximately two hours, which could vary by participant.
