BEHAVIOURAL INTENTION TO REUSE
DIFFERENT TYPES OF CHECKOUTS
The Implications of Social Cues, Hedonic
Shopping Motivation, Eeriness and Habits
II
BEHAVIOURAL INTENTION TO REUSE DIFFERENT TYPES OF CHECKOUTS The Implications of Social Cues, Hedonic Shopping Motivation, Eeriness, and Habits
Master Thesis Marketing Management University of Groningen Faculty of Economics and Business
Department of Marketing
Author 1st Supervisor 2nd Supervisor
R.H. Veldkamp
(s2391422) Dr. J. Van Doorn A. Schumacher, M.Sc.
r.h.veldkamp@student.rug.nl j.van.doorn@rug.nl a.schumacher@rug.nl
Korte Nieuwstraat 4a Nettelbosje 2 Nettelbosje 2
9724 LC Groningen 9747 AE Groningen 9747 AE Groningen
+316 40074117 +31(0)50 36 33657 +31(0)50 3637065
III
IV
MANAGEMENT SUMMARY
With the surge in self-service technologies, retailers have implemented self-scan checkouts in their stores. This shift benefits retailers due to the reductions in labour costs, as employees are replaced by machines. Simultaneously, benefits are presented to customers when using self-scan checkouts, such as the ease of using the self-scan checkout, time savings, and a higher extent of control over the purchase situation. Nowadays, some self-scan checkouts include social cues, which have implications on the behavioural reuse intention of self-scan checkouts. This study shows that the addition of social cues leads to a higher behavioural reuse intention, in comparison to a normal self-scan or a standard checkout (i.e. a checkout including a cashier who scans the products for the customer). Nonetheless, this does not suggest that retailers should implement social cues to all of their self-scan checkouts, nor replace standard checkouts with these types of checkouts. This becomes evident from the notion that habits and eeriness mediate the reuse intention of self-scan checkouts and standard checkouts, but not for self-scan checkouts including social cues. As habits are formed with respect to a certain type of checkout, the reuse intention for that type of checkout will be higher; removing a checkout to which a customer has formed habits, could have negative consequences (for example, a possibly lower customer satisfaction). Moreover, the usage of technology could lead to an evoked feeling of eeriness, which is not present when using a standard checkout. Therefore, different types of checkouts have different advantages and disadvantages. Different types of checkouts can coexist in-store, but they should be reflected upon as complementary, rather than substitutes of one another.
Moreover, hedonic shopping motivation does not significantly affect the main results. However, no distinction was made between different types of hedonic shopping motivation (e.g. as presented by Arnold & Reynolds, 2003). Hence, despite the insignificant effect of hedonic shopping motivation, the possibility that certain types of hedonic shopping motivation interact with different types of checkouts is not excluded. For example, social and idea shopping could still be relevant aspects of hedonic shopping motivation when deciding between different types of checkouts, as they place importance on the customer’s need for social interaction and novelty-seeking.
Keywords: Types of Checkouts, Behavioural Intention to Reuse, Social Cues, Hedonic Shopping
V PREFACE
As our world continues to become increasingly technology-driven, marketing strategies are adapted accordingly (Fernandes & Pedroso, 2017). This has led retailers to rethink their ways to deliver service to their customers via innovative ways (Weijters, Rangarajan, Falk, & Schillewaert, 2007). One of those ways is the introduction of self-scan checkouts, which are now present in many stores in the Netherlands, such as Ikea, Albert Heijn, and McDonalds.
During my previous study ‘International Economics & Business’ it became apparent to me that technology does not only include a “fun side”, but simultaneously strikes fear in the hearts of those subject to the continuous shifts that are currently present in our own world. The introduction of efficient technologies at the cost of employees, has led to polarization: a process where middle-income jobs disappear to a greater degree than low- and high-middle-income jobs. For example, in the United States this has led many people to lose their jobs, which is documented in the movie ‘Inequality for all’, where the middle class express their dissatisfaction with technology, as they lose their jobs, and subsequently, their saved pension money.
This effect is also evident in the introduction of self-scan checkouts, as they remove employees from the interaction, while providing simultaneously higher cost efficiencies, improvements in service quality, and even attract new customers (Elliott, Meng & Hall, 2012). As polarization gives the impression that utilitarian factors (such as improved efficiency) are the sole reason for implementing technologies, I wanted to see if people could also regard implemented technologies as being “fun” (creating hedonic value).
My supervisor, Dr. J. Van Doorn, has helped me in setting up the current study. Hence, I would like to especially thank her for her continuous support and providing me with the necessary feedback, helping me through every step of the process. Moreover, I would also like to thank my family and friends who have supported me throughout the semester.
VI TABLE OF CONTENTS MANAGEMENT SUMMARY ... IV PREFACE ... V INTRODUCTION ... 1 THEORETICAL FRAMEWORK ... 4 I. Literature Review... 4
II. Research Objective and Conceptual Model ... 16
III. Hypotheses ... 18
METHODOLOGY ... 24
I. Research Design... 24
II. Questionnaire Design ... 25
III. Measurements... 25
IV. Plan of Analysis ... 30
RESULTS ... 31
I. Descriptive Statistics ... 31
II. Manipulation Check ... 32
III. Hypotheses ... 33
DISCUSSION ... 37
CONCLUSION ... 39
LIMITATIONS AND FUTURE RESEARCH ... 40
REFERENCES ... 41
APPENDICES ... 46
Appendix A – Questionnaire ... 46
Appendix B – Factor and Reliability Analyses ... 50
Appendix C – Manipulation Check: Variance of Homogeneity, ANOVA, and Post Hoc Test 57 Appendix D – Variance Inflation Factor Scores ... 59
Appendix E – ANCOVA... 60
1
INTRODUCTION
The substantial change from products to services in many economies’ gross domestic product (GDP) accompanied by the increased acceptance of technological advances, such as the rise of the internet, have had a significant effect on marketing strategies (Oghazi et al., 2012; Fernandes & Pedroso, 2017). More specifically, the acceptance and integration of information and communication technology (ICT) in day-to-day business activities is one of the most remarkable long-term trends in the business world (Rust, 2001; Weijters, Rangarajan, Falk, & Schillewaert, 2007). As a result, retailers have been taking into consideration innovative ways to deliver service to their customers (Bobbitt & Dabholkar, 2001; Dabholkar, Bobbitt, & Lee, 2003; Quinn, 1996; Weijters et al., 2007). In this regard, self-service technologies (henceforth, SSTs) have received global interest (Elliott et al., 2013; Fernandes & Pedroso, 2017).
The worldwide interest in SST can be due to the potential strategic value it presents for retailers (Elliott, Meng, & Hall, 2012). SSTs are formalized as “technologies that customers independently use without any interaction with, or assistance from, employees” (Curran & Meuter, 2005, p.103). From a retailer’s perspective, SSTs do not only remove employees from the interaction, but also provide higher cost efficiencies, improvements in the service quality, and could attract new customers who value interaction without an employee (Parasuraman & Grewal, 2000; Elliott et
al., 2012). Nonetheless, Curran et al. (2005) note that service-providing technologies that replace
employees could impose a significant drain on resources if the technology is not universally accepted by customers.
Universal acceptance of SSTs has proven to be ineffective in the past and could be explained by the customer’s initial attitude towards the technology (Dabholkar, Bobbitt, & Lee, 2003). The rejection of SSTs could possibly be explained by the notion that customers were unwilling to change their behaviour to accept the SSTs (Dabholkar et al., 2003). There are people who regard the encounter at a standard checkout as a ‘social experience’ and, therefore, opt for a standard checkout over a self-scan checkout due to the preference of interacting with other people (Zeithaml & Gilly, 1987; Curran et al., 2005). A standard checkout is formalized here as a checkout including a cashier who scans the products for the customer.
2
Nonetheless, customers have increasingly become technologically aware and comfortable, which has consequently led supermarkets to implement self-scan checkouts once more (Hennessy, 1998; Dabholkar et al., 2003). As a result, some people nowadays prefer the self-scan checkout over a standard checkout due to the perceived ease of usage and/or the absence of interaction with employees (Dabholkar, 1996; Meuter, Ostrom, Roundtree, & Bitner, 2000; Dabholkar & Bagozzi, 2002).
3
technology and, consequently, does not increase the behavioural intention to use a self-scan checkout.
Nonetheless, when considering a customer’s shopping experience, not only the utilitarian factors should be taken into account, but also the hedonic ones, as value towards the shopping experience is generated via both, objective perceptions (utilitarian values: the usefulness of the event) and subjective perceptions (hedonic values: the appreciation of the activities) (Holbrook, 1986; Babin, Darden, & Griffin, 1994). Whereas utilitarian shopping motivation incorporates aspects of achievement and/or disappointment over the ability to finalize the shopping task, the hedonic shopping motivation takes into consideration expressions such as enjoyment, escapism, and excitement during the shopping trip (Babin et al., 1994). Moreover, customers often engage in activities due to the pleasure obtained by doing so, irrespective of the benefits they could potentially gain with respect to saving time and/or money (Bateson, 1985a; Marzocchi & Zammit, 2006). Social cues, although possibly a redundant addition based on the utilitarian shopping motivation, could potentially generate value for the customer via the hedonic shopping motivation, because social cues can be reflected upon as ‘the social characteristics of a situation’, and are observed as the opposite of task-related characteristics of a situation (Hu & Jasper, 2006).
The research question takes into account how the implementation of social cues to self-scan checkouts could influence its behavioural reuse intention, whereas hedonic shopping motivation could possibly influence the effectiveness of social cue implementation. Hence, the research question is formalized as follows:
How does the behavioural intention to reuse a self-scan checkout change when social cues are added? How does this relation depend on hedonic shopping motivation?
4
THEORETICAL FRAMEWORK I. Literature Review
The literature review comprises an overview of the most important papers for this thesis, and are based on previous literature on self-service technologies, drivers for self-scan checkout usage, the importance of customers’ need of interaction when deciding between using different types of checkouts, and the possible effects of social cues on self-scan checkout usage. Table I presents the main papers for the literature review, including their topics and main findings.
Previous Research on Self-Service Technologies Usage
5
where the customer is unable to use the service, and is consequently presented with three different options (Meuter et al., 2000). Firstly, the customer may switch between providers; secondly, the customer may switch back to the interpersonal encounter: this could overload the employees as they now have to compensate for the loss of SST, or increases the overall waiting time for the customer (Meuter et al., 2000). Thirdly, the customer might not use the technology at all. Process failure is based on the capacity that the new SST brings to improving service delivery, but is conditional to the degree that the existing service process is effectively implemented (Meuter et
al., 2000). This suggests that a process failure is instigated when a customer’s expectation of the
SST is not realized, leading to significant complications (Meuter et al., 2000). Lastly, customer-driven failure is based on the notion that customers are prepared to account for some of the blame when a dissatisfying encounter with an SST occurs (Meuter et al., 2000). Customer-driven failure thus suggests that customers who use an SST are aware that the type of result experienced by using the SST is influenced by the interaction with the SST (Meuter et al., 2000). SSTs can be reflected upon as fully advantageous when the technology is designed with the customer needs in mind, and when those needs are integrated in the design of the SST (Meuter et al., 2000).
According to Weijters, Rangarajan, Falk, and Schillewaert (2007), retailers should focus on three main criteria if they want to stimulate the customer’s behavioural intention to use an SST. The first criterion is in line with the finding by Meuter et al. (2000), because it focuses on the perceived benefits of using an SST. The second criterion is based on the notion that SSTs are easy to use and have a reliable performance (e.g. no breakdowns that lead to contingency) (Weijters et al., 2007). The third criterion should be realized if the previous two criteria are met: the fun aspect should aid in shaping behavioural intention towards the self-scan checkout (Weijters et al., 2007). The third criterion is in line with the findings by Fernandes et al. (2017), as they state that ‘fun’ is a less important determinant for SST use: it is less important conditional to whether the utilitarian factors (e.g. perceived usefulness and reliability) are realized.
An additional factor that could possibly have an ambiguous effect on SST adoption is novelty (newness) (Weijters et al., 2007). As newness (an aspect of innovativeness) varies as a function of demographic variables (Robertson & Gatignon, 19991; Im, Bayus, & Mason, 2003; Weijters et
al., 2007), newness can lead to either a positive or negative disposition towards SSTs (Blythe,
6
Demographics are important with respect to SST management, due to the notion that demographics are reflected upon as influencing the response towards marketing strategies (Meyers-Levy, 1988; Meyers-Levy & Sternthal, 1991; Mittal & Kamakura, 2001; Weijters et al., 2007). For example, earlier research suggested that young, wealthy, educated males are more likely to adopt SST usage (Dabholkar et al., 2002).
Dabholkar et al. (2002) looked at how SST’s should be promoted based on four characteristics of the target market: efficacy level, the need for interaction with an employee, self-consciousness, and novelty seeking. If the market has a low level of self-efficacy or a high need for interaction, marketers should promote the ease of use or user-friendliness of the SST (Dabholkar et al., 2002). If the market segment has a low level of inherent novelty seeking (an aspect of innovativeness) or a high level of self-consciousness, there is likely to be a high tendency to be reluctant to using SSTs and the focus of promotion should be based on performance or reliability (Dabholkar et al., 2002). Lastly, their study suggests that the fun accompanied with technology should be promoted if the market is characterized by a high level of inherent novelty seeking, self-efficacy, self-consciousness, or a high need of interacting with service employees. The first two (novelty-seeking and self-efficacy) suggest that people are genuinely inclined to try new types of SSTs, especially if they are reflected upon as being fun (Dabholkar et al., 2002). The latter two (high self-consciousness or a high need of interacting with an employee) describe a natural tendency to be reluctant towards SSTs (Dabholkar et al., 2002). For these segments, it is important to subtly emphasize enjoyment that can be derived by using the SST (Dabholkar et al., 2002).
7
Drivers for Self-Scan Checkout Usage
A self-scan checkout is defined as “a terminal incorporating an optical reader that can scan product barcodes and display information about the purchased items, including price, quantity and type; customers can keep a constant tally of how much they have spent and when they have finished shopping they can pay the amount due without placing the merchandise on the checkout counter since they have already scanned the product barcodes” (Marzocchi et al., 2006, p.655). In the past, when self-scan checkouts were introduced in retail stores for the first time, they were met with aversion: the implementation of self-scan checkouts depicted a classic case of an SST that was not accepted after its initial introduction (Dabholkar et al., 2003).
The aversion could possibly be explained by the customers’ absence of trust and confidence in, both, themselves and the SST to perform correctly (Elliott et al., 2013). Nonetheless, nowadays self-scan checkouts are increasingly popular and cannot solely be found in retail stores anymore, but are also present at airport kiosks, theatres, and even libraries (Elliott et al., 2013). This might be due to the fact that the modern consumer is more informed with respect to technology, and more comfortable with using self-scan checkouts (Elliott et al., 2013). However, the study by Elliott et
al. (2013) suggested that technology readiness has only a small direct effect on the intention to use
a self-scan checkout. More importantly, the perceived reliability and fun of using the self-scan checkout moderately mediates the effect that technology readiness has on using the self-scan checkout (Elliott et al., 2013). This mediating effect suggests that a high level of technology readiness does not inherently suggest a high behavioural intention to use self-scanning options (Elliott et al., 2013). More specifically, their study suggested that the perceived reliability and fun towards the self-scan checkout are more important drivers than technology readiness in shaping behavioural intention to use the self-scan checkout (Elliott et al., 2013).
8
which suggests that technologies have a higher likelihood to be accepted if they are reflected upon as practical and/or efficient (i.e. ease of usage and speed of use), rather than appealing to the senses (i.e. fun) (Mordini, 2007). Contradicting the findings by Fernandes et al. (2007), Marzocchi and Zammit (2006) found that ‘perceived control’ over a self-scan checkout and the fun/enjoyment aspect both have a positive effect on the overall service quality of the SST.
Moreover, the study by Mulaomerovic and Trappey (2013) suggested that even culture could potentially play a role towards the intention to use a self-scan. Their study focused on the Taiwanese culture, and found that the collectivistic culture of Taiwan should be considered by retailers when implementing self-scan checkouts (Mulaomerovic et al., 2013). According to their data, Taiwanese students have a high level of social pressure and, in that respect, it is important for retailers to create a positive social opinion about the merits of self-scan checkouts and its usage (Mulaomerovic et al., 2013). Their study suggested that future research should analyse other factors w.r.t. the behavioural intention to use checkouts, such as the need for interaction. The need for interaction could be an important driver when choosing between different types of checkouts.
Driver for Checkout Usage: The Need for Interaction
As service operations are increasingly converted from high-cost manual operations to low-cost automated self-services, there is a negative effect of the new self-service on social bonding and, subsequently, customer loyalty (Selnes & Hansen, 2001). More specifically, the study by Selnes
et al. (2001) found three effects: as stated above, personal service has a positive effect on social
9
On the other hand, if a relationship is complex, suggesting a higher degree of product need complexity, lack of expertise or experience from the customer, personal relationships are necessary to create value for the customer (Selnes et al., 2001). The study suggests that self-service without at least some personal interaction can lead to a negative effect on customer loyalty due to the removal of the social bond mechanism (Selnes et al., 2001). Nonetheless, some customers (although small in size) rather prefer a complete removal of the social bond mechanism, as they see themselves as more effective in executing the service compared to an employee (Meuter et al., 2000). Some customers reflect upon employees as being an annoyance, which could lead to less customer loyalty and an increase in switching behaviour (Meuter et al., 2000). For those customers, self-service technologies may present the appropriate course of action as it makes it possible for customers to produce and consume in their own convenient way (Meuter et al., 2000). Altogether, existent literature suggests that the need for employee interaction is two-sided: some people prefer interaction, others prefer no interaction at all. Elliott et al. (2013) stated that the simple need for interaction could have a pervasive effect on the decision whether to use or not use self-scan checkouts. This is in line with the findings by Curran et al. (2005) who stated that there are people who regard the encounter at a standard checkout as a ‘social experience’ and therefore opt for a standard checkout over a self-scan checkout due to the preference of interacting with other people (Zeithaml & Gilly, 1987; Curran et al., 2005). On the other hand, some people might rather prefer the self-scan checkout over a standard checkout due to the perceived ease of use and/or the absence of interaction with employees (Dabholkar, 1996; Meuter, Ostrom, Roundtree, & Bitner, 2000; Dabholkar et al., 2002).
Recently, self-scan technologies have included social cues, which could potentially be seen as a hybrid: the attributes of a self-scan checkout, while simultaneously including some cues that could be reflected upon as being interactive. It might therefore be interesting to see if the implementation of social cues (i.e. speech-enabled technology) have an effect on the behavioural intention to reuse self-scans, compared to traditional checkouts.
Social Cues as a Potential Driver for Self-Scan Checkout Reusage
10
effect on the experience of self-scan usage. This is due to a prevalent habitability difficulty with regard to implementing spoken language systems (i.e. speech-enabled technologies), which can be explained by the uncanny valley effect (henceforth, UVE) (Moore, 2017). UVE suggests that eeriness and aversion can be triggered via near-humanoid items incorporated in the technology: if human likeness increases for a technology, it can lead to higher affinity, until a point where the increase of human likeness imposes a feeling of creepiness, leading to lower levels of affinity (Moore, 2017). Moore (2017) stated that when the categorical perception is mismatched with the perceptual cues, the combined effect will lead to higher levels of tension. For example, a self-scan checkout (a robotic aspect) incorporating speech (a humanlike aspect) could lead to higher levels of eeriness and tension, due to the mismatch between both aspects. This is in line with the notion that some customers have an uncomfortable feeling when a technology is giving instructions to them (Wisely, 2002; Dabholkar et al., 2003).
A different study by Moore (2012) suggested that the uncomfortable feeling (i.e. eeriness) could be seen as an inner control signal, which subsequently leads a person to take countermeasures: withdraw from the action, attack the entity that evoked the feeling of eeriness, ignore (some of) the conflicting cues, or reduce the alignment evoked by conflicting cues via the integration of the newly found information (this would lead to ‘habituation’). Hence, the first three actions are in line with the statement by Fogg that social cues are unfitting to the technology if its sole purpose is to improve efficiency (social cues could lead to higher levels of tension, which are removed via withdrawal, aggression, or denial). In contrast, the latter action suggests that people could potentially accept speech-enabled technology via habituation. Habituation is “the response decrement as a result of repeated stimulation” (Harris, 1943; Thompson & Spencer, 1966, p.17). Thus, as people become increasingly comfortable with social cues, the initial negative response could potentially be reduced.
11
Table I: Main Papers and Respective Main Findings
Self-Service Technologies
Author Topic Summary of main findings Curran et al. 2005 Factors influencing
customer
satisfaction w.r.t. SST’s
Their study looked at factors that influence (dis)satisfaction w.r.t. SST’s, namely: ease of use, usefulness, risk, and the need for
interaction
Dabholkar et al. 2002 Promotion of factors
influencing
customer adoption of SST’s
Their study looked at what SST factors should be promoted and under which conditions:
- Target market: low self-efficacy or high need of interaction
- Promote the ease of use, or “user-friendliness”
- Target market: low inherent novelty seeking or highly self-conscious
- Promote performance or reliability
- Target market: high inherent novelty seeking, high levels of self-efficacy, high levels of self-consciousness, or a high need for interaction with
employees
12
Meuter et al. 2000 Categories influencing customer
satisfaction w.r.t. SST’s
Their study looked at the most important categories influencing the customer
satisfaction w.r.t. SST’s. According to their study, the categories influencing customer satisfaction positively, are:
- Perceived relative advantage of using SST’s
- Ability to bail out customers from troubling situations
- Expectations meeting realizations The categories influencing customer
satisfaction negatively (influencing customer dissatisfaction), are:
- Technology failures - Process failures
- Customer-driven failures
Reinders et al. 2008 The effects of forced use of SST’s
The study looked at the effect of forced use when implementing an SST. Forced use leads to frustration, and thereby negative effects. The negative effects could be a switch to a competitor, or spreading a negative word-of-mouth.
Weijters et al. 2007 Factors influencing customer adoption w.r.t. SST’s
Their study looked at factors that influence adoption of the SST, and suggested the
following antecedents to be important: ease of use, usefulness, reliability, and fun
Self-Scan Checkouts
13
Dabholkar et al. 2003 Self-scan usage and
the effects of
customer motivation and behaviour
The study looks at the reasons for why customers either use or avoid self-scan
checkouts, and proceeds on existing theory on customer motivation and behaviour with respect to technology-based self-service. Moreover, the study incorporates factors that influence the preference or avoidance, by looking at the attributes of the technology, differences in consumers, and differences in the situational context.
Elliott et al., 2013 Factors influencing the intention to use self-scan checkouts
The study looks at the effect technology readiness has on customers’ intention to use self-scan checkouts. The study suggests that the perceived reliability and the perceived fun of the SST are more important drivers towards the intention to use self-scan checkouts than technology readiness: both (reliability and fun) have a direct and positive effect on the intention to use the self-scan.
Fernandes et al., 2017 Factors influencing
user perceptions of service quality of self-scan checkouts
14
the importance of the various attributes
Marzocchi et al., 2006 Factors influencing
user perceptions of service quality of self-scan checkouts
The study looks at the determinants that positively influence user perceptions of service quality. The study signifies the positive effect of enjoyment and the sense of control on customer satisfaction w.r.t. self-scanning.
Mulaomerovic et al., 2013
The effect of culture on the
implementation of self-scan checkouts
This study focused on Taiwan, and how to implement self-scan checkouts effectively. Apparently, the Taiwanese population is of a collectivistic nature, which suggests that retailers should create a positive social opinion w.r.t. the benefits of using self-scan checkouts The Need for Interaction
Author Topic Summary of main findings Selnes et al., 2001 The effect of
self-service on social bonds and customer loyalty
The study looks at the effect that the transition from traditional checkouts to self-service brings with respect to social bonds and customer loyalty and takes into account the complexity of the relationship between the customer and retailer. Three main findings:
1. Personal service has a positive effect on social bonding, and as a result, a higher level of customer loyalty
15
affects social bonds in low-complexity relationships
3. Undergoing a transition from personal service to self-service positively affects social bonds in high-complexity relationships
The Use of Social Cues
Author Topic Summary of main findings
Moore (2012) Uncanny valley effect based on conflicting cues
In line with the uncanny valley, the study presents a Bayesian model suggesting that conflicting cues lead to perceptual distortions which, in turn, evokes discomfort (eeriness). As a result, this could lead the affected person to take action to reduce the evoked eeriness. Four behaviours have been suggested to reduce the increased tension/eeriness:
1. Withdraw from the action 2. Attack the entity that evoked the
feeling of eeriness, in order to remove the tension
3. Ignore (some of) the conflicting cues 4. Reduce the misalignment evoked by
the conflicting cues, via the integration of new information (this would have ‘habituation’ as a result)
Moore (2017) The habitability problem and speech-enabled robots
16
most evidently as they enter the mass market and seek to provide a common-purpose voice-based interface between humans and
intelligent systems. Two problems occur in configuring the voice-based interface:
1. Visual, vocal and behavioural
affordances of the system needs to be brought to agreement
2. There is a mismatch between the capabilities and expectations of a person and the characteristics and benefits presented by even the most progressive autonomous social agent II. Research Objective and Conceptual Model
The prevalent literature on SST’s and more specifically, self-scan checkouts, have presented an overview of the various antecedents underlying the behavioural intention to use self-scans and the subsequent effect on customer loyalty. The effect of factors such as ease of use, reliability, perceived control, and reliability on the usage of self-scan checkouts have been affirmed by various papers. Nonetheless, the effect of social cues programmed in self-scan checkouts has not been researched previously. As these types of checkouts could be seen as a relatively new technology, novelty could also play a role. An earlier study by Weijters et al. (2007) suggested that newness could possibly have an ambiguous effect on the adoption of self-scan checkouts: due to the notion that newness varies with demographic variables, newness can either lead to a positive or a negative disposition towards the self-scan. This is in line with the findings by Dabholkar et al. (2002) who stated that novelty influences the inclination towards using new types of SST’s, and thus, self-scan checkouts.
17
social interaction over no interaction. Hu et al. (2006) stated that social cues could be seen as the social characteristics of a situation, and therefore, potentially alleviate the loss of interaction when using a self-scan checkout. Therefore, social cues could possibly have an effect on the behavioural intention to reuse self-scan checkouts.
Likewise, the hedonic motive, fun, remains ambiguous in its effect on the usage of self-scan checkouts. The study by Fernandes et al. (2017) has suggested that fun is a less critical determinant of self-scan usage. On the other hand, other studies have signified the importance of fun, where Elliott et al. (2013) stated that fun (combined with the perceived reliability of the self-scan checkout) is an important driver in stimulating behavioural intention to use the self-scan checkout. More specifically, their study suggested that the perceived reliability and fun towards the self-scan checkout are more important drivers than technology readiness in shaping behavioural intention to use the self-scan checkout (Elliott et al., 2013). This is in line with the findings by Marzocchi
et al. (2006) who found that perceived control and the fun/enjoyment aspect both have a positive
effect on the evaluation of the service quality of the SST.
18
This thesis attempts to fill in the existing literature gap by looking at the behavioural intention to reuse various types of checkouts (standard checkout; self-scan checkout; self-scan checkout including social cues), while taking into account how hedonic shopping motivation interacts with the main model. The following conceptual model is designed:
Figure I: Conceptual Model
III. Hypotheses
19
self-scan checkout including social cue = some interaction). As previously stated, the need for interaction could potentially dominate the decision to opt for a self-scan checkout (Elliott et al., 2013) as some individuals reflect upon the standard checkout as a ‘social experience’ and therefore opt for the standard checkout over other types of checkouts (Zeithaml et al., 1987; Curran et al., 2005). Nonetheless, others could possibly opt for a self-scan checkout, due to its lack of interaction and perceived ease of use (Dabholkar, 1996; Meuter et al., 2000; Dabholkar et al., 2002). Therefore, this thesis tests whether the addition social cues could compensate for the loss of interaction when using a self-scan checkout.
Types of Checkouts and Their Respective Behavioural Reuse Intention
This thesis focuses not on the behavioural intention to use a type of checkout, but the behavioural intention to reuse a type of checkout. This is based on the following line of reasoning: when an individual approves of the self-scan technology, it suggests that they reflect upon the checkout process as satisfying and therefore reuses it (Siah & Fam, 2017). Hence, when looking at the intention to use a checkout, it does not present any information on whether the process is regarded as a satisfying one, whereas reuse intention does.
What underlies the reuse intention of a certain type of checkout could be influenced by existing habits. Habits are based on repeated actions, or more specifically: whenever a customer performs a new action, a mental connection is created between the situation and the action, where repetition strengthens and plants this connection in the memory (Wood & Neal, 2009; Lally & Gardner, 2013) which leads other alternative actions to become less accessible in that specific situation (Lally et al., 2013). In turn, when the customer experiences this connection due to the context cue, the habitual response is automatically activated (Lally et al., 2013).
20
Thirdly, habits are difficult to change due to unchanging attributes (Wood et al., 2007). It is for this reason that (different types of) self-scan checkouts might not be a suitable option for most people, as their habits suggest an inclination towards the standard checkout. Therefore, the first hypothesis is:
H1: People are more likely to reuse a standard checkout compared to the two types of self-scan
checkouts, ceteris paribus
The Effect of Eeriness on the Behavioural Intention to Reuse Self-Scan Checkouts
The term “uncanny valley” was first coined by Mori in 1970, where he stated that the final goal of robotics lies in human being themselves: the more a robot resembles a human, the closer it is to this final goal. The problem here lies when people notice that the resemblance is not real, giving them a “strange” feeling (Mori, 1970). The strangeness that is evoked by the humanlike resemblance can represent a negative familiarity: the appearance is humanlike, but there is a negative feeling of familiarity based on the notion that the entity is not human: this depicts the uncanny valley (Mori, 1970).
The uncanny valley theory is mostly involved with appearances, movements and/or behaviours, whereas other parameters such as sound could also possibly lead to a perceptions of uncanniness and hence create a negatively associated affect with the technology (Grimshaw, 2009). This was found by a study which manipulated facial and vocal human realism, where it was hypothesized that a robot combined with a human voice, or a human being with a robotic voice, is regarded with a higher level of eeriness than a robot combined with a robotic voice, or a human being with a human voice (Mitchell, Szerszen, Lu, Schermerhorn, Scheutz, & MacDorman, 2011). Their result was indeed verified as such, which is in line with the findings by Moore (2017).
21
familiarity. Therefore, the addition of social cues could create a negative familiarity feeling due to a mismatch between the voice (the social cues) and the design of the self-scan checkout, which could lead to a lower behavioural intention to use the self-scan. Therefore, the second hypothesis is:
H2: The addition of social cues to a self-scan checkout leads to a lower behavioural intention to
reuse that type of checkout
The Effect of Hedonic Shopping Motivation on the Behavioural Intention to Reuse Checkouts
Earlier research on shopping has established that a person’s hedonic and utilitarian motivations have an impact on the usage of systems, whereas earlier research on human-computer interaction has shown that systems should include, both, hedonic and utilitarian components (O’Brien, 2010). In turn, the hedonic and utilitarian motivations can be used to gain a better understanding of engagement and the purpose and pleasure derived from people using a certain interface (O’Brien, 2010). In general, hedonic and utilitarian motivations and the accompanied engagement with a system, attempt to include the functional (utilitarian) and pleasurable (hedonic) into the experience of users, which leads users to be emotionally, cognitively, and physically involved in the interaction (O’Brien, 2010).
Human motivation theory defines shopping as “an activity that satisfies an individual’s hedonic motivations” (Kim & Hong, 2011, p.315). Shopping preference theory suggests that an individual’s characteristics can be seen as determinants of shopping motives (Sheth, 1983; Kim & Hong, 2011). Therefore, hedonic motivations underlie shopping, where the shopping motives are established by the individual’s characteristics. As previously stated, habits are difficult to change due to unchanging attributes (Wood et al., 2007). It is for this reason that (different types of) self-scan checkouts might not be a suitable option for most people, as their habits suggest an inclination towards the standard checkout. The individual characteristics suggest that people prefer a standard checkout over a self-scan checkout, and obtain more hedonic value by doing so. Therefore, the following hypothesis is established:
H3a: The preference for the standard checkout is more pronounced for people with a higher
22
Hedonism suggests that people allow themselves to be affected by the stimuli they fantasize into being by using their imagination: the aptitude which allows a person to daydream, and subsequently, experience pleasure in the imagination that is not present in reality (Campbell, 1987). Hedonism is the manifestation of longing which leads people to visualize the desire to have or use new products; products that might otherwise, due to their strangeness, have little appeal (Campbell, 1987). Therefore, hedonism, or as considered in this study, hedonic shopping motivation, could lead people to prefer novel products despite the evoked feeling of strangeness (i.e. eeriness). As earlier discussed, a factor that could possibly influence the usage of self-scan including social cues is eeriness. Therefore, hedonic shopping motivation could potentially alleviate the evoked feeling of eeriness, and have a positive effect on the behavioural intention to reuse self-scan checkouts including social cues. Thus, the hypothesis is formalized as follows:
H3b: The preference for the self-scan checkout including social cues is more pronounced for people
with a higher hedonic shopping motivation The Intervening Effects of Habits and Eeriness
Habits and eeriness are established as factors that underlie the main effect: the effect of types of checkouts on their respective behavioural reuse intentions. However, according to earlier studies on habit formation, people’s intentions cannot be seen as the sole describer of behaviour, as it may purely be a “force of habit” rather than their intentions (Ouellette & Wood, 1998; Limayem & Hirt, 2003; Wang, Harris, & Patterson, 2011). Habits are difficult to change, as they are hard to suppress, while simultaneously, the habit of not using something is difficult to overcome (Wang
et al., 2011). Therefore, this thesis also takes into consideration whether habits can be seen as an
intervening variable underlying the relationship tested in hypothesis 1. Thus, the following mediation hypothesis is established:
H4a: There is a mediation such that people will respond to the self-scan checkout with decreased
23
Additionally, Moore (2012) stated that the alignment of conflicting cues could be resolved when newly found information is integrated, which in turn leads to habituation. Given that self-scan checkouts including social cues do not exist for long yet, it is likely that alignment of conflicting cues is not yet established for this type of checkout. Hence, the hypothesis is formalized as follows:
H4b: There is a mediation such that people will respond to the self-scan checkout including social
cues with decreased levels of habit formation (in comparison to the standard checkout), which has a negative effect on the behavioural intention to reuse that type of self-scan checkout
Similar to habits, eeriness could also be an intervening variable. An explanation behind this can be found in the earlier mentioned study by Moore (2012). A central principle in the study was that perceptual tension emerged from conflicting cues with categorical perception, which may lead to a feeling of discomfort (e.g. eeriness) and subsequently leads the person to act in such a way to reduce the effect (Moore, 2012). Hence, evoked eeriness could potentially intervene with the main effect tested in this study: a higher level of evoked eeriness could intervene with reusing a certain type of checkout. Given that standard checkouts have a cashier, the effect of eeriness in that scenario would be less pronounced compared to the self-scan checkout. Thus, the following hypothesis is established:
H5a: There is a mediation such that people will respond to the self-scan checkout excluding social
cues with increased levels of eeriness (in comparison to the standard checkout), which has a negative effect on the behavioural intention to reuse that type self-scan checkout
On the other hand, perceptual tension could be most evident in the case of self-scan checkout including social cues, as the self-scan checkout is a robotic entity, whereas voice (i.e. social cues) is a humanlike feature. Thus, eeriness should become most pronounced when using a self-scan checkout including social cues. Therefore, the following hypothesis is formalized:
H5b: There is a mediation such that people will respond to the self-scan checkout including social
24
METHODOLOGY
The theoretical framework provided the most relevant literature for this thesis, the derived hypotheses, and the conceptual framework. An empirical research will be done in order to test the conceptual framework, given in figure I. This section discusses the methodology, that is: the methods that are used to test the conceptual framework. A distinction is made between how the research is designed, how it is presented to the respondents and via what measurements, and concludes with a plan of analysis for the result section.
I. Research Design
According to the conceptual model, the main effect suggests that each type of checkout has a respective behavioural reuse intention. The secondary effect included in this thesis suggests that hedonic shopping motivation interacts with the effect of the main model, whereas habits and eeriness could potentially mediate the main effect. These variable are therefore included in this thesis and are obtained from earlier studies on these topics. Variables that could possibly change the effectiveness of the main effect and the secondary effect are incorporated in the form of control variables. Control variables are the variables that can be tested for each of the three scenarios (standard checkout; self-scan; self-scan including social cues), and are obtained from existing literature. It is therefore that control variables that solely look at the applicability of self-scan checkouts (e.g. perceived control, ease of usage, etc.) are not included in this study.
A variable that could affect the model’s main- and secondary effect is the ‘novelty-seeking behaviour’ of participants (aspect of innovativeness). The study by Dabholkar et al. (2002) suggested that if a market segment has a low level of inherent novelty-seeking behaviour, there is a high tendency to be reluctant to use new SSTs. Moreover, according to Weijters et al. (2007), newness (and thus novelty-seeking) varies with demographic variables. Therefore, the usage of different types of checkouts could be influenced by both, novelty-seeking behaviour and demographics and are therefore both included in this study as control variables.
25 II. Questionnaire Design
The complete questionnaire can be found in “Appendix A – Questionnaire”. The questionnaire starts with a general introduction, where the participant is thanked for his/her participation, and notifies that the participation will remain completely anonymous and that data is solely used for research purposes. Subsequently, the hedonic shopping motivation is measured. Following, one of the three videos (standard checkout; self-scan; self-scan including social cues) is randomly presented to each participant, followed by a manipulation check based on social presence. This is done in order to test whether the participant has the feeling that the checkout includes some degree of sociability (e.g. human warmth). The manipulation is followed-up by the measurement of the dependent variable: the behavioural intention to reuse. Next, control variables are presented, and the questionnaire concludes with some demographics.
III. Measurements
The measurements are described in the following order: type of checkout (independent variable), behavioural intention to reuse the checkout (dependent variable), social presence (manipulation), hedonic shopping motivation (interaction variable), habits (mediation variable), eeriness (mediation variable), and novelty seeking and demographics (control variables). The complete overview of items per factor, including their mean, standard deviation and Cronbach’s alpha are presented in table II.
Type of Checkout
26
of the total video). These videos are randomly assigned to participant to test for between-subject results. The link to the questionnaire is distributed via personal networks and Facebook.
Behavioural Intention to Reuse
The behavioural intention to reuse is chosen as the dependent variable, and is based on an adapted version of the ‘intention to use’ scale from Dabholkar’s study in 1996, namely “continued interaction” (Shamdasani, Mukherjee, & Malhotra, 2011). The study by Shamdasani et al. (2011) looked at how customers assess service encounters in SST environments via a model that takes into account the antecedents of service quality and their respective effect on service quality. Continued interaction was presented as a measure that takes into account the likelihood of, both, repeat and preferential usage of a technology-based medium over a traditional medium (Shamdasani et al., 2011). As previously explained, habits are formed via repetitions that lead to a mental connection between a situation and an action (Lally et al., 2013). Therefore, as continued interaction increases, habits (the underlying factor driving the main model effect) can be formed towards different types of checkouts. Due to the fact that the study by Shamdasani et al. (2011) looked at the self-service internet technologies, the items for continued interaction are adapted to fit the current research and is paraphrased as ‘behavioural intention to reuse.’ Respondents indicate whether they agree or disagree with the items on a 7-point Likert scale (1 = “Strongly disagree”, 7 = “Strongly agree”).
Social Presence
27
Hedonic Shopping Motivation
Hedonic shopping motivation is based on the hedonic value that is obtained via shopping (the “fun side” of shopping) (Babin et al., 1994). The hedonic shopping motivation is included as a 9-item construct, with a 7-point Likert scale where respondents can indicate whether they “strongly disagree” (= 1) or “strongly agree” (= 7) with the various items.
Habits
Verplanken and Orbell (2003) proclaimed that habits can be seen as psychological constructs, not only based on behavioural frequency in the past, but also on features of habits, such as repetition that occurred in the past, automaticity (which is based on e.g. lack of control or the level of efficiency), and showing one’s identity. The first feature is in line with the study by Lally et al. (2013) who stated that when repetition with respect to a certain action occurs, a mental association is formed that induces the formation of habits. Moreover, the importance of automaticity and showing one’s identity are also suggested by Wood et al. (2007), who stated that the first principle of habit formation is automaticity, and the third principle is based on the difficulty to change habits based on unchanging attributes (such as one’s own identity).
Verplanken et al. (2003) created a 12-item habit index, which is included in this thesis. Respondents indicate on a 5-point Likert scale (1 = “Strongly disagree”, 5 = “”Strongly agree”) whether they agree or disagree with the 12 items.
Eeriness
28
Novelty-Seeking Behaviour
Novelty-seeking, which is regarded as an aspect of innovativeness, is defined as “the desire to seek out new stimuli” (Hirschman, 1980; Dabholkar et al., 2002, p.187). According to previous studies, certain aspects of innovativeness influence the attitude towards technology-based products (Midgley & Dowling, 1978; Hirschman, 1980; Gatignon & Robertson, 1985; Parasuraman, 2000; Dabholkar et al., 2002). Therefore, novelty-seeking behaviour could have an effect on the behavioural intention to reuse self-scans including social cues (the newest type of technology included in this thesis), and needs to be controlled for.
A novelty-seeking behaviour scale is obtained from the study by Dabholkar et al. (2002) which originates from an arousal-seeking scale by Mehrabian and Russell in 1974. Participants indicate whether the items are in line with them, based on a 5-point Likert scale (1 = “Strongly disagree”, 5 = “Strongly agree”).
Demographics
According to Weijters et al. (2007), variables that influence SST use might vary depending on age, gender, and level of education. Another demographic variable that might be of some importance is the area where a person lives in. For example, a rural area might not have the same degree of accessibility to an SST compared to an urban area. Therefore, these variables are added to the questionnaire.
Factor Analysis and Reliability Analysis
29
“automatic unconscious driver” (p.402). Therefore, automaticity of the respondent (factor 1) is more in line with this representation of habits than the unchanging attributes of the respondent (factor 2).
A lower threshold of 0.5 is established for the Kaiser-Meyer-Olkin test (KMO). All variables exceed the lower threshold, which suggests a low proportion of variance between the several items for each factor. Hence, a factor analysis is proven to be useful for the dataset in this thesis.
The Cronbach’s alpha is presented in order to see the internal consistency of each factor. The threshold that is maintained for the Cronbach’s alpha is 0.6. Table III presents the Cronbach’s alpha for each factor, and the Cronbach’s alpha when an item is removed from the factor. All factors have a higher value than 0.6, and therefore, suffice the criterion. Moreover, removing items do not increase the Cronbach’s alpha by much. Thus, except for one of the habits factor, no items are removed. The recoded items are hedonic shopping motivation item 9, and novelty-seeking behaviour item 3 and 5 (changes to the items are presented accordingly in Table II).
Table II: Items’ Mean, Standard Deviation, and Cronbach’s Alpha
Mean S.D. Cronbach’s Alpha if Item Deleted Factor – Social Presence (a = .955)
(1) There is a sense of human contact when using this type of
checkout 2.15 1.415 .957
(2) There is a sense of sociability when using this type of
checkout 1.97 1.164 .938
(3) There is a sense of warmth when using this type of
checkout 1.90 1.213 .908
Factor - Hedonic Shopping Motivation (a = .919)
(1) I reflect upon shopping to be truly a joy 4.33 1.637 .905 (2) I continue to shop, not because I need to, but because I
want to 4.18 1.874 .911
(3) Shopping feels like a true escape 3.24 1.820 .904 (4) I enjoy being immersed in new products 3.51 1.763 .919 (5) I have a good time shopping because I am able to act on
the “spur of the moment” 3.83 1.673 .909
30
(1) Given the transaction I intend to perform, I will definitely
use this type of checkout 5.09 2.131 .835
(2) Given the transaction I intend to perform, I will definitely
continue to use this type of checkout 4.79 1.987 .669 (3) When I need to perform a transaction, I would actively
seek out this type of checkout 4.19 2.224 .843
Factor – Eeriness (a = .908)
(1) This type of checkout is creepy 1.74 1.081 .848 (2) This type of checkout is eerie 1.87 1.126 .818 (3) This type of checkout is unnatural 2.20 1.291 .942 Factor – Habits (a = .887)
Using this type of checkout is something:
(1) I do frequently 3.40 1.401 .873
(2) I do automatically 3.12 1.358 .870
(3) I do without consciously having to remember 2.91 1.364 .882 (4) That makes me feel weird if I do not use it 1.78 1.122 .893
(5) I do without thinking 2.94 1.320 .874
(6) That would require effort not to do it 2.48 1.286 .885 (7) That belongs to my (daily, weekly, monthly) routine 2.86 1.404 .869 (8) I start doing before I realize I’m doing it 2.52 1.279 .875
(9) I would find hard not to do 2.11 1.210 .884
(10) I have no need to think about doing 2.99 1.329 .873
(11) That’s typically me 2.11 1.113 .883
(12) I have been doing for a long time 3.09 1.315 .877 Factor – Novelty-Seeking Behaviour (a = .754)
(1) I am always seeking new experiences 3.58 .894 .701 (2) When things get boring I like to find some new and
unfamiliar experiences 3.49 1.062 .693
(3) I prefer an unpredictable life full of change to a routine
way of life 3.16 1.172 .746
(4) I like to continuously change activities 3.19 1.006 .721 (5) I like meeting others who have new ideas 4.30 .864 .743 (6) I like to experience novelty and change in my daily routine 3.62 .941 .705
IV. Plan of Analysis
31 RESULTS
I. Descriptive Statistics
The sample size for this study involves 111 participants, with an equal distribution of 37 respondents for scenario 1 (standard checkout), 37 for scenario 2 (self-scan checkout), and 37 for scenario 3 (self-scan checkout including social cues). 42% of the sample is male and 58% female. The average age of the sample is 31 years old with a minimum age of 14 and a maximum age of 74 years old. Approximately 55% of the respondents live in an urban area, 20% live in a suburban area, and 25% live in a rural area. The participants’ highest attained level of education are presented in ‘Table III: Distribution Educational Levels’. The total number of participants who filled out the complete demographics is 103 (i.e. 7 respondents did not fill out the complete survey).
Table III: Distribution Educational Levels Highest Level of Education # of Participants
Less than a high school degree 1 High school degree of equivalent 29
Associate degree 8
Bachelor’s degree 42
Master’s degree 24
Doctoral degree 1
Above doctoral degree n.a.
Multicollinearity Check
32
concurrently with the age of the respondent. On the other hand, habits correlate negatively with age (at the 1% level) and where the participant lives (at the 1% level). Novelty correlates positively with education (at the 5% level). Age correlates negatively with, both, the area the respondent lives in (at the 1% level) and the level of education (at the 5% level). As none of the correlations depict a high correlation (the highest correlation is established between habits and age: -.367), the table suggests that multicollinearity does not seem to pose a threat for subsequent analyses.
Table IV: Correlation Table Independent Variables
(1) (2) (3) (4) (5) (6) (7) HSM 1 .036 -.020 -.114 -.150 -.140 -.139 Eeriness .036 1 -.326** -.087 .222* .078 -.088 Habits -.020 -.326** 1 .120 -.367** -.233* .163 Novelty -.114 -.087 .120 1 -.156 -.134 .220* Age -.150 .222* -.367** -.156 1 .342** -.228* Area -.140 .078 -.233* -.134 .343** 1 -.330** Educate -.139 -.088 .163 .220* -.228* -.330** 1
* Correlation is significant at the 0.05 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed)
The next step in evaluating the data is to check whether the manipulation of social presence has worked.
II. Manipulation Check
A manipulation check is conducted in order to see whether respondents reflect upon the different types of checkouts in a correct matter: i.e. to see if the respondents are aware of the (absence of) social presence. As previously stated, the different groups presented with different types of checkouts have been equalized (37 respondents for each scenario). A one-way analysis of variance (henceforth, ANOVA) is performed to see whether the groups differ significantly from one another, whereas the respondents in each group are as similar to one another as possible.
33
108) = 23.272, p < .01), with MSc1 being the mean for scenario 1, MSc2 for scenario 2, and MSc3 for scenario 3. Therefore, the manipulation succeeded in creating different groups.
Nonetheless, despite the notion of differences between the means in groups, the ANOVA does not infer where changes are present between groups. Thus, a Tukey Honest Significant Difference (HSD) post hoc test is done in order to see where changes between groups lie. According to the multiple comparisons, there is a significant negative difference between the standard checkout (scenario 3) and the self-scans (scenario 1 and 2) (MDif3-1 = 1.072, MDif3-2 = 1.229; p < .05). The difference suggests that the standard checkout evokes a higher social presence than either one of the self-scan checkouts. On the other hand, no significant difference in social presence is established between the self-scan checkout including social cues and the self-scan checkout excluding social cues (MDif2-1 = .157; p > .1).
III. Hypotheses
The manipulation check suggested that a difference is established between self-scan checkouts and the standard checkout, based on social presence. Subsequently, the hypotheses are tested. First, the main effect model is tested. The main effect model tests for hypothesis 1 (people are more likely to reuse a standard checkout compared to the two types of self-scan checkouts, ceteris paribus) and hypothesis 2 (the addition of social cues to a self-scan leads to a lower behavioural intention to reuse that type of checkout).
Main Effect Model: One-Way ANCOVA
Prior to the ANCOVA, a VIF-analysis is done to see whether the model is subject to multicollinearity. The VIF-table for the main model is presented in “Appendix D – VIF Scores.” Threshold are established for the VIF-score (lower than 5) and the tolerance-level (higher than .20). All variables do not exceed the boundaries, and therefore, multicollinearity does not seem to pose a threat to the main effect model (presented in “Appendix D1”).
34
The Bonferroni post hoc test shows a significant difference between the self-scan checkout and the self-scan checkout including social cues (MDif2-1 = .620, p < .05). On the other hand, no significant difference was found between the standard checkout and the two self-scan checkouts (MDif3-1 = .115, MDif3-2 = -.505; p > .05). Hence, the self-scan including social cues had a significant higher mean than the self-scan excluding social cues. These two conclusions from the pairwise comparison are in contradiction with, both, hypothesis 1 and 2. Hypothesis 1 tests whether a standard checkout has a higher reuse intention compared to the self-scans. The ANCOVA showed no significant difference in the mean of the behavioural intention to reuse a standard checkout, compared to the self-scans. Hypothesis 2 tests whether the addition of social cues has a negative effect on the intention to reuse that type of checkout. The ANCOVA showed a significant difference between the two types of self-scan checkout, but contradictorily shows that the intention to reuse a self-scan including social cues is higher than the intention to reuse a self-scan excluding social cues.
Interaction Effect Model: Two-Way ANCOVA
Similar to the main effect model, a VIF-analysis is presented for the interaction effect model in “Appendix D2” (including hedonic shopping motivation). All variables suffice the criteria set for the tolerance-level and the VIF scores. Hence, no evidence for multicollinearity is present in the interaction effect model.
35
Mediation Analysis
Lastly, hypotheses 4 and 5 are tested, which hypothesized that eeriness and habits mediate the relation between the type of checkout and its behavioural reuse intention. Therefore, a mediation analysis is done, to see if the relation between the type of checkout and its behavioural reuse intention is mediated by the evoked eeriness and respondents’ habits (Hayes 2013, model 4). Before the mediation analysis, a VIF-analysis is done in order to see whether the model is subject to multicollinearity (presented in “Appendix D3”). Similar to the previous two models, this model also suffices the criteria for the VIF-score and the tolerance level. Moreover, the VIF and the following mediation models depicts two dummy variables: ‘Checkout A’ is a dummy variable which compares the self-scan checkout (1) with the standard checkout (0), whereas ‘Checkout B’ compares the self-scan checkout including social cues (1) with the standard checkout (0). Thus, two models are established, both including the mediation variables habits and eeriness. The interaction effect model suggested that hedonic shopping motivation does not have a significant (interaction) effect with respect to the main model, and is therefore excluded from the mediation analysis. The mediation analyses are presented in “Appendix F- Mediation Analyses.”
Mediation Model A: Standard Checkout Compared to Self-Scan Checkout
Mediation model A is presented in “Appendix F1.” Model A compares the standard checkout with the self-scan checkout excluding social cues. The results depict the expected mediation for eeriness (self-scan checkout greater eeriness lower behavioural intention to reuse the checkout) at the 95% confidence interval (henceforth, CI) (a x b = -.0964; 95% CI: -.2557, -.0206). and the expected mediational path for habits is also shown (self-scan checkout lower habits lower behavioural intention to reuse) at the 95% CI (a x b = -.2658; 95% CI: -.5566, -.0830). Moreover, the mediators cause the direct effect to become insignificant (c’ = -.0089; 95% CI: -.3300, .3122). Therefore, model A suggests a full mediation through habits and evoked eeriness. These results are in line with hypotheses 4a and 5a.
Mediation Model B: Standard Checkout Compared to Self-Scan Checkout Including Social Cues
36
including social cues lower eeriness higher behavioural intention to reuse self-scan including social cues (a x b = .0660; 95% CI: .0032, .1876). Nonetheless, this result can solely be seen as suggestive (rather than conclusive) as the model is close to, but does not reach significance (F(6, 98), = 2.1159, p > 0.05). Habits is shown to be insignificant in showing mediation here (a x b = .1453; 95% CI: -.0296, .4224), with the effect of type of checkout on the habits being insignificant (a = .2916; 95% CI: -.1133, .6965). No mediation is established for model B. Interestingly, model A shows that eeriness and habits are shown to fully mediate the main effect (c’ = 0.0089; 95% CI: -.3300, .3122), whereas mediation is non-existent in model B due to the insignificant mediating effects of eeriness and habits, and the significant direct effect (c’ = .3578; 95% CI: .0405, .6752). Therefore, as eeriness and habits do not mediate the effect when comparing the standard checkout with the self-scan checkout including social cues, hypotheses 4b and 5b are rejected. Table V shows the acceptance/rejection of hypotheses based on the previous analyses. The discussion section sheds more light on these findings.
Table V: Acceptance and Rejection of Hypotheses
37 DISCUSSION
According to the main effect ANCOVA, the behavioural intention to reuse a self-scan including social cues was significantly higher than the intention to reuse a self-scan excluding social cues (in contrast with hypothesis 1 and 2). This finding could potentially be explained by referring back to the notion that newness could have an ambiguous effect on the adoption of self-scan checkouts: newness varies with demographics, which in turn leads to a positive or a negative disposition towards the self-scan checkout (Weijters et al., 2007). According to the descriptive statistics, 66 out of 105 respondents (more than 60%) indicated that their minimum level of education was of a bachelor’s level. Weijters et al. (2007) stated that more highly educated people are expected to value the innovativeness of technology, whereas lower educated people could be more likely to avoid novelty of technologies. Hence, the current study might suggest that the higher behavioural intention to reuse self-scans including social cues could be due to the high educational level and the respective novelty-seeking behaviour of respondents. This could also possibly explain why the behavioural intention to reuse a standard checkout was not shown to be significantly higher than the two types of self-scan checkouts (in contrast with hypothesis 1): the standard checkout holds no degree of newness and is more likely to be favoured by people with a relatively lower educational level (following the line of reasoning by Weijters et al. (2007)).
38
Mediation model B (which compares the self-scan including social cues with the standard checkout) suggested that eeriness and habits were not significantly mediating the main effect (in contrast with hypothesis 4b and 5b). The insignificance of habits could be explained by the fact that self-scan checkouts including social cues are not present in every store, and therefore, makes it difficult to form habits towards this type of checkout. The insignificant result of eeriness could be explained by two reasons. First, avoidance of novelty plays a crucial role in generating the uncanny valley effect (Sasaki, Ihaya, & Yamada, 2017). Consequently, a sample with more highly educated people with higher level of novelty-seeking behaviour could imply that the uncanny valley effect becomes less evident. Second, Berlyne (1950) suggested that when a new stimulus influences a person’s sensory system, a response is evoked which can be described as ‘curiosity.’ As a result, the person is likely to react to the evoked curiosity by performing the same action again in the future (Berlyne, 1950). Hence, the new stimulus of the self-scan checkout (i.e. social cue implementation) might have led to no feeling of evoked eeriness due to the positive effect social cues posed on the sensory system (the creation of curiosity arousal). Altogether, this might suggest that the social cue implementation has led to curiosity, rather than eeriness, which positively influenced the behavioural intention to reuse the self-scan checkout including social cues.
