2.1 The role of new-age technologies in marketing

The landscape of marketing is undergoing a monumental shift due to the advent of new-age technologies such as artificial intelligence (AI), autonomous robots, big data analytics, blockchain, cloud computing, internet of things (IoT), and technology-mediated realities such as AR and VR (Belk et al., 2023). These burgeoning technologies open a realm of opportunities for businesses, including the capacity to anticipate market trends and proactively formulate responses, enabling real-time customer interaction through adaptive engagement strategies, automating processes to minimize human intervention for employees and customers, personalizing communication to meet individual preferences, and fostering sustainable customer engagement (Kumar et al., 2021).

Extended reality, an overarching term for AR and VR, is creatively redefining the dynamics between brands and consumers (Orús et al., 2021). As posited by Sadamali Jayawardena et al. (2023, p. 2), “when the physical environment becomes a part of the user experience, it is AR; otherwise, it is VR.” However, owing to the nuanced differences between AR, VR, and other similar realities, Flavián et al. (2019) proposed a “reality-virtuality continuum” to distinctly delineate the transition from real to virtual environments. According to this model, at the far end of the continuum, past the virtual environment, lies VR, a digitally-constructed artificial world where users can navigate and interact in real-time. Moving along the continuum, we find augmented virtuality, where real content overlays the user's virtual environment. This is succeeded by mixed reality (MR), a term originally encompassing both AR and augmented virtuality. However, as Flavián et al. (2019) clarified, it has been redefined as pure MR, where users immersed in the physical world also have digital content fully integrated into their surroundings. In this configuration, users can interact with both real and digital elements, and these elements can reciprocally interact as well. Nearing the real environment end of the continuum, AR emerges, seamlessly weaving virtual elements into a real-world view in real-time (Sadamali Jayawardena et al., 2023). AR, thus, effectively simulates a physical product examination and shopping experience akin to those provided in physical stores (Huang & Liao, 2015). The augmentation of service via AR creates a unique, personalized encounter for each user, potentially transforming the customer's relationship with the brand.

2.2 The role of emotions in consumer behavior

The stimulus-organism-response (S-O-R) model, a significant contribution to environmental psychology from Mehrabian and Russell (1974), theorizes that environmental stimuli provoke emotions. Though the emotional responses to a particular situation might vary among individuals, Mehrabian and Russell (1974) suggest these emotions can be classified under three orthogonal dimensions: pleasure, arousal, and dominance (PAD). Despite the complex nature of emotions making them challenging to examine, their critical role in marketing and understanding consumer behavior has been repeatedly underscored (e.g., Bagozzi & Pieters, 1998; Bagozzi et al., 1999).

Influential models of consumer behavior also began integrating the concept of anticipated emotions with the advent of the model of goal-directed behavior (Perugini & Bagozzi, 2001). Bagozzi et al. (2016) proposed that consumers often contemplate the positive and negative emotions they might experience as a result of their actions or inactions in hypothetical purchase scenarios. As defined by Bagozzi and Dholakia (2002, p. 8), anticipated emotions are “prefactual appraisals where the individual imagines the affective consequences of goal attainment and goal failure before deciding to act and hence these are predictors for action decision.” In essence, this prefactual state relies on the envisioned success or failure of a goal, emphasizing future outcomes (Bagozzi et al., 2016), as opposed to the actual induction of emotions.

Emotions are typically intense and volatile, often arising from specific stimuli (Bigné et al., 2008). Scherer (2000) argued for an understanding of emotion as a process encompassing several key components. This componential perspective on emotions portrays them as multifaceted processes encompassed in various loosely connected components (Frijda, 1999; Larsen & Fredrickson, 1999; Plutchik, 2001; Scherer, 2000). The primary purpose of this multicomponent model is to establish a more explicit link between the situations that elicit emotions and the patterns of responses they trigger (Scherer, 2000). A set of definitions, which views emotions as a multicomponent construct, is presented in Table 1.

In the context of consumer behavior, Laros and Steenkamp (2005) found that emotions are most commonly conceptualized based on their valence—that is, whether they elicit positive or negative affect. However, research has demonstrated that it is vital to examine emotions with the same valence as they can exert differing influences (Zeelenberg & Pieters, 1999). Bagozzi et al. (1999, p. 187) identified that positive emotions occasionally correlate with “higher levels of physiological arousal, expanded attention, increased optimism, enhanced recall” and are generally tied to goal attainment. This association encourages consumers to pursue plans that were incited by their emotions.

2.3 The role of emotions in AR

With the advent of AR in the domain of marketing, a subsequent wave of academic inquiry soon followed, primarily leveraging the technology acceptance model (TAM) as a key analytical tool (Huang & Liao, 2015; Pantano et al., 2017; Rese et al., 2014, 2017). The TAM, a robust and efficient model, has been validated empirically and iteratively refined to specifically evaluate the acceptance and adoption of AR in marketing contexts. Nonetheless, continuing studies have not only utilized TAM but have also enhanced its scope by incorporating additional antecedent factors that influence AR acceptance and adoption. Noteworthily, several alternative theories have been incorporated, including the flow theory, gratification theory, S-O-R model, and human–computer interaction model (Qin et al., 2021) to frame the adoption behavior and usage of AR technology more comprehensively. Several researchers, recognizing AR as an interactive technology, have also initiated inquiries rooted in technology and media characteristics (Javornik, 2016a). This recognition underscores the intrinsic interactive nature of AR and its potential to form new paradigms of consumer engagement.

In addition, there has been a growing recognition of the need to delve deeper into the emotional aspect of AR. Yim et al. (2017) advocated for a focus on the emotional impacts of AR, emphasizing the exploration of how visual elements presented within an AR context can elicit emotional responses from consumers, in contrast to traditional text-based content. Supporting this focus on emotional engagement, Ghazali et al. (2019) illustrated through their research that flow significantly influenced the intent to continue engaging with AR gaming through the mediating factor of enjoyment. This suggests that when consumers perceive an experience as enjoyable and intriguing, they are more likely to develop a positive behavioral intention to continue with that experience. The researchers used enjoyment as an encompassing concept to signify various positive emotions, thus underscoring the pivotal role of positive emotions in enhancing user engagement with AR. Their study reinforces the need for more nuanced understanding and exploration of the complex interplay between AR and emotions.

In the realm of AR within marketing research, various affective variables associated with AR have been extensively explored. Many of these studies have been characterized by a singular focus on emotional valence, as detailed in Table 2. Among these affective variables, enjoyment is frequently studied and is identified as a major element of AR experiences (Table 2). A plethora of research has indicated that perceived enjoyment or direct enjoyment is often a significant determinant in forming attitudes toward AR, influencing brand attitude and driving purchase or usage intentions (Barhorst et al., 2021; Pantano et al., 2017; Rese et al., 2014, 2017; Smink et al., 2019; Yim et al., 2017; Zanger et al., 2022). However, the concept of enjoyment, as typically applied, is somewhat oversimplified—it is often treated as a catch-all for positive emotions, thereby neglecting the unique influence of other discrete positive emotions.

Nonetheless, Hornung and Smolnik (2022) argue that existing research on emotions in the field of information technology frequently lacks grounding in emotion theories and tends to focus on the basic or discrete emotions exhibited by users. As our study zeroes in on the components of emotions and their elicitation patterns in the context of AR usage, we favor a multicomponential definition of emotions. Drawing on Scherer's (2000) perspective, we define emotions as episodes of coordinated changes in physiological, expressive, and affective components arising from cognitive appraisals of events or thoughts. This definition aligns with our multicomponent model and supports Bagozzi et al.'s (1999) view of emotion's role in marketing, specifically when AR is the stimulus event. We believe this definition provides a comprehensive and nuanced foundation for further investigating the emotional experiences associated with AR in marketing contexts.

To the best of our knowledge, as of the time of writing, six studies have specifically investigated the role of affective responses or emotions in the context of AR in marketing (as detailed in Table 2).

First, Poncin and Ben Mimoun (2014) were among the first to delve into this realm, discovering that an AR magic mirror strategically placed within a store significantly enhanced the store's ambiance, enriching the in-store experience for the customer, which subsequently triggered behavioral intentions. It is noteworthy, however, that their study primarily explored positive emotions related to in-store AR technology, not standalone AR apps. They adapted six positive emotion items from Mehrabian and Russell's seminal work in 1974.

Second, Javornik (2016b) examined affective, cognitive, and behavioral responses as dependent variables. However, the affective responses were mostly geared toward the site or application and the attitude toward the brand, bypassing the analysis of specific emotions or emotional processes. Their study did not primarily focus on emotions or the impact of affective responses on consumer behavior, leaving room for further exploration.

Third, Watson et al. (2018) also studied affective responses, using a variety of positive emotions as their lens. Yet, these emotions were not clearly differentiated, making their study less specific. Their findings revealed that the augmentation of AR influences consumers’ affective states and behavioral intentions. Their work leveraged the Mehrabian and Russell's (1974) environmental psychology model, utilizing an S-O-R paradigm. They adopted seven emotion items on positive affective responses from Chang et al. (2011), which did not divide emotional components, and their study did not investigate the elicitation pattern of a positive affective response.

Fourth, Kowalczuk et al. (2021) study similarly delved into affective, cognitive, and behavioral responses, examining their interaction with AR. However, they equated affective responses to immersion, enjoyment, and product liking. This representation falls short as immersion, measured by absorption, engrossment, and focus, and product liking are not emotions in the conventional sense. Furthermore, the study did not probe into the emotional process itself.

Fifth, Mishra et al. (2021) included emotional appeal as a dependent variable in their research model. However, this construct was only measured by two items (i.e., cheerful and happy), without the inclusion of other emotion components. Furthermore, while their study explored the impact of both AR and VR on consumer responses, emotional appeal was only applied to VR, bypassing AR completely.

Sixth, Zanger et al. (2022) assessed whether AR boosts affective responses, represented by inspiration and enjoyment. While inspiration may not typically be considered an emotion, their findings showed that enjoyment sparked inspiration, which then stimulated positive cognitive responses. Nevertheless, they used enjoyment as a blanket concept to represent a multitude of positive emotions and did not examine the process of emotion elicitation.

Taken collectively, there is a clear need to transition away from the monolithic concept of enjoyment to unravel the complex tapestry of positive emotions. Further research in this field should endeavor to differentiate between the various components of emotions and examine their elicitation processes to fully understand the impact of AR in marketing.

2.4 The measures of emotions

The realm of emotion research offers a plethora of scales designed for the measurement of emotions (Laros & Steenkamp, 2005). Academic discourse in this area reveals a dichotomy of perspectives; some researchers propose the use of unipolar items for emotion measurement while others emphasize the efficacy of bipolar measurements (Bagozzi et al., 1999). Bagozzi (2000) suggests the application of unipolar scales for measuring emotions, reserving bipolar semantic differential scales for the assessment of attitudes. As per the proposition of Perugini and Bagozzi (2001), unipolar emotion items afford respondents the liberty to express varied levels of relevance. Another debate revolves around the choice between dimensional and discrete approaches to best capture and illustrate emotions (Mauss & Robinson, 2009). Despite their inclination toward the dimensional approach, Mauss and Robinson (2009) assert that the absence of a standard emotion measurement protocol aligns with the intricate nature of emotions.

Within the realm of consumer behavior, the long-established frameworks for emotion study are the basic emotions approach and the dimensional theories. The basic emotions approach posits emotions’ classification based on a foundational set of emotions deemed to be “biologically rooted and universally expressed” (Richins, 1997, p. 128). Plutchik (1980) identifies eight primary emotions (i.e., fear, anger, joy, sadness, acceptance, disgust, expectancy, and surprise) that symbolize primary reactions stemming from evolutionary processes (Havlena & Holbrook, 1986). This approach theorizes that all remaining emotions should be cataloged under the umbrella of these basic emotions, and complex emotions are composites of these basic ones (Plutchik, 1980). Izard's (1977) differential emotion theory, which proposes 10 basic emotions, is another tool to scrutinize postpurchase satisfaction. Despite its straightforward application, critics of the basic emotions approach argue that the theory merely assigns labels, neglecting to address the emotional process and its implications (Watson & Spence, 2007).

Dimensional theories, on the other hand, endeavor to categorize emotions based on shared affective dimensions, like valence and arousal, rather than relying on a set of basic emotions (Johnson & Stewart, 2005). Mehrabian and Russell (1974) delineate three independent dimensions for appraising emotional responses: pleasure, arousal, and dominance (PAD). Employing 18 semantic differential items in bipolar dimensions, the PAD scale evaluates emotional reactions induced by marketing stimuli (Richins, 1997). Other researchers such as Holbrook and Batra (1987) and Batra and Holbrook (1990) have employed similar dimensional approaches, with findings pointing to three comparable dimensions of emotions. Moreover, Russell (1980) conceptualized a circumplex model of affect using two bipolar dimensions (i.e., pleasure-displeasure and arousal-sleepiness), encapsulating eight affective concepts. Concurrently, Watson and Tellegen (1985) posited another circumplex model, highlighting a two-factor structure of affect and distinguishing axes as high-positive versus low-positive affect and high-negative versus low-negative affect. Edell and Burke's (1987) seminal model, focused on measuring emotions toward advertising, discovered three interrelated factors: upbeat feeling, negative feeling, and warm feeling. Although dimensional theories adeptly grasp the core dimensions of emotional states, they neglect discrete emotions experienced by consumers and fail to differentiate between emotions with the same valence (Watson & Spence, 2007).

To surmount the limitations of both basic emotion and dimensional approaches, Richins (1997) introduced the consumption emotions set (CES), encompassing 17 emotional clusters through a multidimensional scale. The CES encapsulates a broad spectrum of the most frequently experienced consumption emotions, reflecting an individual's diverse emotional states (Richins, 1997). More recently, Kautish et al. (2021) proposed an online consumption emotion scale, advocating for multi-dimensional models to effectively capture the blend of emotional responses in an online retail context.

Taken collectively, it is clear that the measurement of emotions remains an expansive and nuanced field within psychology and marketing. While various approaches and models have been proposed, each carries its own merits and limitations, reflecting the complexity and diversity inherent in emotional experiences. From basic emotions and dimensional theories to more refined models such as the CES and the online consumption emotion scale, the evolution of these methodologies exemplifies the continuous effort to capture the dynamic nature of emotions. Ultimately, the appropriate measurement tool hinges on the specific research objectives and the context of the study. Thus, it is through this ongoing dialog and iterative refinement that our understanding of emotions can continue to be enriched and deepened.

2.5 Hypotheses development

The pivotal role of emotional responses in the mediating the relationship between the advertising content and the satisfaction of consumers has been well-established by numerous marketing studies (Laros & Steenkamp, 2005). Furthermore, Bezjian-Avery et al. (1998) demonstrated a favorable relationship between the visual attributes of a media format and the evoked emotional responses. Drawing on this, Beaudry and Pinsonneault (2010) proposed that an array of emotions such as excitement, happiness, anxiety, and anger could be experienced either directly or indirectly during the implementation of a novel IT application.

Delving deeper into the context of AR, Kourouthanassis et al. (2015) elucidated a positive relationship between the technological properties of a mobile AR application and the emotional experience derived from its usage. Additionally, Yim et al. (2017) made an intriguing finding that AR, despite often provoking media irritation more than conventional websites, elicited higher media enjoyment among consumers with a less favorable body image. This suggests that AR could potentially invoke a spectrum of emotions, both positive and negative.

Moreover, the recent studies by Zanger et al. (2022) and Watson et al. (2018) further strengthened the argument for AR's capacity to enhance emotional engagement, revealing that product presentations based on AR increase enjoyment and generate stronger positive affective responses, respectively. Similarly, Kowalczuk et al. (2021) confirmed that the usage of AR can incite more positive emotional reactions compared with product presentations on mobile websites. In this context, expressive behaviors hold significant communicative functions and may thus serve as vital components of prototypical emotion (Izard, 1991).

Darwin and Prodger (1998) underscored surprise and astonishment in their enumeration of expressive behaviors linked to positive emotional states. Aligning with this, Javornik et al. (2016) found that surprise often marks consumers’ initial reaction to an AR mirror. Therefore, it is proposed that expressive emotions could constitute the first layer of emotional responses triggered by marketing stimuli.

The study of emotions, grounded in a componential perspective, views these emotional experiences as multidimensional processes. These are manifested across a variety of loosely connected components (Frijda, 1999; Larsen & Fredrickson, 1999; Plutchik, 2001; Scherer, 2000). In his seminal work, Scherer (2000) conceptualizes emotions as a series of coordinated changes across various domains, notably, neurophysiological activation (i.e., the physiological component), motor expression (i.e., the expressive component), and subjective experience (i.e., the affective component).

Primarily, both physiological and expressive components maintain a profound connection with the human physicality. The physiological component is integrally tied to underlying biological processes (Kleinginna & Kleinginna, 1981). Valenstein (1973, as cited in Kleinginna & Kleinginna, 1981) posits that emotional responses often engender robust physiological reactions, linked with both positive and negative states. On the other hand, the expressive component primarily encapsulates those emotions which manifest visibly and are superficially apparent (Kleinginna & Kleinginna, 1981). As elucidated by Darwin and Prodger (1998), certain expressive behaviors correlate with positive emotional states, such as surprise or astonishment, which is universally signified by wide-open eyes and mouth (p. 279).

The affective component, on the other hand, pivots around the realm of subjective feelings (Izard, 2010), offering the most frequently employed definitions of emotions, as embodied in the PAD model of affect. The PAD model emphasizes feelings of arousal and pleasure versus displeasure as key elements of affective emotions. Juslin and Västfjäll (2008) argue that all of these distinct components serve as viable metrics for measuring emotional responses. However, academic discourse persists concerning the synchronization of these components during an emotional response. Consequently, the definition and understanding of emotion may vary depending on the particular components under consideration (Frijda, 1999).

Furthermore, Izard (1991) asserts that certain emotions exhibit a structured order of emergence, underpinned by the intrinsic nature of innate programs. Reinforcing this perspective, O'Shaughnessy and O'Shaughnessy (2003) suggest that the activation of one particular emotion can sequentially trigger a cascade of other emotions within a dynamic system. In terms of this sequence, Moors (2009) underscores that a preceding component can incite changes in subsequent components, even before the former has fully completed its process. Hence, the partial unfolding of one component may instigate transformations in the following components.

Given that expressive emotions, such as surprise, serve as immediate reactions to unexpected stimuli, they might wield a significant influence over other components, including physiological and other emotional forms (Scherer & Fontaine, 2013). This theory is further bolstered by studies illustrating that expressive emotional components are closely associated with other functional responses (Darwin & Prodger, 1998; Keltner et al., 2003).

In light of these discussions, the hypotheses are presented as follows:

Desire, as established by Perugini and Bagozzi (2001), plays an instrumental role in decision-making. This metric essentially gauges whether a consumer is genuinely interested in pursuing an action or merely feels obligated to engage in it. Grounding the crux of decision-making in the belief-desire model, Bagozzi (2007) suggested that the desire to act serves as a crucial intermediary between the driving forces behind action and the intention to perform such an action.

Notably, anticipated emotions have been identified as influential contextual precursors of desire, forging a connection from emotions to desire, and subsequently, intention. In line with the self-regulation theory articulated by Bagozzi (1992), the fulfillment of outcome-desire emerges upon experiencing positive emotions. Such emotions typically result in positive coping responses, such as being inspired to engage in a behavior or exerting effort to sustain it.

Contrastingly, Bagozzi and Pieters (1998) associate anticipatory emotions with the achievement of goals. Likewise, Perugini and Bagozzi (2001) propose that the emotions a consumer experiences beforehand have a direct impact on their desire to engage in goal-oriented behavior. Consequently, we infer that emotions significantly influence a consumer's desire. This inference guides the formulation of the subsequent hypotheses:

The concept of desire, viewed as an individual's intrinsic motivation toward goal attainment, plays a pivotal role in the manifestation of behaviors (Perugini & Bagozzi, 2001). Bagozzi's theory of self-regulation (1992) designates desire as a catalyst for intention, indicating that desire does not just aim at goals but also steers toward goal-directed behaviors or intent to act (Bagozzi, 2000).

In the crux of the decision-making framework of the belief-desire model, it is suggested that action-oriented desire undergoes a transformation into intention within the decision-making process (Bagozzi, 2000). This process elucidates the activation of intentions by desire, cementing the latter as a critical determinant of action (Perugini & Bagozzi, 2001). Perugini and Bagozzi (2004) further demarcate desire from intention by stating that desires, unlike intentions, have a lesser action connection, are conceived in more abstract terms, and span over a longer time horizon. Considering these characteristics, we can use the concept of desire to evaluate an individual's motivation to utilize certain technologies such as AR. Thus, desire becomes a central motivational construct in driving goal-oriented actions (Perugini & Conner, 2000).

Previous research examining the use of AR in marketing has predominantly centered around purchase intention (Beck & Crié, 2018; Hilken et al., 2017; Kowalczuk et al., 2021; Mishra et al., 2021; Watson et al., 2018; Zanger et al., 2022). For instance, Yim et al. (2017) observed a positive relationship between attitude toward AR and intention to purchase. They implied that a positive attitude formed upon using a particular technology that improves the overall shopping experience could enhance purchase decisions. Accordingly, we propose that a heightened desire to use an app could potentially catalyze purchasing behavior.

Our premise is also supported by several dynamics in consumer psychology and marketing. First, frequent app use increases consumers’ exposure to the products promoted by the app, leveraging the psychological principle of the “mere-exposure effect,” wherein consumers’ increased familiarity with products through frequent app use boosts their preference for those products. This psychological effect mediated by mobile app technology hence influences consumers’ purchase intention (Vahdat et al., 2021). Second, the desire to use an AR app may enhance consumers’ perceived value of the overall shopping experience and the brand behind it, resulting in an increased propensity to purchase products from that brand (Beck & Crié, 2018). Third, consumer engagement with technology-mediated marketing such as AI conversational agents has been shown to influence brand purchase positively (McLean et al., 2021). Noteworthily, consumers’ decision to engage (e.g., with an AR app) is often driven by their own needs (Hollebeek et al., 2019). In this regard, a strong desire to use an app can deepen engagement, fostering a sense of commitment and heightening receptiveness toward purchasing in line with the reach-act-convert-engage (RACE) framework (Chaffey & Ellis-Chadwick, 2022). Specifically, a heightened desire to use an AR app derived from personalized experience and engagement behavior offered by AR could stimulate curiosity about the products and sway consumers’ buying decisions (Beck & Crié, 2018). Moreover, the desire to use an app frequently can also expose consumers to social influence through features like user reviews and shared products, which in turn influences their willingness to buy those products (Vahdat et al., 2021).

Nonetheless, it is important to note that while purchase intention reflects the likelihood of future buying behavior (Ajzen, 1991), willingness to purchase indicates a consumer's readiness to commit to a purchase. Poushneh and Vasquez-Parraga (2017) demonstrated that AR positively influences user experience, thereby enhancing their readiness to purchase a specific product. By weaving all these factors together, we propose a theoretical rationale for the connection between the desire to use an app and purchase behavior.

3.1 Study 1

The conceptualization of emotions as a multicomponent construct, as per the research by Larsen and Fredrickson (1999), Plutchik (2001), and Scherer (2000), informs our approach in this study. The emotion scales selected for use are drawn from the CES scale, as proposed by Richins (1997). This choice was informed by the scale's broad application in consumer research, its comprehensive range of categories and emotion items, and its unique ability to differentiate among various forms of emotions with similar valence. Richins (1997) noted that “emotions are context-specific” (p.129), a principle that guided our adoption and adaptation of the CES scale for the specific context of AR. To validate this modified scale, a series of focus groups, expert interviews, pre-tests, and a cross-sectional survey were conducted.

3.1.1 Procedure

The CES scale's eight categories encompassing 22 items of positive emotions and seven categories with 21 items of negative emotions were subjected to focus group analysis to derive modified items suitable for an AR environment. A total of 15 participants, all of whom engage in online shopping at least weekly, took part in one of two focus group sessions, each lasting approximately one hour. The participants were split based on their employment status. Each group consisted of five working adults and two to three university students, respectively.

From these focus group sessions, we replaced lesser-known items with more familiar terms and eliminated two categories from both positive (i.e., romantic love and love) and negative (i.e., envy and loneliness) emotions that seemed incongruent with the experience of using AR. Although there was some debate regarding the removal of the category peacefulness, it was ultimately retained for expert review. The outcome was a revised list comprising 16 positive and 14 negative emotion items.

Subsequently, four experts in marketing and consumer behavior reviewed these scale items to ensure their relevance to the research context and ascertain the content validity of the revised scale. They agreed to remove peacefulness, add items to four categories (i.e., capable and protected in fulfillment, powerful in optimism, bored in discontent, and fearful in fear), and rename two items (depressed to upset, miserable to unpleasant) in sadness to better represent the categories. These category selections align with the multicomponent definition of emotions (Scherer, 2000) adopted in our study.

The final instrument consists of five categories with 17 items for each type of emotion. This instrument was pretested among 70 undergraduate students, identified through convenience sampling, to rectify potential language and cognitive difficulties (Ruel et al., 2016). Minor adjustments were made following this process.

To evaluate the psychometric properties of the revised scale, a cross-sectional survey, recommended by Hinkin (1998), was carried out. We invited undergraduate students from a private university to explore cosmetic products using the Sephora virtual mirror AR app. The participants were given 15 min to familiarize themselves with the system and select their preferred shades of makeup for either eyes or lips by clicking the “Try this shade” button (Figure 1). The app also provided options for users to save their selected “look” or share it with others via a photo (Figure 2). Subsequently, they were asked to complete a survey questionnaire. The data collection was overseen by the principal researcher and a research assistant.

3.2 Study 2

Study 2 was undertaken to scrutinize user emotions toward AR apps, both collectively and on an individual basis predicated on singular valence. The aim was to discern the influence of this variation on the willingness to use AR. We visualized the summarization of Study 2 in Figure 3, in which we administered the model incorporating both positive and negative emotions (Panel A), exclusive positive emotions (Panel B), and solely negative emotions (Panel C).

To gather responses, we implemented a quasi-experiment and subsequently subjected the data to analysis via partial least square structural equation modeling (PLS-SEM), utilizing the SmartPLS 4 software. In our data analysis, we favored the use of variance-based SEM over its covariance-based equivalent. The rationale behind this choice lies in the alignment of PLS-SEM's focus on the expansion and prediction of theory (exploration) with the objectives of our study.

While the roots of consumer emotions research lie in the discipline of psychology, its incorporation into the fields of consumer behavior and technology acceptance to construct a theoretical framework is viewed as exploratory, which necessitates the development of theory. Additionally, our goal extends to predicting the variance elucidated in the critical outcome variable, which in this context is “desire.” These reasons are consistent with the suitability of deploying PLS-SEM.

PLS-SEM also possesses the advantage of not being bounded by normality assumptions, setting it apart from covariance-based SEM. Complementing this, previous research has proposed that variance-based and covariance-based SEM yield similar outcomes (Astrachan et al., 2014). Thus, our selection of PLS-SEM is appropriate and ensures the robustness and validity of our analysis.

3.2.1 Procedure

For this study, two distinct groups participated: the experimental group and the control group. The former engaged with the Lenskart application that allowed users to virtually try on eyewear, whereas the latter browsed Lenskart's mobile website without the benefit of the AR feature. Lenskart, a renowned eyewear retailer based in Delhi, serves as an online portal for eyewear and associated accessories across Asia. Collaborating with Ditto, a US-based start-up, Lenskart began offering a virtual try-on feature in 2016 (Bhalla, 2020).

Participants in the experimental group were instructed to download the Lenskart app onto their smartphones. They were then asked to select the “AR Try On” feature to virtually test eyeglasses or sunglasses that appealed to them. This process required them to take a selfie, which the system would then utilize to identify the frame size suitable for their facial structure (Figure 4).

Conversely, the control group participants were directed to access the Lenskart mobile website via www.lenskart.com on their smartphones. The participants were encouraged to choose any pair of glasses that appealed to them. A “3D Try-on” feature was available on the homepage, where users took a selfie that the system then used to digitally position various frames on the face shown in the product page. Following an average exploration duration of 15 min without interruptions, a questionnaire was provided to both groups via a web link.

3.2.5 Structural model

Upon the successful validation of the measurement model, we employed a bootstrapping procedure, consisting of 5000 iterations, to evaluate the proposed relationships among the constructs. The variance inflation factor (VIF) values ranged from 1.00 to 1.698 for Figure 5 Panel A, 1.00 to 1.55 for Figure 5 Panel B, and 1.00 to 1.309 for Figure 5 Panel C, demonstrating an absence of collinearity concerns, as all values remained well below the conservative threshold of 3.3 (Kock & Lynn, 2012). We also evaluated the R² values for all endogenous variables, comparing them against the benchmark criteria for marketing studies. According to Hair et al. (2016), R² values of 0.75, 0.50, and 0.25 are classified as substantial, moderate, and weak respectively.

The study's results, as illustrated in Panel A, reveal that the application of an AR app significantly triggers both positive and negative expressive emotions. This observation supports H1a and H1b. Subsequently, it was found that while positive expressive emotions evoked both affective and physiological emotions, negative expressive emotions only instigated affective emotions. This evidence corroborates H2a, H2b, and H3a. Two hypotheses—i.e., H3b and H5c—were not explored in the present study due to the insignificant role that negative physiological emotions played as a dimension in the exploratory factor analysis in the previous study.

As for the relationship between the induced positive emotions and consumers’ desire to use AR, both H4a and H4b are validated, with results aligning across both Panels A and B. These findings resonate with Perugini and Bagozzi's (2001) and O'Shaughnessy and O'Shaughnessy's (2003) studies, which concluded that consumers’ emotions significantly influence their pursuit of goal-oriented behavior and stimulate desire. In contrast, the induced negative emotions did not affect the desire to use AR apps, thus refuting H5a and H5b.

These results suggest that the emotional response begins with the expressive component. This concept aligns with the discovery of Javornik et al. (2016) that surprise constitutes consumers' initial reaction to an AR mirror. The results also align with Hinsch et al.'s (2020) research, which observed the “wow effect,” including awe and surprise, generated by the Lego Playground AR app.

However, the present findings diverge from the PAD model of affect developed by Mehrabian and Russell (1974), which posits the simultaneous occurrence of various emotional dimensions. Our research contributes fresh insights to the sequence of emotions, suggesting that certain emotional dimensions, such as arousal (A), might be triggered before the pleasure (P) dimension. This innovative understanding enriches the existing literature by offering a perspective on how one elicited dimension of emotions can stimulate subsequent emotions.

In a collective examination of both positive and negative emotions (Figure 5, Panel A), our study found that positive physiological emotions lead to the desire to use the AR app. However, when positive emotions were analyzed independently (Figure 5, Panel B), this specific emotional component did not influence the desire. Therefore, the effect of H4c remains unverified.

The data from Figure 5 Panel A and Panel C revealed that even though the use of AR apps induces negative expressive emotions, leading to negative affective emotions, these triggered emotions did not consistently affect the desire to use AR apps. This could be explained by Schepers et al.'s (2022) findings that different forms of AI are predominantly associated with positive rather than negative emotions.

To this end, based on the outcomes of Study 2, Study 3 will be conducted with a particular focus on positive emotions (Figure 5, Panel B) to delve deeper into the relationship between positive emotions and behavioral outcomes.

3.3 Study 3

Study 3 was designed as a comprehensive examination of the findings derived from Study 2 with an emphasis on positive emotions (Figure 5, Panel B). The design of this study sought not only to validate our previous findings but also to broaden the scope of the research by introducing four modifications: (1) incorporating a dependent variable, namely, the consumer's willingness to purchase a product; (2) transitioning the focus from a specific product (eyewear) to a brand-centric approach with a well-known furniture brand; (3) broadening the demographic profile by enlisting a nonstudent participant pool; and (4) implementing a tightly controlled lab experiment to replace the previous quasi-experimental design. The SmartPLS 4 software was utilized to analyze all the collected data.

3.3.1 Procedure

Participants allocated to the AR condition experimental group engaged with an application known as Ikea Place, a consumer-friendly AR tool launched by Ikea as a replacement for their former AR feature embedded in the Ikea catalog. The participants downloaded the application and selected the “Try in your place” option, which prompted them to scan their immediate surroundings so the app could comprehend the layout of the space. They were allotted approximately 15 min to familiarize themselves with the various features of the app, including selecting and virtually placing a chosen piece of furniture within their actual environment. The participants also had the flexibility to arrange multiple pieces of furniture and to adjust their positions and orientations within the space (Figure 6).

In contrast, participants in the control group engaged with the mobile website of Ikea (www.ikea.com/my) to select their desired furniture. This group could only view static images of the furniture in isolation or within a pre-set background environment (Figure 7). Upon completion of these tasks, both groups were redirected to an online questionnaire via a shared weblink.

3.3.4 Structural model

Following the approach employed in Study 2, we used a bootstrapping procedure with 5000 samples to scrutinize the proposed relationships among the constructs. The results are presented in Figure 8. The VIF values ranged from 1.00 to 2.06, suggesting the absence of multicollinearity issues (Kock & Lynn, 2012). The findings mirrored those of Figure 5 Panel B in Study 2, with H4c remaining unsupported and further confirming that physiological emotions did not induce a desire to use.

Mirroring the results of Study 2, Study 3 suggested that physiological emotions do not significantly influence the desire to use AR apps, thereby refuting H4c. This finding may be attributable to a common observation in both scenarios, wherein a deficient level of technological embodiment, restricting users’ sense of connectedness with AR, and consequently, inhibiting the desire to engage with it. Flavián et al. (2019) underscored the significance of technological embodiment, asserting that greater embodiment cultivates more immersive experiences by engendering a feeling of physical proximity to the technology. This perspective aligns with the insights presented by Hilken et al. (2018), who stressed the importance of embodiment arising from simulated physical control such as rotating or manipulating the products. They argue that this tangible self-directed interaction constitutes a crucial component of the embodied customer experience that ultimately dictates the success of an AR app. This notion is further substantiated by the study of Hilken et al. (2017), which indicated that customer value perceptions are amplified by facilitating concurrent environmental embedding and simulated physical control.

While environmental embedding (e.g., projected visualization of eyewear) was present in Study 2, the simulated physical control might have been restricted compared with the case of augmentation for self, such as applying make-up. Users could visualize the glasses from multiple perspectives, but they might not have been able to cultivate a tangible understanding of the product. Study 3 provided environmental embedding (e.g., projecting a visualization of a home décor item within a space) but lacked simulated physical control, as it imposed limitations on performing natural movements to accurately adjust the home décor item. Here, environmental embedding was not complemented with simulated physical control, resulting in a deficient overall situated experience. This could potentially influence the motivation to use the AR app, due to perceived physiological limitations. As suggested by Hilken et al. (2017), empowering customers with a sense of physical control is essential to effectively sculpt their online shopping experience.

Noteworthily, we found that desire to use the AR app is positively correlated with the willingness to purchase the product showcased via the AR app. This relationship validates the function of desire in evoking motivational commitment and identifies desire as being immediately influential to actual decision-making (Bagozzi, 2007). Our findings suggest that desire plays a crucial role as a motivator in decision-making processes, thereby affirming previous conclusions that the utilization of AR heightens desire to use an app which in turn increases the intention and willingness to purchase a product (Beck & Crié, 2018). Our results also support the findings in Vahdat et al. (2021), which found that attitude to mobile app use affects the intention to purchase. Additionally, our results shed light on the positive impact of AR on user experience, which consequently influences their willingness to engage in a purchase (Poushneh & Vasquez-Parraga, 2017).

4.1 Theoretical implications

The theoretical implications of this research are manifold, focusing on the enhancement of our understanding of emotions and their impact on consumer behavior in the context of technology use, specifically AR.

First, this examination ventures into new territory by proposing the application of a multicomponent model of emotions within the context of AR, drawing inspiration from Scherer's model (Scherer, 2000). We furnish compelling evidence to support the proposition that the utilization of AR prompts a variety of emotional components. Intriguingly, these include some (like physiological responses) that have often been overlooked in AR-focused marketing studies. We thereby underscore the importance of recognizing and incorporating this complex emotional spectrum when investigating marketing stimuli that influence emotions. Our findings from Study 2 reaffirm this, demonstrating that consumers’ interaction with AR apps elicits a blend of both positive and negative emotions. Building on this emotional exploration, Study 3 provides confirmation that positive emotions directly influence the desire to use AR, which in turn affects the willingness to purchase.

Second, our investigation differentiates between the various emotions associated with the use of AR, disaggregating them to better understand their functions and mechanisms in relation to technology-based products, like AR apps. Existing AR marketing literature often limits itself to single-dimensional affective variables such as enjoyment or entertainment. By contrast, we broaden this perspective by integrating a multitude of emotional categories. This novel approach aids in explaining how different emotional components can motivate the use of AR. Given that positive emotions are known to stimulate creative and innovative actions (Fredrickson, 2004), our identification of various positive emotions linked to AR use can be instrumental in facilitating desired consumer behaviors.

Third, our research makes a noteworthy contribution by mapping the sequence of emotion elicitation tied to AR. While many studies have explored the emergence of emotions or the effects of certain emotions on behavior, there is a lack of clarity regarding the interactive patterns of emotions influencing decision-making processes. Our findings suggest that the expressive component is the initial trigger, preceding other emotional components such as affective and physiological emotions. Given that these expressive and affective components stimulate consumers’ desire to use AR and potentially make a purchase, they may also incite other favorable behaviors, such as increased consumer-brand engagement and word-of-mouth recommendation.

Finally, our study fills an existing gap by integrating emotional antecedents with motivational and behavioral outcomes driven by AR. Notably, prior literature has overlooked the desire to use AR, focusing more on attitudes or intentions. Our research argues that desire, with its distinct characteristics that set it apart from attitudes (Carrus et al., 2008), demands greater scholarly attention. Previous consumer behavior studies have identified positive anticipated emotions as drivers of desire (Bagozzi & Dholakia, 2002; Perugini & Conner, 2000). However, our study adds a new layer by showing that the actual elicitation of emotions, not just anticipated ones, influences desire. By concentrating on real-time induction of emotions, as opposed to projected emotions analyzed in the model of goal-directed behavior (Perugini & Bagozzi, 2001), we capture genuine emotional responses and their impact on behavior. In particular, our study suggests that while AR triggers a mix of positive and negative emotions, only positive emotions influence the desire to use AR. By illuminating these intricate emotional processes, we offer a more comprehensive understanding of consumers’ interactions with AR.

4.2 Practical implications

The practical implications of this study are multi-dimensional and can provide valuable insights for developers and marketers aiming to maximize the potential of AR.

First, this examination furnishes developers and marketers with a more nuanced understanding of the consumer experience associated with AR. The findings could serve as a guidepost for the design and launch of AR apps that deliver substantial value to both consumers and companies over the long term. Acknowledging the process and pattern of interactive emotions connected with AR could enlighten developers and marketers that the elicitation of positive expressive emotions is a vital consideration when designing an AR app.

Second, this exploration suggests that marketers should focus on crafting AR apps that elicit wonder and surprise in consumers. This strategy is likely to trigger other positive emotions and reinforce the desire to use AR. At a product level, such as for Lenskart, newly developed or still-under-design products can be updated in the app for consumers to experiment with virtually. At a brand level, like Ikea, brand-related stimuli can be created and regularly updated to ensure the elicitation of expressive emotions. As indicated by Hinsch et al. (2020), such up-to-date mental accommodations in consumers’ minds could form the foundation for the “wow effect.” Consequently, marketers can leverage this unique AR experience to captivate consumers’ interest in the brand's products and services. Enhancing consumer engagement in this manner can incentivize them to spend more time and money on the app. Pragmatically, marketers can include a call-to-action button for the most frequently virtually tried-on products to boost conversion rates. Given these emotional pathways, there is a significant opportunity for marketers to generate leads and enhance sales.

Third, this investigation highlights the importance of prioritizing the user experience and user interface in AR projects, as advocated by Cognizant (2019). To improve the simulated control of AR as part of the user experience, developers should strive to enhance the app's precision and the natural movements of the virtual objects. As suggested by Flavián et al. (2019), brands using AR apps should assess the degree of technological embodiment that can be integrated into their customers’ experiences.

Fourth, this research underscores the need to minimize negative emotions associated with AR use. While negative emotions do not impact the desire to use AR, reducing them is beneficial. As AR holds the potential to revolutionize the shopping process, developers and marketers should regularly gather user feedback on their experience with the existing features of the brand's AR app. Such feedback can guide improvements in areas of simulated physical control or embodiment, helping to suppress negative emotions while amplifying positive ones, which could lead to a stronger motivation to use the app for purchasing.

Lastly, a report by Boston Consulting Group (BCG) (Bona et al., 2018) revealed that only a small percentage of marketing practitioners from the top 200 advertisers in the US considered AR as a core element of their marketing portfolio. Many are still at the experimental stage of using AR, focusing primarily on brand awareness and consumer attitudes. They often overlook AR's potential to enhance downstream performance, such as purchase intention and sales. The present study, drawing on data from two companies’ first-party AR apps, offers fresh insights on how marketers can foster personalized positive engagement with targeted consumers. Our study reaffirms a progression from mere experimentation with AR to the attainment of bottom-of-funnel objectives like purchase willingness, as implied in the BCG report (Bona et al., 2018).

4.3 Limitations and future research directions

While this study offers significant contributions, it comes with several limitations that provide fertile ground for future research.

First, it is important to note that our study examined only two types of first-party mobile AR apps (i.e., apps owned by the company). Consequently, care should be taken when extrapolating the results to other AR apps, particularly third-party AR apps such as Snapchat. Future research could therefore validate and extend the proposed model to include these third-party AR apps or explore other advanced technologies like VR.

Second, our focus was primarily on three common components of positive emotions. It will be beneficial for future studies to incorporate additional components and further validate the measurement items related to positive emotions.

Third, the relationship between physiological emotions and the desire to use AR yielded inconsistent results across our studies. As such, further investigation into the relationship between the physiological component of emotions, desire, and behavioral outcome is warranted.

Fourth, regarding our samples, the first two studies utilized a convenient sample drawn from a university population. Future research should aim to overcome this limitation by using a random sampling method and expanding the sample to include other populations, thereby enhancing the generalizability of the results. Peterson's (2001) suggestion to replicate results obtained from university students with an adult population is particularly relevant here.

Fifth, our investigation did not fully explore the influence of specific product-related factors such as utilitarian value, hedonic value, and fit with the consumer's needs. These factors, pivotal in shaping consumer purchase decisions, were beyond the purview of our current investigation. The utilitarian value of a product (i.e., its practicality and functionality), the hedonic value (i.e., the pleasure or enjoyment derived from it), and the extent to which a product fits a consumer's needs are all critical dimensions that could influence purchasing behavior. Future research could delve into these areas and assess how they interact with the desire to use an app in shaping purchasing behavior.

Sixth, our study did not fully investigate the role of potential mediators between the desire to use an app and purchasing behavior. It is plausible that multiple mediating factors, such as consumer engagement and experience, play a significant role in this relationship. For instance, heightened engagement with an app may lead to increased familiarity with and preference for the products offered. Similarly, a positive user experience can boost satisfaction levels and potentially influence the decision to make a purchase. Future research could undertake a more nuanced examination of these and other potential mediators to provide a comprehensive understanding of the relationship dynamics at play.

Finally, despite our attempts to balance internal and external validity through the use of both experimental and quasi-experimental settings in our studies, future research could provide more robust results by employing field experiments to test the hypotheses in a natural context. This could help in producing more externally valid and generalizable results.