Human-robot interaction: can cute robots make a difference?
Investigating the relationship between a whimsically cute robot, eeriness and robotacceptance and the moderating role of the type of service setting
Human-robot interaction: can cute robots make a difference?
Investigating the relationship between a whimsically cute robot, eeriness and robotacceptance and the moderating role of the type of service setting
University of Groningen Faculty of Economics and Business
Department of Marketing
MSc Marketing Management Master Thesis
June18, 2018
First supervisor: Dr. J. van Doorn Second supervisor: S. AlBalooshi
By
Roos van Eeden r.n.van.eeden@student.rug.nl
ABSTRACT
The presence of robots is increasing and robots are now behind reception desks, in households and in elderly care. The importance of robot acceptance is becoming increasingly recognized but until recently, studies have mainly focused on the technical aspect of robots. The current study investigates the relationship between a whimsically cute or non-cute robot, eeriness and robot acceptance. A hedonic versus utilitarian service setting is considered as a possible moderator for relation between the robot type and eeriness. An experimental study was conducted, where the type of robot and type of service setting were manipulated. The findings ofthis research indicate that eeriness does have a negative effect on robot acceptance. Moreover, it is found that the hedonic service setting strengthens the negative relationship between the type of robot and eeriness. The study was unable to find the proposed negative relationship between the type of robot and eeriness. The results of this study indicate that the setting in which people are in contact with robots influences their perception of the robot. Furthermore, this research highlights the possible effects of experiencing an eerie robot. When a service robot is used in organizations, the robot should not be too humanlike since this can elicit eerie feelings, which in turn will decrease the robot acceptance.
Keywords: whimsical cuteness, robot acceptance, service robots, eeriness, uncanny valley, human-robot interaction, hedonic service setting, utilitarian service setting, intention to use, attitude
PREFACE AND ACKNOWLEDGEMENTS
This thesis was written for my Master Marketing Management at the University of
TABLE OF CONTENT
1. INTRODUCTION ... 6
2. THEORETICAL FRAMEWORK ... 9
2.1 Robots ... 9
2.2 Robot acceptance ... 10
2.3 The uncanny valley and eeriness ... 11
2.4 Whimsical cuteness ... 14
2.5 Hedonic versus utilitarian service setting ... 16
2.6 Conceptual model ... 17
2.7 Hypotheses ... 18
3. METHODOLOGY ... 22
3.1 Research design and procedures ... 22
3.2 Measures ... 24 3.3 Plan of analysis ... 27 4. RESULTS ... 28 4.1 Descriptive statistics ... 28 4.2 Analyses ... 30 4.3 Hypotheses testing ... 32
4.4 Effects of the control variables ... 37
5. DISCUSSION ... 38
5.1 Overview of results ... 38
5.2 The effect of the type of robot on eeriness ... 39
5.3 The moderating role of the type of service setting ... 40
5.4 The effect of eeriness on robot acceptance ... 41
5.5 Additional results ... 41
5.6 Managerial implications ... 42
5.7 Limitations and future research directions ... 43
6. CONCLUSION ... 45
1. INTRODUCTION
Robots are becoming a bigger part of our life everyday (Beer, Prakash, Mitzner and Rogers, 2011; Kanda, Hirano, Eaton, Ishiguyro, 2004; Thompson, Trafton and McKnight, 2011) and functional robots are proven to be capable and productive helpers in manufacturing (Lawton, 2017). There are also different kinds of social robots that are already used in
practice. Social robots are becoming increasingly present in the service setting, such as robot Pepper. Pepper is a humanoid service robot that provides customersin stores and service settings with information and amusement (Who is Pepper, 2018). In 2014, Nestlé purchased thousand Pepper robots for their stores in Japan (Nestlé employs fleet, 2014). Social robots are also available for private use: robotic dog Aibo provides households with love and affection (Top 10 robots, 2017). If robots continue to be developed, people will more often encounter a robot and engage in human-robot interaction (Hanson, 2006; Van Wynsberghe, 2016). Earlier research on robots has mainly focused on the technical facets of human-robot interaction (Jayawardena, 2010). Moreover, much research has been conducted about the effectiveness of robots in different healthcare institutions (Fricke, Meyer, Wagner, 2017; Hebesberger, Koertner, Gisinger and Pripfl, 2017; Piezzo and Suzuki, 2017; Smarr et al, 2012; Song, Wu, Ni, Li and Qin, 2016).Even though robot effectiveness is largely tested, there is relatively little research on how people experience humanlike robots (Mara and Appel, 2014). Furthermore, the human response to humanoid robots is becoming an increasingly important research topic (Mara and Appel, 2014).
When looking at the human response towards robots, it is shown that children often react positively towards a robot (Short et al., 2014). This might be explained by the fact that robots are less complex and less intimidating than humans (Cabibihan, Javed, Ang and Aljunied, 2013), which might appeal to children. We also know that elderly adults are selective, but open to robots as assistants in the household (Smarr et al., 2012). Even though most adults are curious about engaging with a robot, the encounter also brings out fear and refusal (Hebesberger et al., 2017). People are concerned about a violation of privacy,
information overload and they have little trust in new technologies (AartsandDeRuyter, 2009). These concerns might result in a lower acceptance of robots.
increases as they appear more like a human, until they take a fall in the uncanny valley (Mori, 1970). When entering the uncanny valley, a feeling of eeriness arises and people lose their affinity for the robot or non-human object. It resembles a negative and unpleasant feeling when encountering humanlike robots (Flach et al., 2012; Mara and Appel, 2014), which originates from the instinct to protect oneself from danger (Mori, 1970). Examples are encountering a corpse or a prosthetic hand. Recent literature has mainly focused on the uncanny valley, but did not concentrate on measuring eeriness (MacDorman and Entezari, 2015). People are not yet aware of ways to mitigate eeriness, even though this might influence robot acceptance.
Even though Mori’s (1970) findings gave insights in how robots can be perceived, more research on the acceptance of robots need to be conducted (Beer et al., 2011; Mara et al., 2013). Broadbent, Stafford and MacDonald (2009) found that people’s expectations as well as the robot’s appearance and behavior influence their acceptance in healthcare services.
According to Beer et al. (2011), robot acceptance could potentially be influenced by the robot’s function, its social capabilities and its appearance. Moreover, according to
MacDorman et al. (2005) robots lose their value when humans are not capable of accepting them as social beings because of the way they look. The appearance of robots is therefore of crucial importance when looking at the acceptance of robots. Different appearances of robots might influence whether or not people accept these robots or not. Since the possible effects of robot appearance need more research (Beer et al., 2011; Broadbent et al., 2009), this thesis will focus on this phenomenon. The appearance of robots can be manipulated to make them more appealing to humans. Objects are often designed to be cute in order to charm humans (Pajares Tosca, 2009). A possible solution to increase the acceptance of robots might be to create cute ones. If cuteness can be manipulated when designing a robot, we might be able to explore its effects on outcomes such as eeriness and acceptance.
Whenever cuteness is judged in appearance, it is often negatively criticized or not taken seriously (Morreall, 1991). However, cuteness is an essential part of human evolution (Morreall, 1991). A specific type of cuteness, whimsical cuteness, brings up associations of playfulness and fun (Nenkov and Scott, 2014). These associations seem to be contrasting with the self-protection motive that activate eerie feelings amongst humans. A whimsically cute robot that brings up playfulness might decrease the feeling of eeriness that a person
still lacks examinationof the behavior of consumers (Nenkov and Scott, 2014) and more research is needed.
It is possible that the effects of eeriness are influenced by the type of robot, as well as the type of service setting that the person is in. The type of service setting influences how people experience a service (Hill et al., 2004; Hirschman and Holbrook, 1982; Moore et al., 2005). A service setting can be either hedonic or utilitarian.The utilitarian service setting is associated with task orientation and efficiency, whereas the hedonic service setting is associated with playfulness and fun (Arnold and Reynolds, 2003; Kang and Park-Poaps, 2010). The hedonic service setting seems to match with the playful associations of a
whimsically cute robot. However, no research has explored this possible relationship yet and more research is needed.
So far, no research has investigated how the appearance of social robots can influence robot acceptance. However, this might be crucial in a world where robots become
increasingly present. This research attempts to fill thisgap by uncovering the possible relationship between a whimsically cute robot and robot acceptance. Eeriness is taken into account as a possible mediator, where the hedonic versus utilitarian service setting is a possible moderator on this mediated effect. This thesis hopes to find new insights into the field of robot acceptance and factors influencing robot acceptance and eeriness. By means of this research, the study hopes to answer the following research questions:
2. THEORETICAL FRAMEWORK
In the previous chapter, the topic of robots is introduced, along with the current gaps in the literature and the relevance of investigating robot acceptance. Moreover, the research questions are introduced. In the following chapter, the theoretical findings regarding this topic will be discussed. Consequently, a conceptual model and hypotheses are formed and argued on the basis of existing literature.
2.1Robots
Robots are appearing more in private as well as in businesses. Robots started out as mechanical help in factories, but are now also appearing in front of customers. Additionally, they are increasingly advanced and are able to read our emotions and be our companion. There are robots in all different sizes and forms. There are very mechanical, industrial robots such as a robotic arm. Robotic arms are designed to be safe for humans, as well as fast and precise in the tasks they are designed for (Bichi and Tonietti, 2004). Industrial robots are manufactured with the sole purposeof replacing human labor (Fortunati, Esposito and Lugano, 2015). They are not meant to look like humans and do not have legs or a face (Mori, 1970). The robots need to operate as the machine that they are (Zhao, 2006) and are purely designed for functional purposes (Mori, 1970). There are also more social robots, such as PARO. PARO is a robotic pet seal designed to grant therapy to patients(Sabanovic, Bennet, Chang and Huber, 2013). A social pet such as PARO can indirectly increaseelderly engagement with other people (Sabanovic et al., 2013), offering a possible remedy for loneliness amongst the elderly. Next to PARO, there are also social service robots constructed to inform, assist and guide people, such as robot Spencer in the aviation world (Triebel et al., 2016). This robot tries to escort passengers that arrive at the airport to the passport control. There is also a difference in the appearance of robots. Androids are robots that are highly
humanlike.Humanoid robotshave a more mechanical look whilst still resembling forms of the human body(MacDorman and Ishiguro, 2006). Different robots are designed fordifferent goals and therefore have different benefits. To illustrate, industrial robots are flexible and cost-saving substitutes for machines (Abele, Weigold and Rothenbücher, 2007). Companion robots are intended to help people or to mitigate loneliness (Robotics in healthcare, 2018).
not only to be useful, but also to interact with humans in a natural way (Breazeal, 2003). For the robots to be this intelligent, they require various social and cognitive capabilities
(Anzalone, Boucenna, Ivaldi and Chetouani, 2015). The robots require face and gesture recognition and need to be able to adapt to the human’s response in order to keep the
interaction going (Anzalone et al., 2015). Fong, Nourbakhsh and Dautenhahn (2003) declare that social robots need to be flexible and adaptable in order to interact with humans, since humans differ a lot from each other.
The degree to which a social robot is flexible and adaptable to human behavior differs per robot. Breazeal (2003) identifies four types of social robots: socially evocative, social interface, socially receptive and sociable. A socially evocative robot is a robot that stimulates people to anthropomorphize the robot (Breazeal, 2003), such as for example PARO the pet seal. Social interface takes it a step further, where the robot uses social cues to interact with people (Breazeal, 2003). This means that the robot requires the ability to transfer a message and often uses facial expressions and gestures (Breazeal, 2003). When a robot is socially receptive, the robot itself will also benefit from interacting with humans. Socially receptive robots can learn from humans, but they remain socially passive (Breazeal, 2003; Fong et al., 2003). The last and most advanced type of robot is the sociable robot (Breazal, 2003). This type of robot interacts with humans in an active manner not only to benefit the human, but to benefit itself as well. Sociable robots have their own internal goals (Breazeal, 2003).Although these type of robots are well established in theory, it is often still a challenge for robots to interact with humans on a social basis (Anzalone et al., 2015). People will increasingly engage in human-robot interaction (Hanson, 2006; Van Wynsberghe, 2016)and therefore it is of crucial importance to investigate how robots can interact with humans in order to increase robot acceptance.
2.2 Robot acceptance
overall acceptance of a new technology, such as a robot. Davis et al. (1989) declare that the variables perceived enjoyment, perceived ease of use and perceived usefulness influence the intention to use a technology. Research by Heerink et al. (2008) showed that intention to use directly influences actual usage. This usage shows the acceptance of the robot (Heerink et al., 2008; Heerink et al., 2010). The intention to use arises when someone believes that using an object or technology is useful (Venkatesh and Davis, 2000). After this belief, the person acquires an intention to use the object or technology. People can also acquire an increased intention to use when they identify that something is enjoyable (Shiau and Chau, 2016). Next to intention to use, Kim, Shin and Park (2015) discovered that user attitude has a positive effect on acceptance. Therefore the acceptance of a robot can also be distinguished by looking at the attitude of the user. Attitude can be seen as the extent to which someone judges a
particular behavior or object as favorable or unfavorable (Ajzen, 1991), where we will look at judging the robot. Nomura et al. (2006) indicate that a negative attitude towards robots can influence their behavior towards robots. Furthermore, they mention that experiences with a robot can influence the relationship between attitude and the acceptance of this robot (Nomura et al., 2006).
Research found that visual aspects also influence acceptance. When a system is visually more alluring, this increases perceived ease of use (Van der Heijden, 2003), which increases intention to use and indirectly robot acceptance. Beer et al. (2011) and Broadbent et al. (2009) also declare that the appearance of the robot could influence robot acceptance. According to Broadbent et al. (2009) robot acceptance is dependent of the robot’s appearance, -humanness, -facial expressions, -size, -personality and -adaptability to change behavior to user preferences. The possibility exists that a robot that appears cute will indirectly elicit higher acceptance.
2.3The uncannyvalley and eeriness The uncanny valley
any way. This can be for example an industrial robot, such as a robotic arm. As a robot starts to appear more human-like, feelings of affinity and empathy grow among the viewer (Mara and Appel, 2015; Mori, 1970). However, once a robot resembles a human too closely, the impression of the robot falls into the uncanny valley. In the uncanny valley, a feeling of eeriness arises and people lose their affinity for the non-human object (Mori, 1970). This fall happens because people realize that something they thought was real at first glimpse, turns out to be artificial (Mori, 1970). When a person thinks that he or she touches a real hand but it turns out to be a prosthetic, this can create a feeling of eeriness. Subsequently, when the robot appears even more humanlike, the familiarity of the robot increases again, achieving its highest level of affinity (Mori, 1970). When a robot looks identical to a human being, the robot receives as much affinity as an ordinary human (Geller, 2008).
The uncanny valley may have been inspired by Freud’s paper “The “Uncanny”1”
(Geller, 2008), where Freud (1919) mentioned that people experience an uncanny feeling especially when encountered with a dead body. Mori (1970) also mentioned a corpse as an origin of danger that causes an eerie feeling. Moreover, according to Freud (1919) the
uncanny effect does not derive from a feeling of unfamiliarity. Contrary, the uncanny feelings of Freud descend from something familiar yet strange (Freud, 1919; MacDorman and
Ishiguro, 2006).
Figure 1. The uncanny valley
Eeriness
Eeriness is a negative and unpleasant feeling that people experience when
encountering robots that appear too much as a human (Flach et al., 2012; Mara and Appel, 2014; Mori, 1970). It is a part of the uncanny valley of Mori (1970) and is associated with a mismatch between what humans expect to encounter, and the robot’s actual behavior (MacDorman, 2011). The realization of this mismatch is followed by loss of affinity and a rising feeling of eeriness (Mori, 1970). When a robothas anthropomorphic features butis distinctly not human,this robot does not elicit eerie feelings (Geller, 2008). However, when a robot’s appearance comes to close to a human, the robot is seen as less warm and creepy (Van Doorn et al., 2017). Eeriness is characterized as a sense of unease (Kätsyri et al., 2015) and this feeling is mostly subconscious (MacDorman and Ishiguro, 2006). The eerie feeling resides from the instinct to protect oneself against proximal sources of danger, such as a corps (Mori,1970). It is a defensive response towards something uncanny (Clayton and Leshner, 2015).
might seem normal, but the pace of movement can influence this perception. Thompson et al. (2011) discovered that when avatars move in a humanlike way, they are seen as less eerie and more familiar. This shows that the movement of a non-human object can alter if the object is seen as more or less eerie. A humanlike pace will decrease the feeling of eeriness. Robots are increasingly capable of moving more humanlike. According to Retto (2017), Robot Sophia sets an example of how clumsy movements of robots have come to an end.
Even though the model of Mori is widely used, Hanson-the designer of robot Sophia-is no supporter of the uncanny valley theory. Hanson states that Mori’s uncanny valley model is merely a speculation and not scientifically supported (Geller, 2008). Nonetheless, the model of Mori is widely adopted and tested (Geller, 2008). Semaya and Nagayama (2007) tested the model by Mori (1970) and found supporting evidence that eeriness does arise in the uncanny valley. Nevertheless, they found these results only under the condition that the faces contained irregular features. This is in contradiction with the findings of Geller (2008). Geller (2008) argued that a way to stay out of the uncanny valley and the eerie feeling is by deforming characters. When deforming the characters, this reframes them from being too humanlike (Geller, 2008) and therefore the feeling of eeriness does not arise.
2.4Whimsical cuteness
There might be other ways to mitigate eeriness, apart from deforming them. It is found that non-human objects are designed to be cute in order to attract people (Pajares Tosca, 2009). Moreover, whimsical cute objects have a playful, carefree nature (Nenkov and Scott, 2014), where eeriness has a troubled nature. A whimsically cute robot might therefore serve as a means to mitigate feelings of eeriness.
Cuteness is a visual concept that can be the case for humans, animals as well as for non-human objects (Morreall, 1991). It can be a choice of design to make things cute (Pajares Tosca, 2009). Cuteness is a term that finds its basis in babies, but can be extended to
inanimate objects (Morreall, 1991). Examples of non-human objects that can be cute are cartoon characters or stuffed animals. Morreall (1991) suggests seven babyish features that create cuteness: a large head in relation to the body, a large forehead with eyes set lower than adults, round cheeks, a rounded body, soft body surfaces with thick extremities and a behavior that indicates weakness and clumsiness. These features can be adjusted to fit inanimate
so-called kawaii products are cute because they are small, soft and have pastel colors in their appearance (Hellén and Sääksjärvi, 2013). An example of a classic kawaii character is Hello Kitty (Nittono, 2016).Cute objects containing biomorphic or even anthropomorphic forms in their appearance (DiSalvo and Gemperle, 2003; Marcus, 2002). Anthropomorphic objects are non-living objects that display human-like qualities (DiSalvo and Gemperle, 2003). Since humanoid robots also contain anthropomorphic forms, these cuteness characteristics are transferable to humanoid robots.
Cuteness is not just a term or concept used for children and their toys. Adults can also be targeted by cute things since cute childlike objects are kitsch, which is currently in style (Pajares Tosca, 2009). Some cute concepts are intended for children only (for example Teletubbies), some are meant only for adults (such as Snoopy), and some are able to reach both children and adults (Pajares Tosca, 2009). Cute objects are able to stimulate affectionate and nurturing behavior (Morreall, 1991; Nenkov and Scott, 2014). Cute objects can generate also empathy (Pajares Tosca, 2009). We can identifytwo different types of cuteness:
2003).Agarwal and Karahanna (2000) found that playfulness can indirectly influence intention to use.
2.5 Hedonic versus utilitarian service setting
The playful nature of whimsical cuteness can also be found in the hedonic service setting,in which people can encounter a robot. This playful nature of the service setting might influence people’s intention to use and therefore acceptance of the robot (Agarwal and
Karahanna, 2000; Heerink et al., 2008; Heerink et al., 2010). Service encounters vary from one step services to multiple sequence services (Strombeck and Wakefield, 2008). The duration that a customer is present in the service setting differs from short to longer visits (Strombeck and Wakefield, 2008). When looking at the service setting, we can make a distinction between two types of service settings: the hedonic and the utilitarian service setting. The utilitarian service setting is associated with task orientation and efficiency (Kang and Park-Poaps, 2010; Kim, 2006; Parker and Wang, 2016), whereas the hedonic service setting focuses on playfulness, fun, arousal and joy (Kang and Park-Poaps, 2010; Kim, 2006; Van der Heijden, 2006). Examples of hedonic service settings are massage services, sporting events or amusement parks (Lui, Mattila and Bolton, 2018; Stafford, Stafford and Day, 2002; Strombeck and Wakefield, 2008). The utilitarian service setting focuses on goal directed activities, such as going to the dentist, the bank or health services (Rychalski and Hudson, 2016; Stafford et al., 2002).
The utilitarian service setting is a task-oriented setting (Rychalski and Hudson, 2016). It is a setting in which people often perform a more mundane task. Customers are trying to solve a functional need (Rychalski and Hudson, 2016). In the utilitarian service setting, customers look at the productivity by which they were helped (Moore, Moore and Capella, 2005).The interactions that people obtain in this service setting are not as important as the outcome of the process (Hightower, Brady and Baker, 2002). In the utilitarian service setting there is an external goal, such as increasing productivity or performance (Van der Heijden, 2004). It is a means to an end. Moreover, people in the utilitarian service setting tend to be less emotionally involved than in the hedonic service setting (Hill, Blodgett, Baer and Wakefield, 2004).
The hedonic setting does not work towards an external goal; it is not a means to an end (Van der Heijden, 2004). The interaction is the goal itself. Its value is determined by the fun a person gains from the service (Van der Heijden, 2006).For example, when going to the hairdresser, some people find the interaction with the hairdressers and customers equally as important as the result they gain from their appointment (Hightower et al., 2002). People in the hedonic service setting are often more emotionally involved, since they are more interested in these type of activities (Hill et al., 2004). The hedonic service setting is becoming more and more competitive, which demands firms to raise their game and offer additional value to the customers (Hightower et al., 2002). For example, movie theaters often offer additional 3D services to enhance their service and to differentiate themselves from competitors. Organizations might also adopt a robot in order to gain a competitive advantage. However, the robots need to elicit positive reactions in order to offer additional value.
2.6 Conceptual model
In order to answer the proposed research questions: “To what extent does whimsically cuteness affect eeriness and robot acceptance? And to what extent does the effect of the robot type on eeriness differ between the hedonic and utilitarian service setting?” the following conceptual model is used (see figure 2). The model shows the relationship between a non-cute versus a whimsically cute robot (IV) on robot acceptance (DV). There are concerns that people do not accept robots (MacDorman and Ishiguro, 2006) and research regarding robot acceptance is scarce (Beer et al., 2011).A possible reason why robots are not yet fully accepted might be because robots that appear too humanlike arouse an eerie feeling (Mori, 1970). Therefore, in my conceptual model it is suggest that the relationship between the type of robot and robot acceptance is fully mediated by eeriness (Me). The expectation is that a
whimsically cute robot will have a stronger negative effect on eeriness than a non-cute robot. There is evidence to expect this because a whimsically cute robot has a carefree and playful nature, whilst eeriness is associated with a careful self-protection motive (Mori, 1970; Nenkov and Scott, 2014).Secondly, it is hypothesized that the relationship between the robot type and eeriness is moderated by the type of service setting (Md), where the type of service
by which a task is performed (Moore et al., 2005). Therefore,it is expected that the hedonic service setting will have a stronger effect on the negative relationship between the type of robot and eeriness than the utilitarian service setting. Lastly,it is expected that eeriness has a negative effect on robot acceptance. Robot acceptance is measured by intention to use and attitude. Intention to use has a direct effect on usage, where usage indicates the acceptance of the robot (Heerink et al., 2008; Heerink et al., 2010).Therefore, intention to use is determined to be a measure of robot acceptance. Attitude is a favorable or unfavorable judgment of an object (Ajzen, 1991) that influences the acceptance of a robot (Nomura et al., 2006). Therefore, attitude is also used as a measure of robot acceptance. A negative relationshipis expected between eeriness and robot acceptance, because eeriness is an unpleasant feeling and the intention to use a robot increases when this usage is enjoyable, not unpleasant (Shiau and Chau, 2016). Additionally, this relationship is expected because a robot that is perceived as eerie is likely viewed as unfavorable, where an unfavorable view will lower attitude (Ajzen, 1991).
Figure 2. Conceptual model
2.7 Hypotheses
People can experience eeriness, a negative and unpleasant feeling, when encountering a robot that appears too much as a human (Mori, 1970). This feeling is uncomfortable,
Pepper are described as cute as well (Lee, Kang, Kim, Lee and Kwak, 2016). Products are often designed to be cute with the belief that this will results in a more favorable attitude towards the product(Lee, Chang, Chen, Huang, 2018), where humanoid robots can also be designed to be cute. When the cute appearance of a robot is combined with human-robot interaction, the robot can appeal to people (Cheok and Fernando, 2011). An example can be a cute robot informing and accompanying customers in a store. Whimsically cute objects often have a playful and carefree nature (Nenkov and Scott, 2014) which might decrease the eerie feelings that people experience when encountering a robot. My expectation is that a
whimsically cute robot will remind people of a toyrobot, which will show them that the robot will cause no danger. It is found that cute robots appear inviting and innocent (Morreall, 1991). When people see a robot that is innocent, this will confirm that the robot will not cause a threat. Moreover, an inviting robot will not make people feel uncomfortable and will
therefore not activate an eerie feeling. Hanson (2015) states that when robots appear alive, friendly and cute, people will not perceive them as half-dead and eerie. It is expected that the whimsical cuteness of the robot will activate a carefree nature instead of a self-defense motive, which will result in lower eeriness. A non-cute robot will not influence a person’s perception of his or her environment. Based on this literature, this thesis expects the following:
Hypothesis 1. Eeriness is lower for a cute robot compared to a non-cute robot.
As discussed, robots can be designed to be cute (Pajares Tosca, 2009). Cute non-human objects are able to elicit feelings of empathy and affection among its viewers (Morreall, 1991; Pajares Tosca, 2009). Furthermore, whimsically cute objects elicit associations of fun and playfulness (Nenkov and Scott, 2014). Cute and fun robots suggest that interacting with the robot is a form of entertainment (Lee et al., 2018). According to Van der Heijden (2006), the value that a person derives from the hedonic service setting is the fun that a person gains from the service. The feeling that someone experiences in this setting is more important than performing the task productively (Moore et al., 2005). The hedonic service setting is not focused on an external goal (Van der Heijden, 2004).The interaction and the fun that is experiencedin this setting is the goal itself (Van der Heijden, 2004). A
whimsically cute robot activates feelings of fun (Nenkov and Scott, 2014), where the hedonic service setting also focuses on having fun (Van der Heijden, 2004). The objectives of
whimsical cuteness does not have a vulnerable nature (Nenkov and Scott, 2014).This is in contrast to a feeling or eeriness that arises when people do feel vulnerable, as if they need to defend themselves from danger (Mori, 1970). Therefore, the prediction is that the associations of fun that arise from the cute robot as well as the hedonic service setting, will decrease the perceived danger and therefore the eeriness of the robot. The utilitarian service setting is not associated with fun. This service setting focuses on solving a need or performing a task (Rychalski and Hudson, 2016). The interactions from theservice are not as important as the outcome itself (Hightower et al., 2002) and the process is less emotional (Hill et al., 2004). The utilitarian service setting is not likely to affect the relationship between the type of robot and eeriness as much. Therefore it is expected that the negative relationship between a whimsically cute robot and eeriness is stronger in the hedonic service setting than in the utilitarian service setting. This showcases the second hypothesis:
Hypothesis 2. In the hedonic service setting, a whimsically cute robot has a stronger effect on eeriness than in the utilitarian service setting.
When robots activate an eerie feeling, people might be less willing to accept these robots. Researchers share concerns that humans are not accepting robots yet(MacDorman and Ishiguro, 2006). However, people are increasingly exposed to human-robot interaction, because more robots are still being designed (Hanson, 2006). In consequence, investigating the human response to robots has become increasingly relevant (Mara and Appel, 2014).The acceptance of robots is measured in terms of intention to use the robot and the attitude towards the robot. It is stated that the intention to use increases when this usage is viewed as useful or enjoyable (Shiau and Chau, 2016;Venkatesh and Davis, 2000). This demonstrates that if the human-robot interaction is viewed as either useful or enjoyable, the intention to use the robot will increase. The intention to use directly influences actual usage, which determines the robot acceptance (Heerink et al., 2008; Heerink et al., 2010). Eeriness is an unpleasant feeling that arises from observing an almost completely humanlike robot (Mori, 1970). When the robot is viewed as unnatural and creepy, the robot becomes more eerie (Mori, 1970). These feelings are not enjoyable.It is expected that when a robot is perceived as more eerie, this will result in alower enjoyment. When the perceived enjoyment of interacting with a robot decreases, the intention to use the robot will decrease as well (Shiau and Chau, 2016).
robot is measured by the favorable or unfavorable opinion that a person has about this robot (Ajzen, 1991).It is established that people favor a robot that is not too humanlike (Broadbent et al., 2009), which is in line with Mori’s (1970) statements about the uncanny valley. When people witness a robot that is too humanlike, the robot becomes more eerie (Mori, 1970). When people experience a robot as more eerie, they are likely to acquire an unfavorable opinion about the robot. An unfavorable opinion will decrease a person’s attitude towards the robot (Ajzen, 1991). Accordingly, these eerie feelings will result in an lower attitude towards the robot. Therefore, the lower attitude towards the robot will result in a lower robot
acceptance. Hence the last hypothesis is the following:
3. METHODOLOGY
In the previous chapter, the established theoretical findings and an expected conceptual model were examined. In this chapter, the research design and procedures are discussed, followed by the measures used for the analysis.
3.1 Research design and procedures Research design
A quantitative research is performed in the form of an experimental research design. An important advantage of an experimental research design is that is that it can determine causal effects without participants being biased (Imai, Tingley and Yamamoto, 2013). In order to test the hypotheses, a 2 (whimsically cute robot vs. non-whimsically cute robot) × 2 (hedonic service setting vs. utilitarian service setting) between subjects factorial design is used, resulting in four different scenarios. The participants were randomly assigned to the different conditions, with all elements presented evenly.A scenario-based online questionnaire was developed with Qualtrics. The questionnaire included the four different conditions, as well as items to test all the variables of my conceptual model. Multiple control variables are added, such as age, gender, technology readiness, attitude towards the service provider and repurchase intentions. The questionnaire was distributed by the research platform Prolific.
Procedures
The data collection started with a pre-test via an online anonymous Qualtrics survey to measure the manipulation of the type of robot. This analysis determined which robot would be used as the cute robot and which robot would be used as the non-cute robot in further
analyses. Next, data was collected with a Qualtrics survey through Prolific. Participants were asked to fill in the survey that could be made either on a computer, laptop or mobile device. The survey took 8 minutes to fill in.
The study began with an introductory page that shortly explains the research objective of the study. The introductory page gives the instructions and procedures, ensures the
participant’s anonymity and declares that the participants will receive a monetary
enjoy. In the utilitarian service setting, the participants visit a physiotherapist to get a massage in order to solve back pain. The manipulations can be found in Appendix A. Subsequently, a manipulation check tests whether or not these service setting is perceived as intended. Next, the participant is exposed to either the cute or the non-cute robot as a service agent in this setting (see figure 3). The robot welcomes the participant, asks for the participant’s name and time of appointment, after which the robot accompanies the participant to the reserved room. After being exposed to the robot, the participant is asked to state his or her opinion concerning statements that measure robot acceptance and other possible influencing factors, such as technology readiness and attitude towards the service provider. Moreover, an attention check is incorporated in the test to ensure attentive participation. The attention check contained the following statement: “To show that I am paying attention, I will click on Extremely”. The questionnaire concludes with questions about the participant’s age, gender and education level. The participant is thanked for his or her participation and can exit the survey.
The robotic service agent that is used in this experiment is the humanoid robot Pepper. This robot is 1.2 meters high and has a tablet attached to itself. Pepper is a service robot that is able to move itself and talk with humans via voice or via its tablet. According to its designers at Softbank Robotics (2017), Pepper is a playful character. There is receptionist software designed for Pepper, which means that the robot can welcome visitors, ask for their name and notify the meeting organizer about the arrival of the visitor. These attributes make Pepper a suitable receptionist robot (Pepper the receptionist, 2018; Perera, Pereira, Connell and Veloso, 2017). There is a microphone, loudspeakers, touch sensors and a camera at his forehead (Perera et al., 2017). There are also touch sensors on Pepper’s hand but we will not make use of those functions.
3.2 Measures
The measurement items of the variables are found in related studies and are slightly altered to match the robot setting. Allitems can be found in Appendix A.
Manipulation of service setting: hedonic- versus utilitarianservice setting
To check whether the manipulation of the service setting worked, a ten item scale by Diamantopoulus, Sarstedt, Fuchs, Wilczynski and Kaiser (2012) is used. The measure of Diamantopoulus et al. (2012) is measured on a seven-point Bipolar scale, with an example being: “This robot is… dull/exciting”. The reliability analysis showed that Cronbach’s alpha is .68. All ten items are combined into one new variable to measure the hedonic versus the utilitarian service setting. A high score implies a more hedonic service setting and a lower score implies a more utilitarian service setting.
Whimsical cuteness
the robot more fun, playful or whimsical, they see the robot as more whimsically cute and vice versa.
General cuteness
General cuteness was tested by means of three items by Nenkov and Scott (2014). The items are rated on a seven-point Likert scale ranging from 1 (strongly disagree) to seven (strongly agree). An example item is: “This robot is cute”. A correlation analysis showed that C1 and C2 correlate significantly (r = .65, p = .00), as well as C1 with C3 (r = .78, p = .00) and C2 and C3 (r = .63, p = .00). Reliability analysis showed a Cronbach’s alpha of .87. All items are combined into one new variable.
Kindchenschema cuteness
Kindchenschema cuteness is measured by three items by Nenkov and Scott (2014). The items are rated on a seven-point Likert scale ranging from 1 (strongly disagree) to seven (strongly agree). An example item is: “This robot is vulnerable”. A correlation analysis shows that all significantly correlate with each other, with KC1 and KC2 (r=.20, p=.01), KC1 and KC3 (r=.32, p=.00) and KC2 and KC3 (r=.50, p=.00). Cronbach’s alpha is .61 and therefore all three items are combined into one new variable.
Eeriness
Eeriness is measuredby eight items, with an example question being: “This robot is unnatural”. The items are rated on a seven-point Likert scale, ranging from 1 (strongly disagree) to seven (strongly agree). Two of the items (“the robot is comforting”, and “the robot is relaxing”) are reversed in order to match the other items. A correlation analysis shows that all eight items correlate with each other with a significance level of .00. The Pearson correlations range from .36 to .77. Cronbach’s alpha is .89 and therefore all eight items are combined into one new variable.
Robot acceptance
Robot acceptance is measured by both intention to use and attitude towards the robot. Intention to use is measured on a seven-point Likert scale ranging from 1 (strongly disagree) to seven (strongly agree) based on the research of Heerink et al. (2008). The concept is
visit”. A correlation analysis shows that ITU1 and ITU2 show significance correlation (r = .87, p = .00), as well as ITU1 and ITU3 (r = .95, p= .00) and ITU2 and ITU3 (r = .88, p= .00). Cronbach’s alpha is .96 and therefore all three items are combined into one new variable. The attitude towards the robot is measured by a scale developed by Spears and Singh
(2004).Spears and Sing (2004) offer a five item Bipolar scale, where an example item is: “The robot is unappealing/appealing”. A correlation analysis shows that all items significantly correlate with each other on a significance level of .00. The Pearson correlations range from .69 to .81. The measure of Spears and Singh (2004) has a Cronbach’s alpha of .95. The items of this measure are combined into one new variable that measures the attitude.
Control variables
3.3 Plan of analysis
All responses from the survey on Qualtrics are saved and downloaded into the program SPSS statistics. The dataset is checked, cleaned, and a few items were reversed inorder. To ensure internal validity of the variables, the Cronbach’s alpha is examined. Subsequently, items were combined into new variables. After preparing the dataset,
4. RESULTS
In the previous chapter the sample, procedures and measures were discussed. In the current chapter, the results are presented. First, the descriptive statistics will be presented, after which the analyses and the testingof the hypotheses is explained. Lastly, the significant findings from the covariates are demonstrated.
4.1 Descriptive statistics
The original dataset consisted of 200 respondents. In order to make the data more reliable, two respondents were excluded from the sample: one participant did not give consent and the other participant did not pass the attention check. All remaining 198 respondents gave their consent, passed the attention check and finished the survey.The average participant was 34.78 years old (SD = 12.17), with a minimum age of 18 and a maximum age of 72. Of all participants, 59.6 percent is female and 40.4 percent is male. The majority (62.1 percent) had university as highest education level, whereas 30.3 percent noted secondary education as highest education level. A small percentage of 2.5 percent has primary education as highest education level and 5.1 percent had another education level. Other education levels include military and trade and Masters.
The means, standard deviations and intercorrelations of all study variables are presented in Table 1. There are multiple important significant correlations. Firstly, the correlation table shows that whimsically cuteness and eeriness share a significant moderate negative correlation. This means that when a robot is perceived as more whimsically cute, eeriness decreases and vice versa. The next significant findings are the moderate positive correlation between attitude and whimsically cuteness, and the strong negative correlation between attitude and eeriness. This indicates that when the attitude towards the robot is high, the robot is also perceived as more whimsically cute and vice versa. Lastly, intention to use shows significant correlations with whimsically cuteness, eeriness and attitude. The moderate positive correlation between intention to use and whimsically cuteness indicates that when a robot is perceived as more whimsically cute, the intention to use will also be higher and vice versa. The moderate negative correlation between intention to use and eeriness shows that when a robot is perceived as more eerie, the intention to use is lower and vice versa.
Control variables
Table 1. Descriptive statistics and correlations Variable Mean SD 1 2 3 4 5 6 7 8 1. Age 34.73 12.17 2. Education level 3.37 1.05 -.05 3. Technology readiness 4.54 .89 -.25** -.01 4. Attitude towards service provider 5.12 1.38 -.03 -.05 .19** 5. Repurchase intentions 4.94 1.60 -.06 -.06 .22** .90** 6. Whimsical cuteness 2.67 .90 -.04 .10 .10 .45** .43** 7. Eeriness 3.91 1.23 -.01 -.01 -.26** -.57** -.59** -.41** 8. Attitude 4.83 1.38 .00 .02 .16* .73** .70** .53** -.78** 9. Intention to use 4.34 1.55 -.12 -.01 .30** .80** .85** .43** -.59** .61** N = 198, *p < 0.05, **p < 0.01 4.2 Analyses
Manipulation of robot type: cute versus non-cute
A pre-test was executed to ensure the manipulation of the cute and non-cute robot. In order to manipulate the robot to be whimsically cute and non-cute, the robot Pepper was manipulated by using Photoshop. To make the robot whimsically cute, the head and eyes of the robot are enlarged. To make the robot non-cute, the eyes of the robot are reduced in order to make the proportions of the face more in line with those of a human. In the pre-test, participants were exposed to the original robot, the manipulated whimsically cute robot or to the manipulated non-cute robot. The tablet that is attached to the robot did not display any content. Based on the measures by Nenkov and Scott (2014), participants were asked to rate the robot in terms of three types of cuteness: general cuteness, whimsical cuteness,
The one-way ANOVA was not significant for whimsical cuteness, F(2,24) = 1.57, p = .23. Whimsical cuteness scored higher for the non-cute robot with a mean of 3.24 (SD = .53) than for the cute-robot with a mean of 2.92(SD = .68) and the original robot with a mean of 2.69 (SD = .67).This indicates that the type of robot does not influence whether or not people perceive the robot as more whimsically cute or not. The robots were also tested on general cuteness and kindchenschema cuteness. The one-way ANOVA was not significant for general cuteness, F(2, 24) = .13, p = .88. General cuteness scored the highest for the cute robot with a mean of 3.00 (SD = .71) where the non-cute robot scored a mean of 2.90 (SD = 1.46) and the original robot scored a mean of 1.78 (SD = .81). The one-way ANOVA was also not
significant for kindchenschema cuteness, F(2, 24) = 2.54, p = .10. Kindchenschema scored higher for the non-cute robot with a mean of 3.14 (SD = 1.02) than for the cute robot with a mean of 2.25 (SD = .96) and the original robot with a mean of 2.56 (SD = .41). The non-cute robot scored the highest on both whimsical cuteness and kindchenschema cuteness, which is contradicting with the research by Nenkov and Scott (2014). They indicated that a
whimsically cute robot does not have a vulnerable nature, whilst kindchenschema cuteness does have a vulnerable nature.All findings were not significant. This can be explained by the small sample. Moreover, there were a few comments on the difficulty of the terms, perhaps because the sample consisted of Dutch natives whilst the survey was English. If participants do not fully understand the terms, this might decrease the validity of the research. No items showed significance, but the terms that deemed the most clear were “cute” and “adorable”, therefore it was determined to look at the outcome of this variable; cuteness. Since the cute robot scores the highest on cuteness, this robot is taken into further analyses as the cute robot. The original and the non-cute robot both scored showed a lower score on cuteness. Because the cute robot and the non-cute robot were both manipulated, these two robots were taken into further analyses.
In order to analyze if the manipulation for the cute robot and the non-cute robot was successful in the questionnaire, a one-way ANOVA is performed with the type of robot as factor and the whimsical cuteness variable as dependent list. General cuteness and
perceive the robot as more whimsically cute or not. The one-way ANOVA was also not significant for general cuteness, F(1, 196) = 1.37, p = .24 where the cute robot has a mean of 2.78 (SD = 1.01) and the non-cute robot has a mean of 2.61 (SD = 1.05). The one-way
ANOVA was not significant for kindchenschema cuteness F(1, 196) = .10, p = .75, where the non-cute robot scored slightly higher with a mean of 2.23 (SD = .84) than the cute robot with a mean of 2.19 (SD = .76). Since all results were insignificant, we cannot confirm that the type of robot significantly influences whether or not people perceive the robot as more cute or not. Therefore we can conclude that the manipulation of the type of robot is unsuccessful.
Manipulation of service setting: utilitarian- versus hedonic service
In order to analyze if the manipulation for the hedonic and utilitarian service setting was successful, a one-way ANOVA is performed with the service settings as factor and the manipulation measure by Diamantapolous et al. (2012) as dependent list. The one-way ANOVA was significant, F(1, 196) = 25.97, p = .00. The hedonic service setting has a mean of 3.94 (SD = .67). The utilitarian service setting has a mean of 3.43 (SD = .76). This
indicates that the hedonic service setting is perceived as more hedonic and the utilitarian service setting as more utilitarian. A means plot also shows that the hedonic service setting is perceived as more hedonic and the utilitarian service setting is perceived as more utilitarian This demonstrates that the type of service setting influences if people perceive the service setting as either hedonic or utilitarian. Therefore we can conclude that the manipulation of the service setting is successful.
4.3 Hypotheses testing
Hypothesis 1. Eeriness is lower for a cute robot compared to a non-cute robot.
.24) and a kurtosis of -.54 (SE = .48) for the cute robot and a skewness of -.28 (SE = .24) and a kurtosis of -.43 (SE = .48) for the non-cute robot. The normal Q-Q plots (see Appendix B) and box plots confirmed this. Moreover, there were no outliers. A Levene’s test was executed to verify equality of variances in the sample, which was successful (p = .77).
In order to compare the effect of the type of robot on eeriness,I performed a one-way ANOVA with the type of robot as factor and eeriness as dependent list. The analysis of variance displayed that participants rated the cute robot (M = 3.78, SD = 1.19) lower than the non-cute robot (M = 4.04, SD = 1.26), as can be seen in figure 4. The one-way ANOVA of the type of robot on eeriness was not significant, F(1, 196) = 2.18, p = .14.This means that the effect of a cute robot on eeriness does not significantly differ from the effect of a non-cute robot on eeriness. Therefore, I cannot confirm that eeriness is lower for a cute robot than for a non-cute robot. My first hypothesis is not supported.
Figure 4. Eeriness ratings for the cute and non-cute robot
Hypothesis 2. In the hedonic service setting, a whimsically cute robot has a stronger effect on eeriness than in the utilitarian service setting.
First, the assumption checks are performed. A normality test was executed in order to test if eeriness was normally distributed for the hedonic service setting and the utilitarian service setting.Hypothesis 1 already found that eeriness was normally distributed for both types of robots. A Shapiro-Wilk test showed that eeriness was approximately normally
normal Q-Q plots (see Appendix B) and box plots also confirmed normal distribution of eeriness for both types of service settings. The box plots showed that there were no outliers. Furthermore, the assumption for homogeneity of regression slopes is met (p = .07).
My second hypothesis concerns a moderation effect. To test my second hypothesis, an ANCOVA analysis is performed to determine a statistically significant difference between the type of robot and the type of service setting on eeriness, controlling for age, gender, education level, technology readiness, attitude towards the service provider and repurchase intentions. In this analysis, eeriness is seen as the dependent variable and the type of robot and the type of service setting as fixed factors.The control variables are the covariates. The
ANCOVA analysis (see table 2) showed that there is a significant interaction effect between the type of service setting and the type of robot on eeriness, F(1, 188) = 4.97, p = .03.The model scored an R square of .41 showing that 41 percent of the variance is explained by the model. This shows that the predictor variables do not explain the dependent variable very clearly. The results of the ANCOVA indicate that in the hedonic service setting, the type of robot has a negative effect on eeriness. When the robot is cute, the negative effect between the robot and eeriness is stronger than when there is a non-cute robot.
Table 2. ANCOVA Analysis
Source Type III Sum of
Squares
df Mean Square F Sig. Partial Eta
Squared
Corrected model 121.33 9 13.48 14.40 .00 .41
Intercept 163.10 1 163.10 174.17 .00 .48
Type of robot .89 1 .89 .95 .33 .01
Type of service setting 1.58 1 1.58 1.69 .20 .01
Type of robot * type of service setting 4.65 1 4.65 4.97 .03* .03 Gender .11 1 .11 .11 .74 .00 Age 1.72 1 1.72 1.84 .12 .01 Education level .87 1 .87 .92 .34 .01 Technology readiness 8.07 1 8.07 8.62 .00** .04
Attitude towards service provider 2.68 1 2.68 2.68 .09 .02 Repurchase intentions 7.24 1 7.24 7.73 .01** .04 Error 176.06 188 .94 Total 3330.84 198 Corrected total 297.38 197 R2 = .41 (AdjustedR2 = .38) N = 198, *p < 0.05, **p < 0.01
hedonic service setting of -.06 (SE = .04) of the service setting on the relation between the type of robot on eeriness. Secondly, the bootstrap confidence interval is executed with attitude as dependent variable. The bootstrap confidence interval of the utilitarian service setting did display zero in the range of the LLCI and ULCI (CI = -.12, .35). Therefore we can state with 95 percent certainty that there is no moderating effect of the utilitarian service setting on the relationship between the type of robot and eeriness. The bootstrap confidence interval of the hedonic service setting did not display zero between the LLCI and the ULCI (CI = -.53, -.05). This means that we can state with 95 percent certainty that in the hedonic service setting, there is a moderating effect of -.28 (SE = .12) of the service setting on the relation between the type of robot on eeriness. Since no moderating effect for the utilitarian service setting is found, the effect of the hedonic service setting is stronger. In conclusion, I can confirm my second hypothesis that in the hedonic service setting, a whimsically cute robot has a stronger effect on eeriness than in the utilitarian service setting.
Hypothesis 3. The higher the eeriness, the lower the robot acceptance.
Before performing a regression, the assumption checks are performed. A normality test was executed in order to test if the residuals were normally distributed. Therefore a normality test was performed on eeriness, intention to use and attitude.A Shapiro-Wilk test showed that eeriness was approximately normally distributed (p = .24),with a skewness of -.06 (SE = .17) and a kurtosis of -.58 (SE = .34). The normal Q-Q plot and box plot also confirmed normal distribution of eeriness. The box plot showed that there were no outliers. For intention to use, a normal P-P plot indicated that the residuals were normally distributed. The P-P plot for attitude also indicated that the residuals were normally distributed. Both P-P plots can be found in appendix B. Next, homoscedasticity is checked for both intention to use and attitude. No clear patterns can be distinguished and the points are distributed fairly equal in both scatter plots, where the scatterplot on attitude is slightly less fairly distributed. The assumption of homoscedasticity is met for both intention to use and attitude. Box plots for intention to use and attitude showed that there were no outliers.
has a significant negative effect on intention to use (B = -.14, p = .02). An increase in eeriness will result in a decrease in intention to use. Furthermore, a regression analysis is performed on attitude. The regression analysis shows eeriness has a significant negative effect on attitude (B = -.62, p = .00). This indicates that an increase in eeriness will result in a decrease in attitude towards the robot. Since eeriness has a significant negative effect on both measures of robot acceptance, we can support hypothesis 3.
Additionally, the regression analysis by PROCESS Macro showed that the direct effect from the type of robot on intention to use was not significant (B = -.19, p = .29). The direct effect from the type of robot on attitude was not significant as well (B = -.11, p = .39). These findings suggest a full mediation.
4.4 Effects of the control variables
5. DISCUSSION
In the previous chapter, the analyses and results are presented. In the current chapter, the findings are discussed, as well as the managerial implications and limitations of this research. Moreover, suggestions for future research directions are given.
5.1 Overview of results
The main goal of this research was to investigate the possible relationship between a whimsically cute or non-cute robot and robot acceptance, mediated by eeriness. Moreover, I expected the relationship between the independent variable and mediator to be moderated by the type of service setting. The research questions I hoped to answer were:
“To what extent does whimsically cuteness affecteeriness and robot acceptance? And to what extent does the effect of the robot type on eeriness differ between the hedonic and utilitarian service setting?”
Table 3. Overview of results
Hypothesis Results
H1 Eeriness is lower for a cute robot compared to a non-cute robot Not supported
H2 In hedonic service settings, a whimsically cute robot has a stronger effect on
eeriness than in utilitarian service settings
Supported
H3 The higher the eeriness, the lower the robot acceptance Supported
Additional results
Technology readiness has a negative effect on eeriness Supported Repurchase intentions has a negative effect on eeriness Supported Repurchase intentions has a positive effect on intention to use Supported Attitude towards the service provider has a positive effect on intention to use Supported Attitude towards the service provider has a positive effect on attitude Supported
5.2 The effect of the type of robot on eeriness
Based on the literature review, it was expected that eeriness is lower for a cute robot, compared to a non-cute robot (Cheok and Fernando, 2011; Mori, 1970; Nenkov and Scott, 2014). However, this study was unable to find evidence to support for this hypothesis. No significant relationship was found between the type of robot and eeriness. This means that people do not perceive a robot as more or less eerie when this robot is whimsically cute or non-cute. This outcome might be explained by the fact that the manipulation of the
whimsically cute versus non-cute robot was unsuccessful. Since the type of robot did not significantly influence whether the participants viewed the robot as more or less cute, it cannot be expected that their effect on eeriness will differ significantly.
people of death and therefore decrease eeriness. The current research used two manipulated robots thatboth possessed a slight smile. This might make the robots appear
friendly.According to Hanson (2006), if both robots in this study were perceived as friendly,they are consequently not be perceived as eerie.
5.3 The moderating role of the type of service setting
Based on the theory I expected that the type of service setting would moderate the relationship between the type of robot and eeriness.I argued that in the hedonic service setting, the effect of the type of robot on eeriness is stronger than in the utilitarian service setting (Hightower et al., 2002; Moore et al., 2005; Mori, 1970; Nenkov and Scott, 2014; Rychalski and Hudson, 2016; Van der Heijden, 2006). The results of my analyses confirmed this expectation. I found a significant moderating effect from the hedonic service setting on the relationship between the cute robot and eeriness. I did not find a significant moderating effect for the utilitarian service setting.
The reason why I found the moderating role of the hedonic service setting might be explained by the playful nature that the hedonic service setting and whimsical cuteness have in common (Nenkov and Scott, 2014; Van der Heijden, 2006). The hedonic service setting focuses on the fun that one experiences in the service (Van der Heijden, 2006). Whimsically cuteness activates primed images of fun (Nenkov and Scott, 2014).It is possible that the hedonic service setting stimulates the images of fun that are activated by the cute robot, which makes the effect of the robot on eeriness stronger.
setting is the only service setting to serve as a moderator between the relationship between the type of robot and eeriness.
5.4 The effect of eeriness on robot acceptance
Based on the literature review, it was my expectation that eeriness would have a negative effect on robot acceptance (Ajzen, 1991; Broadbent et al., 2009; Mori, 1970; Shiau and Chau, 2016; Venkatesh and Davis, 2000). The acceptance of the robot is determined by the attitude towards the robot (Ajzen, 1991; Kim et al., 2015) and the intention to use the robot (Heerink et al., 2008; Heerink et al., 2010). The results of this study found sufficient evidence to support my last hypothesis. This indicates that if a robot is perceived as more eerie, the robot acceptance will be lower and vice versa. Furthermore, the results indicate that higher eeriness will result in a lower attitude towards the robot, as well as a lower intention to use the robot. A robot that is perceived as eerie will be avoided if possible, since people do not intend to use it. A robot that elicits eerie feelings will decrease a person’s attitude towards the robot, which means thathis or her robot acceptance decreases. An uncomfortable, eerie experience with one robot can therefore cause a lower acceptance of robots in general. These findings highlight the importance of eeriness. Because social robots are increasingly
interacting with humans in different service settings (Sabanovic et al., 2013; Pepper the receptionist, 2018; Triebel et al., 2016), it is important that humans accept these robots. Social service robots are designed to assist and guide customers (Triebel et al., 2016), but if people perceive them as eerie it is unlike that they will accept them and interact with these robots. Eeriness is therefore a crucial predictor of robot acceptance.
5.5 Additional results
This research found five additional results that all regardthe control variables. It is found that technology readiness as well as repurchase intentions negatively influence eeriness. When people have negative attitudes towards technology, this can bring up negative thoughts regarding the situation (Lazar, Demiris and Thompson, 2016). It is understandable that someone who is not ready for technology will not feel comfortable around a robot.
service, they will rate the robot as less eerie. This is not a surprising outcome. Eeriness is an uncomfortable feeling, which is not likely something that people will come back for. When people do come back for a service, we can assume that they were satisfied with the service and did not experience an uncomfortable, eerie feeling.
Furthermore, this research discovered that repurchase intentions as well as the attitude towards the service provider positively influence the intention to use the robot. When
someone indicates that he or she wants to repurchase a product or service, it is likely that they intend to use this product or service. Moreover, when someone has a positive attitude towards the service provider, this positively influences their intention to use the robot. It is plausible that if someone has a positive attitude towards a service setting, he or shewill be more open to makeuse of its experiences, such as a robot.
Lastly, it is found that the attitude towards the service provider has a positive effect on attitude. The attitude towards the service setting is determined by an evaluation of the
complete service experience, including the robot. The attitude towards the robot measures the evaluation of the robot. Hence, it is reasonable to assume that a positive attitude towards the service provider will also lead to a positive attitude towards the robot. Furthermore, Kim et al. (2015) discovered that attitude has a positive effect on user acceptance. Therefore we can expect that a positive attitude towards the service setting will result in a higher robot acceptance.
5.6Managerial implications
team that specializes in making humanoid- and cute robots that are clearly distinct from human beings.
Furthermore, this study discovered that the hedonic service setting strengthens the negative relationship between the cute robot and eeriness. This finding reveals that
organizations in the hedonic service setting have an advantage over firms in the utilitarian service setting, when it comes to counteracting eerie feelings. Hedonic service settings with robots are encouraged to emphasize the hedonic aspects of their service, such as fun and excitement. When hedonic service settings make use of the robot Pepper, this robot welcomes people and asks for their name. After notifying the meeting organizer about the visitor’s arrival, Pepper can ask the visitor if he or she would like to chat or see him dance (Pepper the receptionist, 2018; Softbank robotics, 2017). Pepper is able to entertain its visitors (Chang, 2017). This could emphasize the hedonic nature of the service setting.
The control variables are found to influence eeriness, intention to use and attitude. Managerial implications can be uncovered regarding the attitude towards the service provider. It is found that attitude towards the service provider has a positive effect on intention to use as well as attitude. Therefore it is a suitable predictor of robot acceptance. Managers can attempt to improve people’s attitude towards the service setting in order to increase the attitude towards and intention to use the robot. For example, managers can ask customers to rate their experience with the service setting and ask for improvements. When the attitude towards the service setting increases, this will also increase robot acceptance.
5.7 Limitations and future research directions
This research is not without limitations. It is important to take these limitations into account when examining the results of this study. Firstly, the manipulation of the type of robot was not successful. This indicates that the type of robot (cute or non-cute) did not significantly influence whether people perceived the robot as more or less whimsically cute. Therefore, we cannot expect to find a significant relationship between the type of robot and eeriness. More research regarding robot aesthetics is needed in order to advance robot appearances (Hanson, 2006). Future research could test the relationship between a
