Retail robots as sales assistants: how speciesism moderates the effect of robot intelligence on customer perceptions and behaviour Free

“Robots declared new ‘species’ by AI expert”—this 2020 headline of a story in The Daily Star, a British tabloid newspaper, may sound rather sensational; however, the presence of frontline service robots in store aisles of retail shops is no longer futuristic (Bogue, 2019; Roozen et al., 2023). Service robots, which are 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), are increasingly being deployed alongside human staff to interact with customers. Examples include the home improvement company Lowe's, which launched “LoweBot” to greet and assist customers (Taylor, 2016), and Giant Food Stores, which introduced “Marty” to point out hazards for store staff to clean up (Green, 2019).

The focus of this research is on humanoid social robots (HSRs), which are autonomous AI-based service robots with humanoid embodiment, intellectual intelligence and social interactivity. Examples of HSRs include SoftBank Robotics “Pepper”, and Hanson Robotics “Sophia”. These robots can listen to customers and react to their needs with expressive body movements. In some Asian countries, retailers have begun to deploy HSRs to interact with customers, for example, UBTECH's Cruzr robot in China, or Toshiba's ChihiraAico, an “uncannily human-like robot” in Japan (Boyd, 2019). In the next few years, the retailing industry is expected to incorporate HSRs to work alongside human staff to create a hybrid human-robot workforce to enhance the in-store customer service experience (Noble et al., 2022; Underwood, 2020). Industry reports estimate that the market size for service robots will grow in double digits, from USD 16.35 billion in 2022 to more than USD 62.35 billion by 2030 (Fortune Business Insights, 2023).

The significant rise of HSRs has captured the attention of marketing academics (e.g. Čaić et al., 2018; Grewal et al., 2021; Rindfleisch et al., 2022). While many studies in robotics deal with robot design, services research has explored the role of HSRs in frontline service operations and the impact of human-robot interactions on consumer experiences (e.g. Čaić et al., 2018; Fuentes-Moraleda et al., 2020). Studies in this area have focused on how HSRs' physical, behavioural, cognitive and emotional characteristics facilitate human-robot interactions. However, Lu et al. (2020) conclude that research remains fragmented, and in some cases, empirical findings are even contradictory. Furthermore, scholars acknowledge that empirical studies of HSRs in a retail environment are nearly non-existent (Belanche et al., 2020; Roozen et al., 2023).

In light of rapid developments and innovations in humanoid robotics, HSRs are moving towards an “advanced created artificial species” (Søraa, 2017), close to being indistinguishable from human beings, showing cognitive, emotional and social abilities. In this context, some researchers suggested investigating how speciesism would influence customer perceptions and acceptance of humanoid robots (Schmitt, 2019, 2020). The concept of speciesism, introduced in the 1970s, explores human cognitive discrimination or prejudice against other non-human species (i.e. animals) (Amiot et al., 2020; Singer, 1973). Research in human-computer interaction has proposed redefining the concept of speciesism and expanding the scope of speciesism from human-animal relationships to human-AI relationships (e.g. Schmitt, 2020). For example, Huo et al. (2023b) found that speciesism moderates the relationship between staff participation and AI anxiety. Despite the notable work by Huo et al. (2023b) on the boundary role of speciesism, there is limited empirical evidence of how humans may prejudicially respond to HSRs. And while speciesism may be widespread, the causes of the phenomenon are not well understood. Building on the existing research gaps, our research aims to explore the role of speciesism as a moderator of the relationship between robot intelligence and customer perceptions and behaviour. We consider perceived competence as the psychological mechanism that mediates this effect. For example, prior research has found that a human (vs robotic) acknowledgement increases customers' loyalty responses; an effect mediated by the perceived competence of the service workers (Frank and Otterbring, 2023). Wu and Huo (2023) also confirms a mediating role of perceived competence between the introduction of service robots in hotels and customer satisfaction.

Our study makes two important contributions. First, our study integrates several literature streams, including human-robot interaction and service robot implementation (Belanche et al., 2020), retailing in a new technology era (Grewal et al., 2021; Noble et al., 2022) and speciesism in human-AI relationships (Huo et al., 2023a; Schmitt, 2019, 2020). We respond to calls to add more theoretical frameworks and empirical studies, as the field of service robots is still in its infancy (Lu et al., 2020). We also add to the novel research on speciesism in human-AI relationships. We demonstrate that the effect of HSR intelligence on consumer purchase decision is moderated by speciesism as a boundary condition, highlighting the importance of group threat viewpoint in human-robot interaction (Huo et al., 2023a). Second, we respond to calls in research to understand how in-store retail robots affect shopping behaviour (Grewal et al., 2021) and show that this interactive effect is mediated by customers' perceived competence towards HSRs while ruling out potential alternative mechanisms (i.e. perceptions of identity threat and anthropomorphism). Thus, our study adds to growing research on consumer-facing retail technologies (Inman and Nikolova, 2017) and the effectiveness of human-robot service encounters (Roozen et al., 2023).

The remainder of this article is organised as follows. The next section reviews the theoretical background and describes the hypotheses development. Next, we describe the method and experimental design, followed by the data analysis and results. We conclude with a discussion of theoretical and practical contributions, limitations and directions for future research.

HSRs are rapidly altering the in-store retailing and customer experience (Noble et al., 2022). According to Aldebaran, formerly known as SoftBank Robotics Europe, typical use cases for HSRs in retail include welcome assistant and receptionist, information agent, sales assistant, concierge, feedback collector, survey conductor and brand ambassador (Aldebaran, 2023). Unlike other shopping-facing technologies, HSRs can engage in complex interpersonal communications and facilitate the in-store consumer shopping experience (Song and Kim, 2022). Different studies highlight that HSRs' appearance (i.e. human-likeness), functionality (i.e. intellectual intelligence) (Bertacchini et al., 2017), social capability (i.e. social intelligence) (Song and Kim, 2020) and behaviour (e.g. autonomy) (Furlough et al., 2021) positively influence consumers' acceptance. However, even though the literature on human-robot interaction and robot acceptance has grown rapidly over the last few years, some scholars point out that the empirical investigation of HSR's effectiveness in retail service domains is still scarce (Song and Kim, 2022). Brengman et al. (2021) found that HSRs are more effective than tablet service kiosks in attracting consumers and converting them into buyers. Roozen et al. (2023, p. 23) argue that “it is still unclear how retail customers react to service encounters with frontline service robots”.

Intelligence refers to the ability to learn from experience and adapt to the environment (Gardner, 1999; Sternberg, 2005; Schlinger, 2003). In AI literature, the focus is on developing machine learning to mimic human intelligence. Huang and Rust (2018) have conceptualised four types of AI intelligence: mechanical, analytical, intuitive and empathetic. More recently, they discussed three AI intelligences for marketing: mechanical, thinking and feeling (Huang and Rust (2022). In this study, we focus on thinking AI, or intellectual intelligence, in terms of the ability to think, learn and solve problems. According to Lv et al. (2023), the most important human characteristic that differentiates humans from animals relates to intellectual intelligence, as the individuals' ability to perform complex cognitive processes and thinking. In the context of human-robot interactions, intellectual intelligence can be defined as consumers' perception of HSR's intelligence as the result of its knowledge, purpose and competence, such as providing product information and in-store customer assistance (Bartneck et al., 2009; Moussawi et al., 2021).

As a result of emerging developments in AI, HSRs increasingly demonstrate intelligent and human-like cognitive behaviours. High levels of intellectual intelligence allow HSRs to perform as human staff members when interacting with customers in retail frontline contexts (Moussawi et al., 2021; Wirtz et al., 2018). In this context, HSRs with high levels of intellectual intelligence can take on the role of sales assistants, helping consumers navigate through the different sections of brick-and-mortar stores and assisting them in finding products. At the same time, they can offer product/sales information to help customers complete their purchase transactions, as any other human sales staff can do (Joe and Song, 2019; Song and Kim, 2020). In this sense, we would expect that consumers, in general, will prefer more versus less intelligent service robots.

Speciesism refers to humans' psychological state that assigns different, unjustified and disadvantageous considerations or treatment that lowers the moral worth of those considered non-human agents (Caviola et al., 2019; Figdor, 2021). The concept explains how humans behave when interacting with other non-human species. Researchers distinguish speciesism from other forms of unjustified discrimination, for example, racism and sexism (Caviola et al., 2019). Studies show that individuals with high speciesism (so-called “speciesists”) tend to hold a belief that humans are superior to other species and have a higher moral status than other species (Figdor, 2021; Huo et al., 2023b). Individuals' speciesism may trigger prejudice towards members of other species, influencing the relationship between members of different species.

The application of speciesism from human-animal relationships has been expanded to human-AI relationships (Huo et al., 2023b). Based on Intergroup Threat Theory (Giger et al., 2019; Stephan et al., 2015), the argument is that humans may subconsciously treat HSRs as a new distinct social species and outgroup members, leading to lower acceptance towards them, particularly when they have less desirable characteristics. Previous research, for example Huo et al. (2023b), has found that speciesism moderates the effect of medical staff participation on AI anxiety, while the effect on AI self-efficacy was not significant (Huo et al., 2023b). The authors conclude that for medical staff with a higher level of speciesism, their anxiety tends to drop when involved in the medical AI development process. The discussion about speciesism is further centred around relative intelligence. For example, research found that prejudicial treatment towards other species is based on perceptions of their intelligence relative to humans (i.e. other species have fewer moral rights because they are less intelligent) (Caviola et al., 2019).

Based on the previous discussion, we predict that speciesism will moderate the effect of HSR intelligence on consumers' judgements towards HSRs. Specifically, we argue that when HSRs present a characteristic that reinforces their outgroup status (i.e. lower intelligence), consumers with higher (vs lower) levels of speciesism are more likely to be negatively biased towards HSRs. On the other hand, highly intelligent HSRs may be perceived as closer to ingroup members because they are able to interact with humans at the same level, an assimilation effect, and therefore more likely to be positively perceived regardless of their level of speciesism. Thus, in the context of retail robots as sales assistants, we hypothesise:

That is, when HSR intellectual intelligence is low, customers' preference to shop using retail HSR sales assistants will be lower for individuals with higher (versus lower) levels of speciesism.

AI-based intellectual intelligence allows HSRs to understand customers' requests and perform in an efficient way. The attribute of intellectual intelligence can be linked with perceptions of competence. Perceived competence relates to the individuals' perception that someone (i.e. another individual) is being proficient or efficient in the execution of a task (Liao et al., 2023). Numerous studies in services research show how customers' perceptions of employee competence positively affect service encounters. A similar effect may occur in human-robot interaction (Belanche et al., 2021; Liao et al., 2023; Wu and Huo, 2023). HSRs interact directly with in-store customers, performing various tasks such as customer service and product recommendations (Bertacchini et al., 2017). The perception of HSR competence relates to the individuals' evaluation that a robot is capable of executing tasks correctly and efficiently. Research shows that the success of HSR-assisted shopping relies on customers' perception of their competence (Leiño Calleja et al., 2023; Liao et al., 2023; Schneider and Kummert, 2021). Furthermore, a competent robot makes customers perceive an enhanced notion of value-for-money (Belanche et al., 2021), thus increasing their willingness to make a purchase. Therefore, we predict that perceived competence mediates the effect of HSR intelligence on consumers' decision-making. Further, we posit that this mediation effect is moderated by speciesism such that this mediation effect is larger with higher levels of speciesism. More formally we hypothesise:

Figure 1 summarises the research model.

The present research includes three online experiments that test the effects of HSRs' intellectual intelligence on individuals' perceived competence, and, consequently, their willingness to choose a store that uses HSRs as part of their customer service. Furthermore, we examine whether speciesism attenuates these effects such that a mediation effect is likely to be observed for individuals low in speciesism but not for those with high levels of speciesism. In study 1, we examine how the effect of HSRs' intellectual intelligence on customers' decision to shop at retail stores using HSRs as sales assistants is moderated by speciesism. Study 2 demonstrates that this effect is mediated by perceived competence. Finally, study 3 integrates findings of studies 1 and 2, and tests the full moderated-mediation model. That is, it tests the moderation of speciesism and the mediation of perceived competence on the effect of HSRs' intellectual intelligence on consumers' decision to shop at stores using HSR sales assistants.

In these studies, participants were asked to consider a situation where they went to a local retail store to purchase a new laptop or a camping gear. All scenarios included an image of a human-like service robot named Sophia approaching them. HSRs like Sophia or Grace are identified as the most advanced human-like, life-like service robots (Reuters, 2021; Saracco, 2021). These robots are designed to replicate human beings in their embodiment, cognitive tasks and social interactions (Saracco, 2021). Sophia and Grace demonstrate that HSRs would be able to be alongside human beings in daily life, creating a multi-species coexistence. Data for all studies were collected on Prolific and analysed with SPSS to perform a logistic regression and PROCESS 4.0 Model 4 and Model 8 (Hayes, 2022) for the mediation and moderated-mediation analysis, respectively.

Study 1 was a scenario-based online experiment. This study aimed to provide evidence for the moderating effect of speciesism on the effect of intellectual intelligence on individuals' preference for HSRs as frontline service agents (H1). Participants were recruited via Prolific in the United States to take part in an online study in exchange for a nominal fee (£1.05). A total of 372 people (50.3% female) aged between 18 and 65 years (M_age = 41.02, SD = 12.28) who completed the study were randomly assigned to one of four conditions in a 2 (prime: speciesism vs neutral) x 2 (intellectual intelligence: high vs low) between-subjects design.  Appendix 3 summarises the demographic details.

To make sure our speciesism prime worked as predicted, we first ran a pre-test with 114 participants recruited from the same online panel. We created a prime based on often-used psychology research methods (e.g. vanDellen et al., 2014) where participants read a short story and engage with the content they read in order to bias their mindset in the researcher's desired direction. Specifically, in the speciesism prime condition, participants read two 200-words paragraphs about human intelligence superiority over chimpanzees and were asked to give a title for each paragraph. In the neutral prime condition, participants were asked to read two 200-words paragraphs about tips on doing laundry and were asked to give a title for each paragraph.  Appendix 1 provides an overview of the priming. After this task, participants reported their level of speciesism using a six-item scale taken from Caviola et al. (2019). As predicted, participants in the speciesism prime condition reported having higher speciesism in comparison to those in the neutral prime condition (M_prime = 5.46 vs M_neutral = 4.82, F(1/112) = 9.78, p = 0.002).

For the manipulation of HSR intellectual intelligence, participants in the main study were shown an electronics store retail scenario after completing the above priming task. In this scenario, participants were asked to consider a laptop purchase scenario with the assistance of an HSR (Sophia). In the high intellectual intelligence condition, the HSR was able to understand and deliver consumers' requests, offering the best product option and demonstrating that it was able to suggest ideal solutions based on the store offerings. On the other hand, in the low intellectual intelligence condition, Sophia demonstrated a partial understanding of the consumers' requests, not being able to answer their questions about different laptop options and suggest the best store offering. The complete scenarios are shown in  Appendix 2.

Purchase decision was the dependent variable. After reading the manipulation scenarios, participants were asked to choose whether they would buy the product from the store (0 = no; 1 = yes) before answering a series of questions. All measures used in the survey were based on existing scales from the literature and used 7-point Likert scales. We adapted a scale from Bartneck et al. (2009) to measure perceptions of the HSR's intellectual intelligence using three items.

Because previous research has suggested that giving HSRs the category of a moral agent or a member of a social category could be considered a threat to human identity (Vanman and Kappas, 2019), and because previous research findings showed that HSRs with greater cognitive abilities elicit aversion due to the threat of AI to the uniqueness of human intelligence (Liang and Lee, 2017; Lv et al., 2023), we controlled for perceived identity threat. Further, literature suggests that anthropomorphism is a key construct in explaining human-robot interactions (e.g. van Doorn et al., 2017). Anthropomorphism refers to the individuals' cognitive process of assigning unique human attributes and characteristics, especially physical appearance or physical reactions, to non-human agents (Abdi et al., 2022; Cheng, 2022; Epley et al., 2007). In this sense, AI-based intellectual intelligence can be understood as a replica of human intelligence, as it reflects humans' cognitive intelligence (Siau and Yang, 2017), making HSRs more human-like (Lv et al., 2023). Therefore, anthropomorphism was also controlled for. A three-item scale adapted from Huang et al. (2021) and Złotowski et al. (2017) was used to measure identity threat, and we used a three-item scale to measure anthropomorphism adapted from Tan et al. (2018). Finally, we asked participants several demographic questions.

Advancements in AI technologies and anthropomorphic robotic design have made it possible to humanise service robots to the point of giving them characteristics that will allow robots to become moral agents (Banks, 2019). However, some scholars suggest that giving HSRs the category of a moral agent or a member of a social category could be considered a threat to human identity (Vanman and Kappas, 2019). According to Intergroup Threat Theory (Giger et al., 2019; Stephan et al., 2015), individuals might perceive that HSRs (outgroup members) could interfere with humans (ingroup members) as they replicate their appearance and behaviours. Prior literature has studied the mediating effects of identity threat. For example, Złotowski et al. (2017) show that identity threat mediates the effects of autonomous robots on attitude towards robots in general and willingness to support robotic research. Other research found that HSRs with greater cognitive abilities elicit aversion due to the threat of AI to the uniqueness of human intelligence (Liang and Lee, 2017; Lv et al., 2023).

Literature suggests that anthropomorphism is a key construct in explaining human-robot interactions (e.g. van Doorn et al., 2017). Anthropomorphism refers to the individuals' cognitive process of assigning unique human attributes and characteristics, especially physical appearance or physical reactions, to non-human agents (Abdi et al., 2022; Cheng, 2022; Epley et al., 2007). Based on the need for social connection (Epley et al., 2007), individuals tend to anthropomorphise non-human agents, including HSRs (Blut et al., 2021), to replicate human-to-human interactions in different social settings. In this context, AI-based intellectual intelligence can be understood as a replica of human intelligence, as it reflects humans' cognitive intelligence (Siau and Yang, 2017), making HSRs more human-like (Lv et al., 2023). Anthropomorphism has been used to help scholars to study the acceptance of HSRs (Lin et al., 2020). But, previous literature has shown that robot anthropomorphism can be a double-edged sword (Holthöwer and van Doorn, 2023). Studies have found that anthropomorphic characteristics make human-computer interactions more enjoyable, but that they can also elicit feelings of eeriness. Thus, anthropomorphism has been both positively and negatively associated with customers' HSR usage (Blut et al., 2021; Cheng, 2022; Lin et al., 2020).

The results of the manipulation check supported the effectiveness of the manipulation of intellectual intelligence. One-way ANOVA revealed that participants perceived lower intellectual intelligence in the low than in the high HSR intellectual intelligence condition (M_lower = 2.39 vs M_higher = 5.27, F (1, 374) = 435.80, p < 0.001).

We ran a logistic regression to test the interaction effect of intellectual intelligence and speciesism on consumers' purchase decision. We also used perceptions of anthropomorphism and identity threat as control variables. Consistent with our hypotheses, we found a significant effect of intellectual intelligence on choice – X²(1) = 39.01, p < 0.001 – and a significant interaction effect of intellectual intelligence and speciesism on choice – X²(1) = 4.59, p = 0.032. As shown in Figure 2, when intellectual intelligence was high, participants chose to purchase at the store with the HSR Sophia regardless of their level of speciesism (39% vs 42%, p = 0.549). However, when HSR intellectual intelligence was low, participants were less likely to buy at the store when they had a higher level of speciesism (3%) than those with a lower level of speciesism (12%, p = 0.024).

Our results are consistent with the notion that humans will prefer to interact with a higher intelligent being than one with lower intelligence. That said, in the real world, most service robots will be of lower to moderate intelligence. In these cases, people with higher speciesism will be less likely to tolerate another species that is not at their level in the retail environment. In other words, humans are most likely to tolerate robots that are more like themselves and can relate and engage with them at the same level. But when this does not happen, they are more likely to derogate the HSR and less likely to make a purchase at this store.

The findings in study 1 indicate that speciesists (i.e. individuals with high levels of speciesism) prefer HSRs with high levels of human-like intellectual intelligence when interacting with HSRs working as frontline retail sales assistants. In this context, speciesists will avoid interacting with HSRs showing lower levels of human-like intellectual intelligence; they perceive that they are not interacting with an agent that can engage with them at the same human level to enhance their shopping experience. High levels of speciesism influence individuals to relate, interact and value other agents demonstrating similar behavioural or social characteristics and limiting encounters with agents not demonstrating these traits. Current HSRs working in retail are programmed to demonstrate moderate to lower human-like intellectual intelligence, as high levels of intelligence can trigger fear or rejection during HRIs (Vollmer, 2018); however, employing HSRs demonstrating lower levels of intellectual intelligence can affect frontline retail service performance for customers that are speciesists.

The goal of study 2 was to test the mediation effect of perceived competence on choice (H2). Similar to study 1, participants were recruited via Prolific in the United States for a nominal fee (£ 1.05) and were shown a similar retail encounter scenario, but this time, they witnessed the purchase of camping gear in a retail store. A total of 119 participants (56.3% female) ranging from 18 to 65 years old (M = 35.04, SD = 11.07) completed the study (see  Appendix C for the demographic details). Participants were randomly assigned to one of two conditions (intellectual intelligence: high vs low) of a between-subjects design. In this study, participants were asked to consider a shopping scenario where they were aiming to buy some camping equipment with the assistance of an HSR (Sophia). In the high intellectual intelligence condition, Sophia was able to understand and deliver consumers' requests, offering the best product option and demonstrating that it was able to suggest ideal solutions based on the store offerings. On the other hand, in the low intellectual intelligence condition, Sophia demonstrated partial to no understanding of the consumers' requests, not being able to answer their questions about different camping equipment options and suggest the best store offering (see  Appendix B for the complete scenarios). After reading the scenario, participants were asked to answer the same set of questions as in study 1. Table 1 includes the scale items of all constructs with reliability and validity results. Table 2 shows the variables' correlation matrix.

As expected, HSR intellectual intelligence was perceived differently depending on the conditions provided. Participants perceived the HSR to be less intelligent in the low condition than in the high condition (M_lower = 2.21, SD_lower = 1.23, M_higher = 4.86, SD_higher = 1.52, F(1, 117) = 118.04, p < 0.001). This result shows that our manipulation worked as expected.

We have hypothesised that perceived competence will mediate the relationship between HSR intellectual intelligence and the choice to use HSRs in a retail environment. To test this mediation hypothesis, we performed PROCESS model 4 analysis (Hayes, 2022). We used the same dummy variable coding approach as in study 1 (i.e. 0 = low HSR intellectual intelligence condition, 1 = high HSR intellectual intelligence condition) and the same dependent variable. Perceived competence was included as a mediator. While we have controlled for identity threat and anthropomorphism in study 1, it is still possible that these may serve as an alternative account for the proposed effects. Therefore, we also included identity threat and anthropomorphism as competing mediators (using a parallel mediation analysis) in order to rule out potential alternative accounts for the mechanism. Finally, age, education, gender and income were included as covariates.

The results reveal that the effect of HSR intellectual intelligence on perceived competence (β = 3.40; t = 13.97, p < 0.001) and anthropomorphism (β = 0.47; t = 2.12, p = 0.036) was significant. However, this effect was not significant for identity threat (β = −0.40; t = −1.33, p = 0.185). Furthermore, the effects of HSR intellectual intelligence (β = 2.82; Z = 2.83, p = 0.004), perceived competence (β = 0.77; Z = 2.91, p = 0.003) and anthropomorphism (β = 0.83; Z = 2.00, p = 0.044) on purchase decision were all statistically significant. But, consistent with our hypothesis, the effect of HSR intellectual intelligence on purchase decision through perceived competence was the only statistically significant indirect effect (index: 2.61, 95%, CI: 1.24 to 11.48), with 5,000 bootstrap samples. Importantly, the mediation effects of identity threat (95%, CI: −0.35 to 0.65) and anthropomorphism (95%, CI: −0.02 to 1.86) were not statistically significant.

In this context, participants perceiving high levels of HSR intellectual intelligence have high levels of perceived competence. On the contrary, when participants experience low levels of HSR intellectual intelligence, they perceive lower levels of perceived competence. These results provide support for H2. The regression scores for the mediation are shown in Table 3.

Consistent with novel literature (e.g. Wu and Huo, 2023), study 2 shows the significant mediating effects of individuals' perceived competence in a retail HRI context; however, to our knowledge, this is the first study that tests the indirect effects of individuals' perceived competence on purchase decision when interacting with a highly human-like intelligent HSR during a frontline encounter. In this context, the present study demonstrates that when highly intelligent HSRs work as frontline retail sales assistants and interact with customers, individuals' HSR perceived competence has a positive statistically significant indirect effect on their purchase decision. Consumers' expectations are that frontline retail HSRs are highly competent (Christoforakos et al., 2021) which depends on the HSR's ability to understand the customers' requests and resolve them efficiently in order to improve the in-store purchase experience (Bertacchini et al., 2017). Finally, study 2 results indicate that under the context and conditions in the present study, the effects of anthropomorphism and perceived identity threat do not have a statistically significant indirect effect on customers' purchase decisions.

Similar to study 2, this study was a scenario-based online experiment replicating the purchase of camping gear at a retail store, assisted by Sophia. This study aimed to provide evidence for the moderating effect of speciesism on the effect of intellectual intelligence on individuals' HSR use, showing that this effect is mediated by individuals' perceived competence (H3). As in previous studies, participants were recruited via Prolific in the United States to take part in an online study in exchange for a nominal fee (£ 1.05). A total of 396 people (49.2% identified as female) between 18 and 55 years (M_age = 31.51, SD = 9.70) completed the study (see  Appendix 3 for the demographic details). Participants were randomly assigned to one of the two conditions that were used for study 2 (intellectual intelligence: high vs low; see  Appendix 2 for the scenarios).

All measures were based on study 1 and study 2. Similar to the previous studies, we asked participants several demographic questions. The reliability and validity of all scale items are shown in Table 4, and the correlation matrix is shown in Table 5.

The results of the manipulation check supported the effectiveness of the scenarios. Participants perceived less HSR intellectual intelligence in the low HSR intellectual intelligence condition than in the high HSR intellectual intelligence condition (M_lower = 2.87, SD_lower = 1.42, M_higher = 5.26, SD_higher = 1.33, F(1, 394) = 299.29, p < 0.001). One-way ANOVA revealed different use intentions depending on the HSR intellectual intelligence. Participants in the high intellectual intelligence condition showed higher HSR use than those in the low intellectual intelligence condition (M_yes = 5.37, SD_yes = 1.10, M_no = 2.80, SD_no = 1.44, F(1, 394) = 395.67, p < 0.001).

Following our predictions in hypothesis 3, the level of speciesism moderates the relationship between HSR intellectual intelligence and perceived competence such that an effect is predicted for low but not for high HSR intelligence. Specifically, when HSR intellectual intelligence is low, we expect individuals with higher levels of speciesism (vs low) to rate the HSR as less competent and display lower HSR acceptance.

To test the moderated mediation hypothesis, we conducted PROCESS 4.0 Model 8 analysis (Hayes, 2022). Similar to previous studies, we used a dummy variable coding approach to include the experimental treatments as the independent variables in the model estimation (0 = low HSR intellectual intelligence condition vs 1 = high HSR intellectual intelligence condition). Results indicate that HSR intellectual intelligence does have a significant positive effect on perceived competence (β = 1.99, t = 3.96, p = 0.001). Furthermore, we can find that speciesism negatively affects perceived competence (β = −0.26, t = −3.38, p < 0.0001). Finally, there was a significant interaction effect of intellectual intelligence and speciesism on perceived competence (β = 0.25, t = 2.50, p < 0.01), showing that individuals with high levels of speciesism would perceive the HSR less competent when in the lower intelligent condition, which affects their in-store purchase decision. Figure 3 shows the significant interaction between perceived competence and speciesism.

Consistent with H3, the results reveal that the moderated mediation effect is significant. The regression scores obtained using PROCESS Model 8 are shown in Table 6.

Study 3 shows that individuals' high level of speciesism moderates the relationship between HSR intellectual intelligence and perceived competence on customers' purchase decision when intellectual intelligence is low. In this context, speciesists reduce their perception of HSR competence during interactions with frontline retail HSRs when they perceive that they have lower intellectual intelligence. As previously argued, speciesists prefer to interact with HSRs that replicate human-like levels of intelligence, especially when HSRs are assigned to assist their in-store shopping journey, as any human frontline retail staff, and they will avoid interactions with HSRs with low intellectual intelligence. This expectation aligns with the intended usage of HSRs in frontline retail services, as retail robots as sales assistants aim to enhance customers' purchase experience (Bertacchini et al., 2017; Lv et al., 2023; Underwood, 2020) and individuals' perceived competence of HSRs performance in service contexts (Wu and Huo, 2023).

Retailers increasingly endeavour to implement AI innovations such as HSRs as sales assistants to enhance the customer experience. Previous research has focused on factors that positively influence the acceptance of HSRs in retail and service industries (Bertacchini et al., 2017; Bradwell et al., 2021; Jeong and Ha, 2020). Scholars point out that it is necessary to shift from a “mechano-centric” view to a “human-oriented” one (Stock and Nguyen, 2019). Specifically, human traits and biases need to be studied to understand the compatibility between humans and robots better. The objective of this research is to address the call to investigate the role of speciesism as a moderating variable and its influence on HSR acceptance (Schmitt, 2019, 2020). According to Caviola et al. (2019), speciesism is related to biased (negative) behaviour towards members of other species. Speciesists, that is individuals with high levels of speciesism, feel superior over non-human agents, which can lead to inequality between humans and agents of other species. Furthermore, speciesism creates a socio-ideological belief that validates the creation of a species hierarchy, where humans are over other non-human agents, even though they demonstrate human-like characteristics. Finally, speciesists dehumanise other non-human agents, allowing them to feel superior, not fearing them (Caviola et al., 2019).

Based on our results, we conclude that speciesism reduces individuals' perceived competence in HSRs in frontline retail services when they do not perform human-like, that is high levels of intellectual intelligence. Speciesists are willing to have interactions with HSRs that demonstrate high levels of intellectual intelligence during their in-store shopping journey as they perceive that the HSR is competent enough in its role as frontline retail assistants, similar to human retail staff. In this context, it is possible to infer that when HSRs demonstrate high levels of intellectual intelligence while working as frontline retail staff, they are not only performing as expected to enhance customers' in-store purchase experience, but they are performing as human replicas and can match the speciesists expectations of human-like intelligence, making them feel that the HSR is matching their human superiority, as it can interact in with the same level of intellectual intelligence as theirs.

The present research makes several important contributions to theory. First, our study integrates several literature streams, including human-robot interaction, retailing in a new technology era (Grewal et al., 2021; Noble et al., 2022) and speciesism in human-AI relationships (Huo et al., 2023a; Schmitt, 2019, 2020). Scholars acknowledge that the field of service robots is still in its infancy (Lu et al., 2020) and that more theoretical frameworks and empirical studies of HSRs in a retail environment are needed (Belanche et al., 2020; Fuentes-Moraleda et al., 2020; Roozen et al., 2023). This research developed a new theoretical model that integrates service robot attributes, consumer perceptions and the construct of speciesism. To the best of our knowledge, this is the first empirical study demonstrating the effects of speciesism in HRIs in frontline retail services when in-store customers interact with highly intelligent HSRs.

Second, our study adds to growing research on consumer-facing retail technologies (Inman and Nikolova, 2017) and the effectiveness of human-robot service encounters (Roozen et al., 2023). HSRs can be programmed with cognitive and emotional characteristics to assist customers during the different stages of the in-store purchase journey (Bertacchini et al., 2017), creating an innovative retail experience (Fuentes-Moraleda et al., 2020). In recent years, several studies have aimed to understand the impact of HSRs in retail. Wang et al. (2022) found that participants primed with robots (vs humans) were more likely to engage in exploratory consumption behaviours. But the impact was affected by the degree of service robots' intelligence and moderated by consumers' subjective knowledge. Song and Kim (2022) study the usefulness, social capability and appearance of retail service robots and whether anxiety towards robots inhibits the relations between these facilitators and human-robot interaction. The authors confirm the important role of pre-existing anxiety as a moderator. Nonetheless, empirical studies of HSRs in retail remain scarce and it is “still unclear how retail customers react to service encounters with frontline service robots” (Roozen et al., 2023, p. 23). We offer new insights into how in-store retail robots affect purchase decision. Our results show that perceived competence is a far more important mediator than identity threat and anthropomorphism. We thus support prior studies that have confirmed the critical role of competence. For example, Leiño Calleja et al. (2023) show that HSRs generally elicit lower levels of customer orientation due to lower competence compared to human frontline employees. Yet, the authors found that robots can be perceived as competent as human employees under certain circumstances. Furthermore, our study supports Liao et al. (2023), who reveal that perceived competence is an antecedent variable that affects people's acceptance of robots.

Third, this research is one of the first that integrates the concept of speciesism in the discussion of HSR effectiveness in frontline retail services. We extend the novel research on speciesism in human-AI relationships. Scholars point out that the species bias against technology is not simply “technology phobia”, but that speciesism seems to be the result of a fundamental, categorical comparison of humans and machines (Huo et al., 2023a). We support prior research that recognises the role of speciesism as a boundary condition (Huo et al., 2023b). Our results demonstrate the importance of group threat viewpoint in human-robot interaction (Huo et al., 2023a). We further offer a potential perspective for future research in human-robot collaboration (HRC), where robot coworkers will work alongside human retail staff (Noble et al., 2022; Paluch et al., 2022; Tuzovic and Paluch, 2023). While our study focused on customer perceptions, the results are relevant for the transformation of workplaces and the reluctance of employees to use AI technology. We contribute to the novel application of speciesism in HRC literature that may help understand how different personality traits influence how frontline employees cooperate with HSRs (Le et al., 2023). This knowledge will help managers recruit and motivate employees to cooperate better with AI technologies (Huo et al., 2023b).

The present research offers several managerial implications for marketers and retailers. First, retailers need to be able to identify what HSR characteristics will influence the acceptance of HSRs in frontline retail settings, as the biases generated by speciesism would influence individuals' interactions with HSRs. The findings offer insights for retailers to potentially “hire” HSRs that replicate the intellectual intelligence of human staff. HSRs replicating, even improving, cognitive human retails assistants will allow customers to experience an enhanced in-store shopping experience, making them feel that they are interacting with well-trained, efficient retail staff.

Second, retailers may develop educational marketing campaigns and communication strategies on what to expect when interacting with HSRs, which will help to establish performance expectations. For example, retailers should promote HSRs' role as shopper assistants (Bertacchini et al., 2017) and how consumers could interact with HSRs and what to expect from them. Demonstrating how highly intelligent HSRs intend to support and enhance the in-store shopping experience will strengthen consumers' perceived competence (Mende et al., 2019).

Third, the research findings suggest that it is critical to pay more attention to customers with higher human uniqueness perception (see Huo et al., 2023a). Our findings provide a basis for segmentation and targeting. For example, Huo et al. (2023b, p. 6) suggest to “select employees with high-level speciesism as much as possible to actively make their employees involved in the development process of medical AI”. While the study was conducted in a different context, the segmentation of customers with high-level speciesism offers retailers better strategies for developing targeted messaging. The knowledge of consumers' speciesism thus could noticeably improve in-store experience with retail robots as sales assistants. In this context, retailers can assign human staff to work alongside HSRs and educate consumers on how to interact with highly developed HSR assistants, demonstrating how this shopper-facing technology is there to help improve their experience.

Our research has several limitations that offer avenues for future research. First, the present study included participants from the United States. Consequently, our results cannot be generalised to other societies that are in close contact with HSRs, like Japan, where the general public interacts with different types of robots that benefit their daily lives (Johnston, 1999; Lufkin, 2020). On the contrary, Western individuals may have a negative bias towards robots (Johnston, 1999). On the other hand, Asian individuals perceive robots in a positive way, as they are presented as allies to defeat enemies of humanity, for example, Alita (Alita: Battle Angel), Briareos (Appleseed Alpha), Gundam Robot (Gundam movies), Mazinger Z (Mazinger series), helping humans to live better and enjoy life, for example, Asimo by Honda, Qrio by Sony, Kirobo (Robi) by Robo Garage and HRP-3 by Kawada Industry (Encyclopedia of Japan, 2020). Hence, there is a possibility that our sample had a pre-existing negative perception towards HSRs replicating humans because of their culture and social backgrounds, benefitting our results.

Second, the context of the research was limited to the retailing industry. Future research may extend the study to other service sectors. For example, HSRs have been adopted in hospitality and healthcare services. Scholars should thus investigate if the role of speciesism differs depending on the nature of the service, for example, utilitarian versus hedonic services.

Furthermore, this study used online-based scenarios. There is a possibility that the effects of speciesism could be different when individuals interact in real life with a HSR demonstrating different levels of intellectual intelligence. However, the online experiments allowed us to easily manipulate intellectual intelligence levels in a retail scenario from a well-known, virtually impossible-to-access, most advanced HSR named Sophia.

With regard to robot attributes, we decided to focus on intellectual intelligence since AI has gradually acquired the ability to “think as humans” (Huo et al., 2023b). Recent research has distinguished multiple intelligences, ranging from mechanical and thinking to feeling AI (Huang and Rust, 2018). Future research may consider expanding our theoretical model and manipulating HSR's ability to show empathy.

The present study was the first step in investigating the moderating effects of speciesism on perceived competence and, consequently, making a purchase decision. We did not measure personality in this study. Past research has identified human personality as a predictor of consumer attitudes and robot acceptance in the human-robot interaction literature – but with mixed results (Esterwood et al., 2021). Thus, future research should investigate the relationship between speciesism and different personality traits in the context of human-AI relationships and HSRs as artificial species. Moreover, future studies should focus on understanding intergroup relationships (Vanman and Kappas, 2019), specifically if HSRs outperform humans. Empirical research is still scarce on the ethical challenges around human-robot multi-species coexistence and their impact on current social norms.

The first and second authors contributed equally to this work.

In the next page, you will read a short article. Please take your time to read the paragraphs in full. In the sections that follow, you will be asked to provide some information relating to this article.

Results from magnetic resonance imaging (MRI) brain scans of both baby chimps and baby humans revealed that unlike chimps, humans undergo a massive explosion of connections between brain cells in the first two years of life, which plays such a key role in human development, affecting children's IQ, social life and long-term response to stress.

Humans are nothing more than a profoundly evolved species, and the accumulation of evidence is ever-growing accumulation. As the most evolved species, and only one capable of grasping the concept of morality through rational and analytical thought, we have inhabited and dominated the planet in a very short time span. Precisely, it is our capabilities to engage in this level of thought that separates humans from other species. One of the human brain's most prized regions is the overdeveloped cerebral cortex; it represents over 80% of our brain mass and is thought to contain 100 billion neurons.

The cerebral cortex is associated with complex, higher thinking, such as decision-making, executive control, emotional regulation and language. Even though the human brain makes up about 2% of body weight, it consumes more than 25% of our body's overall energy. For this reason, even the most highly developed chimpanzee cannot reason with the average human possessing a 95–105 IQ score.

Please give a title for the short article you read above.

In the next page, you will read a short article. Please take your time to read the paragraphs in full. In the sections that follow, you will be asked to provide some information relating to this article.

Humans' ability to control fire brought a semblance of day to night, helping our ancestors to see in an otherwise dark world and keep nocturnal predators at bay. The warmth of the flames also helped people stay warm in cold weather, enabling us to live in cooler areas. And of course, it gave us cooking, which some researchers suggest influenced human evolution – cooked foods are easier to chew and digest, perhaps contributing to reductions in human tooth and gut size.

Humans are certainly unique in the level of abstractness with which we can reason about others' mental states and create art. We have our advanced language skills to thank for that. We may see evidence of basic linguistic abilities in chimpanzees, but we are the only ones writing things down. We (humans) tell stories, we dream, we imagine things about ourselves and others and we spend a great deal of time thinking about the future and analysing the past.

While we see the roots of many human behaviours in our closest relatives, chimpanzees, we are the only ones who peer into their world and write books about it.

Please give a title for the short article you read above.

In the next page, you will read a short article. Please take your time to read the paragraphs in full. In the sections that follow, you will be asked to provide some information relating to this article.

Laundry is a necessary chore that everyone has to do at some point. It can be a tedious task, but it doesn't have to be. By following a few simple guidelines, you can make laundry a more efficient and enjoyable activity. The first step is to sort your laundry. This will help you to wash similar items together and prevent stains from colors. You can sort your laundry by color, fabric, type and soil level.

Once your laundry is sorted, you can choose the appropriate washing cycle. The washing cycle will depend on the type of fabric, the soil level and the desired results. For example, delicate fabrics should be washed on a gentle cycle, while heavily soiled items may require a heavier cycle. It is also important to use the correct amount of detergent. Too much detergent can leave a residue on your clothes while too little will not clean them effectively. You can find the recommended amount of detergent on the product's packaging.

After washing, you should rinse your clothes thoroughly. This will help to remove any leftover detergent and prevent your clothes from feeling stiff. Finally, you need to dry your clothes out. You can dry your clothes on a clothesline, in a dryer or in a combination washer/dryer. If you are drying your clothes in a dryer, be sure to choose the correct drying cycle.

Please give a title for the short article you read above.

In the next page, you will read a short article. Please take your time to read the paragraphs in full. In the sections that follow, you will be asked to provide some information relating to this article.

Following these simple guidelines will help you to wash your laundry effectively and efficiently.

By taking the time to sort your laundry, choose the correct washing cycle and use the correct amount of detergent, you can save time and energy.

Here are some additional tips for laundry:

1. Pre-treat stains before washing.

2. Use cold water whenever possible.

3. Air-dry your clothes whenever possible.

4. Wash towels and bedding separately.

5. Clean your washing machine and dryer regularly.

By following these tips, you can keep your clothes looking and smelling their best.

Laundry hacks can also help you to save time and energy. Here are a few laundry hacks that you can try:

1. Use vinegar as a fabric softener.

2. Add baking soda to the washing machine to boost cleaning power.

3. Use a dryer ball to reduce static cling.

4. Wash your clothes inside out to protect the colors.

5. Hang your clothes to dry instead of using the dryer.

These laundry tips are easy to implement and can save you money in the long run.

Try it out.

Please give a title for the short article you read above.

Table A1