Robots, including in-store physical robots and virtual chatbots, have been increasingly employed to substitute human employees in customer service provision. These customer-facing service robots (SRs) can be defined as “system-based autonomous and adaptable interfaces that interact, communicate and deliver service to an organisation’s customers” (Wirtz et al., 2018, p.909). For these roles, SRs are often programmed to simulate customer-facing employees, engaging in different tasks with different intelligences and mimicking human-human interaction (Huang and Rust, 2021; Pantano and Scarpi, 2022). As a part of this, SRs may also be designed with physical and/or non-physical humanlike features, which can be stereotypically feminine or masculine, and which can affect customers’ emotional, cognitive, and behavioural responses (e.g., Blut et al., 2021; Tay et al., 2014). Despite a growing number of empirical studies, there is currently a lack of comprehensive synthesis of the extant literature related to the impacts of anthropomorphic design and gendering of SRs on consumer behaviour, which is vital to determine how further research can make meaningful contributions.
Anthropomorphism in SR literature refers to either perceived human-likeness (Blut et al., 2021) or design attributes of a robot (Lu et al., 2020; Wirtz et al., 2018). Anthropomorphism can affect customers’ reactions towards SR interactions in various service contexts, including public service (van Pinxteren et al., 2019), home service (Letheren et al., 2021), hospitality (Lv et al., 2021; Yam et al., 2021; Yoganathan et al., 2021), and retail (Whang and Im, 2021). Customers’ resultant emotional, cognitive, and behavioural responses can be positive or negative. For example, customers served by humanoid robots have a higher tendency of adoption (Blut et al., 2021; Sheehan et al., 2020) and re-use intention (Moriuchi, 2021); higher satisfaction (Yam et al., 2021); more positive evaluation (Li and Sung, 2021); closer emotional connection (Araujo, 2018); and more willingness to purchase (Yoganathan et al., 2021). On the other hand, researchers reveal that anthropomorphism of robots has the potential to harm human-robot interactions, causing greater discomfort and negative attitudes toward the robot (e.g., Kim et al., 2019; Mende et al., 2019). However, research on anthropomorphism of SRs has focused more on positive than negative effects such that less is known about the latter, especially the outcomes brought by the uncanny valley theory in customer-SR interactions (Wirtz et al., 2018).
Robot gendering refers to assigning gender onto a robotic platform through characteristics such as name, voice, and physique (Bryant et al., 2020; Robertson, 2010). Prior research examining effects of anthropomorphism of SRs often conflates gender with anthropomorphism by using gender-stereotypical features or characteristics to convey human-likeness (Crolic et al., 2021; Yam et al., 2021). However, gender-stereotypes have long been recognised in a technology context (e.g., Nass et al., 1997) and can impact how customers perceive and evaluate services (Pitardi et al., 2022a). For instance, Tay et al. (2014) demonstrate that participants show more positive evaluation and greater acceptance of a female-gendered healthcare robot and a male-gendered security robot due to gender-occupational role stereotypes. Broadly speaking, male-gendered robots are perceived to deliver a sense of intelligence, dominance, and competence while female-gendered robots provide a feeling of affection, communality, and particularly warmth (Ahn et al., 2022; Borau et al., 2021; Eyssel and Hegel, 2012; Seo, 2022) which plays a crucial role in customer-SR interactions (Choi et al., 2021). These findings raise important questions regarding how researchers and practitioners imbue robots with humanlike features.
Research on SRs in specific situations is beginning to emerge. During service failure, customers attribute more blame to a human employee than a SR (Belanche et al., 2020a; Leo and Huh, 2020) due to the robot’s perceived low level of agency (Gray et al., 2007). However, it is less clear how blame would be attributed to a humanlike versus non-humanlike SR, considering the impact of perceived thoughtfulness and/or agency brought by human-likeness (Waytz et al., 2014; Yam et al., 2021). While humanoids are more able to recover service failure than non-humanoids (Choi et al, 2021), it is unclear whether there would be any differences caused by gendering. Another important context to note is embarrassing service encounters (e.g., body measurement) in which a humanlike SR can have unfavourable effects on consumer embarrassment due to its resemblance with humans (Pitardi et al., 2022b). However, it is unclear whether human-robot gender congruity versus incongruity (Pitardi et al., 2022a) would have more or less favourable effects in embarrassing service encounters.
The field of SR research is in its infancy (Wirtz et al., 2018) and prior research is fragmented (Lu et al., 2020). Pioneering studies have systematically identified key dimensions of robot-delivered frontline service (Wirtz et al., 2018) and synthesised the impacts of SRs on customers and service employees (Lu et al., 2020), indicating the non-negligibility of anthropomorphic design in customer-SR interactions. Although van Pinxteren et al. (2020) innovatively conducted a systematic review of humanlike communications in conversational agents, a comprehensive literature analysis of anthropomorphism and gender of SRs and their impacts on human-robot service interactions is lacking. Aiming to overcome the gap and highlight the importance of discussing human-likeness of non-human service agents, we will undertake a systematic review to investigate the effects of anthropomorphism and gendering of SRs. The paper will provide a definition of gender of a SR which is missing in previous studies, describe the relationship between anthropomorphism and gender, and clarify the role of anthropomorphism and gender in human-robot service interactions. It will also identify knowledge gaps and key areas of interest via a research agenda for humanlike SRs. The paper will contribute to management and marketing literature as well as the SR literature by answering several research calls for understanding a robot’s design in general (Lu et al., 2020) and the robot’s humanlike characteristics (Blut et al., 2021; Choi et al., 2021; Xing et al., 2022) including gender (Belanche et al., 2020b; Choi et al., 2021) in particular. It could also provide theoretical support for organisations to deliver better service for consumers.
This research paper formulates role of virtual service robots in view of relationship marketing. The paper draws special attention to some underlying premises of adoption of virtual service robots in managing relationships. Although, robots in service encounters witnessed substantial attention from scholars in recent years, more attention is needed especially related to adoption of service robots in managing customer relationships. Post examining practitioner literature from the IT industry (global consulting firms and CRM vendors), the paper elaborates on three research avenues. First, identifying customer barriers and motivators in case of service robot adoption. Second, to examine robots as interaction partners in actual field setting and third, to evaluate different characteristics of relationships styles and how these might diminish or enhance customer experience. In closing, based on the study’s findings theoretical and practical implications are offered.
Artificial intelligence (AI) is expected to change what we know about customer satisfaction and service quality. This review reveals that in addition to the typical antecedents of satisfaction described in the literature (i.e. expectations, performance, and disconfirmation), new drivers of customer satisfaction with AI-driven services have emerged. While service quality remains one of the most important determinants of satisfaction, AI has influenced the traditional dimensions of service quality and this influence can be seen in the new dimensions of service quality and the new models for measuring AI service quality.
Service companies are striving to improve the customer experience and provide personalized services (Grewal, 2020). For example, companies are introducing innovative digital technologies (e.g., artificial intelligence and robotics) to improve customer service encounters (Lariviere et al., 2017; Marinova et al., 2017; Wirtz & Jerger, 2016; Wirtz et al., 2021). Customers are therefore becoming more accustomed to robots in their daily lives (Kunz et al., 2019) and perceive robots as independent social entities with an automated social presence (van Doorn et al., 2017), consequently referred to as social service robots. Social service robots are one example of such a frontline technology that companies employ to enhance the service encounter.
Engaging in social interactions with service robots has a significant impact at all phases of the customer’s service experience. Indeed, robot proactivity (i.e., robots mimicking human frontline employees’ proactive behaviors) elicits different customer reactions compared to human-human interaction, such as positive/negative responses or various levels of satisfaction (Krishna et al., 2019; Pitardi et al., 2021). Understanding which emotions service robots evoke is particularly important for designing the service encounter, especially in the pre-purchase phase (Tielman et al., 2014; Lerner et al., 2015). This phase is delicate because here the customer anticipates the final purchase choice through a series of considerations and evaluations (Ashman et al., 2015; Lemon & Verhoef, 2016). In this phase, a service robot can help customers to better identify their own needs and find appropriate product options, by pointing out and picking items (Roggeveen & Sethuraman, 2020).
Despite the development of research on service robots during the service encounter, further research is needed during the pre-purchase phase in retail contexts (Ashman et al., 2015). Although several researchers have conceptualized human-robot interactions as a moment of co-creation of value in the service encounter and have acknowledged the social presence of robots along with the benefits to the customer experience itself (Fernandes & Oliveira, 2021), there is still a need for empirical research on the effects of the proactive behavior of different types of robots and the unexpected emotional consequences of this type of interaction (Lu et al., 2020; Puntoni et al., 2021). Therefore, the present study aims to investigate customers’ reactions (i.e., facial expressions) to two types of social service robot behaviors (approaching vs non-approaching). Specifically, this study aims to determine whether the approaching behavior (or the sole presence) of a social service robot elicits positive or negative customer emotions. With two additional conditions including a human, the research study determines whether the reactions elicited by different approaches of the service robot are consistent with customers’ responses to relationships with front-line employees.
This study adheres to the concept of human-robot shared experience (Gaggioli et al., 2021) and responds to the call for a shift in research from robots perceived solely as objects to robots perceived as active participants in the co-creation of experiences (Rancati & Maggioni, 2021). Therefore, we use a non-functionally oriented service setting. Unlike previous studies, the service robot is not primarily a goal-directed tool that performs a specific task (e.g., guiding the consumer’s choices), but rather serves in a social context to enhance the service experience.
For this study, we used the social robot Misty II (https://www.mistyrobotics.com/). 150 participants (75M, 75F) volunteered for the study. In addition to participants’ demographics, we recorded their facial expressions during the service encounter to determine participants’ emotional states during the object observation. Additionally, we registered a self-reported measure of emotions. We capture the Positive Affect and Negative Affect Schedule (PANAS) (Terraciano et al., 2003) which assesses participants’ experience before and after the object observation.
In this study, we investigate whether the presence of another social agent (human/robot) increases the aesthetic experience of a painting when visiting an art museum. Our experimental study applies a between-subjects design and participants were randomly assigned to one of five conditions: I) robot approaching condition; II) robot only present; III) human partner approaching condition; IV) human partner only present; V) customer alone.
In the approaching conditions I and III, the social service robot or the human partner interacts with the study participants, showing some degrees of agency. The robot or human physically approaches the participants (reduces spatial distance) and looks at the participants while observing an object. In the non-approaching conditions II and IV, the social service robot or the human partner is only present; they do not approach the participants as in conditions I and III. The service robot and the human partner behave similarly, comparable to other bystanders. They show body and eye movements, but they do not demonstrate an intent to engage with the participants, for example, in order to communicate their observation of the object. Condition V is a control condition in which participants observe the object alone. In all conditions, the visual stimulus selected as the observed object is “The Starry Night” by Vincent van Gogh. It was chosen because it has been shown to elicit complex emotions (Chirico et al., 2021). We can also assume that the popularity of the painting does not induce a “surprise effect” due to its initial exposure.
We expect that conditions I and III, which relate to the approaching behavior (i.e., agency) of the robot and the human, will differ in the facial expression analysis or in the PANAS, thereby establishing a positive effect in the customer viewing the object. Similarly, we expect that conditions II and IV, in which the robot or the human is only present (i.e., presence), will not differ. Finally, we expect that conditions I and III will differ from each other as well as from the passive control condition V. Additional results will be presented during the workshop.
Our study implements previous studies on artificial intelligence and extends previous research with a multidisciplinary approach that combines neuro-tools (facial expressions) with self-reports in the study of positive and negative emotions, proposing that consumers perceive service robots (approaching vs non-approaching differently), and examines whether this affects the service encounter during the pre-purchase phase.
Extended Abstract
Due to the rapid digital transformation, service robots are increasingly replacing human service agents (Lewnes, 2021). As a particular form of service robot powered by natural language processing to simulate human conversations (either written or spoken) (Hoyer et al., 2020), chatbots have been put in the spotlight – due to the need for fast contactless online customer service (Stoilova, 2021). They can be implemented on a website, in a messenger, in a mobile App, in a robot, or elsewhere providing users with assistance and information through text, voice (e.g., Siri and Alexa), and a novel way – video. As the first touchpoint for customers on 7/24, service chatbots are widely adopted by firms for saving costs and improving online service. However, when issues get more complex and require individual attention or recovery service, human employees still need to be involved in supporting the problem-solving process with their emotional and social skills. Research reveals that not all applications in service delivery are widely adopted or successful (Davenport et al., 2020). In some cases, the statements suggesting the benefits of chatbots are “self-evident” (Alami et al., 2020). The changes in consumer behaviour and the challenges of service strategies have raised the need for research (Lu et al.,2019). From a theoretical perspective, a comprehensive understanding of consumers’ perceptions of service chatbot adoption is required (Gursoy et al., 2019).
We review the most well-known theories and the most frequent independent variables used in technology adoption studies (see Appendix A). Regarding service chatbots’ adoption, current literature arguably falls short in at least three research areas. First, traditional technology adoption models, such as TAM and UTAUT, were well developed but in the context of technologies whose features and abilities are vastly different from modern AI (Artificial Intelligence) (Lu et al., 2019). Some critical factors are inappropriate for investigating service robots, for instance, perceived ease of use and effort expectancy. Service chatbots possessing human intelligence powered by AI will not hinder adoption (Gursoy et al., 2019), as consumers do not need to learn how to operate chatbots. This limitation suggests that updated constructs and models are required to understand service chatbots’ adoption, i.e., Performance Efficacy (Lu et al., 2019). Beyond the ability to perform a task (i.e., usefulness), performance efficacy addresses specific areas where service chatbots are more competent than human agents (i.e., accuracy and efficiency). Second, prior studies mainly investigated from a technology perspective to understand consumer behaviour. However, adoption is influenced by not only technology characteristics, but also user characteristics (i.e., innovativeness). Drawing on the Diffusion of Innovation Theory and Social Identity Theory, the greater strength of identity makes consumers more sensitive to information relevant to identity, more likely to adopt identity-relevant products and more likely to engage in behaviours that directly implicate the identity (Reed et al., 2012). Consumer innovativeness sheds light on the individual differences when considering service targeting and positioning. Third, scholars call for research on consumer responses to different service types (Paluch and Wirtz, 2020), given service tasks can be distributed between human and robots. According to the Task Technology Fit theory, individuals will adopt a technology based on matching the task requirement and technology characteristics (Goodhue and Thompson, 1995). It is possible that, although consumers perceive technology as being advanced, they do not adopt it if they think it is unfit for their task and cannot improve their performance (Junglas et al., 2008). Little research has been conducted concerning consumers’ perceptions on task technology fit of chatbot services.
In line with the research gaps in the literature, this ongoing research brings TAM, DOI and TTF theories into the chatbot service domain for the first time, aiming to investigate consumers’ perception of service chatbots adoption. An integrated research model (Fig. 1) was proposed to test the key determinants: Perceived Performance Efficacy, Consumer Innovativeness, Trust, Perceived Autonomy, and Perceived Task Technology Fit. Prior usage experience and knowledge of chatbots were control variables. All measurement items were adapted from literature to ensure reliability and measured on a 5-point Likert scale to enable flexibility and prevent the target audience from being too neutral and overwhelmed (Colman et al., 1997).
Fig.1 Research model
This study employed a survey to collect data for examining the research model. A total of 360 responses were collected in January 2022, and 356 responses were identified as valid and then used for data analysis. Structural Equation Modelling was used to test the research model based on the systematic process developed by Hair et al. (1998). We used software SPSS 27 and Amos 27 as data analysis tools.
The initial findings of this ongoing research mostly support the significant effects of the proposed model of consumers’ service chatbot adoption. The results theoretically contribute to the knowledge of Technology Adoption and Task Technology Fit theories in the service marketing domain. Performance efficacy has the strongest and most direct effect on the intention of service chatbot adoption. Whether a chatbot’s performance can meet consumers’ expectations is the most critical factor of adoption in online service. Most importantly, the study confirms the effect of perceived task technology fit on chatbot adoption for the first time. Findings suggest consumers have their preconceived belief on the appropriateness of tasks for service chatbots. The research also found more evidence on the relationship between consumers’ innovative characteristics and adoption behaviour. Surprisingly, prior knowledge and usage experience of chatbots don’t hinder the intention of service chatbot adoption in this study. For practical implications, findings benefit marketing practitioners in deciding on effective service technology strategies for improving consumers’ online service experience. When firms develop and implement service chatbots, they must be aware of consumers’ higher expectations of AI performance. Despite consumers’ concern on chatbot performance, firms still can provide satisfying chatbot services by targeting the right consumers and delivering the right service tasks. For those consumers who are inclined to resist the adoption of new technologies, firms must provide different service options for them to avoid service failure.
1. Purpose
Academics are increasingly concerned about how AI and service robots are shaking up the business world (Chang and Kim, 2022; Haenlein et al., 2022; Hollebeek et al., 2021). Most studies have examined the likelihood of service robots replacing human service personnel, and research into the robot characteristics that customers care about has focused on the one-on-one interactions between customers and robots. However, less studies have explored how the interaction between customers and service personnel affects the customers’ intentions and general experiences of receiving services from service robots.
In this study, we explored the contexts in which customers require robot services from the perspective of integrated services. Specifically, we examined the contexts in which robots can compensate for the skills that human service personnel lack to provide the highest quality of service to customers. We assumed that the customers’ attachment styles with service personnel either strengthen or weaken their perceptions of the usefulness, social capability, and appearance of robots, which in turn affect their perceptions of the quality of robot services and their intentions to replace human service personnel with service robots. Attachment styles, which originated from attachment behavior theory (Bowlby, 1979), refer to people’s adjustment of their own behavior to establish relationships with their primary caregivers during childhood according to the messages that such caregivers convey to them. Ainsworth et al. (1978) divided attachment styles into three types: secure, anxious, and avoidant. Attachment styles are relevant to specific interpersonal relationships.
Mende, Bolton, and Bitner (2013) applied the theory of attachment styles to examine the relationship between business companies and customers. The results obtained indicated that customers with avoidant attachment styles consider certain companies untrustworthy and themselves as undeserving of favor from service personnel. They may also believe that they can look after themselves and do not need help from companies. Customers with anxious attachment styles exhibit uncertainty toward their self-worth and are unsure of whether companies would provide them with the help they need in a timely manner because the behaviors of others are unpredictable. Consequently, they have a low sense of control toward the services provided by companies. To provide high-quality services, these relationships between customers and service personnel can be supplemented using robots. The research framework is presented in Fig. 1.
2. Methodology
A questionnaire was distributed to the customers of service companies that use service robots. The respondents were confirmed to have received robot services before the questionnaire, such as hotel tours, store services, smart ordering services, and hospital services. A total of 345 responses were received, 58% and 42% of which were provided by female and male respondents, respectively. In total, 40.9% of all respondents were aged 21–30, and 73.3% had received college or university education (see Table 1). Regarding the reliability and validity of the responses, all the questionnaire dimensions exhibited Cronbach’s α higher than 0.88, average variance extracted no lower than 0.65, and factor loadings attaining the level of significance. Discriminant validity analysis (Table 2) indicated that the square root of the average variance explained for each dimension was higher than the correlation coefficient between each pair of dimensions, indicating satisfactory overall variable reliability and validity.
3. Findings and conclusions
Regression analysis revealed the following results (see Table 3):
(1) The usefulness, social capability, and appearance of robots reinforce the quality of their services, which in turn increases the customers’ intentions to replace human service personnel with service robots. Thus, H1–H4 are supported (p < 0.01). Customers using service robots care about whether these robots provide services with an increased efficiency, answer questions appropriately, and are attractive.
(2) Customer attachment avoidance considerably reinforces the effect of robot usefulness and social capability on the quality of their services but weakens that of their appearance on the quality of their services. Accordingly, H5 is partially supported (p < 0.01). In other words, the usefulness and social capability of service robots are required aspects for human service personnel to deal with customers with attachment avoidance. However, more attractive robots are regarded as more questionable in terms of the quality of their services.
(3) Customer attachment anxiety considerably weakens the effect of robot appearance on the quality of their services but does not substantially weaken that of their usefulness and social capability on the quality of their services. Thus, H6 is partially supported (p < 0.01). Customers with attachment anxiety are unlikely to be concerned about the appearance of service robots. Accordingly, appearance may not be a robot feature complementary to human service personnel.
4. Originality/Value
This study explored how service robots complement human service personnel in providing a satisfactory customer experience from the perspective of integrated services and the relationship between customers and service personnel. According to various perspectives, the attachment styles of customers with service personnel substantially affect their perceived importance of the quality of robot services, including their usefulness, social capability, and appearance. Business owners should note that customers who have avoidant attachment styles with human service personnel are particularly sensitive to the usefulness and social capability of service robots; that is, instead of the appearance of robots, they are more concerned about whether robots can efficiently solve their problems, hence compensating for the lack of efficiency in human service personnel. The results of this study provide a key reference to academics and service industries for improving their customers’ service experience by using robots or artificial intelligence with customer-centered thinking. Promoting the interactions among customers, human service personnel, and service robots is of utmost importance.
1. PHENOMENON UNDER INVESTIGATION
Service robots are increasingly implemented in service industries, including the hospitality industry (Ivanov & Webster, 2017; Naumov, 2019). Service robots taking over frontline tasks will inevitably impact guests’ experience. While the success of robots is frequently evaluated based on their task execution, in hospitality settings, robots’ impact on guest experience is of predominant interest. Recently, researchers contributed insights based on hypothetical scenarios (Belanche et al., 2021; Choi et al., 2020; Hoang & Tran, 2022), or analysis of review data (Huang et al., 2021). Consequently, there is a need for research in real-life experiments, capturing the effects that human-robot interaction has on guest experience compared to traditional human-human interaction. Our study presents insights from cross-context field experiments to measure the impact of service robots on the hospitality guest experience.
2. THEORETICAL FOUNDATIONS & POTENTIAL CONTRIBUTIONS TO THE FIELD
This study contributes to two streams of discussion. First, we contribute to the debates about digital transformation in the hospitality industry. Digital technologies such as artificial intelligence, automation, and robots are transforming service industries. Due to the importance of the ‘frontline’, researchers position the hospitality industry as a fruitful research context to further understand digital transformation in service industries (Buhalis et al., 2019; Wirtz et al., 2018). To this end, we follow recent studies and conceptualise digital transformation as a “socioeconomic change across individuals, organizations, ecosystems, and societies that is shaped by the adoption and utilization of digital technologies” (Dabrowska et al., 2022, p. 3). Specifically, we study the co-existence and interdependence of humans and service robots on an individual level (Dabrowska et al., 2022; Tung & Law, 2017).
Second, we contribute to the literature at the crossroads of service robots and guest experience. Customer experience management has become the most promising approach for creating loyalty and other positive result variables and has been addressed in academic literature and practice through a wide array of industries (Homburg et al., 2017; Lemon & Verhoef, 2016). Designing, managing, and measuring experiences has become a key topic in the experience driven hospitality industry. As such, this requires understanding of stakeholder experiences along the touchpoints within the entire customer journey, their distinct roles and behaviours and the influencing factors (Kranzbühler et al., 2018). These insights address possible cause-effect relationships for effective interventions on the touchpoint level in the journey, such as the encounter with a service robot.
Given the need to understand the impact of service robots on the overall guest experience along the entire customer journey, we complement recent qualitative and conceptual studies (Belanche et al., 2021; Choi et al., 2020; Ivanov & Webster, 2017), and specifically contribute to the call for comparing “customer experiences across different contexts using quantitative research” (Huang et al., 2021). By measuring the touchpoint experiences ‘in the moment’, as the real time experiments of this study do, we contribute to the understanding of the guest experience as a response to managerial stimuli (Becker & Jaakkola, 2020). We provide insights how the application of service robots impacts hospitality experience, ultimately affecting guest satisfaction and loyalty (Pijls et al., 2017).
3. RESEARCH QUESTION
To what extent does a robot-human service encounter, compared to a human-human encounter, impact the guest’s hospitality experience?
4. METHODOLOGY
We conducted two experiments addressing hospitality company-owned touchpoints (Lemon & Verhoef, 2016), i.e., human encounters with a service robot.
Experiment A (n=135) concerns the implementation of an information provision robot in a hotel lobby. Experiment B (n=151) evaluates the guest experience of robots that deliver food and beverages in a fast-food restaurant. In both experiments, guests were randomly assigned to either a human-human or a robot-human interaction. Data on the guest experience were collected via surveys provided to guests soon after experience of the interaction. We build on established constructs to assess hospitality experience in a service context (Pijls et al., 2017). Hence, we measured experiential factors inviting, care and comfort and relate it to overall experience and satisfaction.
5. FINDINGS
In the information provision experiment (A), in comparison to human-human interaction, robots have a significant negative effect on the experiential factor ‘inviting’ and the outcome variable of perceived ‘overall hospitality’. In the restaurant experiment (B), food delivery robots influence the overall hospitality guest experience positively on the dimensions of ‘overall satisfaction’ and ‘overall experience’ of hospitality.
6. DISCUSSION
Our findings from real-life experiments suggest a confirmation of earlier hypothetical studies arguing that guests deem robots more capable of functional, repetitive work such as transporting items and not of social, interactive tasks, such as speaking with guests (Ivanov & Webster, 2019).
Moreover, the two field experiments measure the impact of robot-human (compared to human-human) service encounters on touchpoint hospitality experiences and the evaluative outcomes of it. Interestingly, the direction of the impact is different in both experiments. Also, the experiential dimensions involved are different. In both studies an impact on the overall outcomes in terms of satisfaction and experience has been found. The different directions of the outcome underline that the robot-human encounter experience is subjective and context-specific. We provide empirical evidence for earlier conceptual contributions, underlining the importance of considering customer, situational, and sociocultural contingencies (Wirtz et al., 2018). Future research should investigate key contingencies and specify different responses in relationship to different stimuli in human-robot interaction.
Lastly, our experimental study joins discussions on digital transformation in the hospitality industry. We deliver empirical insights how service robots can enhance (or reduce) guest experience in the frontline (Liu & Hung, 2021). Considering the relevance of guest experiences in hospitality settings, we present an early attempt to understand how to design effective service robots and where to implement them in frontline touchpoints. Thereby, we enable hospitality professionals to better design and manage effective robot driven service encounters. Our study provides first cues regarding guests’ utilization of service robots, thereby presenting first insights into the adoption of digital technologies in a hospitality context (Dabrowska et al., 2022; Tussyadiah et al., 2020).
Comply or resist? The use of service robots for law enforcement
Qingxuan Zhang and Liliana L. Bove
The University of Melbourne, Australia
The global pandemic increased the need for monitoring and enforcement of COVID-related policies such as mandatory vaccinations, face mask-wearing, travel restrictions, and quarantine which were largely carried out by law enforcement officers (Kugler et al. 2021). This increase in demand for law enforcement officers, in addition to their greater risk of exposure brought about by their close contact with the public (Jennings and Perez, 2020) resulted in many countries, such as Australia (Opie, 2022), the US (Lee, 2021), and Canada (Piapot et al., 2022) experiencing severe staff shortages. At times officer shortage was so acute that police stations were forced to close (The Age, 2021). A potential solution to this chronic shortage is the use of robots to monitor and enforce COVID compliance.
The use of robots in the law enforcement sector was estimated to be US$ 1.6 billion in 2020 and is expected to increase to US$ 4.2 billion by 2027 (GIA, 2021). With recent advancements in technology, robots can carry out tasks that were previously only reserved for human law enforcement officers. For instance, robots, equipped with state-of-the-art cameras, sensors, tracking technology, and connectivity capabilities, can now conduct autonomous surveillance without the presence of human officers (Market Trends, 2021).
Despite the large number of studies that explore how service robots can enhance customer comfort (Lin et al., 2022), bring enjoyment (Van Pinxteren et al., 2019), and elicit compensatory consumption (Mende et al., 2019) in multiple settings, scant attention has been paid to contexts in which the primary role of the service robot is to gain customer compliance. The overall service experience is largely unexplored, and the unintended consequence of such robot uses on consumers are unclear.
In response to these gaps, this study aims to explore citizen attitudes and emotional reactions towards service robots in law enforcement settings. Drawing on a Norwegian study where fishers had a lower acceptance rate of modern control activities (e.g., remote monitoring with drones) compared to traditional means of control (e.g., physical inspection by coastguards) (Diekert et al., 2021), we anticipate citizens will feel and act similarly towards service robots in law enforcement settings.
Of further interest is understanding citizen behavior in law enforcement settings where the rules and regulations are deemed to be unfair and not legitimate. Underpinned by reactance theory, we predict that when an individual’s freedom is threatened, the threatened behavior becomes more attractive (Brehm, 1966) and reactance, manifested in negative emotions such as offense (Heide et al., 2007), hostility, aggression, anger and frustration (Clee and Wicklund, 1980; Miron and Brehm, 2006) is aroused. In these situations, the affected individual may be motivated to restore freedom by seeking opportunism to misbehave towards the source of the threat (Dillard and Shen, 2005) i.e., the robot.
Using in-depth interviews of Chinese citizens who have experienced interactions with service robots charged with COVID-related control activities, we will explore how robot characteristics such as level of anthropomorphism and/or biomimetricity impact customer emotional attitudes, and compliance or reactance behaviors.
Robots as law enforcers
Singapore appears to have led the way in the use of robots for law enforcement possibly due to the high level of trust their citizens place on the government (Rieger and Wang, 2022). It first trialed the use of a robot dog called Spot, which carried a loudspeaker to broadcast coronavirus-related messages to enforce social distancing in parks (BBC, 2020). It also deployed a Boston Dynamics robot dog with sensors and speakers at a reservoir to warn people about the laws against loitering and gathering (Su, 2020). More recently, Singapore introduced a pair of patrol robots, dubbed Xavier, to augment the work of public officers by helping to enforce COVID-19 protocols and deter other undesirable civic habits (e.g., smoking in banned areas) in a busy shopping district (Barrett, 2021). When Xavier detects any undesirable behaviors, it sends real-time alerts to the command-and-control center, and officers can choose to respond in person or remotely by displaying an appropriate message on the robot’s interactive dashboard (Adams, 2021; HTX, 2021).
Similarly, humanoid anti-pandemic robots, dubbed Jasiri, are put to work at an international airport in Nairobi, Kenya (Reuters, 2021). Jasiri, checks passengers’ temperatures, tells those not wearing masks to put them on, and enforces social distancing rules with a camera mounted on an extendable neck (Reuters, 2021).
Methodology
Given the scarcity of research as to how consumers react to robots used in law enforcement settings exploratory research in the form of in-depth interviews will be used. A phenomenological approach will be taken to obtain a first-person description of personal experience with a law-enforcement robot.
The first context investigates several eastern Chinese cities where robot dogs, referred to as Preserved Egg, a famous Chinese dish, patrol the streets three or four times a day and help the authorities spread Covid-related messages, such as calling for mandatory COVID-19 tests or encouraging people to vaccinate (McMorrow, 2022). The second context is in several major airports in China where service robots like Xavier are used to check travelers’ body temperature and QR codes related to past traces.
Participants will be recruited through snowball sampling beginning with personal contacts and then following up with informant referrals (Noy, 2008). They will also be encouraged to use their native language to allow freer, flowing conversation. The messaging app WhatsApp will be used as the platform for communication because of its free access and assured anonymity to participants with its end-to-end encryption (WhatsApp, 2022). While videoconferencing allows the interview to be recorded, due to the perceived sensitivity of the topic, informants will be asked if interviews can be restricted to audio recording to allow verbatim transcription. Participants will also be assured of anonymity, referred only to by their pseudonym during the interview. If audio recording approval is not given, extensive notes will be taken during the interview.
To ensure a generalizable sample across various age groups, educational backgrounds, and risk preferences, individuals such as elderly people who do not use WhatsApp will be interviewed in person via an intermediary family member who is a WhatsApp user.
References
Adams, E. (2021). Singapore is testing robots to patrol the streets for “undesirable” behavior like smoking. USA TODAY. Retrieved August 5, 2022, from https://www.usatoday.com/story/tech/news/2021/09/07/robots-now-patrol-streets-singapore-undesirable-behaviors/5753334001
Barrett, C. (2021). “Exterminate bad habits”: Meet Xavier, Singapore’s COVID-compliance robot. The Sydney Morning Herald. Retrieved August 6, 2022, from https://www.smh.com.au/world/asia/exterminate-bad-habits-meet-xavier-singapore-s-covid-compliance-robot-20210908-p58q1c.html
BBC (2020). Robot dog enforces social distancing in city park. BBC. Retrieved August 5, 2022, from https://www.bbc.com/news/av/technology-52619568
Brehm, J. W. (1966). A theory of psychological reactance. New York: Academic Press.
Clee, M. A. & Wicklund R. A. (1980). Consumer behavior and psychological reactance. Journal of Consumer Research, 6 (4), 389-405.
Diekert, F., Nøstbakken, L., & Richter, A. (2021). Control activities and compliance behavior—Survey evidence from Norway. Marine Policy, 125.
Fang, S., Han, X., & Chen, S. (2022). The Impact of Tourist–Robot Interaction on Tourist Engagement in the Hospitality Industry: A Mixed-Method Study. Cornell Hospitality Quarterly. https://doi.org/10.1177/19389655221102383
GIA. (2021). Security and Law Enforcement Robots. (n.d.). Www.marketresearch.com. Retrieved August 9, 2022, from https://www.marketresearch.com/Global-Industry-Analysts-v1039/Security-Law-Enforcement-Robots-14406125/
Heide, J. B., Wathne, K. H. & Rokkan, A. I. (2007). Interfirm monitoring, social contracts, and relationship outcomes. Journal of Marketing Research, (August), 425-433.
Jennings, W. G., & Perez, N. M. (2020). The Immediate Impact of COVID-19 on Law Enforcement in the United States. American journal of criminal justice: AJCJ, 45(4), 690–701. https://doi.org/10.1007/s12103-020-09536-2
Kugler, M. B., Oliver, M., Chu, J., & Lee, N. R. (2020). American law enforcement responses to COVID-19. J. Crim. L. & Crimin. Online, 111, 19.
Lee, R. (2021). Officials say there’s a nationwide police shortage and it’s affecting Polk County as well. The Ledger. Retrieved August 6, 2022, from https://www.theledger.com/story/news/2021/09/18/nationwide-polk-county-florida-police-shortage-experts-say-its-not-because-covid-19/8164249002/
Market Trends. (2021). How are robotics applications facilitating law enforcement. www.analyticsinsight.net. Retrieved August 15, 2022, from https://www.analyticsinsight.net/how-are-robotics-applications-facilitating-law-enforcement/#:~:text=In%20law%20enforcement%2C%20these%20robots
McMorrow, R. (2022). The robot dogs policing Shanghai’s strict lockdown. Financial Times. Retrieved August 8, 2022, from https://www.ft.com/content/5c437146-2d18-466b-84af-24a47b32de59
Noy, Chaim. (2008), “Sampling Knowledge: The Hermeneutics of Snowball Sampling in Qualitative Research,” International Journal of Social Research Methodology, 11 (4), 327–44.
Opie, R. (2022) “It’s failing the community of South Australia”: Union calls for policing model overhaul. ABC News. Retrieved August 7, 2022, from https://www.abc.net.au/news/2022-06-20/sa-police-union-calls-for-policing-model-overhaul/101166688
Piapot, N., Freeze, C., & Woo, A. (2022). COVID-stricken police forces across Canada struggle with shortage of officers. The Globe and Mail. Retrieved August 7, 2022, from https://www.theglobeandmail.com/canada/article-covid-stricken-police-forces-across-canada-struggle-with-shortage-of/
Reuters. (2021). Airport robots give hi-tech boost to Kenya’s Covid-19 fight. Deccan Herald. Retrieved August 8, 2022, from https://www.deccanherald.com/international/airport-robots-give-hi-tech-boost-to-kenyas-covid-19-fight-951888.html
Richter, A., & Grasman, J. (2013). The transmission of sustainable harvesting norms when agents are conditionally cooperative. Ecological Economics, 93, 202–209.
Rieger, M.O., Wang, M. Trust in Government Actions During the COVID-19 Crisis. Soc Indic Res 159, 967–989 (2022). https://doi.org/10.1007/s11205-021-02772-x
Su, E. (2020). Roaming “robodog” tells Singapore park goers to keep their distance. The Sydney Morning Herald. Retrieved August 5, 2022, from https://www.smh.com.au/world/asia/roaming-robodog-tells-singapore-park-goers-to-keep-their-distance-20200509-p54rcs.html
The Age. (2021). Calls for police to return to Melbourne as COVID response triggers station closures. Tpav.org.au. Retrieved August 7, 2022, from https://tpav.org.au/news/news-coverage/2021-news-coverage/calls-for-police-to-return-to-melbourne-as-covid-response-triggers-station-closures
WhatsApp. (2022). WhatsApp.com. Retrieved August 7, 2022, from https://www.whatsapp.com/privacy
Robots, including in-store physical robots and virtual chatbots, have been increasingly employed to substitute human employees in customer service provision. These customer-facing service robots (SRs) can be defined as “system-based autonomous and adaptable interfaces that interact, communicate and deliver service to an organisation’s customers” (Wirtz et al., 2018, p.909). For these roles, SRs are often programmed to simulate customer-facing employees, engaging in different tasks with different intelligences and mimicking human-human interaction (Huang and Rust, 2021; Pantano and Scarpi, 2022). As a part of this, SRs may also be designed with physical and/or non-physical humanlike features, which can be stereotypically feminine or masculine, and which can affect customers’ emotional, cognitive, and behavioural responses (e.g., Blut et al., 2021; Tay et al., 2014). Despite a growing number of empirical studies, there is currently a lack of comprehensive synthesis of the extant literature related to the impacts of anthropomorphic design and gendering of SRs on consumer behaviour, which is vital to determine how further research can make meaningful contributions.
Anthropomorphism in SR literature refers to either perceived human-likeness (Blut et al., 2021) or design attributes of a robot (Lu et al., 2020; Wirtz et al., 2018). Anthropomorphism can affect customers’ reactions towards SR interactions in various service contexts, including public service (van Pinxteren et al., 2019), home service (Letheren et al., 2021), hospitality (Lv et al., 2021; Yam et al., 2021; Yoganathan et al., 2021), and retail (Whang and Im, 2021). Customers’ resultant emotional, cognitive, and behavioural responses can be positive or negative. For example, customers served by humanoid robots have a higher tendency of adoption (Blut et al., 2021; Sheehan et al., 2020) and re-use intention (Moriuchi, 2021); higher satisfaction (Yam et al., 2021); more positive evaluation (Li and Sung, 2021); closer emotional connection (Araujo, 2018); and more willingness to purchase (Yoganathan et al., 2021). On the other hand, researchers reveal that anthropomorphism of robots has the potential to harm human-robot interactions, causing greater discomfort and negative attitudes toward the robot (e.g., Kim et al., 2019; Mende et al., 2019). However, research on anthropomorphism of SRs has focused more on positive than negative effects such that less is known about the latter, especially the outcomes brought by the uncanny valley theory in customer-SR interactions (Wirtz et al., 2018).
Robot gendering refers to assigning gender onto a robotic platform through characteristics such as name, voice, and physique (Bryant et al., 2020; Robertson, 2010). Prior research examining effects of anthropomorphism of SRs often conflates gender with anthropomorphism by using gender-stereotypical features or characteristics to convey human-likeness (Crolic et al., 2021; Yam et al., 2021). However, gender-stereotypes have long been recognised in a technology context (e.g., Nass et al., 1997) and can impact how customers perceive and evaluate services (Pitardi et al., 2022a). For instance, Tay et al. (2014) demonstrate that participants show more positive evaluation and greater acceptance of a female-gendered healthcare robot and a male-gendered security robot due to gender-occupational role stereotypes. Broadly speaking, male-gendered robots are perceived to deliver a sense of intelligence, dominance, and competence while female-gendered robots provide a feeling of affection, communality, and particularly warmth (Ahn et al., 2022; Borau et al., 2021; Eyssel and Hegel, 2012; Seo, 2022) which plays a crucial role in customer-SR interactions (Choi et al., 2021). These findings raise important questions regarding how researchers and practitioners imbue robots with humanlike features.
Research on SRs in specific situations is beginning to emerge. During service failure, customers attribute more blame to a human employee than a SR (Belanche et al., 2020a; Leo and Huh, 2020) due to the robot’s perceived low level of agency (Gray et al., 2007). However, it is less clear how blame would be attributed to a humanlike versus non-humanlike SR, considering the impact of perceived thoughtfulness and/or agency brought by human-likeness (Waytz et al., 2014; Yam et al., 2021). While humanoids are more able to recover service failure than non-humanoids (Choi et al, 2021), it is unclear whether there would be any differences caused by gendering. Another important context to note is embarrassing service encounters (e.g., body measurement) in which a humanlike SR can have unfavourable effects on consumer embarrassment due to its resemblance with humans (Pitardi et al., 2022b). However, it is unclear whether human-robot gender congruity versus incongruity (Pitardi et al., 2022a) would have more or less favourable effects in embarrassing service encounters.
The field of SR research is in its infancy (Wirtz et al., 2018) and prior research is fragmented (Lu et al., 2020). Pioneering studies have systematically identified key dimensions of robot-delivered frontline service (Wirtz et al., 2018) and synthesised the impacts of SRs on customers and service employees (Lu et al., 2020), indicating the non-negligibility of anthropomorphic design in customer-SR interactions. Although van Pinxteren et al. (2020) innovatively conducted a systematic review of humanlike communications in conversational agents, a comprehensive literature analysis of anthropomorphism and gender of SRs and their impacts on human-robot service interactions is lacking. Aiming to overcome the gap and highlight the importance of discussing human-likeness of non-human service agents, we will undertake a systematic review to investigate the effects of anthropomorphism and gendering of SRs. The paper will provide a definition of gender of a SR which is missing in previous studies, describe the relationship between anthropomorphism and gender, and clarify the role of anthropomorphism and gender in human-robot service interactions. It will also identify knowledge gaps and key areas of interest via a research agenda for humanlike SRs. The paper will contribute to management and marketing literature as well as the SR literature by answering several research calls for understanding a robot’s design in general (Lu et al., 2020) and the robot’s humanlike characteristics (Blut et al., 2021; Choi et al., 2021; Xing et al., 2022) including gender (Belanche et al., 2020b; Choi et al., 2021) in particular. It could also provide theoretical support for organisations to deliver better service for consumers.
THEORETICAL BACKGROUND
How would consumers feel when service robots fail and what are the behavioral consequences? When people perceive that a negative event cannot be resolved and altered in the future, they feel helpless (Gelbrich, 2010). Past research has documented that consumers show more negative responses when a service failure is attributed to stable causes (Van Vaerenbergh et al., 2014; Weiner 1985). In the service technology context, Belanche et al. (2020) found that consumers tend to make stronger attribution of stability to service failures caused by service robots than by humans. This could be because people tend to perceive automated systems as less flexible and adaptable (Leo & Huh, 2020) and require a substantial amount of time and effort to fix functional issues (Joyeux & Albiez, 2011). Thus, consumers may expect that a service robot’s error is more likely to be stable (i.e., occur constantly in the future), thereby feeling helpless (Gelbrich, 2010; Luse & Burkman, 2022).
Extending this notion, we argue that consumer helplessness will be contingent upon the failure type they encounter. Based on the theory of mind perception (Gray et al., 2007), past research suggests that while humans are seen as having greater minds (both high in agency and experience), robots are typically perceived as having moderate levels of agency but low levels of experiential mind (Gray & Wegner, 2012). In other words, people expect service robots to possess a great level of functional capabilities, while lacking social-emotional capabilities, which are key determinants of outcome and process failures, respectively (Choi et al., 2021; Smith et al., 1999). Also, while mechanical and thinking intelligence can be easily replaced by robots, feeling intelligence is hardly carried out by them (Huang & Rust, 2021). As a result, consumers who encounter a process failure by a robot view that the issue can hardly be fixed, resulting in a high level of helplessness. In contrast, an outcome failure by a service robot can be perceived as easily resolved by, for example, a quick system check or machine learning (Heller, 2019). In other words, a process (vs. outcome) failure by a robot is perceived to be nearly irresolvable, eliciting stronger helplessness.
Further, the helpless consumers are more likely to engage in negative word-of-mouth, but less likely to directly complain to the company because they doubt the service provider can remedy the service failure while still needing to vent their negative emotions (Gelbrich, 2010). Thus, we hypothesize:
H1. Consumers are more likely to engage in NWOM, but less likely to directly complain to the organization when service process (vs. outcome) failure is caused by a robot.
H2. The effect of robot’s service failure type (process vs. outcome) on complaint behavior (NWOM, direct complain) will be mediated by helplessness.
How can managers let consumers believe that their robot is capable of fixing social-emotional functions and expect the same error would not occur in the future? Empirical evidence has shown that subtle cues (e.g., appearance) make robots to be perceived as warmer, and consumers tend to respond less negatively to their mistakes (Xu & Liu, 2022; Yam et al., 2021). Moreover, adding warmth cues to robots can enhance perceived social-emotional skills (Choi et al., 2021). Therefore, we propose that adding warmth cues to robots can make consumers to believe their social mistakes (i.e., process failures) could be resolved, mitigating helplessness.
H3. In a process failure, the effect of robot’s failure on coping strategy mediated by helplessness will be mitigated when the robot has warmth cues. There will be no such effect of warmth cue in an outcome failure.
METHODOLOGY
Study 1
To test H1-2, Study 1 will employ a 2 (failure type: process vs. outcome) between-subjects experimental design. Participants will first read a scenario describing a service failure at a restaurant deploying service robots. Failure type will be manipulated by adapting previous empirical study scenarios (e.g., Choi et al. 2021). After reading the scenario, participants will be asked to indicate their intention to complain directly to the manager (Evanschitzky et al., 2011) and NWOM intention (Zeithaml et al., 1996). Also, helplessness will be measured (Gelbrich, 2010), followed by anger and frustration as alternative explanations (Gelbrich, 2010), and perceived severity of the service failure and propensity to complain as potential control variables.
Study 2
To test H3, Study 2 will employ a 2 (warmth cues on a robot: present vs. absent) by 2 (failure type: process vs. outcome) between-subjects experimental design. Participants will first read a scenario describing a process or outcome failure by a service robot at a hotel (Choi et al. 2021). Warmth cues will be manipulated by adding a broad smile and cute appearance to the robot image, while such cues will not be applied in the cue absence condition, following Liu et al. (2022). Pretest will be conducted to verify the stimuli. After reading the scenario, the same items will be measured as in Study 1.
EXPECTED CONTRIBUTION
When a robot fails, consumers may not complain directly to the firm but spread negative words particularly when they experience a process failure. We argue that this is due to consumer helplessness such that consumers do not expect the problem to be resolved. Thus, managers should pay extra attention in monitoring how their robots “behave” towards customers. By adding warmth cues to robots, managers can expect direct complaints from customers, which allow them to take immediate action and ultimately improve service quality.
The current research expects to add to the literature on service robot failure by examining the role of helplessness, which has been underexplored in the service robotics research, and its unique consequences (Gelbrich, 2010).
References
Belanche, D., Casaló, L. V., Flavián, C., & Schepers, J. (2020). Robots or frontline employees? Exploring customers’ attributions of responsibility and stability after service failure or success. Journal of Service Management, 31(2), 267-289.
Choi, S., Mattila, A. S., & Bolton, L. E. (2021). To err is human (-oid): how do consumers react to robot service failure and recovery?. Journal of Service Research, 24(3), 354-371.
Evanschitzky, H., Brock, C., & Blut, M. (2011). Will you tolerate this? The impact of affective commitment on complaint intention and postrecovery behavior. Journal of Service Research, 14(4), 410-425.
Gelbrich, K. (2010). Anger, frustration, and helplessness after service failure: coping strategies and effective informational support. Journal of the Academy of Marketing Science, 38(5), 567-585.
Gray, H. M., Gray, K., & Wegner, D. M. (2007). Dimensions of mind perception. science, 315(5812), 619-619.
Heller, M. (2019), “Machine Learning Algorithms Explained,” InfoWorld (May 9), https://www.infoworld.com/article/3394399/machine-learning-algorithms-explained.html.
Huang, M. H., & Rust, R. T. (2021). Engaged to a robot? The role of AI in service. Journal of Service Research, 24(1), 30-41.
Leo, X., & Huh, Y. E. (2020). Who gets the blame for service failures? Attribution of responsibility toward robot versus human service providers and service firms. Computers in Human Behavior, 113, 106520.
Liu, X. S., Yi, X. S., & Wan, L. C. (2022). Friendly or competent? The effects of perception of robot appearance and service context on usage intention. Annals of Tourism Research, 92, 103324.
Luse, A., & Burkman, J. (2022). Learned helplessness attributional scale (LHAS): Development and validation of an attributional style measure. Journal of Business Research, 151, 623-634.
Smith, A. K., Bolton, R. N., & Wagner, J. (1999). A model of customer satisfaction with service encounters involving failure and recovery. Journal of marketing research, 36(3), 356-372.
Van Vaerenbergh, Y., Orsingher, C., Vermeir, I., & Larivière, B. (2014). A meta-analysis of relationships linking service failure attributions to customer outcomes. Journal of Service Research, 17(4), 381-398.
Gray, K., & Wegner, D. M. (2012). Feeling robots and human zombies: Mind perception and the uncanny valley. Cognition, 125(1), 125-130.
Weiner, B. (1985). An attributional theory of achievement motivation and emotion. Psychological review, 92(4), 548-573.
Xu, X., & Liu, J. (2022). Artificial intelligence humor in service recovery. Annals of Tourism Research, 95, 103439.
Yam, K. C., Bigman, Y. E., Tang, P. M., Ilies, R., De Cremer, D., Soh, H., & Gray, K. (2021). Robots at work: People prefer—and forgive—service robots with perceived feelings. Journal of Applied Psychology, 106(10), 1557-1572.
Zeithaml, V. A., Berry, L. L., & Parasuraman, A. (1996). The behavioral consequences of service quality. Journal of marketing, 60(2), 31-46.