Does metaverse improve recommendations quality and customer trust? A user-centric evaluation framework based on the cognitive-affective-behavioural theory

Abumalloh, Rabab Ali; Nilashi, Mehrbakhsh; Halabi, Osama; Ali, Raian Article Does metaverse improve recommendations quality and customer trust? A user-centric evaluation framework based on the cognitiveaffective-behavioural theory Journal of Innovation & Knowledge (JIK) Provided in Cooperation with: Elsevier Suggested Citation: Abumalloh, Rabab Ali; Nilashi, Mehrbakhsh; Halabi, Osama; Ali, Raian (2024) : Does metaverse improve recommendations quality and customer trust? A user-centric evaluation framework based on the cognitive-affective-behavioural theory, Journal of Innovation & Knowledge (JIK), ISSN 2444-569X, Elsevier, Amsterdam, Vol. 9, Iss. 4, pp. 1-16, https://doi.org/10.1016/j.jik.2024.100569 This Version is available at: https://hdl.handle.net/10419/327471 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Does metaverse improve recommendations quality and customer trust? A user-centric evaluation framework based on the cognitive-affectivebehavioural theory Rabab Ali Abumalloh a , Mehrbakhsh Nilashi b,d , Osama Halabi a , Raian Ali c, a Department of Computer Science and Engineering, Qatar University, Doha, Qatar b UCSI Graduate Business School, UCSI University, 56000, Cheras, Kuala Lumpur, Malaysia c College of Science and Engineering, Hamad Bin Khalifa University, Qatar d Centre for Business Informatics and Industrial Management (CBIIM), UCSI Graduate Business School, UCSI University, Kuala Lumpur, Malaysia ARTICLE INFO Article History: Received 10 September 2023 Accepted 5 September 2024 Available online 30 October 2024 ABSTRACT Recommendation agents (RAs) have proven to be effective decision-making tools for customers, as they can boost trust and loyalty when customers shop online. They can analyse large amounts of data using machine learning algorithms and predictive analytics capabilities to provide highly relevant recommendations to users. In previous studies, several approaches have been implemented to refine and assess the effectiveness of these agents. As a new form of virtual reality universe, metaverses can be seen as a new venue for improvements in the performance of online RAs. By exploiting the capabilities of the metaverse and incorporating data about the user’s behaviour and preferences, the performance of these systems can be enhanced in terms of the accuracy, diversity, and novelty of the generated recommendations. The metaverse can provide visually appealing and interactive recommendations, and there are several potential factors that can affect the customer’s experience. The cognitive-affective-behavioural theory is used to develop the proposed research model. This study investigates the impact of the capabilities of the metaverse on three quality factors of RAs: diversity, accuracy, and novelty. The influence of the quality of the recommendations on affective trust and the influence of affective trust on customer loyalty are also examined. In addition, as this is an emerging technology, perceived privacy plays a crucial role in maintaining users’trust and confidence. Hence, the moderating influence of perceived privacy on the relationship between the quality and affective trust of RAs is examined. The moderating impact of product knowledge on the relationship between the individual perception of trust and loyalty is investigated. Data were acquired from 288 Malaysian respondents and analysed using the PLS-SEM method. The findings of this study show that the capabilities of the metaverse have favorable impacts on several quality factors of the recommender system, including accuracy, diversity, and novelty. Furthermore, these quality factors impact the perceived quality of RAs, which in turn impacts customer trust and loyalty. Perceived privacy acts as a moderator on the relationship between the quality of recommendations and the individual’s perception of trust. © 2024 The Author(s). Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Keywords: Metaverse Recommendation agents Accuracy Novelty Diversity Perceived privacy Industrial and innovation JEL classification: C12: Hypothesis Testing: General M15: IT Management O39: Technological Change: Other Introduction An automated recommendation system, also called a recommendation agent (RA), analyses the data associated with a user and suggests products, services, and content based on the user’s personal preferences (Bagherifard et al., 2017;García-S anchez et al., 2020; Gharahighehi et al., 2021;Nilashi et al., 2020). There are many applications for these agents in domains such as e-commerce (Khan et al., 2021), media (Herce-Zelaya et al., 2020;Nilashi et al., 2023), travel (Renjith et al., 2020), news (Karimi et al., 2018), and entertainment (Airen & Agrawal, 2023). RAs are also playing an increasingly crucial part in individuals’decision-making procedures (Scholz et al., 2017;Wang et al., 2022), especially in the context of online retail and e-commerce. An RA can offer valuable assistance to customers in the highly competitive environment of e-commerce (Guo et al., 2014), where there is a vast array of options that often overloads the user’s capabilities (Liu et al., 2022). In this case, an RA can assist the user in Corresponding author. E-mail addresses: [email protected] (R.A. Abumalloh), [email protected] (M. Nilashi), [email protected] (O. Halabi), [email protected] (R. Ali). https://doi.org/10.1016/j.jik.2024.100569 2444-569X/© 2024 The Author(s). Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Journal of Innovation & Knowledge 9 (2024) 100569 Journal of Innovation &Knowledge https://www.journals.elsevier.com/journal-of-innovation-and-knowledge finding products that will meet their specific needs and preferences more rapidly and easily. These agents can analyse vast amounts of data with the help of machine learning algorithms and predictive analytic capabilities, and can provide highly targeted and relevant recommendations that can increase customer loyalty and satisfaction. An RA works by analysing user data (e.g., browser history records, search queries, and past purchases) in order to build a profile of the user. To generate the list of recommendations for the user, the RA uses various algorithms such as collaborative filtering (CF) (Kuo & Li, 2023), content-based filtering (CBF) (Afoudi et al., 2021), and hybrid approaches (Biswas & Liu, 2022). In a CBF algorithm, the suggested items are similar to items the user has previously purchased, while a CF algorithm recommends items that are common among similar users (Afoudi et al., 2021). The fundamental idea behind CBF is to suggest new products based on their attributes and the historical preferences of the customer (Kuo & Cheng, 2022). Hybrid recommendation approaches combine both CF and CBF to provide more precise and personalised recommendations (Tewari, 2020). The use of RAs has increased customer engagement, retention, and revenue to many well-known online shopping platforms, such as Amazon.com, Alibaba.com, and Shein.com. However, the efficiency of an RA is determined by the quality of the data and the algorithms used to generate recommendations. Users who are provided with inaccurate or irrelevant recommendations may lose trust in the agent and the company as a whole. The primary objective of an RA is to provide customised selections to users based on their preferences, interests, and past behaviour, where several interrelated factors will impact on individuals’perceptions of the quality of the generated suggestions. The accuracy of an RA is very important, as this has a direct impact on the quality of recommendations provided to users (Zanon et al., 2022). Inaccurate recommendations will negatively impact the users’trust in these systems (Nilashi et al., 2016), which can lead to a decrease in an individual’s engagement and usage behavior. Previous studies have shown that the accuracy of recommendations can have a significant impact on their overall performance. Scholars have worked on other various aspects of the design of RAs in order to improve user experiences such as diversity and novelty. Diversity represents the system’s capacity to offer a variety of products, with the aim of broadening the consumers’preferences (Zanon et al., 2022). Another important factor is the novelty of the recommendations: an item that exactly matches the needs of the customer is hardly a suitable choice if the customer is already very familiar with it (Kunkel & Ziegler, 2023). A successful RA should be able to nudge users out of their "comfort zone", allowing them to enjoy novel, diverse, yet still pleasant experiences (Gravino et al., 2019). The accuracy of recommendations is affected by the quality of data. Data collection methods, preprocessing techniques, and integration processes all affect the quality of the data used in RAs, meaning that it is essential to ensure that the data used are precise, relevant, and up-to-date. Previous studies in the literature have reported that the sparsity problem (a lack of information about users’preferences, interests, and behaviour) has a major effect on the accuracy of recommendations (Heidari et al., 2022;Joorabloo et al., 2021). Most RAs suffer from this problem, as only a small percentage of items are rated by users and there are limited numbers of interactions between users and items. Furthermore, biased recommendations that favour popular items over less popular ones can result from a sparse dataset, potentially leading to a loss of diversity in recommendations (Geng et al., 2023). RAs may not be able to accurately detect the users’needs and preferences, and addressing this issue in the context of recommendation systems is an important area of research. The level of user engagement with an RA can be improved with the help of emerging digital technologies (Margaris et al., 2018). Previous research has suggested that the use of social media data to enhance the performance of an RA can yield improved results (Nilashi et al., 2023), and to achieve this, RAs can be connected to various online communities, meaning that users of social networking sites have many options for reviewing and rating products purchased online (Vazquez et al., 2023). Textual feedback can be used as a complement to user ratings, in order to enhance the quality of the data and the accuracy of recommendations, thereby ensuring that they are tailored to the preferences of the users (Nilashi et al., 2023). In addition, connecting RAs to popular social media sites may encourage more people to use them by making it easier for them to recommend services to their friends and followers. As an emerging technology, the metaverse offers new opportunities for the enhanced operation of RAs and the delivery of more correct and relevant suggestions to users. As a result of the intersection between the real and virtual worlds in the metaverse, unique mechanisms for both generating and perceiving value have emerged (Mancuso et al., 2023). With the integration of several innovations within the metaverse (El Hedhli et al., 2023), the recommendations can take novel and unexpected forms in terms of design, presentation, and implementation. Insights into user preferences and attitudes can be gained through the metaverse’s virtual setting, which allows users to interact in real time, both with each other and with digital items. The capabilities of the metaverse in terms of data availability and enrichments can enhance the performance of RAs, which accordingly influences the user’s overall experience. However, further investigations are essential in order to understand the possible impacts of the capabilities of the metaverse on the ways in which users perceive the different quality factors of RAs. The promising capabilities offered by the metaverse are associated with significant privacy concerns, which can have a significant impact on the customer’s experiences and the adoption of technology. Privacy concerns have been linked to several outcome variables in the literature, including the acceptance of technology in healthcare (Dhagarra et al., 2020), trust, intention, and attitude towards search engines (Palanisamy, 2014), and customer loyalty toward online shopping (Wong et al., 2019). The metaverse primarily offers customers granular services based on multiple dimensions of interaction, in which data are collected from numerous venues, unlike the data gathered from a single venue by conventional internet apps (S. Zhang et al., 2023b). In general, businesses gather customers’private information to execute transactions as well as to better comprehend their needs, requirements, and choices, so that they can tailor their advertising campaigns to them (Degutis et al., 2023). Privacy concerns grow with the implementation of RAs, as the volumes of data gathered and the user profiles expand, to provide classifiers with more data for learning and inference (Slokom et al., 2021). The huge volumes of personally identifiable information that will be exposed if the privacy of a Meta application is compromised will be much greater than for conventional Internet applications, thus posing a severe risk to consumers’privacy (Liu et al., 2021;Wang et al., 2021). In a survey conducted by Statista (2022a), 87 % of American citizens said they would be worried about their privacy if Facebook were to be successful in building the metaverse. In addition, 50 % of those surveyed said they were concerned that hackers would be able to impersonate other people too easily, and 41 % believed it would be too difficult to preserve their true identity in the metaverse. Due to these uncertainties, the acceptance of RAs in the metaverse will require a certain level of familiarity with and comprehension of the products involved. Customers must be knowledgeable about the products they interact with, as the metaverse’s capabilities are developed further. In traditional RAs, the level of knowledge of the product will impact how the customers perceive the quality of the recommendations (Yoon et al., 2013). Based on the above discussion, we can draw up the following research questions: i. How do metaverse capabilities influence various quality factors of recommender agents (RAs)? R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 2 ii. How does perceived privacy affect the relationship between recommendations quality and customer affective trust? iii. How does product knowledge influence the relationship between customer affective trust and customer loyalty? The remainder of this research is structured as follows: Section 2 introduces the Cognitive-Affective-Behavioural theory as the theoretical foundation of the study. Section 3 presents an overview of metaverse capabilities in retailing. Hypothesis development is detailed in Section 4. The study methodology is outlined in Section 5, and the results are discussed in Section 6. Finally, Section 7 presents the conclusion and the research contributions of the study. Theoretical background Cognition-affect-behavioural model Together with the theory of planned behavior (Ajzen, 1991), the technology acceptance theory has been deployed as a theoretical base in several contexts to appraise an individual’s early acceptance of information systems, innovations, and tools (Davis et al., 1985). However, early acceptance cannot truly reflect the system’s performance, as a long-standing relationship between the users and the system is the true measure of the system’s performance (Bhattacherjee, 2001). Researchers have therefore considered a three-stage sequence of continuous intention, from quality features to longstanding commitment (Frank et al., 2014). Lazarus (1991) explored the impact of the appraisal of interior and contextual variables on the affective response, which in turn influences coping actions. Drawing on Lazarus (1991) framework of appraisal, Bagozzi (1992) investigated the link between attitude and behaviour based on the intercession of affective variables between the cognitive appraisal and the coping response. Bagozzi (1992) stressed the prominence of the motivational link between attitude and behavior in forming a user’s desire to perform an action. Perusing this line of research, Eagly and Chaiken (1993) presented the attitude formation theory, in which attitudes are formed solely based on the impact of three folds: cognition, affect, and behavior. Expanding on this theoretical framework, user behavior is assessed using a three-tiered chain model. The knowledge that users gain about the technology through cognitive variables will formulate their views. As the consumer engages with the technology, cognitive variables are framed to develop beliefs. The level of experience with the innovation creates a feeling of emotional desire that frames the emotive perceptions. This theory highlights the intervention of emotive perceptions between beliefs and actions. Users’emotional perceptions can be influenced by their positive or negative evaluation of the technology, which in turn shapes their affective response (Kwon & Vogt, 2010). In this research, we adopt the cognition-affect-behaviour model to explore the factors that influence a customer’s loyalty toward an RA in the metaverse across three structured layers: (i) the appraisal of the quality of the recommendations (cognitive), represented by diversity, novelty, and perceived accuracy; (ii) the affective component, represented by trust; and (iii) the behavioural component, represented by loyalty. Cognitive overload theory Individuals may feel stressed and mentally exhausted if their cognitive load exceeds a particular limit (Pang & Ruan, 2023). Information overload may result in negative effects such as information anxiety, information fatigue, and tension (Islam et al., 2022). According to cognitive load theory, humans have limited mental capacities, and a cognitive load that is too high will impact their ability to acquire information about products or services, leading them to have negative opinions towards a good or service. Individuals are described as ‘cognitive misers’who frequently hesitate to put forth more effort than is essential. RAs have proven their effectiveness in addressing the problem of information overload in traditional online environments. The problem of information overload has been linked to e-commerce in several contexts (Fang et al., 2021) where the environment in which the customer processes the information influences their perceptions of products and services. In situated cognition theory, rather than serving as an intangible mental process, information processing is thought to take place internally and actively based on the environment in which the customer is located (Semin & Garrido, 2015). Users can better understand the value of a product or service when their experiences enable them to relate intangible facts to actual interactions and real-world events (Fan et al., 2020). Online shoppers often have difficulty in seeing how things will fit into their own unique environments (Jung et al., 2015), which increases their mental workload. Hence, if the cognitive load is too great, users will experience negative emotions as a result of the gap between their desires and the influence of their surroundings, which will have a detrimental impact on their ability to make decisions (Liu & Goodhue, 2012). In the metaverse, this challenge can be addressed by providing an immersive environment where users can interact with products or services in a realistic and contextually relevant manner. Metaverse recommendations can help individuals overcome the problem of information overload in different ways, as they can provide a greater resemblance to individuals’interactions and interpretations of objects in the real world. These recommendations allow customers to overlay virtual information onto their actual surroundings, thus reducing the cognitive load required to understand how a product will fit into their personal contexts. Metaverse capabilities in retailing Although there is no single agreed-upon description of the metaverse, most experts concur that it is a network of interconnected virtual settings in which individuals can engage to mimic activities in the real world (Park & Lim, 2023;Zallio & Clarkson, 2022). As the name suggests, the metaverse is not only a virtual venue but a point at which the real and virtual worlds meet. The metaverse is also made up of a collection of different platforms and innovations that work together to provide users with an experience that unites both the real and the virtual realms (Hazan et al., 2022). Through a combination of diverse technical infrastructures, a fully developed metaverse can create a parallel environment for cultural exchange and human engagement (Stephens, 2021). Developments in the metaverse have been facilitated by the huge advancements in the areas of artificial intelligence (AI), big data (BD), virtual reality (VR), and augmented reality (AR), which are all used to enhance the actual environment and enhance the consumer’s experience in the metaverse (Park & Lim, 2023). All of these technological innovations are crucial for creating and enhancing immersive realities, as they allow individuals to engage with telepresence, which often affects the level of immersion in the metaverse environment (Giang Barrera & Shah, 2023). The metaverse should provide individuals with a realistic experience and enable them to be immersed in the virtual world (Dionisio et al., 2013). The capabilities of the metaverse are expected to cause shifts in the shopping experiences of individuals in both the digital and physical retail sectors by providing them with engaging and immersive knowledge about goods and services (Klaus & Kuppelwieser, 2023; Serravalle et al., 2023). Since they can break down the social and physical barriers between customers and brands, the metaverse and other fully immersive technologies have the opportunity to stand alone as innovative advertising instruments (Chekembayeva et al., R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 3 2023). Luxury businesses can use social media to communicate with clients in the metaverse and apply digital advertising strategies to enhance their performance (Pangarkar et al., 2023). VR, as a promising tool for experiencing the metaverse, is growing in popularity among businesses (Sadamali Jayawardena et al., 2023); for example, Gucci has introduced ‘Gucci Town,’an interactive environment within the realm of the metaverse. The interactive components of Gucci Town include minigames, browseable art gadgets, and the Gucci store, in which consumers can purchase clothes for their avatars. In order to create highly realistic and enjoyable shopping experiences that are appealing to customers, some merchants have started working with game designers, who were early pioneers in the realm of retailing in the metaverse (Yoo et al., 2023). For instance, Uniqlo collaborated with video game developers from Mojang Studios to create a collection of T-shirts with the Minecraft theme, which were later made available in the physical world as well as in online Minecraft marketplaces (Waters, 2020). The existence of avatars is a crucial element that differentiates the experience in the metaverse from earlier online environments. Metaverse systems provide 3D avatars that reflect the individual’s personality, and can be viewed by other individuals, in contrast to current virtual worlds where individuals are identifiable by their cardinalities and photos. Platforms have integrated avatars with various capabilities to improve online interactions, due to the importance of these avatars in the realm of the metaverse (Kim et al., 2023). Recent studies indicate that virtual influencers on social media achieve a significantly higher engagement rate compared to human influencers, and this trend is expected to intensify with the emergence of the metaverse and the increased prevalence of VR (El Hedhli et al., 2023;Li et al., 2023). Research model and hypotheses Capabilities of the metaverse and the novelty and quality of RAs The degree to which recommendations are perceived as new, surprising, or unexpected by a user is referred to as novelty (Ali Abumalloh et al., 2020). Providing a suitable level of novelty is a difficult task for RAs, as it necessitates striking a counterbalance between the investigation of new products and the use of previously established preferences, while also maintaining a specific accuracy level. The usage of the metaverse can improve the novelty of recommendations by providing a rich and diverse environment in which user interactions and preferences can take place. In the metaverse, users will be able to interact with other users in ways that are not possible in the real world (Dong et al., 2023). In addition, users of the metaverse will be able to communicate with each other in a variety of ways, such as through chatting and sharing information regarding products, thereby opening up new avenues for both retailers and consumers (El Hedhli et al., 2023). These interactions have the potential to generate a huge volume of information regarding the users’preferences, behaviours, and relationships, thereby enabling the metaverse to exploit users’previous preferences. The data collected in this way can then be employed in the datasets of an RA to optimize the novelty of the selections. For instance, an RA could make use of information gleaned from user interactions in a digital shopping mall to make recommendations for new and interesting products that the user has probably never seen before. The metaverse has the capability to emerge as an innovative venue, particularly in the context of recommendations generated to users. This innovation is reflected by the data, which are collected in different forms to enable the generation of more novel recommendations. The novelty factor will enable enhanced experience through the generation of unique recommendations, which in addition to matching the users’needswillopennew choices for them. For instance, product recommendations in the metaverse can consider the real-time interactivity provided by the metaverse, and provide 3D visual effects of the products in addition to tactile sensations, unlike conventional systems that can only display images of the goods or hyperlinks of the results to the user (Wei et al., 2023). Overall, exploiting the capabilities of the metaverse will improve the novelty of recommendations. From studies in the literature that have explored the novelty of the recommendations in conventional ecommerce platforms, it is clear that the degree to which recommendationsarenovelisasignificant factor determining the quality of recommendations produced by recommender systems (Ali Abumalloh et al., 2020). We therefore propose the following two hypotheses: H1: The capabilities of the metaverse have a positive impact on the novelty of recommendations. H2: The novelty of recommendations has a positive impact on the quality of recommendations. Capabilities of the metaverse and the accuracy and quality of RAs The quality of an RA can be examined in two ways: through a system-centric or a user-centric evaluation (Cremonesi et al., 2013). In a system-centric assessment, researchers are interested in the design of and improvements to the predictive techniques used in the RA, which are used to build a product suggestion list, optimise the performance of the RA, and hence give consumers a more fulfilling experience (Knijnenburg, 2012). However, the predictive accuracy of the recommendation does not necessarily mean a fulfilling user experience, and it is essential to consider the contrast between the recommendations presented to users and the actual users’choices, since not all suggestions will become user choices (Hazrati & Ricci, 2022). In contrast, the quality of the RA is determined in a user-centric assessment using data gathered from individuals that have engaged with the system via different approaches. The quality of the recommendations determined by system-centric methodologies may produce conflicting findings compared to the results of user-centric metrics of the quality of the suggestions. Scholars have recently stated that the objective of the RA should move beyond making accurate predictions, thus indicating the importance of considering users’ perceptions of the quality of the generated recommendations (Pu et al., 2011). In this research, we will focus on users’perceptions of the accuracy of recommendations, based on the extent to which consumers believe the presented recommendations are in agreement with their tastes. The capabilities of the metaverse give rise to innovative approaches towards providing accurate recommendations, which are facilitated through the collection of users’data from diverse channels, thereby aiding the RA in generating recommendations that accurately match users’tastes and preferences. The capabilities of the metaverse can facilitate user-centered, conscious experiences by successfully adapting visualisations that meet user requirements, especially in terms of timing, location, and information presentation (Wei et al., 2023). The capabilities of the metaverse can be used to create accurate automated recommendations based on user behavioural data, such as clicks, tracks, and gaze movements (Lee, 2022). Furthermore, through an empirical analysis, the capabilities of the metaverse can enable the arrangement of visual recommendations in a way that does not obstruct the sight of other components of the system (Wei et al., 2023). Users can interact with and control these recommendations directly or through non-player characters. Intelligent robots, for example, may offer guidance, engage with people through visualisation-based interaction, and help users perceive the metaverse world. In addition, previous studies in the context of classical RAs have reported that perceived accuracy impacts the overall quality of the RA (Ali Abumalloh et al., 2020;Nilashi et al., 2016). Accordingly, we propose the following two hypotheses: H3: The capabilities of the metaverse have a positive impact on the accuracy of recommendations. R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 4 H4: The accuracy of recommendations has a positive impact on the quality of recommendations. Capabilities of the metaverse and the diversity and quality of RAs To achieve the most effective compromise in terms of the choice between accuracy and diversity, a growing number of researchers have treated the tasks of an RA as an optimisation issue, with accuracy and diversity metrics as conflicting goals (Karabadji et al., 2018). Diversification is a popular topic within research on RAs, as it assists in addressing the issue of overfitting and enhances user experience. A high level of diversity in RAs means that the system makes a variety of product selections that are customised to the user’s requirements and choices (De Biasio et al., 2023). To achieve high-quality recommendations, a diverse set of suggestions is essential. However, diversification impacts users differently depending on their personalities: people who are open to trying new things tend to favour a wider range of recommendations. Diversifying recommendations to boost user satisfaction by representing the complete range of consumer preferences is one of the primary objectives of studies on the topic of diversity (Kunaver & Po zrl, 2017). RAs that offer a diverse set of recommendations can make users more satisfied in several ways (Beel et al., 2013): firstly, diverse recommendations can introduce users to new products that they might not have discovered on their own; and secondly, a variety of recommendations can prevent users from getting bored with the system (Nilashi et al., 2016). This can increase users’trust by providing them with more options that are aligned with their interests. The metaverse can help an RA to suggest diverse products and unusual experiences that users might like, as it provides the RA with more information about what users prefer. This information is gathered from user interactions and engagement, and can be used to suggest new and diverse products. Moreover, the metaverse indirectly influences the diversity of recommendations by exposing users to a broader range of experiences, which can shape users’preferences, leading to more diverse product suggestions. Incorporating metaverse capabilities into an RA can therefore result in a richer and more satisfying user experience. Diversity in recommendations can be reflected through the capabilities of the metaverse in the form of diverse sets of products, diverse interaction experiences with the recommendations, and diverse forms of visualisation of recommendations. In addition, the innovativeness of the metaverse can help to address the problems of data sparsity and cold start, which have been linked to the diversity of recommendations in several studies (Reshma et al., 2016). To overcome these issues, the metaverse provides several forms of data collection that can aid in improving the diversity of recommendations, including dynamic data, cross-experience data, and data collected from avatars representing users (Wei et al., 2023). It is therefore likely that including the capabilities of the metaverse will increase the variety of recommendations in RAs. Building on results in the literature in regard to the impact of diversity on the quality of an RA (Nilashi et al., 2016), we propose the following hypotheses: H5: The capabilities of the metaverse have a positive impact on the diversity of recommendations. H6: The diversity of recommendations has a positive impact on the quality of recommendations. Quality, affective trust, and loyalty The majority of online retailers now use RAs to help individuals make purchases and to lessen the cognitive strains associated with information overload. However, only a limited number of studies have explored users’loyalty in the setting of e-commerce RAs (Ali Abumalloh et al., 2020;Yoon et al., 2013). Given the fierce competition in the digital world, loyalty or continuing intention has been highlighted as an indicator of the effectiveness of digital venues (Kaur et al., 2020;Tseng et al., 2018). Authors in the literature have explored individuals’interactions with online stores over time using objective metrics that can be measured by the purchase proportion; however, purchase proportion metrics related to loyalty have been critiqued for being only a partial estimation of consumers’commitment (Anderson & Srinivasan, 2003), meaning that it is important to examine the elements that can encourage people to remain loyal to online shopping sites and technologies focusing on different behavioral and attitudinal indicators. We adopted the cognitive-affective-behaviour model, referred to here as the ABC model, to investigate customers’loyalty toward an RA in the metaverse. This is a basic model that outlines a three-part structure of attitudes, detailing how people evaluate products and services through cognitive, affective, and behavioral components (Eagly & Chaiken, 1998). In the literature, these three components have been explored with a focus on various appraisal, attitudinal, and behavioural factors; for instance, Sandyopasana and Alif (2020) adopted this model to explore the factors of quality (cognition), trust and satisfaction, switching barriers (affective), and loyalty (behavior), whereas another study by Q. Zhang et al. (2023a) explored the perception of value (cognition), satisfaction (affective), and loyalty (behaviour) in the context of mobile payments. In our model, the quality dimension of an online RA can provoke an affective trust toward the RA, which can lead to loyalty towards the RA. Affective trust is defined as feelings of confidence towards the service provider, which are provoked by the care and concern the service provider extends (Johnson & Grayson, 2005). In this research, loyalty is considered a behavioural construct that is reflected by the intention to use the system again for the purchase process (Abumalloh et al., 2020). Loyalty is also reflected by the intention to react in a positive manner toward online merchants. The behavioural intention towards longstanding engagement with the system is induced by consumers’feelings of confidence (affective trust). Thus, following the ABC framework and building on the extant literature, we present the following hypotheses: H7: The quality of recommendations has a positive influence on the affective trust in the RA. H8: Consumers’affective trust has a positive influence on users’loyalty to the RA. Privacy, recommendations quality, and trust Although the metaverse offers immense potential advantages, it is imperative to acknowledge that the issues of privacy and security require attention (Fernandez & Hui, 2022), and that concerns have been consistently raised in regard to virtual systems in the literature (Barth et al., 2022;Jeon & Lee, 2022;Lata & Singh, 2022;Schomakers et al., 2021;Soumelidou & Tsohou, 2021). These concerns are more prevalent in the context of the metaverse. In a survey carried out in the United States, a significant proportion of respondents (71 %) reported apprehension over their privacy and security in the metaverse (Statista, 2022b). Moreover, 43 % of all participants harboured major concerns over the potential theft of their real identity in the metaverse. A further substantial proportion (approximately 41 %) held the belief that safeguarding their data within the metaverse would pose formidable difficulties. Privacy concerns have been explored in the literature with a focus on both their antecedents and consequences, including in the context of the metaverse. Several studies have indicated that privacy concerns in the metaverse stem from the collection and sharing of users’ R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 5 data (Alkaeed et al., 2023;Canbay et al., 2022). The metaverse is subject to several privacy risks, such as insecure design, broken authentication, data injection, phishing, unauthorised access, data theft, eavesdropping, and personal information leakage (Huang et al., 2023). Privacy concerns have been explored in contexts beyond the metaverse in the literature, with a focus on their significant impact on users’behavior across various settings. The authors of (Sheehan & Hoy, 1999) reported that privacy concerns led to more conservative behavior in terms of personal information sharing. In another study (Phelps et al., 2001), privacy concerns were found to impact purchase behaviour and the purchase decision process (catalogue purchasing habits). Anic et al. (2019) also indicated that privacy concerns impacted both the fabrication of personal information and willingness to share information. In a study by Bansal and Zahedi (2008), privacy concerns were found to moderate the relationship between the quality of privacy statements and trust. In this study, people with high levels of privacy concerns were found to rely on the adequacy of the privacy policy statement. In our research, we aim to explore how privacy concerns moderate the relationship between RA quality and customer trust. Hence, we hypothesise that perceived privacy has a moderating impact on the relationship between the recommendation quality and customer trust: H9: Perceived privacy has a moderating impact on the relationship between the quality of recommendations and customer affective trust. Product knowledge, trust, and loyalty In the realm of online commerce, goods are frequently categorised into two main types: search items, which are distinguished by the ability to verify their attributes before purchase, and experience items, where the focus is on the experiential aspect, which can only be evaluated after purchase (Yoon et al., 2013). Goods classified as ’searchable’are those that offer comparatively straightforward ways to verify and review their attributes prior to completing a purchase; in contrast, the characteristics of experiential items cannot be examined or verified easily before they are consumed. Movies often serve as prime examples of experiential goods. Numerous studies have found that consumers are more likely to heed the advice of RAs for experience items than for search items, as the evaluation of experience items is more complex than for search items (Aggarwal & Vaidyanathan, 2003). The user’s experience and knowledge of the product are important when evaluating different attributes of recommendations, particularly for experiential goods or services. In their research, Xiao and Benbasat (2007) discovered that users with higher product expertise tended to have less favorable evaluations of a CF RA. Perera (2000) explored the interaction effects between different types of RAs (CBF versus CF) and users’knowledge about product classes. The findings revealed that customers with less knowledge of the products had more positive affective reactions, such as satisfaction and affective trust, towards CF RAs compared to CBF RAs. Customers’level of knowledge about the product can also influence their perceptions of the recommendation, since experienced consumers may rely less on RA recommendations due to their extensive knowledge of the product. However, this may not be the case in the context of the metaverse, as advanced algorithms and artificial intelligence approaches are frequently used to offer consumers tailored recommendations, based on information collected from users. Hence, customers are given recommendations that are more accurate and in line with their preferences and needs when they have a better level of product knowledge. Previous studies have explored the relationships between product knowledge, quality, satisfaction, trust, and loyalty in traditional ecommerce. For instance, product knowledge was found to have a moderating impact on the relationship between recommendation quality and satisfaction in a study by Yoon et al. (2013), and on the relationship between restaurant stimuli (e.g., quality) and diners’ emotions in a study by Peng and Chen (2015). In our study, we aim to examine the impact of product knowledge on the relationship between trust and loyalty. Based on the above discussion, we hypothesise that: H10: Product knowledge has a moderating impact on the relationship between affective trust and customer loyalty. Based on the above discussion, and the proposed hypotheses, we present the initial research model in Fig. 1. Method of the study Data collection In this study, we targeted participants in Malaysia through several modes, including WhatsApp, LinkedIn, and Facebook. At the beginning of the survey, participants were provided with the following link: https://www.youtube.com/watch?v=0OtIM0L7lLo. The questionnaire includes three main parts: the first part consists of two screening questions, the second part contains demographic data, and the third part includes the main survey items. Those who did not meet the screening criteria were excluded from the survey. To ensure data completeness, all survey questions were set as required for completion. Referring to the sample size, we followed the 10 times rule of thumb by Barclay et al. (1995), which suggests that the sample size should be at least 10 times the larger of: (1) the highest number of formative indicators used to measure a single construct, or (2) the highest number of structural paths directed at a particular construct in the structural model. This rule of thumb effectively means that the minimum sample size should be 10 times the maximum number of arrowheads pointing at a latent variable in the PLS path model. Participants were asked to indicate their level of familiarity with recommender agents and their knowledge of using the metaverse in the retail industry. Responses from participants who indicated they were not at all familiar with either of these topics were excluded from the study. The main survey comprises nine sections designed to measure research hypotheses. Each item in the survey is assessed using a 5point Likert scale (see Appendix A). The survey items allow participants to express their attitudes toward the variables subjectively. The data collection process spanned approximately four months, from January 2023 to April 2023. We received 288 valid responses, which were used for analysis. The demographic data analysis is presented in Table 1. As shown in Table 1, the majority of respondents were male, with the most common age range being 36−40. Most participants reported an average income of $751−$1500. Additionally, most respondents were moderately familiar with both the recommender system and the use of the metaverse. Empirical results We used SmartPLS to conduct analyses on both the structural and measurement models to ensure the validity and reliability of our research model. We adopted a structured approach to evaluate the quality of the collected data and the significance of the hypotheses using PLS-SEM. This method is robust, as it can handle complex models involving multiple constructs, indicators, and layers of relationships, making it well-suited for the context of our study (Hair Jr et al., 2020). It can also handle both reflective and formative measurement models. Structural Equation Modeling (SEM) is used to assess the relationships between independent and dependent variables (Hair et al., 2013). The quantitative research community recognizes SEM as a reliable method for factor analysis and path analysis. To ensure the survey yielded R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 6 accurate and valid results, we conducted three main checks on the outer model using the tool: Convergent Validity (CV), Internal Consistency (IC), and Discriminant Validity (DV). For the CV test, all survey indicators exhibited outer loadings above 0.4. Hence, following the guideline set by Hair et al. (2013), we decided to retain all items for further analysis. The next step in evaluating Convergent Validity (CV) involves the average variance extracted (AVE). The AVE test requires that the correlation between items within the same factor must meet a minimum threshold of 0.5. All factors in the proposed model satisfied this AVE criterion. To assess Internal Consistency (IC) of the outer model, both Composite Reliability (CR) and Cronbach’s Alpha (CA) tests were employed. Each factor in the study model needed to achieve a minimum threshold of 0.7 for both tests. The analysis confirmed that the IC of the outer model was supported (see Table 2). Several tests were conducted to assess the Discriminant Validity (DV) of the model: the Heterotrait-Monotrait Ratio of Correlations Fig. 1. Initial research model. Table 1 Demographic results of the participants. Item N = 288 Frequency Percent Gender Female 69 24.0 Male 219 76.0 Age Under 30 9 3.1 30 −35 52 18.1 36 −40 81 28.1 41 −45 64 22.2 46 −50 63 21.9 51 and over 19 6.6 Income 0−500$ 21 7.29 $501 - $750 94 32.64 $751−$1500 113 39.24 $1501-$2000 31 10.76 Above 2000 29 10.07 Favorite E-Commerce Website Alibaba 4 1.4 Amazon 46 16.0 eBay 3 1.0 Lazada 62 21.5 Mudah.my 36 12.5 Namshi 4 1.4 Shein 4 1.4 Shopee 27 9.4 Taobao 66 22.9 Other 36 12.5 Level of Familiarity with the Recommender System High Familiarity 117 40.63 Moderate Familiarity 157 54.51 Low Familiarity 14 4.86 Level of Familiarity with the Usage of the Metaverse in the Retail Industry High Familiarity 125 43.4 Moderate Familiarity 147 51.04 Low Familiarity 16 5.56 Table 2 Constructs reliability and validity. Item Outer loadings Cronbach’s Alpha Composite Reliability AVE Accuracy 0.722 0.825 0.550 ACC1 0.641 ACC2 0.548 ACC3 0.856 ACC4 0.869 Customer Loyalty 0.820 0.893 0.735 CL1 0.872 CL2 0.824 CL3 0.875 Diversity 0.731 0.882 0.788 DIV1 0.888 DIV2 0.888 Novelty 0.785 0.903 0.823 NOV1 0.898 NOV2 0.916 Perceived Privacy 0.826 0.884 0.657 PPV1 0.783 PPV2 0.807 PPV3 0.836 PPV4 0.816 Product Knowledge 0.719 0.840 0.637 PRK1 0.834 PRK2 0.800 PRK3 0.758 Recommendation Quality 0.844 0.895 0.681 RECQ1 0.844 RECQ2 0.797 RECQ3 0.847 RECQ4 0.811 Trust 0.890 0.948 0.900 TRU1 0.954 TRU2 0.944 Metaverse Capabilities 0.722 0.844 0.643 MC1 0.810 MC2 0.832 MC3 0.763 R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 7 (HTMT), the Fornell-Larcker (FL) criterion, and Cross Loadings (CL). The DV test evaluates the degree of differentiation between research factors. The HTMT criterion measures the average correlations between indicators across different constructs. The FL test ensures that the correlation between each factor and other factors in the model is less than the square root of the AVE for that factor. The CL test requires that the outer loadings of indicators for each factor be greater than their cross-loadings. The results of the DV tests are detailed in Tables 3,4−5. In the next stage, the relationships between research variables were examined. Evaluating the research model involves testing the proposed hypotheses using path coefficient analysis, which is a crucial step. Additional tests for the inner model include examining coefficients of determination and effect size. The final inner model is presented in Figs. 2 and 3, and Tables 6 and 7. To assess the model’s paths, a bootstrapping procedure was conducted (Hair et al., 2013). The results confirm the significance of all research hypotheses within the model. Analysis of the inner model showed that the impact of metaverse capabilities on the accuracy of RAs is the strongest among the research paths, with a coefficient of 0.650. This is followed by the influence of trust on user loyalty, which has a coefficient of 0.552. The model’s predictive accuracy was assessed using the R-squared test, which evaluates the proportion of variance in the endogenous variable explained by the exogenous variables (Hair et al., 2013). Rsquared values range from 0 to 1, with higher values indicating greater predictive accuracy. In this study, R-squared values range from 0.111 to 0.492. Given that the research focuses on consumeroriented topics, specifically on understanding trust and loyalty among consumers, an R-squared value of 0.2 is considered significant (Hair et al., 2013). The results indicate that the accuracy, novelty, and diversity of RAs account for 48.8 % of the variance in recommendation quality. Additionally, the research model explains 46.7 % of the variance in user loyalty. Furthermore, the model accounts for 49.2 % of the variance in users’trust. The moderation effect indicates that the presence of a third variable can either strengthen or weaken the relationship between an endogenous factor and an exogenous factor (Hair et al., 2013). This study primarily investigates how perceived privacy affects the relationship between RA quality and trust. Our goal is to determine the significance of the moderator’s impact on this relationship. To achieve this, we employed a two-stage approach to operationalize the interaction effect using SmartPLS (Hair et al., 2013). The analysis results (Fig. 4) revealed that perceived privacy moderates the relationship between RA quality and consumer trust. Specifically, a higher level of perceived privacy is associated with a stronger positive relationship between recommendation quality and customer trust, as indicated by a significant beta coefficient of 0.092. Conversely, the moderation effect of product knowledge on the relationship between trust and loyalty was not supported. Table 3 Heterotrait-monotrait ratio (HTMT). Construct ACC CL TRU DIV NOV PPV PPV RECQ MC Accuracy Customer Loyalty 0.771 Customer Trust 0.716 0.742 Diversity 0.669 0.850 0.751 Novelty 0.687 0.735 0.605 0.769 Perceived Privacy 0.880 0.876 0.760 0.838 0.807 Product Knowledge 0.840 0.568 0.409 0.562 0.576 0.601 Recommendation Quality 0.709 0.742 0.693 0.735 0.695 0.803 0.461 Metaverse Capabilities 0.867 0.605 0.563 0.458 0.520 0.685 0.579 0.557 Table 4 Fornell-Larcker criterion. Construct ACC CL TRU DIV NOV PPV PPV RECQ MC Accuracy 0.742 Customer Loyalty 0.604 0.857 Customer Trust 0.579 0.636 0.949 Diversity 0.482 0.659 0.607 0.888 Novelty 0.517 0.595 0.513 0.585 0.907 Perceived Privacy 0.702 0.724 0.655 0.652 0.655 0.811 Product Knowledge 0.597 0.447 0.334 0.415 0.439 0.477 0.798 Recommendation Quality 0.577 0.617 0.609 0.581 0.573 0.673 0.370 0.825 Metaverse Capabilities 0.650 0.467 0.452 0.333 0.389 0.529 0.423 0.437 0.802 Table 5 Cross loadings results. Items ACC CL DIV MC NOV PPV PRK RECQ TRU ACC1 0.641 0.357 0.314 0.381 0.330 0.345 0.567 0.300 0.354 ACC2 0.548 0.354 0.324 0.296 0.336 0.400 0.319 0.306 0.349 ACC3 0.869 0.526 0.411 0.584 0.429 0.665 0.477 0.537 0.508 ACC4 0.856 0.523 0.385 0.592 0.436 0.601 0.445 0.508 0.484 CL1 0.574 0.872 0.582 0.423 0.565 0.640 0.433 0.550 0.583 CL2 0.461 0.824 0.529 0.353 0.422 0.621 0.345 0.518 0.520 CL3 0.510 0.875 0.581 0.422 0.534 0.600 0.365 0.517 0.528 DIV1 0.422 0.577 0.888 0.281 0.520 0.545 0.331 0.524 0.551 DIV2 0.434 0.592 0.888 0.309 0.518 0.613 0.405 0.508 0.527 MC1 0.568 0.405 0.262 0.810 0.299 0.420 0.384 0.341 0.332 MC2 0.534 0.351 0.273 0.832 0.309 0.413 0.299 0.392 0.411 MC3 0.456 0.368 0.266 0.763 0.330 0.441 0.334 0.317 0.344 NOV1 0.448 0.498 0.484 0.379 0.898 0.529 0.368 0.464 0.371 NOV2 0.488 0.577 0.573 0.329 0.916 0.654 0.426 0.571 0.551 PPV1 0.703 0.608 0.495 0.519 0.519 0.783 0.414 0.560 0.548 PPV2 0.585 0.602 0.502 0.484 0.552 0.807 0.415 0.485 0.515 PPV3 0.501 0.625 0.583 0.372 0.575 0.836 0.361 0.601 0.561 PPV4 0.478 0.504 0.532 0.335 0.471 0.816 0.354 0.531 0.492 PRK1 0.497 0.327 0.298 0.348 0.341 0.346 0.834 0.258 0.241 PRK2 0.514 0.415 0.384 0.370 0.383 0.473 0.800 0.351 0.312 PRK3 0.406 0.310 0.296 0.284 0.318 0.293 0.758 0.261 0.234 RECQ1 0.505 0.499 0.465 0.397 0.511 0.598 0.345 0.844 0.493 RECQ2 0.402 0.511 0.417 0.319 0.405 0.528 0.236 0.797 0.399 RECQ3 0.498 0.534 0.527 0.373 0.449 0.569 0.340 0.847 0.522 RECQ4 0.487 0.496 0.498 0.349 0.515 0.528 0.290 0.811 0.573 TRU1 0.563 0.623 0.599 0.443 0.535 0.655 0.337 0.607 0.954 TRU2 0.535 0.582 0.550 0.413 0.434 0.585 0.296 0.546 0.944 R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 8 Heidari, N., Moradi, P., & Koochari, A. (2022). An attention-based deep learning method for solving the cold-start and sparsity issues of recommender systems. KnowledgeBased Systems, 256, 109835. doi:10.1016/j.knosys.2022.109835. Herce-Zelaya, J., Porcel, C., Bernab e-Moreno, J., Tejeda-Lorente, A., & Herrera-Viedma, E. (2020). New technique to alleviate the cold start problem in recommender systems using information from social media and random decision forests. Information Sciences, 536, 156–170. doi:10.1016/j.ins.2020.05.071. Hong, I. B., & Cho, H. (2011). The impact of consumer trust on attitudinal loyalty and purchase intentions in B2C e-marketplaces: Intermediary trust vs. seller trust. International Journal of Information Management, 31(5), 469–479. doi:10.1016/j. ijinfomgt.2011.02.001. Huang, W.-y., Schrank, H., & Dubinsky, A. J. (2004). Effect of brand name on consumers’ risk perceptions of online shopping. Journal of Consumer Behaviour, 4(1), 40–50. doi:10.1002/cb.156. Huang, Y., Li, Y. J., & Cai, Z. (2023). Security and privacy in metaverse: A comprehensive survey. Big Data Mining and Analytics, 6(2), 234–247. Hwang, S., & Kim, S. (2018). Does mIM experience affect satisfaction with and loyalty toward O2O services? Computers in Human Behavior, 82(2018), 70–80. doi:10.1016/j.chb.2017.12.044. Islam, A. K. M. N., M€ antym€ aki, M., Laato, S., & Turel, O. (2022). Adverse consequences of emotional support seeking through social network sites in coping with stress from a global pandemic. International Journal of Information Management, 62, 102431. doi:10.1016/j.ijinfomgt.2021.102431. Jeon, H., & Lee, C. (2022). Internet of things technology: Balancing privacy concerns with convenience. Telematics and Informatics, 70, 101816. doi:10.1016/j.tele.2022.101816. Johnson, D., & Grayson, K. (2005). Cognitive and affective trust in service relationships. Journal of Business Research, 58(4), 500–507. Joorabloo, N., Jalili, M., & Ren, Y. (2021). A probabilistic graph-based method to solve precision-diversity dilemma in recommender systems. Expert Systems with Applications, 184, 115485. doi:10.1016/j.eswa.2021.115485. Jung, T., Chung, N., & Tom Dieck, M. C. (2015). The determinants of recommendations to use augmented reality technologies - The case of a Korean theme park. Tourism Management, 49,75–86. doi:10.1016/j.tourman.2015.02.013. Karabadji, N. E. I., Beldjoudi, S., Seridi, H., Aridhi, S., & Dhifli, W. (2018). Improving memory-based user collaborative filtering with evolutionary multi-objective optimization. Expert Systems with Applications, 98, 153–165. doi:10.1016/j. eswa.2018.01.015. Karimi, M., Jannach, D., & Jugovac, M. (2018). News recommender systems −Survey and roads ahead. Information Processing & Management, 54(6), 1203–1227. doi:10.1016/j.ipm.2018.04.008. Kaur, H., Paruthi, M., Islam, J., & Hollebeek, L. D. (2020). The role of brand community identification and reward on consumer brand engagement and brand loyalty in virtual brand communities. Telematics and Informatics, 46, 101321. doi:10.1016/j. tele.2019.101321. Khan, Z., Hussain, M. I., Iltaf, N., Kim, J., & Jeon, M. (2021). Contextual recommender system for E-commerce applications. Applied Soft Computing, 109, 107552. doi:10.1016/j.asoc.2021.107552. Kim, D. J., Ferrin, D. L., & Rao, H. R. (2008). A trust-based consumer decision-making model in electronic commerce: The role of trust, perceived risk, and their antecedents. Decision Support Systems, 44(2), 544–564. doi:10.1016/j.dss.2007.07.001. Kim, D. J., Song, Y. I., Braynov, S. B., & Rao, H. R. (2005). A multidimensional trust formation model in B-to-C e-commerce: a conceptual framework and content analyses of academia/practitioner perspectives. Decision Support Systems, 40(2), 143–165. doi:10.1016/j.dss.2004.01.006. Kim, D. Y., Lee, H. K., & Chung, K. (2023). Avatar-mediated experience in the metaverse: The impact of avatar realism on user-avatar relationship. Journal of Retailing and Consumer Services, 73, 103382. doi:10.1016/j.jretconser.2023.103382. Kim, J., Jin, B., & Swinney, J. L. (2009). The role of etail quality, e-satisfaction and e-trust in online loyalty development process. Journal of Retailing and Consumer Services, 16(4), 239–247. doi:10.1016/j.jretconser.2008.11.019. Klaus, P. P., & Kuppelwieser, V. (2023). A glimpse of the future retail customer experience −Guidelines for research and practice. Journal of Retailing and Consumer Services, 73, 103205. doi:10.1016/j.jretconser.2022.103205. Knijnenburg, B. P. (2012). Conducting user experiments in recommender systems. In Proceedings of the sixth ACM conference on recommender systems (pp. 3−4). http:// delivery.acm.org.library.iau.edu.sa/10.1145/2370000/2365956/p3-knijnenburg. pdf?ip=91.227.24.65&id=2365956&acc=ACTIVE SERVICE&- key=4A313E3191DFF965.CD6AA1E1785C7CCE.4D4702B0C3E38B35.4D4702B0C3E38B35&CFID=849210383&CFTOKEN=51692793&__acm__=1515437624_. Knijnenburg, B. P., Willemsen, M. C., Gantner, Z., Soncu, H., & Newell, C. (2012). Explaining the user experience of recommender systems. User Modeling and UserAdapted Interaction, 22(4-5), 441–504. doi:10.1007/s11257-011-9118-4. Komiak, S. Y. X., & Benbasat, I. (2006). The effects of personalization and familiarity on trust and adoption of recommendation agents. MIS Quarterly, 30(4), 941–960. doi:10.2307/25148760. Kunaver, M., & Po zrl, T. (2017). Diversity in recommender systems −A survey. Knowledge-Based Systems, 123, 154–162. doi:10.1016/j.knosys.2017.02.009. Kunkel, J., & Ziegler, J. (2023). A comparative study of item space visualizations for recommender systems. International Journal of Human-Computer Studies, 172, 102987. doi:10.1016/j.ijhcs.2022.102987. Kuo, R. J., & Cheng, H.-R. (2022). A content-based recommender system with consideration of repeat purchase behavior. Applied Soft Computing, 127, 109361. doi:10.1016/j.asoc.2022.109361. Kuo, R. J., & Li, S.-S. (2023). Applying particle swarm optimization algorithm-based collaborative filtering recommender system considering rating and review. Applied Soft Computing, 135, 110038. doi:10.1016/j.asoc.2023.110038. Kwon, J., & Vogt, C. A. (2010). Identifying the role of cognitive, affective, and behavioral components in understanding residents’attitudes toward place marketing. Journal of Travel Research, 49(4), 423–435. Lata, S., & Singh, D. (2022). Intrusion detection system in cloud environment: Literature survey & future research directions. International Journal of Information Management Data Insights, 2,(2) 100134. doi:10.1016/j.jjimei.2022.100134. Lazarus, R. S. (1991). Cognition and motivation in emotion. American Psychologist, 46 (4), 352–367. doi:10.1037/0003-066X.46.4.352. Lee, U.-K. (2022). Tourism using virtual reality: Media richness and information system successes. Sustainability, 14(7), 3975. Lenzini, G., van Houten, Y., Huijsen, W., & Melenhorst, M. (2010). Shall i trust a recommendation? Towards an evaluation of the trustworthiness of recommender sites: 5968 (pp. 121−128). LNCS. doi:10.1007/978-3-642-12082-4_16. Li, H., Lei, Y., Zhou, Q., & Yuan, H. (2023). Can you sense without being human? Comparing virtual and human influencers endorsement effectiveness. Journal of Retailing and Consumer Services, 75, 103456. doi:10.1016/j.jretconser.2023.103456. Liang, Y., Zhang, X., Wang, H., & Liu, M. (2024). Users’willingness to adopt Metaverse drawing on flow theory: An empirical study using PLS-SEM and FsQCA. Heliyon, e33394. doi:10.1016/j.heliyon.2024.e33394. Lin, H.-H., & Wang, Y.-S. (2006). An examination of the determinants of customer loyalty in mobile commerce contexts. Information & Management, 43(3), 271–282. doi:10.1016/j.im.2005.08.001. Liu, B. Q., & Goodhue, D. L. (2012). Two worlds of trust for potential e-commerce users [21_Publication in refereed journal]. Humans as Cognitive Misers, 23(4), 1246– 1262. doi:10.1287/isre.1120.0424. Liu, J., Shi, C., Yang, C., Lu, Z., & Yu, P. S. (2022). A survey on heterogeneous information network based recommender systems: Concepts, methods, applications and resources. AI Open, 3,40–57. doi:10.1016/j.aiopen.2022.03.002. Liu, Q., Hao, Z., Peng, Y., Jiang, H., Wu, J., Peng, T., Wang, G., & Zhang, S. (2021). SecVKQ: Secure and verifiable kNN queries in sensor−cloud systems. Journal of Systems Architecture, 120, 102300. doi:10.1016/j.sysarc.2021.102300. Mancuso, I., Messeni Petruzzelli, A., & Panniello, U. (2023). Digital business model innovation in metaverse: How to approach virtual economy opportunities. Information Processing & Management, 60,(5) 103457. doi:10.1016/j.ipm.2023.103457. Margaris, D., Vassilakis, C., & Georgiadis, P. (2018). Query personalization using social network information and collaborative filtering techniques. Future Generation Computer Systems, 78, 440–450. doi:10.1016/j.future.2017.03.015. McNee, S.M., Riedl, J., & Konstan, J.A. (2006). Being accurate is not enough: how accuracy metrics have hurt recommender systems. CHI’06 extended abstracts on Human factors in computing systems, Nilashi, M., Ali Abumalloh, R., Samad, S., Minaei-Bidgoli, B., Hang Thi, H., Alghamdi, O. A., Yousoof Ismail, M., & Ahmadi, H. (2023). The impact of multi-criteria ratings in social networking sites on the performance of online recommendation agents. Telematics and Informatics, 76, 101919. doi:10.1016/j. tele.2022.101919. Nilashi, M., Asadi, S., Abumalloh, R. A., Samad, S., & Ibrahim, O. (2020). Intelligent recommender systems in the COVID-19 outbreak: The case of wearable healthcare devices. Journal of Soft Computing and Decision Support Systems, 7(4), 8–12. Nilashi, M., Jannach, D., Ibrahim, O.b., Esfahani, M. D., & Ahmadi, H. (2016). Recommendation quality, transparency, and website quality for trust-building in recommendation agents. Electronic Commerce Research and Applications, 19,70–84. doi:10.1016/j.elerap.2016.09.003. O’Donovan, J., & Smyth, B. (2005). Trust in recommender systems. Oliver, R. L. (1999). Whence consumer loyalty? Journal of Marketing, 63,33–44. doi:10.2307/1252099. Ozdemir, S., Zhang, S., Gupta, S., & Bebek, G. (2020). The effects of trust and peer influence on corporate brand—Consumer relationships and consumer loyalty. Journal of Business Research, 117, 791–805. doi:10.1016/j.jbusres.2020.02.027. Palanisamy, R. (2014). The impact of privacy concerns on trust, attitude and intention of using a search engine: an empirical analysis. International Journal of Electronic Business, 11(3), 274–296. Pang, H., & Ruan, Y. (2023). Can information and communication overload influence smartphone app users’social network exhaustion, privacy invasion and discontinuance intention? A cognition-affect-conation approach. Journal of Retailing and Consumer Services, 73, 103378. doi:10.1016/j.jretconser.2023.103378. Pangarkar, A., Patel, J., & Kumar, S. K. (2023). Drivers of eWOM engagement on social media for luxury consumers: Analysis, implications, and future research directions. Journal of Retailing and Consumer Services, 74, 103410. doi:10.1016/j.jretconser.2023.103410. Park, H., & Lim, R. E. (2023). Fashion and the metaverse: Clarifying the domain and establishing a research agenda. Journal of Retailing and Consumer Services, 74, 103413. doi:10.1016/j.jretconser.2023.103413. Peng, N., & Chen, A. H. (2015). Diners’loyalty toward luxury restaurants: the moderating role of product knowledge. Marketing Intelligence & Planning, 33(2), 179–196. Perera, R. E. (2000). Optimizing human-computer interaction for the electronic commerce environment. Journal of Electronic Commerce Research, 1(1), 23–44. Phelps, J. E., D’Souza, G., & Nowak, G. J. (2001). Antecedents and consequences of consumer privacy concerns: An empirical investigation. Journal of Interactive Marketing, 15(4), 2–17. doi:10.1002/dir.1019. Pu, P., Chen, L., & Hu, R. (2011). A user-centric evaluation framework for recommender systems. In Proceedings of the fifth ACM conference on recommender systems. R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 15 Renjith, S., Sreekumar, A., & Jathavedan, M. (2020). An extensive study on the evolution of context-aware personalized travel recommender systems. Information Processing & Management, 57,(1) 102078. doi:10.1016/j.ipm.2019.102078. Reshma, R., Ambikesh, G., & Thilagam, P. S. (2016). Alleviating data sparsity and cold start in recommender systems using social behaviour. 2016 international conference on recent trends in information technology (ICRTIT). Sadamali Jayawardena, N., Thaichon, P., Quach, S., Razzaq, A., & Behl, A. (2023). The persuasion effects of virtual reality (VR) and augmented reality (AR) video advertisements: A conceptual review. Journal of Business Research, 160, 113739. doi:10.1016/j.jbusres.2023.113739. Sandyopasana,T.,&Alif,M.G.(2020).Influence of perceived quality of mobile payment application towards loyalty. Understanding digital industry (pp. 163−167). Routledge. Scholz, M., Dorner, V., Schryen, G., & Benlian, A. (2017). A configuration-based recommender system for supporting e-commerce decisions. European Journal of Operational Research, 259(1), 205–215. doi:10.1016/j.ejor.2016.09.057. Schomakers, E.-M., Biermann, H., & Ziefle, M. (2021). Users’preferences for smart home automation −investigating aspects of privacy and trust. Telematics and Informatics, 64, 101689. doi:10.1016/j.tele.2021.101689. Semin, G. R., & Garrido, M. V. (2015). Socially situated cognition. Theory and explanation in social psychology (pp. 283−302). The Guilford Press. Serravalle, F., Vanheems, R., & Viassone, M. (2023). Does product involvement drive consumer flow state in the AR environment? A study on behavioural responses. Journal of Retailing and Consumer Services, 72, 103279. doi:10.1016/j.jretconser.2023.103279. Sheehan, K. B., & Hoy, M. G. (1999). Flaming, complaining, abstaining: How online users respond to privacy concerns. Journal of Advertising, 28(3), 37–51. Slokom, M., Hanjalic, A., & Larson, M. (2021). Towards user-oriented privacy for recommender system data: A personalization-based approach to gender obfuscation for user profiles. Information Processing & Management, 58,(6) 102722. doi:10.1016/j. ipm.2021.102722. Soumelidou, A., & Tsohou, A. (2021). Towards the creation of a profile of the information privacy aware user through a systematic literature review of information privacy awareness. Telematics and Informatics, 61, 101592. doi:10.1016/j. tele.2021.101592. Statista. (2022). Potential security concerns according to metaverse enthusiasts in the United States as of August 2022. Retrieved 2023 from https://www.statista.com/sta tistics/1346474/us-metaverse-security-concerns-2022/ Statista. (2022). Concerns posed by the metaverse according to adults in the United States as of December 2021. Retrieved March from https://www.statista.com/statistics/ 1288065/united-states-adults-concerns-about-the-metaverse/ Stephens, D. (2021). The metaverse will radically change retail. Business of fashion. . Retrieved June, 6 from https://www.businessoffashion.com/opinions/retail/themetaverse-will-radically-change-retail/. Tewari, A. S. (2020). Generating items recommendations by fusing content and useritem based collaborative filtering. Procedia Computer Science, 167, 1934–1940. doi:10.1016/j.procs.2020.03.215. Tsai, Y. C., & Yeh, J. C. (2010). Perceived risk of information security and privacy in online shopping: A study of environmentally sustainable products. African Journal of Business Management, 4(18), 4057. Tseng, F.-C., Pham, T. T. L., Cheng, T. C. E., & Teng, C.-I. (2018). Enhancing customer loyalty to mobile instant messaging: Perspectives of network effect and self-determination theories. Telematics and Informatics, 35(5), 1133–1143. doi:10.1016/j. tele.2018.01.011. Vazquez, E. E., Patel, C., Alvidrez, S., & Siliceo, L. (2023). Images, reviews, and purchase intention on social commerce: The role of mental imagery vividness, cognitive and affective social presence. Journal of Retailing and Consumer Services, 74, 103415. doi:10.1016/j.jretconser.2023.103415. Wang, S., Zhang, P., Wang, H., Yu, H., & Zhang, F. (2022). Detecting shilling groups in online recommender systems based on graph convolutional network. Information Processing & Management, 59,(5) 103031. doi:10.1016/j.ipm.2022.103031. Wang, Y., Pan, J., & Chen, Y. (2021). Fine-grained secure attribute-based encryption. Advances in cryptology − crypto 2021: 41st annual international cryptology conference, crypto 2021, virtual event (p. 41) August 16−20, 2021, ProceedingsPart IV. Waters, M. (2020). Why video games are the next retail frontier. Retrieved June 5, 2023 from https://www.modernretail.co/retailers/why-video-games-are-the-nextretail-frontier/ Wei, L., Wang, X., Wang, T., Duan, Z., Hong, Y., He, X., & Huang, H. (2023). Recommendation systems for the metaverse. Blockchains, 1(1), 19–33. Wetsch, L. R. (2013). Trust, satisfaction and loyalty in customer relationship management: an application of justice theory: 4 (pp. 29−42). The Haworth Press, Inc.. doi:10.1300/J366v04n03_03. Wong, W. P. M., Tan, K. L., Ida, A. K., & Lim, B. C. Y. (2019). The effect of technology trust on customer e-loyalty in online shopping and the mediating effect of trustworthiness. Journal of Marketing Advances and Practices, 1(2), 38–51. Xiao, & Benbasat (2007). E-commerce product recommendation agents: Use, characteristics, and impact. MIS Quarterly, 31(1), 137–209. doi:10.2307/25148784. Yoo, K., Welden, R., Hewett, K., & Haenlein, M. (2023). The merchants of meta: A research agenda to understand the future of retailing in the metaverse. Journal of Retailing, 99(2), 173–192. doi:10.1016/j.jretai.2023.02.002. Yoon, V. Y., Hostler, R. E., Guo, Z., & Guimaraes, T. (2013). Assessing the moderating effect of consumer product knowledge and online shopping experience on using recommendation agents for customer loyalty. Decision Support Systems, 55(4), 883–893. doi:10.1016/j.dss.2012.12.024. Zallio, M., & Clarkson, P. J. (2022). Designing the metaverse: A study on inclusion, diversity, equity, accessibility and safety for digital immersive environments. Telematics and Informatics, 75, 101909. doi:10.1016/j.tele.2022.101909. Zanon, A. L., Rocha, L. C. D.d., & Manzato, M. G. (2022). Balancing the trade-off between accuracy and diversity in recommender systems with personalized explanations based on linked open data. Knowledge-Based Systems, 252, 109333. doi:10.1016/j. knosys.2022.109333. Zhang, Q., Ariffin, S. K., Richardson, C., & Wang, Y. (2023). Influencing factors of customer loyalty in mobile payment: A consumption value perspective and the role of alternative attractiveness. Journal of Retailing and Consumer Services, 73, 103302. doi:10.1016/j.jretconser.2023.103302. Zhang, S., Wang, Y., Luo, E., Liu, Q., Gu, K., & Wang, G. (2023). A traceable and revocable decentralized multi-authority privacy protection scheme for social metaverse. Journal of Systems Architecture, 140, 102899. doi:10.1016/j.sysarc.2023.102899. R.A. Abumalloh, M. Nilashi, O. Halabi et al. Journal of Innovation & Knowledge 9 (2024) 100569 16