Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
Consumer Sentiment and Hotel Aspect Preferences Across Trip Modes and Purposes
Previous Article in Journal
Research on the Matching Effect of Social Media Advertising Appeal and Narrative Person: Evidence from China
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
JTAER
Volume 19
Issue 4
10.3390/jtaer19040144
jtaer-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Signaling Effects in AI Streamers: Optimal Separation Strategy Under Different Market Conditions

by
Ying Yu
1 and
Yunpeng Yang
2,*
1
Faculty of Humanities and Social Sciences, Macao Polytechnic University, Macao 999078, China
2
Antai College of Economics and Management, Shanghai Jiao Tong University, Shanghai 200030, China
*
Author to whom correspondence should be addressed.
J. Theor. Appl. Electron. Commer. Res. 2024, 19(4), 2997-3016; https://doi.org/10.3390/jtaer19040144
Submission received: 4 August 2024 / Revised: 29 October 2024 / Accepted: 31 October 2024 / Published: 3 November 2024
Browse Figures
Figure 1
<p>Timing of model.</p> "> Figure 2
<p>The high-quality firm’s profit variation with the fraction of uninformed consumers.</p> "> Figure 3
<p>The high-quality firm’s profit variation with the fraction of uninformed consumers under different marginal costs for acceptance improvement.</p> ">
Versions Notes

Abstract

:
The fusion of livestreaming e-commerce and AI technology is booming, and many firms have started to replace human streamers with AI streamers. Despite their popularity, the acceptance of AI streamers by consumers varies widely and the signaling effects of AI streamers still remain unclear. We build an analytical model and compare scenarios where the acceptance level is either exogenously given or endogenously determined, highlighting the implications for firms’ optimal separation strategy. Our findings suggest that in markets with moderate information asymmetry, using both price and acceptance level as joint signals can be more profitable for high-quality firms. Conversely, in highly asymmetric markets, firms must incur additional costs to distinguish their high-quality products, regardless of the signaling strategy employed. Our paper provides strategic insights for firms aiming to leverage AI streamers in diverse market conditions.

1. Introduction

The burgeoning advancements in artificial intelligence (AI) and augmented reality over recent years have catalyzed a far-reaching transformation across a multitude of business domains [1,2]. Within the dynamic sphere of livestreaming e-commerce [3], a novel approach has emerged [4,5,6], leveraging AI technology to replace traditional human streamers [7,8]. This innovative model, referred to as AI-oriented livestreaming, is fast becoming a prevailing trend and has been extensively experimented with by various enterprises [9]. This burgeoning paradigm shift is exemplified by leading entities such as the esteemed sports brand Nike and the chic British fashion brand Boohoo, who have, respectively, created their own AI-driven streamers [10]. For example, the sports brand Nike has integrated AI streamers to facilitate interactive shopping experiences, allowing consumers to explore products in virtual settings personalized to their preferences and needs. These AI streamers support a 24/7 online presence, enabling global customer access at any time. Similarly, the British fashion brand Boohoo has implemented AI-powered streamers to showcase diverse styling options, addressing varying consumer tastes and promoting their products with a degree of consistency and visual appeal. These applications have allowed both brands to maintain continuous customer interaction, reducing reliance on human resources and potentially increasing consumer engagement and brand loyalty. The advent of these AI streamers has permitted these companies to maintain round-the-clock, uninterrupted livestreaming, at a significantly reduced cost [11,12,13,14]. This strategic move not only enhances the overall shopping experience for consumers by making it more engaging and immersive but also mitigates potential risks associated with the personal conduct or controversies of human streamers [15].
However, the adoption of AI-oriented livestreaming e-commerce is not without its challenges [5,16]. As AI streamers gain popularity, user adoption and acceptance become critical factors in their success. Despite the numerous benefits AI streamers offer, users may be hesitant to embrace this new technology due to various reasons [17]. The adoption of AI-oriented livestreaming e-commerce can lead to functional failures, such as inaccurate demand forecasting, unsmooth transaction processes, and inadequate after-sales service [1,18,19]. These issues can result from the inability of AI streamers to provide personalized services like human streamers, causing consumers to have a more negative attitude towards them. For instance, consumers might encounter invalid responses, wrong information, or a lack of empathy from AI streamers, as seen in the case of Sony’s virtual streamer, Sony-N, which led to customer complaints and service failures [20]. Sony’s experience with its AI streamer, Sony-N, illustrates some of the potential limitations. Sony-N encountered several issues, including difficulties in responding empathetically to customer queries, occasional misinformation, and a limited capacity for personalized interaction. These challenges led to a decline in customer satisfaction and generated complaints about service quality, affecting Sony’s brand image. This example highlights that while AI streamers offer scalability and cost-efficiency, their lack of adaptive, human-like qualities can lead to consumer hesitancy and potentially negative perceptions. These contrasting examples underscore both the opportunities and risks associated with AI streamers in livestreaming e-commerce, reinforcing the need to understand the factors that drive their acceptance among consumers.
Considering both the benefits and limitations of AI streamers, it is natural to question whether it is advantageous for firms to adopt AI streamers over traditional ones. In the ever-evolving landscape of AI-driven media [21,22,23], the question of streamers’ ability to convey quality information remains pertinent [24]. With the influx of AI streamers into the market, it becomes imperative to assess whether these digital entities can effectively perform the same role as traditional information disseminators [4,8,19,25]. Traditional livestreamers often play a crucial role in signaling product quality [26,27]. In the context of AI streamers, can quality information about products be effectively conveyed? What are the effects of consumers’ acceptance level for AI streamers? What are firms’ optimal strategies to separate themselves from low-quality service? These questions still remain unknown and our paper tries to unveil the mechanisms behind them. By examining the signaling effects of AI streamers, this paper seeks to provide strategic insights for firms looking to optimize their livestreaming strategies and maximize the value of AI technology in diverse market conditions.
The rest of our paper is organized as follows. Section 2 provides a detailed review of the related literature and our potential contributions to the literature. Section 3 describes the basic setup in our analytical model. Section 4 elaborates the equilibrium in our model. Section 5 further extends our model to the case of endogenous acceptance levels. Section 6 concludes our paper with the main findings, managerial implications, limitations, and further directions.

2. Literature Review

Livestreaming e-commerce has emerged as a dynamic and transformative area of research, integrating entertainment, social interaction, and commercial transactions in real-time digital environments [28,29,30]. This literature review examines key themes within livestreaming e-commerce, the evolving impact of AI streamers, and the strategic implications of signaling effects in marketing.

2.1. Livestreaming E-Commerce: Consumer Behavior and Market Dynamics

Recently, scholarly research on livestreaming e-commerce has burgeoned, encompassing multidisciplinary inquiries into its theoretical underpinnings, practical implications, and regulatory challenges [31,32]. Livestreaming e-commerce platforms have revolutionized traditional retail paradigms by offering interactive and immersive shopping experiences directly through digital channels [33,34]. Academic investigations have delved into various aspects of live commerce, including consumer behavior analysis, brand marketing dynamics [35], platform ecosystem studies, and regulatory policy evaluations [36,37,38]. This section synthesizes research exploring the factors influencing consumer behavior within livestreaming contexts, the role of livestreaming influencers, and regulatory considerations shaping this burgeoning industry.
Consumers increasingly turn to livestreaming influencers or streamers for guidance and product recommendations, drawn by their perceived authenticity, expertise, and ability to create engaging content that blends entertainment with commercial transactions [39,40,41,42]. Research underscores the significant impact of social interactions and peer recommendations within livestreaming environments, shaping consumer perceptions and purchase intentions [43,44]. The interactive nature of livestreaming platforms allows consumers to participate actively in product demonstrations, Q&A sessions, and live interactions with streamers, fostering a sense of community and immediacy that traditional retail experiences often lack [45,46].
Regulatory frameworks also play a crucial role in shaping the livestreaming e-commerce landscape, addressing issues such as consumer protection, advertising standards, and platform governance [36,37,38]. Scholars have examined the implications of these regulations on market entry barriers, competition dynamics, and the operational practices of livestreaming platforms and influencers alike. Understanding these regulatory frameworks is essential for stakeholders navigating the complexities of a rapidly evolving digital marketplace where innovation often outpaces existing regulatory frameworks.

2.2. AI Streamers in Livestreaming E-Commerce: Opportunities and Challenges

The integration of AI streamers represents a transformative development within livestreaming e-commerce, promising to enhance user engagement, operational efficiency, and content personalization [47]. This section reviews research on the benefits of AI streamers, challenges related to consumer acceptance, and the implications for emotional engagement and trust-building in digital interactions.
Some studies support the benefits of adopting AI streamers. Incorporating virtual humans can set the platform apart from competitors, showcasing its technological capabilities and innovative approach. AI streamers do not experience fatigue, illness, or schedule conflicts, allowing for consistent and reliable streaming schedules [48]. AI streamers can also be customized in appearance and behavior to fit various themes, narratives, and audience preferences, offering endless creative possibilities [49]. As for the global reach and operation costs, AI streamers can easily be localized for different languages and cultures, reducing long-term costs associated with hiring, training, and managing human streamers [50].
On the other hand, however, many scholars also argue the potential risks and limitation when adopting AI streamers. One challenge is audience acceptance. While younger audiences may embrace virtual humans [4,8,16,51], older or more traditional viewers might be less receptive. They may find this new form of streaming unfamiliar or unappealing [9]. Further, the emotional connection in AI streamers seem to be weak. AI streamers often lack the emotional intelligence to connect with consumers on a personal level. This can make interactions feel impersonal and transactional, which diminishes the overall user experience [20]. Building the same level of emotional connection and trust that human streamers naturally create can be challenging for virtual characters [17]. While virtual humans can express emotions, viewers may find it harder to connect emotionally compared to real humans, lowering emotional resonance. The emotional expressions of virtual characters may appear less natural or lack depth [15]. In livestreaming e-commerce, the quality of interaction is crucial. Establishing trust and authenticity in AI-driven interactions remains a critical area of concern, as these virtual entities may struggle to replicate the nuanced social cues and empathetic responses characteristic of human interactions. Human streamers can engage with the audience through spontaneous and dynamic conversations, while AI streamers often follow pre-programmed scripts, limiting their ability to adapt to real-time audience feedback [52].
As a result, the acceptance of AI streamers by consumers varies widely. Many consumers may be skeptical or uncomfortable with AI-driven interactions, especially if they perceive the technology as intrusive or lacking authenticity. This can lead to reduced engagement and lower viewer retention rates [24]. From a strategic standpoint, navigating the integration of AI streamers requires a careful consideration of consumer preferences, technological capabilities, and the ethical implications surrounding data privacy and algorithmic transparency. The development of AI streamers represents a paradigm shift in how digital content creators engage with audiences, posing both opportunities for innovation and challenges in maintaining meaningful user connections.

2.3. Signaling Effects in Livestreaming E-Commerce: Mechanisms and Implications

Signaling effects are pivotal in shaping consumer perceptions, behaviors, and purchase decisions within livestreaming e-commerce platforms [51,52,53,54,55,56,57]. This section examines how pricing strategies, advertising dynamics, and word-of-mouth interactions function as signals, shaping brand credibility, consumer trust, and strategic positioning in digital marketplaces.
The literature has extensively discussed the transmission mechanisms of signaling effects, including but not limited to the following pathways: (1) price: Both low prices and high prices can exert signaling effects [53]. Platform entry may enhance the signaling effect of prices [54]. (2) Advertising: Early research has focused more on the signaling effect of advertising costs [55]. As advertising forms evolved, scholars found that advertisements could also convey information by promoting search behavior [56], and even skippable ads can have a signaling effect [57]. (3) Warranty: high-quality firms can differentiate themselves by offering quality assurance contracts or related derivative contracts, such as shipping insurance [58]. (4) Word-of-mouth: with the development of information technology, the signaling effect of word-of-mouth has garnered considerable attention from scholars [59]. (5) Multidimensional combinations: examples include the combination of price and advertising [60], the combination of word-of-mouth and advertising [61], and the bundling of new and old products [62].
Effective signaling strategies enable brands to convey quality, value, and reliability to consumers within competitive digital marketplaces [52,53]. Transparent pricing policies and promotional strategies not only attract consumer attention but also bolster brand integrity and foster long-term relationships in livestreaming e-commerce environments [54,55]. Furthermore, leveraging word-of-mouth interactions and influencer endorsements enhances signaling effects, harnessing the power of social networks to bolster brand visibility and credibility [56,57]. Understanding the mechanisms behind signaling effects is crucial for marketers aiming to optimize promotional strategies and capitalize on consumer preferences for authenticity, value, and social endorsement within livestreaming e-commerce platforms. As digital communication technologies continue to evolve, these strategies offer brands new opportunities to engage with consumers through targeted messaging and interactive content experiences, thereby influencing purchasing decisions and cultivating brand loyalty in dynamic digital ecosystems [63].
The above literature provides us helpful insights in understanding livestreaming e-commerce and the information role of AI streamers [29,30]. However, previous studies do not fully explore the mechanisms of livestreamers in information transmission, especially in terms of product quality information. Specifically, AI streamers also largely differ from traditional human streamers, and consumers may have varying acceptance levels to this new form of livestreamers. How quality information is transmitted in AI streamers still remains unknown. Our paper integrates the characteristics of AI streamers into the signaling game, highlighting the effects of consumers’ acceptance level towards AI streamers and firms’ optimal strategies. Our model can shed light on the informative role of AI streamers and provide valuable insights for both firms and consumers, equipping them with actionable knowledge to navigate the complexities of livestreaming commerce effectively in the AI era.

2.4. The Factors Influencing the Adoption and Effectiveness of AI Streamers

The rise of AI streamers in livestreaming e-commerce has prompted extensive interest, as these digital presenters offer a unique blend of cost-efficiency, operational flexibility, and scalability compared to human streamers. This literature review explores the adoption of AI streamers across various organizations and analyzes the specific features that contribute to their success or hinder their acceptance. Key success and failure factors are identified, focusing on interaction quality, adaptability, trust, and emotional resonance—factors that collectively influence consumer acceptance and engagement.
One core feature impacting AI streamer success is interaction quality, which encompasses both the technical performance and perceived responsiveness of the AI streamer. The literature indicates that effective AI streamers deliver clear, accurate information and enable smooth, immersive interactions, which can enhance consumer trust and facilitate informed purchase decisions. For example, Khan et al. (2023) [64] highlight how technical attributes, such as seamless information delivery and quick response times, reinforce consumer satisfaction and trust. Additionally, Nike’s AI streamer, which incorporates augmented reality for product visualization, demonstrates the importance of providing users with real-time interactive capabilities that enhance product appeal and offer a personalized experience.
Adaptability also emerges as a pivotal factor in AI streamer success. Studies emphasize that consumers respond more positively to AI streamers that can tailor their recommendations based on real-time user data [3]. Boohoo’s AI streamers, for example, leverage data-driven algorithms to offer styling suggestions aligned with individual consumer preferences. This adaptability has proven effective in meeting the dynamic and diverse needs of consumers, which is critical for fostering long-term engagement [24]. Conversely, a lack of adaptability, as seen in Sony-N’s AI streamer, limited the AI’s ability to adjust responses to nuanced consumer inquiries, resulting in customer frustration and diminished trust in the platform [65].
Trust and emotional resonance also significantly affect AI streamer acceptance, as consumers seek a sense of connection and assurance in their shopping experiences. Research by Zhang et al. (2023) [52] underscores that AI streamers must establish credibility and reliability to gain consumer confidence, especially when replacing human streamers who typically bring personal charisma and empathy to interactions. While AI technology can reduce the risk of human error, such as product misrepresentation or behavioral inconsistencies, the absence of authentic empathy and emotional intelligence often deters consumers from fully engaging with AI streamers. The failure of Sony-N, as discussed in several studies [24], highlights how the lack of genuine emotional engagement—evident in situations where the AI provided inaccurate or emotionally tone-deaf responses—can lead to consumer disengagement.
Lastly, the consumer perception of novelty and functionality in AI-driven experiences is a critical dimension discussed in the literature. As noted by Jim and Lim (2021) [65], initial consumer interest in AI streamers may arise from the novelty of the technology; however, sustained engagement often relies on the AI’s ability to enhance functional aspects of the shopping experience, such as convenience, product discovery, and efficient service [7]. The positive reception of AI streamers in companies like Nike and Boohoo points to the value of functional enhancement in driving consumer acceptance. Yet, research also indicates that when AI systems fall short of operational efficiency, novelty alone is insufficient to retain consumers, as observed in instances where poor performance led to negative attitudes and reluctance to adopt AI-based services [66].
In summary, the literature suggests that AI streamer success largely depends on factors that directly affect consumer experience and trust. Firms aiming to implement AI streamers may benefit from prioritizing these factors to mitigate challenges and leverage the unique advantages AI streamers offer. Our study builds on these findings by exploring how AI streamers can optimally signal quality under different levels of market information asymmetry, providing strategic insights for firms seeking to maximize profitability and consumer acceptance. Unlike previous studies that primarily emphasize the operational or experiential aspects of AI streamers, our study introduces an analytical model that explores the firm’s optimal separation strategy, providing a new theoretical perspective on market signaling in this domain. This approach not only deepens the understanding of AI streamers as quality signals in competitive environments but also offers actionable insights for firms seeking to maximize profitability through strategic adjustments in signaling based on market conditions, thereby advancing both theoretical and practical discussions on AI adoption in livestreaming e-commerce.

3. Methodology

3.1. Aims, Hypothesis, and Constructs

Our primary aim is to analyze how firms can use pricing and acceptance levels as joint signals to convey product quality effectively. For this, we investigate the optimal signaling strategy for firms employing AI streamers to convey product quality effectively under different market conditions. Our research questions center on how firms can use both pricing and acceptance levels as signals in a livestreaming e-commerce context, specifically addressing the following questions: How can high-quality firms effectively signal their quality using AI streamers, particularly under different levels of market information asymmetry? How does the endogenous determination of AI streamer acceptance level affect consumer perception and the effectiveness of signaling strategies?
Our model relates to both the Signaling Theory and Technology Acceptance Model (TAM). Originating from Spence (1973) [67], Signaling Theory explains how entities in markets with asymmetric information communicate unobservable qualities through observable signals. In our study, pricing and acceptance level serve as dual signals, where acceptance level acts as an additional, consumer-facing signal of quality that firms can influence. This aligns with Signaling Theory by enabling high-quality firms to convey unobservable quality through observable, interpretable indicators (price and acceptance) in a competitive market environment. The constructs of acceptance level and interaction quality draw from the TAM, where perceived usefulness and ease of use are key to understanding user acceptance [68,69]. Acceptance level captures how AI streamer interaction is perceived by consumers, acting as a proxy for ease of use and willingness to engage. Interaction quality is crucial in signaling value and reliability in AI-mediated settings, addressing both perceived usefulness and emotional connection, as outlined by the TAM. Grounded in Signaling Theory and the TAM, we hypothesize the following:
Hypothesis 1:
In moderately asymmetric markets, combining price and acceptance level as signals enhances profitability for high-quality firms compared to using price alone.
The combination of price and acceptance level draws from Signaling Theory, which suggests that using multiple signals together can effectively communicate quality in markets where information is moderately unclear. Here, price conveys a traditional signal of high quality, while acceptance level, reflecting consumers’ readiness to interact with AI, serves as a second, reinforcing signal. This idea is also supported by the TAM, which states that people adopt new technologies based on perceived usefulness and ease of use, both of which can be inferred from high acceptance levels.
Hypothesis 2:
In highly asymmetric markets, high-quality firms incur substantial costs to maintain a separate quality signal, regardless of the chosen signaling strategy.
When consumers know very little about product quality, firms with high-quality products need to invest heavily in signaling efforts to stand out. In these cases, no matter what method they choose (price, acceptance level, or both), firms face high costs to communicate their quality effectively because the audience requires more persuasion. Signaling Theory posits that in highly asymmetric markets, where consumers lack knowledge or trust in the technology, firms must spend more to ensure that their signals are both visible and credible. Here, signaling quality becomes challenging, as firms need to overcome significant skepticism or unfamiliarity. Even with a combined signal of price and acceptance level, high-quality firms often face high costs in establishing the necessary consumer confidence and market acceptance.
We chose our constructs based on insights from Signaling Theory and the TAM, selecting variables that directly impact AI streamer efficacy in signaling product quality. These include the following: (1) acceptance level: representing consumer readiness to interact with AI streamers, it is a key determinant in both Signaling Theory and the TAM, as it influences perceived ease of use and trust. (2) Pricing: chosen as a primary quality signal, it is aligned with Signaling Theory’s proposition that price can serve as an indicator of quality in uncertain markets. (3) Interaction quality: This factor stems from the consumer behavior literature, as interactions critically shape consumer engagement and are essential for establishing the reliability of AI streamers as quality signals. Through these constructs, we seek to elucidate the conditions under which firms can effectively leverage AI streamers to optimize profitability and maintain a competitive edge, drawing on theoretical frameworks that emphasize the role of consumer acceptance and trust in AI-driven commerce.

3.2. The Model

Our model considers a firm i that has private information about its product quality q i , where i ∈ L , H such that q H > q L . Consumers may have different levels of knowledge about the product quality q i since consumers vary in their ability to accurately assess the product quality. For instance, a consumer who has bought a similar product can have a greater awareness of the quality. By contrast, a consumer who has never bought a similar product might have no knowledge about the product quality. The former is a typical case of an “informed consumer”, in contrast to the “uniformed consumers” in the latter case. For simplicity, we consider two kinds of consumers in the market, i.e., the informed and the uninformed. Suppose the proportion of uninformed consumers is λ , where we assume 0 < λ < 1 . These consumers have no information about the product quality. The rest of the consumers are informed, representing a fraction of 1 − λ . Informed consumers can precisely tell the product quality. The size of λ shows the degree of information asymmetry in the market. When λ increases, it means the level of information asymmetry intensifies.
Define q ^ i as the consumers’ belief about the product quality. The informed consumers can precisely discern product quality, regardless of the signals they observe, i.e., q ^ i = q i . The uninformed consumers, however, cannot exactly tell the quality. They update their beliefs about the product quality based on the signals they observe. Assume that the uninformed consumers hold a prior belief that the probability of high quality is β 0 and the probability is 1 − β 0 for low quality. Under this prior belief, the consumers’ initial expected quality of the product is q ¯ = β 0 q H + ( 1 − β 0 ) q L . The uninformed consumers update their prior belief β 0 to the posterior belief β ( p ) (abbreviated as β ) based on the price p of the product set by the firm. In our benchmark model, we focus on the case where only the price signals and we will further discuss other possible signals in the model extension. Thus, under the posterior belief, the expected product quality becomes q ^ i = β q H + ( 1 − β ) q L .
Although AI streamers can be customized in appearance and behavior to fit various themes, narratives, and audience preferences, offering endless creative possibilities, some older or more traditional consumers might be less receptive, potentially limiting the audience [24]. They may find this new form of streaming unfamiliar or unappealing, limiting the potential audience for virtual human livestreams. Moreover, building the same level of emotional connection and trust that human streamers naturally create can be challenging for virtual characters [15]. While virtual streamers can express emotions, consumers may find it harder to connect emotionally compared to real humans, lowering emotional resonance. The emotional expressions of virtual characters may appear less natural or lack depth [20,36]. Based on these concerns, we incorporate the parameter a to describe the consumer acceptance of AI streamers, with a > 0 . Necessarily, when a becomes larger, consumers are more willing to accept AI streamers and consumers can also obtain more utility. The utility of the consumers can be expressed as follows:
u i = a q ^ i + t − p i
where the parameter t denotes the consumers’ preference. A higher value of t means the consumers prefer the product more and they are more likely to buy the product. For simplicity, we assume the parameter t follows a uniform distribution in [ 0 , 1 ] . Parameter p i represents the pricing set by a firm i . When the utility function satisfies u i > 0 , the consumers will purchase the product; otherwise, the consumers will not buy it.
Define N i as the number of paid consumers in firm i , and combining this with the consumer utility function, we can obtain the number of paid consumers:
N i = 1 − p i a + q ^ i
And the profit of the firm can be expressed as
Π i = p i 1 − p i a + q ^ i
We summarize the timing of the model in Figure 1. At the beginning of the model, nature determines the firm’s quality in terms of its products and the degree of information asymmetry in the market. After that, the firm strategically decides the settings of the AI streamers and prices for the products. In the benchmark case, the acceptance level of AI streamers is exogenously given; in the model extension, the firm can endogenously determine the acceptance level. Consumers can update their beliefs about the products based on the signals they observe. Our model compares two different scenarios for signals: single signal (only price serves as the signal) or joint signal (both price and acceptance level serve as the signal). Finally, consumers make purchase decisions based on their posterior beliefs. The key notations of the model are provided in Table 1.

4. Equilibrium Analysis

We start the equilibrium analysis from a benchmark case where there is no information asymmetry. After that, we analyze the case of information asymmetry and the separating equilibrium in this case.

4.1. The Benchmark Case: Information Symmetry

Under information symmetry, all the consumers have exact knowledge of the information regarding product quality, i.e., q ^ i = q i . In this case, both the high-quality and low-quality firms make decision based on their true product quality. In this benchmark case, we discuss the subgame perfect Nash equilibrium, where consumers make their purchase decisions according to the accurate information about the product quality. We can obtain the equilibrium results in this benchmark case as follows:
Proposition 1.
In the case of information symmetry, it is optimal for the firm  i  to set p i = 1 2 a + a q i  with maximized profits of  Π i * = 1 4 a + 2 a q i + a q i 2 , where i ∈ { L , H } .
From Proposition 1, we can know that the high-quality firm sets a price higher than the low-quality firm, because of its advantage in product quality, i.e., q H > q L . As a result, the high-quality firm can also obtain higher profits than the low-quality firm. This result is intuitive and also corresponds to previous studies [70]. The consumers’ acceptance level for AI streamers, however, does not affect the firm’s strategy and profits. This is because in this benchmark case, we do not incorporate the differences in acceptance level between the high-quality and the low-quality firms. We will discuss the above differences in our model extension.

4.2. The Case of Information Asymmetry

In the case of information asymmetry, there are two possible choices for the low-quality firm: either set its price based on its actual quality level or mimic the pricing level as the high-quality firm does. In this section, we only consider the case of a single signal, where only the price ( p ) serves as the signal. As a comparison, the section on model extension will further discuss the case of the joint signal, where both the price and acceptance level ( p , a ) jointly serve as the signal.
To avoid being regarded as a low-quality firm by consumers, the high-quality firm has to provide a reasonable price to separate itself from the low-quality firm. Both the informed and uninformed consumers will perceive the product as high quality, provided that the high-quality firm performs this separation strategy. In this case, all the consumers will buy the product at a relatively higher price. Nevertheless, if the low-quality firm mimics the high-quality firm by setting p L = p H , the uninformed consumers ( λ ) are likely to regard the low-quality firm as the high-quality firm, i.e., q ^ L = q H . In contrast, the informed consumers ( 1 − λ ) can still tell the truth, i.e., q ^ L = q L , and they will not purchase the products. Let Π L M denote the profits that the low-quality firm can obtain by mimicking the high-quality firm, and then we have
Π L M = λ p H 1 − p H a + q H
Hence, we can find that the low-quality firm is faced with a decrease in purchase demand when it tends to mimic the high-quality firm. This mechanism ensures the possibility of separating the equilibrium, where the low-quality firm has no incentives for imitation, allowing the high-quality firm to signal product quality through pricing to differentiate from the low-quality firm. In the separating equilibrium, the constraint for incentive compatibility has to be satisfied:
Π L M ≤ Π L ∗
where Π L ∗ denotes the low-quality firm’s maximized profit when it does not mimic the high-quality firm. Note that in our model, the high-quality firm has no incentives to mimic the low-quality firm, since the high-quality firm cannot obtain higher profits through imitation. And the constraint for individual rationality requires both types of firms to be profitable to engage in the market. Combining these constraints, we can obtain the following proposition:
Proposition 2.
In the case of information asymmetry, the separating equilibrium is as follows:
When 0 < λ < 1 + q L 1 + q H 2 , it is optimal for the firm i  to set p i ∗ = a 2 1 + q i , with maximized profits of  Π i * = a 4 1 + q i 2 , where i ∈ { L , H } .
When 1 + q L 1 + q H 2 < λ < 1 , it is optimal for the high-quality firm to set p H ∗ = a 2 1 + q H − λ ( 1 + q H ) 2 − ( 1 + q L ) 2 λ  and p L ∗ = a 2 1 + q L , with maximized profits of  Π H * = a 1 + q L 2 4 λ  and Π L * = a 4 1 + q L 2 .
Proposition 2 indicates that the low-quality firm does not imitate the high-quality firm in the case of a low λ , i.e., where there are fewer uninformed consumers in the market. In this case, the market information asymmetry is low. The high-quality firm can simply set the price in its preferred scenario (similar to the case of information symmetry). In this range, the high-quality firm can without cost separate itself from the low-quality firm.
When the fraction of uninformed consumers grows larger, however, it becomes profitable for the low-quality firm to mimic the high-quality firm, such that the high-quality firm cannot implement a costless separation strategy. As a result, the high-quality firm has to set a lower price than that of the symmetric information scenario, incurring extra costs for separating itself form the low-quality firm, i.e., a costly separation strategy.
Figure 2 illustrates the equilibrium in Proposition 2. In this numerical example, we take a = 1 , q H = 1 , and q L = 0.5 . We can find that when the degree of information asymmetry ( λ ) is at a relatively low level, the high-quality firm can achieve a costless separation. When λ continues to increase and becomes larger than a certain threshold, the high-quality firm can only accomplish a costly separation, where an increase of λ can incur a lower level of profits.

5. Model Extension: Endogenous Acceptance Level

This section extends our model to the case where the firm can affect the consumers’ acceptance of AI streamers, i.e., the parameter a can be endogenously determined by the firm. We use parameter k i to capture the marginal costs of improving the consumers’ acceptance, with k i > 0 for i ∈ { L , H } . We assume k L > k H because the high-quality firm tends to have a technical advantage [71]. For technical convenience, we assume that
k H > k L 1 + q H 2 1 + q H 2 − q H − q L 2 + q H + q L 1 + q H 2 + 1 + q L 2 1 + q L 4
This assumption ensures the possibility for the low-quality firm to mimic the acceptance level set by the high-quality firm. Otherwise, when k H becomes extremely small compared to k L , it can be quite expensive and disadvantageous for the low-quality firm to imitate, which is not our focus. Based on the above setup, the profit function of the firm can be expressed as the following:
Π i ( p i , a i ) = p i 1 − p i a i + q ^ i − k i a i 2
Referring to the setup of cost structures in previous studies [72], we take the quadratic form of the cost for improving the acceptance level because the marginal costs tend to be higher when a i stays at a relatively high level. Based on the above setup, we can draw the following proposition:
Proposition 3.
When the acceptance level  a  is endogenously determined, the separating equilibrium is as follows:
When 0 < λ ≤ k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 , it is optimal for the firm i  to set p i ∗ = ( 1 + q i ) 3 16 k i ,  a i ∗ = ( 1 + q i ) 2 8 k i , with maximized profits of  Π i * = ( 1 + q i ) 4 64 k i , where i ∈ { L , H } .
When k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 < λ < 1 , it is optimal for the high-quality firm to set
p H ∗ = λ 1 + q H 2 − λ 2 1 + q H 4 − 1 + q L 4 1 + q H 16 k L ,
a H ∗ = λ 1 + q H 2 − λ 2 1 + q H 4 − 1 + q L 4 8 k L ,
with maximized profits of
Π H ∗ = 1 + q H 4 64 k L 2 λ − λ 2 − 1 + q L 1 + q H 4 2 k L − k H λ − λ 2 − 1 + q L 1 + q H 4 .
And the low-quality firm sets p L ∗ = ( 1 + q L ) 3 16 k L , a L ∗ = ( 1 + q L ) 2 8 k L , with maximized profits of Π L * = ( 1 + q L ) 4 64 k L .
From Proposition 3, we can find that the firm can simultaneously adjust both the pricing and acceptance level to reflect its quality type and achieve separation. In the case of fewer uninformed consumers (low value of λ ), the optimal pricing and acceptance level are the same as the equilibrium in the case of symmetric information. Hence, the high-quality firm can achieve costless separation. When the number of uninformed consumers increases (high value of λ ), however, the high-quality firm has to pay more efforts to distinguish itself from the low-quality firm, incurring costly separation.
Figure 3 shows the high-quality firm’s profit variation with the fraction of uninformed consumers under the joint signal of price and acceptance level, which is a visual explanation of Proposition 3. In Figure 3, the blue line denotes the profit of the high-quality firm in the separating equilibrium. By comparing subgraphs Figure 3a–d, we can find that an increase in k H will incur the occurrence of a costly separation. In other words, when the high-quality firm is equipped with higher marginal costs in improving acceptance level, the high-quality firm has less advantages to distinguish itself from the low-quality firm, inducing the costly separation.
Based on the results in Proposition 3, we can also draw the following corollary:
Corollary 1. 
When k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 < λ < 1 , there are the following properties: (1) ∂ a H ∗ ∂ λ < 0 , ∂ p H ∗ ∂ λ < 0 , ∂ Π H ∗ ∂ λ < 0 ; (2) ∂ a H ∗ ∂ k L < 0 , , ∂ p H ∗ ∂ k L < 0 , ∂ Π H ∗ ∂ k L < 0 .
Corollary 1 demonstrates the impacts of different parameters on the equilibrium acceptance level, pricing, and profits. Corresponding to the previous study [73], Corollary 1 shows that the information asymmetry can cause negative effects on the profits of the high-quality firm. As the number of uninformed consumers increase, i.e., the information asymmetry becomes more serious, the high-quality firm has to provide a lower level of the acceptance level to enable consumers to tell it apart from the low-quality firm, and its equilibrium price is lower, inducing a decrease in the high-quality firm’s profits.
Further, by summarizing the above results and compare with the equilibrium in the case of a single signal, we can obtain the following proposition:
Proposition 4.
By comparing the single price signal  ( p )  with the joint signal of price and acceptance level ( p , a ) , we can draw the following conclusions:
When 0 < λ ≤ 1 + q L 1 + q H 2 , both the single signal and the joint signal can be costless.
When 1 + q L 1 + q H 2 < λ ≤ k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 , it can be costly to use the single signal but costless to use the joint signal.
When k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 < λ < 1 , both the single signal and the joint signal can be costly.
From Proposition 4, we can find that when there are fewer uninformed consumers in the market, regardless of the single-signal or the joint-signal strategy, the high-quality firm can always achieve costless separation. In this case, it is unprofitable for the low-quality firm to imitate, as otherwise it would lose large proportion of informed consumers. Hence, the high-quality firm does not have to make efforts in improving acceptance level to separate itself from the low-quality firm. To achieve separation, the high-quality firm only needs to set a proper pricing level, as in the case of information symmetry. Our result corresponds to a previous study [53], which shows that having a large fraction of informed consumers is sufficient for the price to signal quality information.
The case differs when the number of uninformed consumers increases to a moderate level. It can be beneficial for the high-quality firm to use the joint signal of price and acceptance level to achieve separation. Through the joint signal, the high-quality firm can earn higher profits than the single signal in the separating equilibrium. To achieve separation, it can be costly to use the single signal but costless to use the joint signal. However, when the number of uninformed consumers continues to increase to a higher level, it will be costly to achieve separation, regardless of using a single or joint signal.

6. Conclusions

The asymmetry of quality information in livestreaming e-commerce significantly constrains product trading, where firms utilize pricing and the assistance of livestreamers to signal quality. This paper investigates the signaling effects of AI streamers in livestreaming e-commerce, considering their dual role as a novel form of livestreaming with inherent benefits and limitations. Since AI streamers are a new form of livestreaming with both benefits and limitations, we specifically analyze consumers’ attitudes towards AI streamers and how their acceptance level affects the equilibrium. Our model explores the firm’s optimal separation strategy and compares two possible scenarios, namely a single-signal and joint-signal scenario. In the former, the acceptance level is exogenously given and only the product price serves as the signal, while in the latter, the acceptance level is endogenously determined by the firm and both the price and acceptance level jointly serve as the signal. By comparing these two cases, our paper demonstrates the firm’s optimal signaling strategy in different market environments.
Our study yields several key findings: Firstly, under conditions of asymmetric quality information, firms can differentiate quality through varied pricing and acceptance levels. In a separating equilibrium, prices and profits for low-quality firms align with those under information symmetry, while those for high-quality firms depend on market asymmetry. Secondly, in moderately asymmetric markets, using a joint signal (price and acceptance level) can be cost-effective compared to a single-signal strategy. High-quality firms benefit from signaling quality through both price and acceptance levels, enhancing profitability. Thirdly, under high degrees of information asymmetry, high-quality firms face substantial costs in separating themselves from low-quality competitors, irrespective of signaling strategy. This asymmetry consistently impacts high-quality firms’ profitability, exacerbated by higher marginal costs to improve acceptance levels.
To support these findings, we identify several practical strategies firms can implement to increase AI streamer acceptance, thereby enhancing profitability: (1) Improve interactive capabilities. Firms can increase consumer engagement with AI streamers by incorporating advanced recommendation systems and emotion recognition capabilities. Real-time personalized recommendations based on consumer interactions, coupled with sentiment analysis, enable AI streamers to adapt their responses according to user emotions, fostering a more responsive and engaging experience. (2) Enhance transparency and consistency. Consumer acceptance can also be boosted by transparently communicating the AI streamer’s capabilities and limitations. Setting realistic expectations for the AI’s role, particularly in terms of empathy and knowledge depth, can reduce consumer frustration. Additionally, firms can maintain up-to-date responses by regularly refreshing training data to enhance response consistency, further building trust. (3) Establish a continuous user feedback loop. Implementing real-time feedback mechanisms allows consumers to rate or comment on their experiences with AI streamers. By capturing these data, firms can refine the AI’s performance, addressing consumer preferences dynamically. Furthermore, A/B testing different interactive features allows firms to continuously optimize the AI’s engagement style. (4) Employ a hybrid approach with human interaction options. In complex or high-value transactions, AI streamers can seamlessly transfer interactions to human agents, ensuring consumer needs are met when issues exceed AI capabilities. For product launches or peak sales events, a hybrid approach enables AI streamers to handle repetitive queries while human agents focus on VIP clients, optimizing customer service resources and satisfaction. (5) Emphasize the unique advantages of AI streamers. Marketing the AI streamer’s 24/7 availability and efficiency can appeal to consumers who value immediate support. Additionally, AI streamers offer a highly scalable platform for delivering data-driven personalization, which is less feasible with human streamers. Highlighting these advantages can position AI streamers as a superior option for certain consumer segments.
Through the analysis above, we can also draw several implications for practice. When a firm with a high-quality product tends to distinguish itself from the low-quality firm, its strategy in pricing and AI streamers should be based on the degree of information asymmetry. If the degree of market asymmetry is relatively low, the firm can simply achieve the separation through the price strategy, without any extra costs; when the degree of asymmetry reaches a moderately high level, it can be a wise choice for the firm to signal quality information through the joint signal, namely the combination of the pricing and acceptance level. When the degree of asymmetry reaches a very high level, however, it is inevitable for the firm to pay extra costs to achieve the separation. In conclusion, understanding these dynamics enables firms to strategically leverage AI streamers in livestreaming e-commerce, adapting signaling strategies to varying market conditions to optimize profitability and maintain competitive advantage.

7. Limitations and Future Research

There are some limitations in this paper. First, the study’s reliance on theoretical modeling may limit its practical applicability, as real-world markets are often more complex and involve additional variables that may impact consumer behavior. While the hypotheses presented are grounded in established theories, they may not fully capture the nuances of consumer interactions with AI streamers across diverse contexts and cultural backgrounds. Future research could benefit from empirical testing of these hypotheses to assess their applicability in various market environments. Second, this research focuses on a limited set of variables—price, acceptance level, and signaling strategy—without accounting for other potentially influential factors such as branding strength, consumer personality traits, or situational elements like economic conditions. By expanding the scope of analysis to include these additional factors, future studies could offer a more comprehensive understanding of the factors that contribute to the success or failure of AI streamers in e-commerce. Third, the model assumes relatively static market conditions and does not account for dynamic shifts in technology, consumer preferences, or regulatory changes, which are increasingly relevant in the rapidly evolving field of AI-driven commerce. AI technology is advancing quickly, and the way consumers perceive and interact with it is likely to change in response to innovations and societal adaptations.
In future directions, research can deepen our understanding of AI streamer effectiveness and offer empirical validation of our theoretical model through varied research designs. First, quantitative research would offer a rigorous approach to empirically test the model, particularly through survey-based and experimental methods. Surveys targeting consumer attitudes and perceptions towards AI streamers could explore relationships among critical variables, such as profit margins, willingness to buy, purchase intentions, price perception, situational factors, and actual purchase decisions. For instance, examining how consumer attitudes influence the perceived value of products promoted by AI streamers could reveal correlations between acceptance levels and the willingness to pay a premium, further validating the joint-signaling model. Second, qualitative research approaches, such as in-depth interviews and focus groups, could capture nuanced perceptions of AI streamers, shedding light on their effects on trust, brand loyalty, and purchase intentions. Such studies would also allow firms to understand the emotional factors that impact consumer openness to AI-promoted products, providing complementary insights to quantitative findings. This qualitative angle is particularly useful for uncovering perceptions that may drive long-term consumer engagement or hesitancy, informing strategies to increase acceptance. Third, future research can also focus on an experimental study, with a control group as a viable design to assess causality and model validation. By comparing groups exposed to AI and human streamers under similar conditions, firms can directly observe the impact of AI presence on factors such as trust, perceived quality, and purchase intentions. This approach enables the controlled testing of the signaling effects and provides a basis for evaluating the tangible benefits of AI streamers. Fourth, longitudinal research would also add valuable insights by tracking changes in consumer attitudes and purchase behaviors over time. As consumers become more familiar with AI streamers, their acceptance may evolve, potentially affecting the long-term profitability of firms using this strategy. Observing this progression across different demographic groups would provide a robust understanding of acceptance trends and their business implications. By combining these research methods, future studies can more comprehensively assess the practical impact of AI streamers and substantiate our findings across diverse market contexts.

Author Contributions

Conceptualization, Y.Y. (Ying Yu) and Y.Y. (Yunpeng Yang); methodology, Y.Y. (Ying Yu); software, Y.Y. (Ying Yu); validation, Y.Y. (Ying Yu) and Y.Y. (Yunpeng Yang); formal analysis, Y.Y. (Ying Yu); investigation, Y.Y. (Ying Yu) and Y.Y. (Yunpeng Yang); resources, Y.Y. (Ying Yu); data curation, Y.Y. (Yunpeng Yang); writing—original draft preparation, Y.Y. (Ying Yu) and Y.Y. (Yunpeng Yang); writing—review and editing, Y.Y. (Ying Yu) and Y.Y. (Yunpeng Yang); visualization, Y.Y. (Ying Yu); supervision, Y.Y. (Yunpeng Yang); project administration, Y.Y. (Yunpeng Yang); funding acquisition, Y.Y. (Ying Yu) and Y.Y. (Yunpeng Yang). All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the China Postdoctoral Science Foundation (2024M750571), the Major Projects on Philosophy and Social Science Research of the Ministry of Education of the People’s Republic of China (20JZD010), the National Natural Science Foundation of China (72472101, 72031006), the Humanities and Social Science Fund of the Ministry of Education of China (24YJC630187), the Soft Science Research Project of Shanghai (24692108000), the Humanities Young Talent Cultivation Program at Shanghai Jiao Tong University (2023QN004), and the Startup Fund for Young Faculty at SJTU (SFYF at SJTU). We gratefully acknowledge the above financial support.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are contained within the article (refer to Appendix A). The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author/s.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A

Proof of Proposition 1: 
Under information symmetry, by calculating the first-order-condition (FOC) of the firm’s profit function with respect to the pricing, we have
∂ Π i ∂ p i = 1 − 2 p i a + q i .
By solving ∂ Π i ∂ p i = 0 , we can obtain the optimal pricing
p i ∗ = 1 2 a + a q i ,
for any i ∈ { L , H } . Correspondingly, the optimal profits of the firm i are
Π i ∗ = 1 4 a + 2 a q i + a q i 2 .
□
Proof of Proposition 2: 
Under information asymmetry, if the high-quality firm wants to distinguish itself from the low-quality firm, the high-quality firm has to set the pricing such that the low-quality firm is unprofitable to imitate. This constraint for incentive compatibility requires the following inequality to be true:
p H λ 1 − p H a + q H − 1 4 a + 2 a q L + a q L 2 ≤ 0 .
Combining with the firm’s profit function, we can find the following results:
When 0 < λ < 1 + q L 1 + q H 2 , it is optimal for the firm i to set p i ∗ = a 2 1 + q i with maximized profits of Π i * = a 4 1 + q i 2 , where i ∈ { L , H } .
When 1 + q L 1 + q H 2 < λ < 1 , it is optimal for the high-quality firm to set p H ∗ = a 2 1 + q H − λ ( 1 + q H ) 2 − ( 1 + q L ) 2 λ and p L ∗ = a 2 1 + q L , with maximized profits of Π H * = a 1 + q L 2 4 λ and Π L * = a 4 1 + q L 2 . □
Proof of Proposition 3: 
When the acceptance level a is endogenously determined, if the high-quality firm wants to separate itself from the low-quality firm, the high-quality firm needs to simultaneously set the pricing and acceptance levels to serve as the joint signal. To ensure the low-quality firm is unprofitable to imitate, the constraint for incentive compatibility requires the following inequality to holdheld, i.e.,
p H λ 1 − p H a H + q H − k L a H 2 − 1 + 4 q L + 6 q L 2 + 4 q L 3 + q L 4 64 k L ≤ 0
Combining with the firm’s profit function, we can find the following results:
When 0 < λ ≤ k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 , it is optimal for the firm i to set p i ∗ = ( 1 + q i ) 3 16 k i , a i ∗ = ( 1 + q i ) 2 8 k i , with maximized profits of Π i * = ( 1 + q i ) 4 64 k i , where i ∈ { L , H } .
When k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 < λ < 1 , it is optimal for the high-quality firm to set
p H ∗ = λ 1 + q H 2 − λ 2 1 + q H 4 − 1 + q L 4 1 + q H 16 k L ,
a H ∗ = λ 1 + q H 2 − λ 2 1 + q H 4 − 1 + q L 4 8 k L .
Correspondingly, the maximized profits of the firm are given by the following:
Π H ∗ = 1 + q H 4 64 k L 2 λ − λ 2 − 1 + q L 1 + q H 4 2 k L − k H λ − λ 2 − 1 + q L 1 + q H 4 .
In this case, it is optimal for the low-quality firm to set p L ∗ = ( 1 + q L ) 3 16 k L , a L ∗ = ( 1 + q L ) 2 8 k L , and the low-quality firm can earn the maximized profits of Π L * = ( 1 + q L ) 4 64 k L . □
Proof of Corollary 1: 
From Proposition 3, in the case of k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 < λ < 1 , we have
∂ a H ∗ ∂ λ = 1 + q H 2 − 1 + q H 4 λ − 1 + q L 4 + 1 + q H 4 λ 2 8 k L < 0
∂ p H ∗ ∂ λ = 1 + q H 2 1 + q H 2 − 1 + q H 4 λ − 1 + q L 4 + 1 + q H 4 λ 2 8 k L < 0
∂ Π H ∗ ∂ λ = 1 + q H 2 32 k L 2 1 + q H 4 λ 2 − 1 + q L 4 λ ( 1 + q H ) 2 − 1 + q H 4 λ 2 − 1 + q L 4 k L ( 1 + q H ) 2 − k H λ ( 1 + q H ) 2 + k H 1 + q H 4 λ 2 − 1 + q L 4 < 0
And
∂ a H ∗ ∂ k L = − 1 + q H 2 λ − 1 + q H 4 λ 2 − 1 + q L 4 8 k L 2 < 0
∂ p H ∗ ∂ k L = − 1 + q H 2 λ − 1 + q H 4 λ 2 − 1 + q L 4 8 k L 2 1 + q H 2 < 0
∂ Π H ∗ ∂ k L = − λ − 2 q H λ − q H 2 λ + − 1 + q L 4 + 1 + q H 4 λ 2 32 k L 3 k L + 2 k L q H + k L q H 2 − k H λ − 2 k H q H λ − k H q H 2 λ + k H − 1 + q L 4 + 1 + q H 4 λ 2 < 0
 □
Proof of Proposition 4: 
From Proposition 2, we can infer that when λ > 1 + q L 1 + q H 2 , it can be costly for the high-quality firm to separate itself from the low-quality firm, while costless in the case of λ < 1 + q L 1 + q H 2 . From Proposition 3, we can infer that when λ > k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 , it can be costly for the high-quality firm to separate itself from the low-quality firm, while costless in the case of λ <   k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 . Note that
    k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 − 1 + q L 1 + q H 2 = k L 2 + k H 2 1 + q L 1 + q H 4 2 k H k L − 1 + q L 1 + q H 2   > k L 2 + k H 2 2 k H k L 1 + q L 1 + q H 2 − 1 + q L 1 + q H 2   ≥ 0 ,
where the first inequality is equivalent to k L 2 + k H 2 1 + q L 1 + q H 4 > k L 2 + k H 2 1 + q L 1 + q H 2 , rewritten as k L 2 1 − 1 + q L 1 + q H 2 ≥ 1 + q L 1 + q H 2 k H 2 1 − 1 + q L 1 + q H 2 , which is true due to k L k H > 1 > 1 + q L 1 + q H ; the second inequality origins from k L 2 + k H 2 ≥ 2 k L k H .
Hence, we can summarize the following results:
When 0 < λ ≤ 1 + q L 1 + q H 2 , both the single signal and the joint signal can be costless.
When 1 + q L 1 + q H 2 < λ ≤ k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 , it can be costly to use the single signal while costless to use the joint signal.
When k L 2 1 + q H 4 + k H 2 1 + q L 4 2 k H k L 1 + q H 4 < λ < 1 , both the single signal and the joint signal can be costly. □

References

  1. Zhou, C.; Li, H.; Zhang, L.; Ren, Y. Optimal Recommendation Strategies for AI-Powered E-Commerce Platforms: A Study of Duopoly Manufacturers and Market Competition. J. Theor. Appl. Electron. Commer. Res. 2023, 18, 1086–1106. [Google Scholar] [CrossRef]
  2. Light, B.; Johari, R.; Weintraub, G. Quality Selection in Two-Sided Markets: A Constrained Price Discrimination Approach. Oper. Res. 2024, 72, 1928–1957. [Google Scholar] [CrossRef]
  3. Chen, N.; Yang, Y. The Role of Influencers in Live Streaming E-Commerce: Influencer Trust, Attachment, and Consumer Purchase Intention. J. Theor. Appl. Electron. Commer. Res. 2023, 18, 1601–1618. [Google Scholar] [CrossRef]
  4. Yao, R.; Qi, G.; Wu, Z.; Sun, H.; Sheng, D. Digital Human Calls You Dear: How Do Customers Respond to Virtual Streamers’ Social-Oriented Language in e-Commerce Livestreaming? A Stereotyping Perspective. J. Retail. Consum. Serv. 2024, 79, 103872. [Google Scholar] [CrossRef]
  5. Gao, J.; Zhao, X.; Zhai, M.; Zhang, D.; Li, G. AI or Human? The Effect of Streamer Types on Consumer Purchase Intention in Live Streaming. Int. J. Hum. –Comput. Interact. 2024, 40, 1–13. [Google Scholar] [CrossRef]
  6. Yan, H.; Chen, Z.; Yang, Y. State-Owned Capital and Quality of Green Innovation: Evidence from Chinese Listed Private Firms. Heliyon. 2024, 10, e28179. [Google Scholar] [CrossRef]
  7. Saleem, I.; Hoque, S.M.S.; Tashfeen, R.; Weller, M. The Interplay of AI Adoption, IoT Edge, and Adaptive Resilience to Explain Digital Innovation: Evidence from German Family-Owned SMEs. J. Theor. Appl. Electron. Commer. Res. 2023, 18, 1419–1430. [Google Scholar] [CrossRef]
  8. Lu, B.; Li, G.; Ge, J. Effects of Streamer Effort and Popularity on Livestream Retailing Performance: A Mixed-Method Study. Electron. Commer. Res. 2023, 23, 1–31. [Google Scholar] [CrossRef]
  9. Bharadwaj, N.; Ballings, M.; Naik, P.A.; Moore, M.; Arat, M.M. A New Livestream Retail Analytics Framework to Assess the Sales Impact of Emotional Displays. J. Mark. 2022, 86, 27–47. [Google Scholar] [CrossRef]
  10. Cao, P.; Liu, S. The Impact of Artificial Intelligence Technology Stimuli on Sustainable Consumption Behavior: Evidence from Ant Forest Users in China. Behav. Sci. 2023, 13, 604. [Google Scholar] [CrossRef]
  11. Yang, Y.; Chen, H.; Liang, H. Did New Retail Enhance Enterprise Competition during the COVID-19 Pandemic? An Empirical Analysis of Operating Efficiency. J. Theor. Appl. Electron. Commer. Res. 2023, 18, 352–371. [Google Scholar] [CrossRef]
  12. Nan, G.; Ding, N.; Li, G.; Li, Z.; Li, D. Two-Tier Regulation Models for the User-Generated Content Platform: A Game Theoretic Analysis. Decis. Support Syst. 2023, 175, 114034. [Google Scholar] [CrossRef]
  13. Morath, F.; Münster, J. Online Shopping and Platform Design with Ex Ante Registration Requirements. Manag. Sci. 2018, 64, 360–380. [Google Scholar] [CrossRef]
  14. Zhang, Z.; Chen, Z.; Wan, M.; Zhang, Z. Dynamic Quality Management of Live Streaming E-Commerce Supply Chain Considering Streamer Type. Comput. Ind. Eng. 2023, 182, 109357. [Google Scholar] [CrossRef]
  15. Xie, J.; Wu, H.; Liu, K.; Cui, Y.; Zhang, X. Is Virtual Streamer Useful? Effect of Streamer Type on Consumer Brand Forgiveness When Streamers Make Inappropriate Remarks. J. Retail. Consum. Serv. 2024, 79, 103899. [Google Scholar] [CrossRef]
  16. Zhang, X.; Shi, Y.; Li, T.; Guan, Y.; Cui, X. How Do Virtual AI Streamers Influence Viewers’ Livestream Shopping Behavior? The Effects of Persuasive Factors and the Mediating Role of Arousal. Inf. Syst. Front. 2023, 25, 1–32. [Google Scholar] [CrossRef]
  17. Hu, H.; Ma, F. Human-like Bots Are Not Humans: The Weakness of Sensory Language for Virtual Streamers in Livestream Commerce. J. Retail. Consum. Serv. 2023, 75, 103541. [Google Scholar] [CrossRef]
  18. Cheng, C.; Luo, J.; Zhu, C.; Zhang, S. Artificial Intelligence and the Skill Premium: A Numerical Analysis of Theoretical Models. Technol. Forecast. Soc. Change 2024, 200, 123140. [Google Scholar] [CrossRef]
  19. Fan, Q.; Dai, Y.; Wen, X. Is Smarter Better? A Moral Judgment Perspective on Consumer Attitudes about Different Types of AI Services. J. Theor. Appl. Electron. Commer. Res. 2024, 19, 1637–1659. [Google Scholar] [CrossRef]
  20. Xu, Y.; Ruan, Y. AI and Human Broadcasters: Relative Impact on Consumer Engagement in Live Streaming Commerce. Electron. Commer. Res. Appl. 2023, 62, 101335. [Google Scholar] [CrossRef]
  21. Amaldoss, W.; Du, J.; Shin, W. Media Platforms’ Content Provision Strategies and Sources of Profits. Mark. Sci. 2021, 40, 527–547. [Google Scholar] [CrossRef]
  22. Xu, S.; Mou, Y.; Ding, Z. The More Open, the Better? Research on the Influence of Subject Diversity on Trust of Tourism Platforms. Mark. Intell. Plan. 2023, 41, 1213–1235. [Google Scholar] [CrossRef]
  23. Jeon, D.-S.; Kim, B.-C.; Menicucci, D. Second-Degree Price Discrimination by a Two-Sided Monopoly Platform. Am. Econ. J. Microecon. 2022, 14, 322–369. [Google Scholar] [CrossRef]
  24. Peng, Y.; Wang, Y.; Li, J.; Yang, Q. Impact of AI-Oriented Live-Streaming E-Commerce Service Failures on Consumer Disengagement—Empirical Evidence from China. J. Theor. Appl. Electron. Commer. Res. 2024, 19, 1580–1598. [Google Scholar] [CrossRef]
  25. Jiao, Y.; Sarigöllü, E.; Lou, L.; Huang, B. How Streamers Foster Consumer Stickiness in Live Streaming Sales. J. Theor. Appl. Electron. Commer. Res. 2023, 18, 1196–1216. [Google Scholar] [CrossRef]
  26. Lu, B.; Chen, Z. Live Streaming Commerce and Consumers’ Purchase Intention: An Uncertainty Reduction Perspective. Inf. Manag. 2021, 58, 103509. [Google Scholar] [CrossRef]
  27. Yu, Y.; Liu, Y. Fly or Fall with the Wings of Live Streamers? Quality Signaling and Platform Governance in Live Streaming Commerce. Manag. Decis. Econ. 2024, 45, 1547–1556. [Google Scholar] [CrossRef]
  28. Safadi, H.; Watson, R.T. Knowledge Monopolies and the Innovation Divide: A Governance Perspective. Inf. Organ. 2023, 33, 100466. [Google Scholar] [CrossRef]
  29. Song, S.; Xu, Y.; Ma, B.; Zong, X. Understanding How Streamer’s Self-Presentation in E-Commerce Live Streaming Affects Consumers: The Role of Persuasion Knowledge. J. Theor. Appl. Electron. Commer. Res. 2024, 19, 1922–1942. [Google Scholar] [CrossRef]
  30. Zhao, X.; Yang, S. Does Intelligence Improve the Efficiency of Technological Innovation? J. Knowl. Econ. 2023, 14, 3671–3695. [Google Scholar] [CrossRef]
  31. Sun, Y.; Shao, X.; Li, X.; Guo, Y.; Nie, K. How Live Streaming Influences Purchase Intentions in Social Commerce: An IT Affordance Perspective. Electron. Commer. Res. Appl. 2019, 37, 100886. [Google Scholar] [CrossRef]
  32. Zhu, L.; Liu, N. Game Theoretic Analysis of Logistics Service Coordination in a Live-Streaming e-Commerce System. Electron. Commer. Res. 2023, 23, 1049–1087. [Google Scholar] [CrossRef]
  33. Chen, N.; Yang, Y. The Impact of Customer Experience on Consumer Purchase Intention in Cross-Border E-Commerce——Taking Network Structural Embeddedness as Mediator Variable. J. Retail. Consum. Serv. 2021, 59, 102344. [Google Scholar] [CrossRef]
  34. Yang, Y.; Chen, N.; Chen, H. The Digital Platform, Enterprise Digital Transformation, and Enterprise Performance of Cross-Border E-Commerce—From the Perspective of Digital Transformation and Data Elements. J. Theor. Appl. Electron. Commer. Res. 2023, 18, 777–794. [Google Scholar] [CrossRef]
  35. Huang, Z.; Yan, X.; Deng, J. How Social Presence Influences Consumer Well-Being in Live Video Commerce: The Mediating Role of Shopping Enjoyment and the Moderating Role of Familiarity. J. Theor. Appl. Electron. Commer. Res. 2024, 19, 725–742. [Google Scholar] [CrossRef]
  36. Xu, Q.; Shao, Z.; Zhang, L.; He, Y. Optimal Livestream Selling Strategy with Buy-Online-and-Return-in-Store. Electron. Commer. Res. Appl. 2023, 61, 101307. [Google Scholar] [CrossRef]
  37. Gardete, P.M.; Guo, L. Prepurchase Information Acquisition and Credible Advertising. Manag. Sci. 2021, 67, 1696–1717. [Google Scholar] [CrossRef]
  38. Park, H.J.; Lin, L.M. The Effects of Match-Ups on the Consumer Attitudes toward Internet Celebrities and Their Live Streaming Contents in the Context of Product Endorsement. J. Retail. Consum. Serv. 2020, 52, 101934. [Google Scholar] [CrossRef]
  39. Fu, J.-R.; Lu, I.-W.; Chen, J.H.F.; Farn, C.-K. Investigating Consumers’ Online Social Shopping Intention: An Information Processing Perspective. Int. J. Inf. Manag. 2020, 54, 102189. [Google Scholar] [CrossRef]
  40. Teresa Borges-Tiago, M.; Santiago, J.; Tiago, F. Mega or Macro Social Media Influencers: Who Endorses Brands Better? J. Bus. Res. 2023, 157, 113606. [Google Scholar] [CrossRef]
  41. Wu, Y.; Niu, G.; Chen, Z.; Zhang, D. Purchasing Social Attention by Tipping: Materialism Predicts Online Tipping in Live-Streaming Platform through Self-Enhancement Motive. J. Consum. Behav. 2022, 21, 468–480. [Google Scholar] [CrossRef]
  42. Xu, X.; Huang, D.; Shang, X. Social Presence or Physical Presence? Determinants of Purchasing Behaviour in Tourism Live-Streamed Shopping. Tour. Manag. Perspect. 2021, 40, 100917. [Google Scholar] [CrossRef]
  43. Guo, L.; Hu, X.; Lu, J.; Ma, L. Effects of Customer Trust on Engagement in Live Streaming Commerce: Mediating Role of Swift Guanxi. Internet Res. 2021, 31, 1718–1744. [Google Scholar] [CrossRef]
  44. Hwang, K.; Zhang, Q. Influence of Parasocial Relationship between Digital Celebrities and Their Followers on Followers’ Purchase and Electronic Word-of-Mouth Intentions, and Persuasion Knowledge. Comput. Hum. Behav. 2018, 87, 155–173. [Google Scholar] [CrossRef]
  45. Yu, E.; Jung, C.; Kim, H.; Jung, J. Impact of Viewer Engagement on Gift-Giving in Live Video Streaming. Telemat. Inform. 2018, 35, 1450–1460. [Google Scholar] [CrossRef]
  46. Zhou, J.; Zhou, J.; Ding, Y.; Wang, H. The Magic of Danmaku: A Social Interaction Perspective of Gift Sending on Live Streaming Platforms. Electron. Commer. Res. Appl. 2019, 34, 100815. [Google Scholar] [CrossRef]
  47. Lin, Q.; Jia, N.; Chen, L.; Zhong, S.; Yang, Y.; Gao, T. A Two-Stage Prediction Model Based on Behavior Mining in Livestream e-Commerce. Decis. Support Syst. 2023, 174, 114013. [Google Scholar] [CrossRef]
  48. Chang, W. The Effectiveness of AI Salesperson vs. Human Salesperson across the Buyer-Seller Relationship Stages. J. Bus. Res. 2022, 148, 241–251. [Google Scholar] [CrossRef]
  49. Davenport, T.; Guha, A.; Grewal, D.; Bressgott, T. How Artificial Intelligence Will Change the Future of Marketing. J. Acad. Mark. Sci. 2020, 48, 24–42. [Google Scholar] [CrossRef]
  50. Luo, X.; Qin, M.S.; Fang, Z.; Qu, Z. Artificial Intelligence Coaches for Sales Agents: Caveats and Solutions. J. Mark. 2021, 85, 14–32. [Google Scholar] [CrossRef]
  51. Zhou, Z.; Wen, S.; Li, T.; Zhang, X.; Chi, M. Virtual Streamer and Destination Visitation: An Attractiveness Transfer Perspective. J. Destin. Mark. Manag. 2024, 33, 100922. [Google Scholar] [CrossRef]
  52. Zhang, Y.; Wang, X.; Zhao, X. Supervising or Assisting? The Influence of Virtual Anchor Driven by AI–Human Collaboration on Customer Engagement in Live Streaming e-Commerce. Electron. Commer. Res. 2023, 4, 1–24. [Google Scholar] [CrossRef]
  53. Bagwell, K.; Riordan, M.H. High and Declining Prices Signal Product Quality. Am. Econ. Rev. 1991, 81, 224–239. [Google Scholar]
  54. Qiu, Y.; Rao, R.C. Can Merchants Benefit from Entry by (Amazon-Like) Platform If Multiagent Prices Signal Quality? Mark. Sci. 2024, 43, 778–796. [Google Scholar] [CrossRef]
  55. Kihlstrom, R.E.; Riordan, M.H. Advertising as a Signal. J. Political Econ. 1984, 92, 427–450. [Google Scholar] [CrossRef]
  56. Mayzlin, D.; Shin, J. Uninformative Advertising as an Invitation to Search. Mark. Sci. 2011, 30, 666–685. [Google Scholar] [CrossRef]
  57. Chen, Y.; Liu, Q. Signaling through Advertising When an Ad Can Be Blocked. Mark. Sci. 2022, 41, 166–187. [Google Scholar] [CrossRef]
  58. Zhang, C.; Yu, M.; Chen, J. Signaling Quality with Return Insurance: Theory and Empirical Evidence. Manag. Sci. 2022, 68, 5847–5867. [Google Scholar] [CrossRef]
  59. Mayzlin, D. Promotional Chat on the Internet. Mark. Sci. 2006, 25, 155–163. [Google Scholar] [CrossRef]
  60. Milgrom, P.; Roberts, J. Price and Advertising Signals of Product Quality. J. Political Econ. 1986, 94, 796–821. [Google Scholar] [CrossRef]
  61. Joshi, Y.V.; Musalem, A. When Consumers Learn, Money Burns: Signaling Quality via Advertising with Observational Learning and Word of Mouth. Mark. Sci. 2021, 40, 168–188. [Google Scholar] [CrossRef]
  62. Dana, J.D. Bundling Can Signal High Quality. Int. J. Ind. Organ. 2020, 69, 102579. [Google Scholar] [CrossRef]
  63. Yang, W.; Zhu, C.; Yang, Y. Does Urban Digital Construction Promote Economic Growth? Evidence from China. Economies 2024, 12, 59. [Google Scholar] [CrossRef]
  64. Khan, A.U.; Rafi, M.; Zhang, Z.; Khan, A. Determining the Impact of Technological Modernization and Management Capabilities on User Satisfaction and Trust in Library Services. GKMC Glob. Knowl. Mem. Commun. 2023, 72, 593–611. [Google Scholar] [CrossRef]
  65. Jin, Z.; Lim, C.-K. Structural Relationships among Service Quality, Systemic Characteristics, Customer Trust, Perceived Risk, Customer Satisfaction and Intention of Continuous Use in Mobile Payment Service. JSMS J. Syst. Manag. Sci. 2021, 11, 48–64. [Google Scholar] [CrossRef]
  66. Gao, W.; Jiang, N.; Guo, Q. How Do Virtual Streamers Affect Purchase Intention in the Live Streaming Context? A Presence Perspective. J. Retail. Consum. Serv. 2023, 73, 103356. [Google Scholar] [CrossRef]
  67. Spence, M. Job Market Signaling. Q. J. Econ. 1973, 87, 355–374. [Google Scholar] [CrossRef]
  68. Davis, F.D. Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Q. 1989, 13, 319–340. [Google Scholar] [CrossRef]
  69. Davis, F.D.; Bagozzi, R.P.; Warshaw, P.R. User Acceptance of Computer Technology: A Comparison of Two Theoretical Models. Manag. Sci. 1989, 35, 982–1003. [Google Scholar] [CrossRef]
  70. Xing, A.; Wang, H. Pricing and Sample Set Strategies of Data Providers under Quality Information Asymmetry. J. Oper. Res. Soc. 2024, 75, 278–296. [Google Scholar] [CrossRef]
  71. Manfredi Latilla, V.; Frattini, F.; Messeni Petruzzelli, A.; Berner, M. Knowledge Management and Knowledge Transfer in Arts and Crafts Organizations: Evidence from an Exploratory Multiple Case-Study Analysis. J. Knowl. Manag. 2019, 23, 1335–1354. [Google Scholar] [CrossRef]
  72. Liu, Y.; Chen, H.; Duan, R. Whom to Benefit? Competing Platforms’ Strategic Investment in Recommender Systems. Electron. Commer. Res. Appl. 2022, 56, 101210. [Google Scholar] [CrossRef]
  73. Markopoulos, P.M.; Aron, R.; Ungar, L.H. Product Information Websites: Are They Good for Consumers? J. Manag. Inf. Syst. 2016, 33, 624–651. [Google Scholar] [CrossRef]
Jtaer 19 00144 g001
Figure 1. Timing of model.
Figure 1. Timing of model.
Jtaer 19 00144 g001
Jtaer 19 00144 g002
Figure 2. The high-quality firm’s profit variation with the fraction of uninformed consumers.
Figure 2. The high-quality firm’s profit variation with the fraction of uninformed consumers.
Jtaer 19 00144 g002
Jtaer 19 00144 g003
Figure 3. The high-quality firm’s profit variation with the fraction of uninformed consumers under different marginal costs for acceptance improvement.
Figure 3. The high-quality firm’s profit variation with the fraction of uninformed consumers under different marginal costs for acceptance improvement.
Jtaer 19 00144 g003
Table 1. Notations for the parameters in the model.
Table 1. Notations for the parameters in the model.
NotationDescription
i Firm type, i ∈ { L , H }
q i Product quality, with q H > q L
λ The fraction of uninformed consumers in the market
β 0 Consumers’ prior belief for the product being high quality
β Consumers’ posterior belief for the product being high quality
q ^ i The expected product quality under posterior belief
p The pricing of the product
a Consumers’ acceptance of AI streamers
k The marginal costs for increasing the consumers’ acceptance
t Consumers’ preference
N i The number of consumers who purchase the products of firm i
Π i The profit of the firm i
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Yu, Y.; Yang, Y. Signaling Effects in AI Streamers: Optimal Separation Strategy Under Different Market Conditions. J. Theor. Appl. Electron. Commer. Res. 2024, 19, 2997-3016. https://doi.org/10.3390/jtaer19040144

AMA Style

Yu Y, Yang Y. Signaling Effects in AI Streamers: Optimal Separation Strategy Under Different Market Conditions. Journal of Theoretical and Applied Electronic Commerce Research. 2024; 19(4):2997-3016. https://doi.org/10.3390/jtaer19040144

Chicago/Turabian Style

Yu, Ying, and Yunpeng Yang. 2024. "Signaling Effects in AI Streamers: Optimal Separation Strategy Under Different Market Conditions" Journal of Theoretical and Applied Electronic Commerce Research 19, no. 4: 2997-3016. https://doi.org/10.3390/jtaer19040144

APA Style

Yu, Y., & Yang, Y. (2024). Signaling Effects in AI Streamers: Optimal Separation Strategy Under Different Market Conditions. Journal of Theoretical and Applied Electronic Commerce Research, 19(4), 2997-3016. https://doi.org/10.3390/jtaer19040144

Article Metrics

Back to TopTop