Research on Adoption Intention Toward Intelligent Messaging Service: From Self-Determination Theory Perspective

Abstract

Empowered by artificial intelligence and 5G technologies, intelligent messaging service instead of the existing short messaging service could provide an omni-channel service, thus achieving higher interconnection for mobile users. In this paper, we adopted mixed methods research and explored the psychological factors that affect adoption intention to adopt intelligent messaging services among mobile users based on self-determination theory. After semi-structured interviews, we constructed a partial least squares structural equation model from the perspectives of intrinsic and extrinsic motivations. In addition, openness and perceived complexity were also introduced as an extended dimension. Through an online survey, 548 valid questionnaires were obtained. The results show that intrinsic motivation has a greater effect on adoption intention. Specifically, attitude, perceived autonomy, perceived relatedness, and perceived system quality have significant positive impacts on the adoption intention of intelligent messaging, while perceived complexity has a negative direct impact on adoption intention. Although perceived competence and perceived media richness have no significant effects on adoption intention, an indirect effect on adoption intention through attitude was observed. Notably, perceived interactivity and openness have no effect on adoption intention. Through this study, we aim to provide guidance for developers to focus on mobile users’ psychological needs regarding upgraded interactive channels, which can accelerate the construction of an omni-channel media environment.

1. Introduction

Empowered by ICT, mobile media has gradually become the dominant form since the Web 2.0 era, which has accelerated the construction of a mobile ecosystem to enhance the level of interaction between commercial/government institutions and mobile users. Multiple interactive channels have been developed to provide online services for mobile users, e.g., mobile applications, instant messaging tools, vertical or horizontal web portals, and social media platforms. In this context, due to its advantages of high reach rate, low cost, strong reminder ability, and portability, the short messaging service (SMS) is still an active channel for both individual and enterprise users. It has been reported that the global SMS market was worth USD 64,489.2 billion in 2021 and is slated to reach USD 84,904.89 million by 2027, with a compound annual growth rate of 3.8% from 2022 to 2027 [1]. American consumers exchanged 2 trillion messages in 2021 [2]. In China, the volume of texting had reached 1874.81 billion by the end of 2022 [3], and the associated revenue has increased by 2.7% over the last year [4]. For personal users, despite the impact of Internet-based over-the-top (OTT) products, people still need SMS as a channel to communicate with others at some point. For enterprise users, interaction by SMS can help a company to obtain the best advertising effect and reach diverse users.

With the emergence of AI and 5G technologies, SMS has been upgraded to intelligent messaging, which is an omni-channel instant message application via the original SMS portal. Intelligent messaging can provide both individual and business users with a wealth of mobile resources and functions via higher-level intelligent interactivity, e.g., chatbot Q&A, online purchases, mobile payments, booking, government affairs service, etc. According to messaging trends released by infobip, the mobile messaging industry witnessed a dramatic shift in 2024. RCS traffic increased 550% due to the development of AI, brand favorites, and Apple iOS-supported services. RCS entered the top 10 list of omni-channel combinations, especially boosting marketing and advertising services [5]. At present, the 5G Rich Communication Suite (5G RCS), which had been in the pre-commercialization stage and supported by various types of mobile terminals, is a popular form of intelligent messaging service in China.

It is worth noting that intelligent messaging not only supports individual users but also provides a new platform for business users. Various data types (e.g., text, images, audio, video) can be transmitted through the mobile Internet to enhance the information interaction between target users and service providers (e.g., companies and government departments). The function of AI-enabled chatbots in intelligent messaging further improves the effectiveness of conversation with target users, e.g., instant replies, reminders, or answers. The user interface and functionality of intelligent messaging are similar to those of mobile apps, including service accounts, cards, floating windows, and so on, realizing a messaging channel that includes group chats, videos, read receipts, and other advanced features—so-called “message as a platform”. Mobile users no longer need to download or install an external installation package and instead enjoy one-stop service with an application-like experience, which facilitates 24 h instant human–computer interaction on smartphones. The efficiency of two-way communication and the response speed can be guaranteed.

We can see that the usage of SMS has been widely studied in the mobile communication age, and its updated application, i.e., intelligent messaging, is in the pre-commercialization phase in China. The AI-enabled functions and all-in-one platform will provide mobile users with more convenience and a better interaction experience. However, to date, few researchers have investigated the adoption intention of intelligent messaging services. Similar applications, such as intelligent or AI-chatbot services, have been studied in terms of, e.g., impact factors for user satisfaction [6,7] and user loyalty [8] after wide commercial use. It can be seen that most prior studies focused on system-centric factors and used traditional models such as the technology acceptance model (TAM) or the unified theory of acceptance and use of technology (UTAUT). However, little is known about how intrinsic psychological needs drive the adoption of intelligent messaging. To rectify this, based on self-determination theory, we aim to explore the factors influencing the adoption of intelligent messaging among mobile users. To this end, the following research questions (RQs) are proposed:

• RQ1: What are the psychological factors that influence intelligent messaging service adoption intention among mobile users?
• RQ2: How do influencing factors affect intelligent messaging service adoption intention among mobile users?

Similar research questions are popular in the study of emerging technologies and applications, especially in the field of mobile scenarios, in which the factors of users’ behavioral intentions have been explored with respect to virtual reality applications [9], social media [10], personal financial management software [11] and so on. The contributions of this paper are three-fold and can be summarized as follows:

• We explore the antecedents of intelligent messaging adoption intention through mixed methods research (i.e., qualitative and quantitative research). From the perspective of intrinsic and extrinsic motivation, self-determination theory (SDT) is used as a framework instead of traditional theoretical models (e.g., TAM and UTAUT), as it focuses on psychological processes and human motivational behavior.
• Semi-structured interviews were conducted among intelligent messaging service users to find out the SDT extrinsic motivation and extended dimension factors. Two extended variables, i.e., openness to experiencing personality traits and perceived complexity of the diffusion of innovation theory, are summarized from qualitative research and then introduced into the structural equation model.
• We have found that the effects of intrinsic motivation are greater than those of extrinsic motivation on adoption intention, which is in line with the expectations of SDT. Our research could learn the psychological motivation and adoption intention regarding intelligent messaging among mobile users in the pre-commercial stage, which is expected to enhance “user-centered” awareness for technical developers and service providers, which can further accelerate the large-scale commercialization of intelligent messaging and construction of the omni-channel media environment.

The rest of this paper is organized as follows: Section 2 presents the literature review about the usage of SMS and SDT. Section 3 explains the proposed theoretical model and describes the hypotheses among constructs. Section 4 and Section 5 present the research methods and data analysis results. Section 6 discusses the main findings and summarizes insights. Finally, Section 7 concludes this paper.

2. Literature Review

2.1. The Usage of SMS

As a basic interactive outlet on mobile phones, SMS plays a vital role in two-way communication in the mobile age. It has been widely used in the fields of disease prevention, agricultural production, government and business affairs, among others, which demonstrates its popularity and usefulness among institutional and individual users.

For disease prevention, Le and Holt pointed out that the spiritually based SMS texting intervention may be a culturally appropriate and cost-effective approach to promoting cervical cancer early detection information to church-attending African American women [12]. In addition, through randomized controlled trials, Tull et al. found that sending SMS reminders to parents/guardians resulted in a greater acceptance of the HPV vaccine among adolescents who participated in school vaccinations [13]. In addition, Gibson et al. further showed that SMS reminders need to be combined with incentives (money) to significantly improve immunization coverage and timeliness in a setting with high baseline immunization coverage levels [14].

For agricultural production, most Trinidad farmers think that SMS is an easy way to obtain information and communicate with extension officers [15]. Beza et al. showed that the intention to adopt mobile SMS technology for agricultural data provision was predicted by the factors of performance expectancy, effort expectancy, price value, and trust [16]. In the context of government affairs, Shareef et al. revealed that the key factors affecting the development of public administration attitudes towards the provision of services through mobile SMS are time and location, relevance, and reliability [17].

For the businesses, SMS is regarded as an important advertisement channel to reach the mobile users. Lin and Chen found that attitudes towards SMS advertising and perceived behavioral control have positive and significant effects on consumers’ usage intentions [18]. Tseng et al. indicated that, if consumers explicitly allow the receipt of text messaging advertising from companies, they are more likely to read that content carefully, which further strengthens the effect of messaging content relevance, informativeness, entertainment, and interactivity on the attitudes toward the SMS advertising [19].

Bakr et al. also pointed out that accurate targeting and personalization are crucial to ensure that SMS advertising is useful and relevant to recipients, where perceived value, ad trust, and channel acceptance positively and significantly influence attitudes toward SMS advertising [20]. Sharma et al. further suggested that SMS advertising perception has a significant positive effect on purchase intention, which is mediated by advertising value and attitudes towards SMS advertising [21].

During the COVID-19 pandemic, as a complementary channel, SMS was frequently used to distribute epidemic-protection information or reminders to mobile users. Yu et al. found that the respondents had positive attitudes toward public-interest SMS from government and official institutions [22]. Lai et al. revealed that SMS was an efficient and cost-effective method of flow control during COVID-19 [23]. In addition, Loubet et al. pointed out that the SMS tracking platform can be used as an early warning system for COVID-19 patients who have clinical conditions [24].

2.2. AI-Empowered Messaging Services

As the core function in intelligent messaging services, intelligent or AI chatbot service has been widely studied; examples are impact factors for user satisfaction [6,7] and user loyalty [8] after wide commercial use. For user satisfaction, Hsu et al. found that the recovery and conversational quality of the AI chatbot affected the customers’ satisfaction and then deepened their loyalty [6]. Rese et al. revealed the effects of the gender of chatbot avatars effects on service recovery. The higher perceived humanness and service recovery improved the users’ satisfaction [7]. Users’ e-brand loyalty was enhanced by AI chatbot service quality and consumer trust played a mediating role in this relationship [8]. Researchers have also found that the characteristics of a chatbot play an important role; for example, Sidlauskiene et al. posited that an anthropomorphic verbal design for AI-driven chatbots impacted on the perceived product personalization [25]. Hsu et al. examined the contextual factors of customers’ perceptions and chatbot adoption intentions and then revealed suitable tasks for chatbots and mobile apps [26]. Wang et al. conducted an experiment and mentioned that the communication styles of the chatbot (e.g., social-oriented) increased the enhanced consumers’ satisfaction with service recovery [27]. Gnewuch et al. showed that including social cues (e.g., name, human-like avatar) in the design enhances the social presence of chatbots and usage intention towards them [28]. From self-determination theory, Jiménez-Barreto et al. confirmed the direct influence of self-determined interaction on customer experience, participants’ attitudes, and satisfaction with the airline chatbot [29].

For the 5G rich communication services (RCSs), Yu et al. conducted an offline experiment and then found that satisfaction, perceived convenience, and performance expectancy have a direct impact on behavioral intention in the usage of 5G RCS among university students [30]. In a word, SMS still exerts an important role in daily life. It is valuable to further study the psychological factors that influence users’ adoption of emerging intelligent messaging, which will help to improve and popularize intelligent messaging services in general.

2.3. Self-Determination Theory

Self-determination theory (SDT) is regarded as an effective contemporary framework to explore human motivational behavior, which reveals the usage intention for different types of technologies from the intrinsic and extrinsic perspectives [31]. SDT is applied in psychology to study motivation, personality development, and wellness [32], e.g., the studies of public organizations [33], sports [34], nature conservation [35], social media [36,37], and so on. Notably, SDT assumes that humans are active, growth-oriented organisms who naturally tend to integrate their psychological elements into a unified sense of self and integrate themselves into the middle of larger social structures [38].

As an important part of SDT, autonomous motivation consists of both intrinsic motivation and extrinsic motivation. People can identify with the value of activity and integrate it into their sense of themselves [39]. Intrinsic motivation means satisfaction tends to be inherent in performing an activity rather than some separable consequence [40]. In other words, when the individual is intrinsically motivated, they will act for fun or a challenge rather than because of external stimulation, stress, or reward. A direct corollary of the SDT is that people will tend to pursue goals, domains, and relationships that allow/support the satisfaction of their needs [38]. Therefore, the understanding of human motivation requires considering the intrinsic psychological needs for autonomy, competence, and relatedness [38]. When those needs are met, the individuals tend to be intrinsically motivated and satisfied [41]. The strong link between intrinsic motivation and the satisfaction of needs for autonomy and competence has been clearly documented.

Osei et al. found that perceived relatedness, perceived autonomy, and perceived competence positively influence students’ behavioral intentions and actual usage of e-learning [42]. Lee proposed that autonomy and relatedness have positive effects on perceived well-being [43]. In addition, Demircioglu suggested that employees using social media for working had higher self-determination, concluding that autonomy and competence can increase job satisfaction [41]. In this way, we choose three basic psychological needs, i.e., perceived competence, perceived autonomy, and perceived relatedness to measure the individuals’ intrinsic motivations.

In order to obtain separable results at any time, extrinsic motivation refers to the construct that activities are carried out only for their instrumental value instead of enjoying the activities themselves [40]. In other words, extrinsic motivation is the characteristic possessed by a particular thing or event itself, and individuals will adopt the thing or take action for its instrumental value. In our research, intelligent messaging is the upgrade of traditional SMS with improved system quality, as well as more obvious interactivity and media richness due to the application of artificial intelligence technology and rich communication services. Therefore, we select perceived media richness, perceived interactivity, and perceived system quality as extrinsic motivation through our semi-structured interview results shown in

Table A2. Interview coding results.

Parent Node	Child Nodes (Number)	Example
PMR	High media richness (10)	S12: “5G RCS (Intelligent Messaging) can provide richer information. When learning news, the provided information help me understand comprehensively.”
		S16: “Intuitively, it includes integrated media forms (e.g., pictures, audio, video) and enable interaction.”
		S17: “5G RCS (Intelligent Messaging) has richer content!”
		S21: “5G RCS (Intelligent Messaging) conveys sufficient and more attractive contents, which is better than traditional (applications) in terms of media richness.”
	Fast response and informative (24)	S1: “The answer to question of 5G RCS (Intelligent Messaging) is fast and very convenient.”
		S9: “It (Intelligent Messaging) integrates various applications and functions. When looking for functions and services, it quickly responds my needs and saves searching times.”
		S12: “To me, I think it is very fast than 4G. The amount of information is quite large, i.e., it can convey a lot of information in short time, which is the most significant advantage.”
	Various types of services (17)	S8: “IM summarizes more services, e.g., car services, like 3D version of 4S store experience.”
		S12: “5G RCS contents are relatively extensive and wide coverage.”
		S14: “Compared with traditional SMS, 5G RCS can handle more information related to government and commercial affairs.”
PI	Intelligent Response (6)	S9: “It looks like I’m talking to an AI Chatbot, who can respond to my needs directly without searching on mobile phone and help find what I want.”
		S10: “Whatever you ask, it will answer your question immediately, which is much more sensitive than some customer service.”
		S19: “When typing a question, 5G RCS will automatically capture the core problem, and smart interactions is realized.”
	Improved Intelligence (4)	S1: “It has developed to the point that everyone uses it. Do not answer questions that are not what you want to ask. It will give you an answer right away.”
		S19: “My suggestion is the smarter the better, it can improve human–computer interaction. ”
PSQ	One-stop service (8)	S3: “It has a lot of functions. You can find all services in a place without other software.”
		S9: “5G RCS (Intelligent Messaging) provide one-stop service directly. Telling your needs through the dialog box, it will solve everything for you.”
		S19: “Via 5G RCS (Intelligent Messaging) portal, as long as you send a demand, e.g., recharging the phone bill, it may come out directly, a one-stop service.”
	User Friendly (6)	S4: “I think 5G RCS (Intelligent Messaging) is more convenient for some elderly people. Because 5G RCS (Intelligent Messaging) can deliver contents directly.”
		S9: “I feel (Intelligent Messaging) may be more friendly to older people like our parents… it takes SMS form…so everyone will be more familiar with it with more user-friendly.”
		S19: “Like the old people in my family, they may not use mobile Apps, but they can type. 5G RCS (Intelligent Messaging) will provide more convenient service.”
	Reduce software and save memory (16)	S3: “It can save and instead of mobile Apps on phone. I estimate 5–6 Apps can be reduced.”
		S4: “It can reduce software redundancy. Because with the direct portal, some services must be used on Apps or applets will not be used any more. Those software can be uninstalled.”
		S15: “It avoids repeatedly downloading many official accounts and Apps. e.g., health code”
	Low learning cost (4)	S11: “The operation of 5G RCS is same as previous SMS. The learning cost is very low.”
		S13: “The operation mode of 5G RCS is not different with WeChat official account. Learning cost is quite low… In fact, people are already familiar with this interactive way.”
		S18: “Usage habits do not need to be cultivated”
OPEN	Out of curiosity (3)	S9: “Yes. It’s novelty. For me, 5G RCS provide novel experience…”
		S15: “I’m more willing to try and prefer to some new applications…”
		S21: “5G RCS is new thing. I prefer to have a try with these kind of emerging applications.”
PCOM	Hard to search information (7)	S2: “(In 5G RCS) I need to find applications, which is laborious and troublesome.”
		S17: “You need to enter the original SMS portal and cannot click directly on desktop. one more step operation compared with Apps.”
		S20: “If you use 5G RCS every day, some disadvantages exists, e.g., search the history contents from address book or inbox.”
	Worry about delays (4)	S5: “If many people usage, application crash will occur. Just like some Apps now.”
		S14: “I’m more worried about whether there exists any problems or error delays when 5G RCS operated.”
	Slow response (2)	S13: “Its response speed is quite slower than mature OTT applications like WeChat.”
		S21: “I do not think the speed is fast, sometimes it’s quite slow to open pictures.”
AINT	Willing to use (21)	S7: “Yes, because it has a lot of functions in it, it’s very convenient, you do not have to bother to find. It’s very clear, and you can see it when you click in.”
		S11: “I am willing to use it, because I think it is beneficial to experience and indeed service responds seem faster.”
		S6: “Yes, because it does not take up memory”
	Intended to use first (16)	S12: “I may give priority to use it. Because a platform with all functions can reduce my time and energy of swinging and choice among different platform in a short time”
		S14: “I will give priority to it. I think the point that attracts me more is that IM can save mobile phone memory.”

.

3. Theoretical Model and Hypotheses

In order to ascertain the mobile users’ experience of the intelligent messaging service, we conducted a semi-structured interview from July to October 2022. Based on self-determination theory and interview results, we construct a theoretical model from the perspectives of intrinsic motivation (including perceived competence, perceived autonomy, and perceived relatedness) and extrinsic motivation (including perceived media richness, perceived interactivity, and perceived system quality) by integrating the factors of openness and perceived complexity. We aim to reveal how these factors influence the attitude and adoption intention of the emerging intelligent messaging. The relationships among the constructs are shown in Figure 1. The definition of each construct is shown in

Table 1. Construct definition.

Construct	Definition	Reference
PC	The degree to which an individual reflects and perceives the ability to use intelligent messaging.	[44]
PA	The degree to which individuals feel free to use intelligent messaging.	[38]
PRN	The degree of relevance that individuals perceive to other organizations in the process of using intelligent messaging.	[38]
PMR	The media richness of intelligent messaging perceived by individuals, such as personalization and diversification.	[45]
PI	The degree of two-way communication, timeliness, and interpersonal simulation of intelligent messaging experienced by individuals.	[46]
PSQ	The degree to which individuals perceive the quality of system characteristics such as operability, interface aesthetics, and reliability of intelligent messaging.	[47]
OPEN	The personality dimension that characterizes someone who is intellectually curious and tends to seek new experiences and explore novel ideas.	[48]
PCOM	The individual’s perception of how difficult it is to use intelligent messaging.	[49]
ATT	The individual’s judgment on the function of intelligent messaging, etc.	[50]
AINT	The intensity of an individual’s intention to adopt intelligent messaging.	[51]

.

3.1. Attitude and Adoption Intention

According to the theory of planned behavior, attitude has an important impact on behavioral motivation [50], through the judgment of whether this behavior is good or bad. Similarly, in the technology acceptance model, attitudes also have a positive effect on the intention to adopt new technologies, which is defined as the measure of the strength of one’s intention to perform a specified behavior [51].

Lavidas et al. found that attitudes positively affected undergraduates’ Google Scholar adoption intentions [52]. Bitrián et al. also confirmed the relationship between users’ attitudes toward personal financial management apps and usage intention is positive and significant [11]. Similarly, Cheong and Park pointed out that attitude toward mobile Internet is the most significant factor in predicting behavioral intention [53]. Furthermore, Li showed that individuals’ attitudes toward using mobile health services could positively affect their intention to use such services in the future [54]. Therefore, we proposed the following:

H1. 

Attitude has a positive impact on the adoption intention of intelligent messaging.

3.2. Intrinsic Motivation

As the classical antecedents of intrinsic motivation in SDT, the relevant research hypotheses about perceived competence, perceived autonomy and perceived relatedness are illustrated as follows.

3.2.1. Perceived Competence

Perceived competence involves a sense of mastery that is best fulfilled in a well-structured environment that provides optimal challenges, positive feedback, opportunities for growth, and a feeling that one can succeed and grow [44]. In other words, when an individual uses an application that provides positive feedback and provides opportunities for benign communication, the individual can gain great perceived competence.

Touati and Baek found that perceived competence had a positive effect on the attitude towards a mobile learning game, which was the main predictor of enjoyment [55]. Similarly, Lee et al. found that perceived competence positively affected students’ intention to use online knowledge-sharing systems through performance expectancy [56]. Additionally, in the area of mobile health, Liu et al. pointed out that perceived competence played a positive role in influencing individuals’ routine intention to use the mobile health service [57]. Almathami et al. also indicated that consumers’ perceived competence positively impacted their motivation toward the use of tele-consultation systems [58]. Therefore, we proposed the following:

H2a. 

Perceived competence has a positive impact on the attitude of intelligent messaging.

H2b. 

Perceived competence has a positive impact on the adoption intention of intelligent messaging.

3.2.2. Perceived Autonomy

Perceived autonomy refers to the volition that the organismic desire to self-organize experience or behavior and be concordant with its own integrated sense of self, which is an essential aspect of healthy human functioning [38].

Chatzisarantis et al. found that perceived autonomy support positively affected attitudes toward health behavior, with significant positive effects on both intention and health behavior [59]. In terms of mobile learning, Raman et al. found that autonomy had the greatest positive impact on behavioral intention on a web-based course management platform called Moodle [60]. Similarly, Khalid et al. also found a positive and significant effect of perceived autonomy on behavioral intention to use Massive Open Online Courses (MOOCs) in Thailand [61]. In addition, Huang et al. suggested that perceived autonomy had a positive effect on the intention to use virtual reality technology in virtual learning environments [62]. In addition, Lau and Ki also found that autonomy positively affected consumers’ intention to continuously use VR fashion apps [9]. Therefore, we propose the following:

H3a. 

Perceived autonomy has a positive impact on the attitude of intelligent messaging.

H3b. 

Perceived autonomy has a positive impact on the adoption intention of intelligent messaging.

3.2.3. Perceived Relatedness

Perceived relatedness refers to the desire to feel connected to others (e.g., be loved and cared for), which is a fundamental need for individuals [38]. In other words, relatedness focuses on improving intimate communication between individuals and organizations.

Tsai et al. found that for exercisers, perceived relatedness had a positive impact on attitude, which further influenced behavioral intention to use a real-time, interactive mobile application about running, while, for spectators, perceived relatedness had no effect on attitude [63]. Grayson et al. also indicated relatedness was a significant predictor of the intention to use bicycle lanes, which had a positive and significant impact [64]. In terms of learning, Khan et al. pointed out that perceived relatedness had a positive and significant effect on students’ behavioral intentions with respect to Massive Open Online Courses (MOOCs) [65]. Similarly, Nikou and Economides found that perceived relatedness positively affected students’ intention to use mobile-based assessment through perceived usefulness and perceived ease of use [66]. In addition, Luo et al. also showed that perceived relatedness could also positively affect college students’ continued intention with respect to online self-regulated learning (SRL) through perceived enjoyment [67]. Relying on omni-channel capabilities and enriched media forms, intelligent messaging can strengthen the intimate communication between mobile users and potential service providers (e.g., institutions, vendors, etc.), enhancing users’ perceived relatedness. Therefore, we propose the following:

H4a. 

Perceived relatedness has a positive impact on the attitude of intelligent messaging.

H4b. 

Perceived relatedness has a positive impact on the adoption intention of intelligent messaging.

3.3. Extrinsic Motivation

Due to the lack of consensus on the constructs of extrinsic motivation among previous studies, we need to conduct qualitative research to refine the specific constructs under the extrinsic motivation framework. Similar procedures have been carried out in the studies of organizational information adoption [68], augmented reality [69], and online supermarkets [70]. Semi-structured interviews help to further refine the constructs in the model and grasp the connotation of the survey questions [70]. Therefore, according to the characteristics of intelligent messaging and the results of the semi-structured interviews in Appendix A Table A2, we summarized perceived media richness, perceived interactivity, and perceived system quality as extrinsic motivation, which is consistent with its definition [40].

3.3.1. Perceived Media Richness

The concept of media richness is derived from the media richness theory (MRT), which provides a foundation for understanding human behavior involving electronic communication media [71]. A primary goal of media richness theory (MRT) is to determine which technologies best reduce uncertainty and equivocality in various business settings [72]. “Richness” is defined as the ability to enable users to convey information and thus facilitate the acquisition of shared meaning and understanding within a given time interval [73]. Therefore, a medium that possesses rich information and high vividness could improve the effectiveness of communication [74]. Four criteria for media richness were proposed in [45], including feedback, multiple cues, language variety, and personal focus. Intelligent messaging can provide various types of information, e.g., video, audio, and files, which increases its media richness, a characteristic that was perceived and considered to be useful by interviewees. Therefore, we use the dimension of perceived media richness.

Zhou et al. found that perceived media richness could positively predict Chinese university students’ attitudes towards e-learning during COVID-19 through perceived ease of use, and further improve their behavioral intention [75]. In addition, Lin and Chen found that perceived media richness not only positively affected users’ attitudes towards an augmented reality question-answering system for mobile cloud computing directly, but also positively affected attitudes through perceived usefulness [76]. Safdar et al. also found that perceived media richness positively affected students’ intention to use a cloud-based virtual learning environment in rural areas [77]. Similarly, Lai and Chang pointed out that media richness had a positive impact on users’ intention to use dedicated e-book readers, as well as a positive impact through perceived usefulness [78]. Therefore, we propose the following:

H5a. 

Perceived media richness has a positive impact on the attitude of intelligent messaging.

H5b. 

Perceived media richness has a positive impact on the adoption intention of intelligent messaging.

3.3.2. Perceived Interactivity

The concept of perceived interactivity originates from social presence theory, which helps mimic real interpersonal communications, thus reducing the sense of distance between two parties [79]. Whether online or face-to-face communication, perceived interactivity is considered an important factor in the information exchange process [80]. Timeliness and engagement are important [81]. Lin and Chang defined perceived interactivity as the extent to which users perceive their experiences as a simulation of interpersonal interaction and feeling that they are in the presence of another user in a social setting [46]. In addition, the interactive innovation of social media offers two-way communication, which helps to promote faster the adoption process since it attains a critical mass of users more quickly [82]. Aided by a chatbot, intelligent messaging can better simulate human interaction and make users feel as similar to communicating with real people. Intelligent messaging provides users with additional channels and timely increases two-way communication timely between users and governmental/commercial organizations. Perceived interactivity is a good measure of how users feel about intelligent messaging rather than a broader measure of its usefulness. For this reason, we used the dimension of perceived interactivity.

Yim and Yoo found that when using digital menus, consumers’ perceived interactivity positively affected attitudes toward digital menus [83]. In addition to the direct effect of perceived interactivity on attitude, Ahn et al. also found that perceived interactivity would have a positive impact on users’ attitudes towards sports websites through entertainment and information indirectly [84]. Willoughby and L’Engle also suggested that the perceived interactivity of a sexual health SMS service not only was positively associated with positive attitudes toward the service, but also positively influenced the repeated use of the service [85]. Similarly, Shin et al. found that perceived interactivity positively affected both attitudes and intention to use smart TVs [86]. In addition, Xu et al. also suggested that higher perceived interactivity increased users’ intention of future participation in online social interactions [87]. Trendiness and interactivity were all significant determinants of adoption intention [88]. Therefore, we propose the following:

H6a. 

Perceived interactivity has a positive impact on the attitude of intelligent messaging.

H6b. 

Perceived interactivity has a positive impact on the adoption intention of intelligent messaging.

3.3.3. Perceived System Quality

System quality refers to distinctive characteristics of information systems [89], which is concerned with the level of technical success in the production of information that performs tasks or daily activities [90]. On the Internet and in cyberspace, system quality measures desired characteristics, such as usability, availability, reliability, adaptability, and response time, which are valued by users [47]. When a mobile operating system is difficult to use and the interface is inefficiently designed, users may perceive that the corresponding service provider is unable to provide a quality service [91].

As an update and upgrade of SMS, forms for official accounts, cards, and floating windows are added inside, which satisfies usage habits. In addition, perceived system quality can more intuitively reflect the respondents’ views on the function of intelligent messaging, which is more specific than simply examining whether intelligent messaging is useful or easy to use. Expressions of perceived system quality were also stated in the interview. Therefore, we choose perceived system quality as a dimension of extrinsic motivation.

Lee et al. found that system quality positively affected customer attitudes towards virtual reality [92]. Additionally, Halim and Elbadrawy found that system quality positively affected attitudes toward e-learning systems during COVID-19 through perceived usefulness [93]. Similarly, Calisir et al. showed that system quality had a positive effect on blue-collar workers’ attitudes toward e-learning systems through perceived ease of use [94]. In addition, Du et al. pointed out that teachers’ continuous usage intention of VR technology for classroom teaching was indirectly positively affected by the system quality through the task technology fit [69]. Alzahrani et al. found the system quality of a digital library had a positive effect on university students’ usage intention in Malaysia [95]. Notably, Wut et al. indicated that system quality positively affected male students’ behavioral intention to use discussion forums on electronic learning platforms [96]. Therefore, we propose the following:

H7a. 

Perceived system quality has a positive impact on the attitude of intelligent messaging.

H7b. 

Perceived system quality has a positive impact on the adoption intention of intelligent messaging.

3.4. Extended Variables

In the semi-structured interviews, some interviewees mentioned that their curiosity motivated their intelligent messaging service adoption intention. In addition, some interviewees said that there existed complicated situations when using intelligent messaging (as shown in Table A2). Therefore, we introduced two extended independent variables—openness to experience and perceived complexity—to expand the richness and improve the interpretability of the model.

3.4.1. Openness to Experience

The big five personality traits model has emerged as an influential model for understanding the relationship between individuals’ personalities and behaviors [97]. Therein, openness to experience is a personality dimension that characterizes someone who is intellectually curious and tends to seek new experiences (or novel ideas) [48]. Personality traits reflect a stable set of individual characteristics and tendencies, which can remain relatively stable even across cultures [98]. Individuals who score high on openness not only listen to new ideas but also change their own beliefs after new experiences [99].

Denden et al. found that openness positively affected students’ attitudes toward gamified learning environments, which further positively affected intention to use [100]. In addition, Liang et al. indicated that consumers’ intentions toward using mobile self-checkout in fashion retail stores were positively predicted by openness to experience [101]. Chipeva et al. also pointed out that openness was a significant predictor of behavioral intention, which had a positive effect on intention [102]. In addition, Irfan and Ahmad proposed that openness positively moderated the linkage between information acquisition and intention to use 5G technology [103]. Therefore, we propose the following:

H8. 

Openness to experience has a positive impact on the adoption intention of intelligent messaging.

3.4.2. Perceived Complexity

According to innovation diffusion theory, complexity refers to the degree to which an innovation is considered relatively difficult to understand and use from the user side [49,104], especially for new technologies and applications. If technology users find a system difficult to use, they may stop using it [10].

Lawson-Body et al. stated that the lower the perceived complexity of using e-government services, the more likely that e-government services would be adopted by veterans [104]. Lean et al. also pointed out that perceived complexity had a significant negative relationship with the intention to use the e-government service [105]. Similarly, Ramkumar and Jenamani showed that perceived complexity was a negative predictor of organizational buyers’ behavioral intention to accept e-procurement services [106]. Sullivan and Koh also suggested that perceived complexity was the main inhibitor of social media continuance intention, which has a negative effect on continuance intention [10]. Therefore, we proposed that

H9. 

Perceived complexity has a negative impact on the adoption intention of intelligent messaging.

4. Research Method

4.1. Data Collection

In this work, we conduct mixed-methods research, which combines the qualitative and quantitative methods together. For the qualitative study, from July to October 2022, we conducted a semi-structured interview with a total of 21 interviewees, including 11 males and 10 females from 18 to 47 years old. The demographic information is shown in Appendix A

Table A3. Demographic information of the interviewees.

No.	Gender	Age	Occupation
S1	Male	47	Electrician
S2	Female	18	Student (Undergraduate)
S3	Female	23	Student (Graduate)
S4	Female	22	Student (Undergraduate)
S5	Female	28	Student (Ph.D.)
S6	Female	19	Student (College)
S7	Female	20	Freelance
S8	Male	40	Civil servant
S9	Female	23	Student (Undergraduate)
S10	Female	21	Student (Undergraduate)
S11	Male	24	Student (Graduate)
S12	Female	21	Student (Undergraduate)
S13	Male	42	Software practitioner
S14	Female	24	Freelance
S15	Male	32	Sales engineer
S16	Male	42	ICT company manager
S17	Male	32	ICT company manager
S18	Male	45	ICT company manager
S19	Male	33	Securities company manager
S20	Male	41	ICT company manager
S21	Male	37	ICT company staff

. The interviewees were recruited from both online channels (e.g., posting recruitment advertisements on social media and public accounts) and the offline channels, who used to operate and experience the intelligent messaging service (5G RCS application) recently. We aimed to learn the attitude and usage experience of mobile users who used intelligent messaging service. Based on quasi-grounded theory, we employed thematic analysis using NVivo 11. The researchers analyzed the raw data from semi-structured interview and then derived the themes. The codes were categorized under these themes as shown in Table A2. The analytic procedure was following [107].

For the quantitative study, each participant would read the introduction and applied scenarios about intelligent messaging, and then watch a 4 min 37 s video about the existing forms and operating steps of intelligent messaging (including the operations of instant messaging communication, news, search, payment, medical care, traveling order, etc.) at the beginning of the questionnaire. Those stimulus materials were extracted from tele-operator advertising or other popular science articles/videos, which are relatively easy to understand. We also set a duration limit and allowed participants to have a sufficient period to view those materials. Finally, the participants completed the questionnaires based on their understanding of intelligent messaging and actual use.

The questionnaire was compiled and distributed through social media and the Credamo survey platform (https://www.credamo.com/) from the period of 21 February to 17 March 2023, and contains 53 items with a seven-point Likert scale (rating from “1” strongly disagree to “7” strongly agree), except for verification/demographic questions in Appendix A

Table A1. Items of constructs and sources.

Items		Sources
Perceived Competence (PC)
PC1: I feel a sense of accomplishment from using intelligent messaging to handle affairs.
PC2: I feel that I can handle affairs by using intelligent messaging effectively.		[123]
PC3: I think I have the ability to use intelligent messaging proficiently after I have been using it for a while.	*	[124]
PC4: Overall, I’m capable to use intelligent messaging effectively and easily to communicate with others and handle affairs.		[58]
Perceived Autonomy (PA)
PA1: I feel free to express my personal thoughts and complete online tasks when I use intelligent messaging.
PA2: I feel like I can pretty much use intelligent messaging as I want to.
PA3: I feel that intelligent messaging gives me a variety of services to choose from.	*	[124]
PA4: I think the features and services provided by intelligent messaging match my daily usage habits		[123]
PA5: I think I can freely choose whether to use intelligent messaging to handle affairs.	*
Perceived Relatedness (PRN)
PRN1: I feel that intelligent messaging provides an open channel for me to communicate with government, business and other organizations.		[123]
PRN2: I feel more relevant when using intelligent messaging to communicate with organizations such as government, business and other organizations.
PRN3: I feel that I can establish a more friendly communication relationship with other people, government organizations and institutions by using intelligent messaging.
PRN4: When I use intelligent messaging, I feel a lot of closeness and intimacy with the service provider (company, platform).		[125]
PRN5: I think it is possible to contact the organization through intelligent messaging to handle important affairs	*	[126]
Perceived Media Richness (PMR)
PMR1: I feel that the features of intelligent messaging allow me to give and receive timely feedback.	*	[127] [128] [129]
PMR2: I think intelligent messaging can customize the service to my own personal requirements.	*
PMR3: I feel that intelligent messaging provides users with multiple message types (e.g., text, picture, voice, location, etc.) in online communication.	*
PMR4: I think intelligent messaging can provide me with a variety of rich media information.
PMR5: I think intelligent messaging can help me to communicate quickly.
Perceived Interactivity (PI)
PI1: I feel using intelligent messaging for interactive communication looks like chatting with human customer service.		[130]
PI2: I feel like using intelligent messaging gives me a sense of conversation.
PI3: I perceive that intelligent messaging can capture my information needs keenly.
PI4: I think I can get better responses and feedback by using intelligent messaging.		[85]
PI5: I think intelligent messaging will help easy communication with other organizations in the future.	*	[131]
Perceived System Quality (PSQ)
PSQ1: I think the interface layout of intelligent messaging is clear and user-friendly.		[92]
PSQ2: I think intelligent messaging is very functional.		[132]
PSQ3: I think intelligent messaging can provide a reliable service.	*
PSQ4: I think intelligent messaging has a better operability.	*	[133]
PSQ5: I think intelligent messaging allows loading all kinds of rich media information more quickly.
Openness to Experience (OPEN)
I see myself as someone who
OPEN1: is original, comes up with new ideas.		[134]
OPEN2: is curious about many different things.	*
OPEN3: is ingenious, a deep thinker.
OPEN4: has an active imagination.
OPEN5: is inventive.
OPEN6: values artistic, aesthetic experiences
OPEN7: prefers work that is routine (Reverse encoding)	*
OPEN8: likes to reflect, play with ideas	*
OPEN9: has few artistic interests (Reverse encoding)
OPEN10: is sophisticated in art, music, or literature
Perceived Complexity (PCOM)
PCOM1: If I were to adopt intelligent messaging, it would be complicated to learn.		[135]
PCOM2: If I were to adopt intelligent messaging, it would be frustrating to operate.
PCOM3: I think using intelligent messaging requires a certain amount of time to learn.		[136]
PCOM4: I believe that intelligent messaging is more troublesome to use than other software.
PCOM5: Overall, I do not think intelligent messaging is easy to operate.		[137]
Attitude (ATT)
ATT1: What do you think the function of intelligent messaging is: Bad (1)—Good (7)		[92] [138]
ATT2: How do you like using intelligent messaging to handle affairs? Unfavorable (1)—Favorable (7)
ATT3: How much do you support the large-scale commercial use of intelligent messaging? Low (1)—High (7)
ATT4: In general, what is your attitude towards intelligent messaging? Negative (1) —Positive (7)
Adoption Intention (AI)
AI1: If conditions permit, I would like to adopt intelligent messaging for online business and communication in the future.		[139] [87]
AI2: The likelihood that I will adopt intelligent messaging in the future is very high.	*
AI3: I am interested in adopting intelligent messaging to communicate with others and organizations.
AI4: I would like to be the first to use intelligent messaging.
AI5: I think I can accept intelligent messaging quickly.

Note: * indicates the deletion of the items in pre-test.

. Similar to well-known survey platforms (e.g., Amazon Mechanical Turk), Credamo is a popular online survey platform widely used in China. It provides a reliable crowdsourcing service, allowing researchers to create survey projects that are distributed to and completed by registered users in various regions of China [108,109]. Due to the huge database, Credamo randomly selects samples based on the IP addresses of participants. Each individual in the group has an equal chance of being selected, ensuring the representativeness of the sample [110]. Therefore, in this sense, our questionnaire collection process was a kind of probability sampling, which ensures the randomness of the sample as much as possible.

First, we distributed the original questionnaire via social media (e.g., WeChat) for the pre-test from 21 February to 27 February 2023. As a mini-scale study, pre-testing is designed to test the feasibility of the methodology, data collection process, and data analysis [88]. Similar to previous studies in online learning [67], online medicine [80], and social media [46,87], we obtained convenience samples through social media for pre-testing to evaluate the logical consistency, comprehensibility, sequence of items, and contextual relevance of the questionnaire. We further modified and deleted some items (marked with * in Table A1) to ensure the composite reliability.

After that, in the formal study, we distributed the revised questionnaire widely through the Credamo survey platform from 1 March to 17 March 2023, in order to obtain newly collected probability samples. The convenience samples in the pre-test were not included in the formal study.

We finally collected 715 questionnaires. After excluding invalid samples that (1) did not pass the verification questions and (2) had a lower response time (less than 300 s), 548 valid questionnaires were obtained, which was greater than 10 times the largest number of paths and was considered sufficient for PLS-SEM analysis [111]. The recovery rate was 76.6%. After preliminary screening of the collected questionnaires, we performed the normality test jointly with the Kolmogorov–Smirnov test and Shapiro–Wilk test in IBM SPSS Statistics 19. We found that the significance levels of all items were less than 0.05, which showed the collected data had a skewed distribution. PLS-SEM can be used in a distribution-free case and allows the analysis of skewed data [111]. Therefore, we chose the PLS-SEM method and then analyzed the performance using Smart PLS 3.3.2 software.

4.2. Common Method Bias

We analyzed the common method bias (CMB) by Harman’s single factor test in SPSS and the results showed that CMB did not affect the reliability due to the variance explained by the first factor is 28.771%, which is lower than 40% [112]. Moreover, according to the suggestion of Liang et al. in [113], we used PLS for the verification of common method variance (CMV). Specifically, we included a general method factor in the PLS model, whose indicators include all the principal constructs’ indicators, and calculated each indicator’s variances substantively explained by the principal construct and by the method. As shown in

Table A4. Common method bias analysis.

Construct	Indicator	Substantive Factor Loading R1	$\mathbf{R}{1}^2$	Method Factor Loading R2	$\mathbf{R}{2}^2$
ATT	ATT1	0.7687 ***	0.5909	−0.0837 *	0.007
	ATT2	0.6597 ***	0.4352	0.2466 ***	0.0608
	ATT3	0.7323 ***	0.5362	−0.0579	0.0034
	ATT4	0.8059 ***	0.6495	−0.0847 *	0.0072
AI	AI1	0.7322 ***	0.5362	−0.0242	0.0006
	AI3	0.7088 ***	0.5023	0.0881	0.0078
	AI4	0.7855 ***	0.617	0.0721	0.0052
	AI5	0.7196 ***	0.5179	−0.1444 *	0.0209
PC	PC1	0.7380 ***	0.5446	0.2394 ***	0.0573
	PC2	0.7455 ***	0.5558	−0.0535	0.0029
	PC4	0.7162 ***	0.5129	−0.2005 ***	0.0402
PA	PA1	0.7782 ***	0.6055	−0.0385	0.0015
	PA2	0.7383 ***	0.5451	−0.0561	0.0032
	PA4	0.7115 ***	0.5062	0.0965	0.0093
PRN	PRN1	0.6543 ***	0.4281	−0.0091	0.0001
	PRN2	0.7583 ***	0.575	−0.0208	0.0004
	PRN3	0.6984 ***	0.4878	0.0569	0.0032
	PRN4	0.7440 ***	0.5535	−0.0253	0.0006
PMR	PMR4	0.7918 ***	0.6269	0.0225	0.0005
	PMR5	0.7778 ***	0.605	−0.0231	0.0005
PI	PI1	0.8233 ***	0.6778	−0.0389	0.0015
	PI2	0.7694 ***	0.5921	−0.0573	0.0033
	PI3	0.6530 ***	0.4265	0.0615	0.0038
	PI4	0.7050 ***	0.497	0.0479	0.0023
PSQ	PSQ1	0.6616 ***	0.4377	−0.0833	0.0069
	PSQ2	0.6969 ***	0.4857	−0.0084	0.0001
	PSQ5	0.7748 ***	0.6003	0.0791	0.0063
OPEN	OPEN1	0.7581 ***	0.5747	0.0001	0
	OPEN10	0.7367 ***	0.5428	−0.0112	0.0001
	OPEN3	0.7328 ***	0.537	0.0307	0.0009
	OPEN4	0.6908 ***	0.4773	−0.0551	0.003
	OPEN5	0.7542 ***	0.5688	0.0286	0.0008
	OPEN9	0.7458 ***	0.5562	0.0143	0.0002
	OPEN6	0.6613 ***	0.4374	−0.013	0.0002
PCOM	PCOM1	0.7255 ***	0.5264	0.2226 ***	0.0495
	PCOM2	0.7969 ***	0.635	0.0541	0.0029
	PCOM3	0.7475 ***	0.5587	0.0693	0.0048
	PCOM4	0.7176 ***	0.515	−0.2423 ***	0.0587
	PCOM5	0.6650 ***	0.4422	−0.1090 *	0.0119
Average		0.7329	0.539	−0.0003	0.01

Note: * p < 0.05, *** p < 0.001.

, most method factor loadings are not significant. The average substantively explained variance of the indicators is 0.7329, while the average method-based variance is 0.0100. The ratio of substantial variance to method variance is about 54:1. Due to the low level of significance and small magnitude of method variance, the issue of CMV in this study is not serious.

4.3. Demographic Characteristics

We used screened data for demographic analysis, whose characteristics are shown in

Table 2. The demographic characteristics (N = 548).

Characteristic	Category	Frequency	Percentage
Gender	Male	256	53.28
	Female	292	46.72
Age	18–25 years old	137	25.00
	26–35 years old	292	53.28
	36–45 years old	92	16.79
	46–60 years old	27	4.93
	Over 60 years old	0	0.00
Education	Junior high school and below	0	0.00
	High School	19	3.47
	College	49	8.94
	University undergraduate	400	72.99
	Master’s degree and above	80	14.60
Job	Student	106	19.34
	Professionals	47	8.58
	Government/institutions officer	64	11.68
	Service industry	19	3.47
	Freelancers	10	1.82
	Workers	7	1.28
	Staff	281	51.28
	Other	14	2.55
Area	Eastern China	222	40.51
	Northern China	68	12.41
	the Central of China	82	14.96
	Southern China	91	16.61
	Southwestern China	48	8.76
	Northwestern China	10	1.82
	Northeastern China	27	4.93
Current type of SMS	Text+hyperlink	500	91.24
	Card type	403	73.54
	5G RCS	256	46.72
	Other	1	0.18

. In terms of gender, the ratio of male to female participants in the survey was close to 1:1. In terms of age distribution, the survey had the most participants aged 26 to 35 years old, accounting for 53.28%, followed by participants aged 18 to 25 years old, and no participants over the age of 60 years old. From the perspective of education level, among the participants in this survey, there were no participants who were in junior high school or below, and the largest number of participants comprised those with a bachelor’s degree, accounting for 72.99%. In terms of occupation, the participants were mainly staff, accounting for 51.28%, followed by students, accounting for 19.34%. Participants came from 27 provinces, municipalities, and autonomous regions in China, covering 7 areas across the country, of which Eastern China accounted for the most, accounting for 40.51%, followed by Southern China, accounting for 16.61%.

For the messaging types used, the “text + hyplink” ranked the top with a total of 500 people, accounting for 91.24%. Nearly half of the participants used the emerging form of 5G RCS. In

Table 3. The demographic characteristics of monthly rent (N = 548).

Characteristic	Num.	Min.	Max.	Mean	Median	SD
Monthly rent (CNY)	548	8	600	100.08	85	71.851

, the average monthly phone rent fee of the participants in the past three months ranged from CNY 8 to CNY 600 with a large variation. The average monthly phone bill of each participant was about CNY 100.

5. Data Analysis

5.1. Measurement Model

In this paper, we use the partial least squares structural equation model (PLS-SEM) method to analyze the measurement model via Smart PLS 3 software. As a variance-based method, PLS-SEM will estimate path coefficients to maximize the $R^2$ values of the endogenous constructs [111]. Compared with other methods (e.g., LISREL, AMOS), PLS-SEM requires relatively small sample sizes and fewer measurement scales, which is suitable for exploratory studies and performs well in prediction [114].

According to the Hulland criterion, the factor loading for each problem item is recommended to be larger than 0.5 and preferably greater than 0.7 when performing the PLS-SEM analysis [115]. Consistency reliability can be assessed by Cronbach’s $\alpha$ and composite reliability (CR), which should be at least 0.6 [116] and better above 0.7 [117], respectively. However, Cronbach’s $\alpha$ is sensitive to the number of items in the scale and generally tends to underestimate internal consistency reliability, which may be used as a conservative measure of internal consistency reliability [111]. Due to the limitations of Cronbach’s $\alpha$, it is more appropriate to apply a different internal consistency reliability measure, which is referred to as composite reliability [111]. In this manuscript, we use Cronbach’s $\alpha$ and composite reliability to jointly evaluate consistency reliability.

The average variance extracted value (AVE) aims to test the convergent validity with the threshold of above 0.5 [118], which means that the construct explains more than 50% of the variance of its indicators on average. In addition, the discriminant validity is tested by the Fornell–Larker criterion. The square root of the AVE values needs to be higher than the latent variable correlation [119]. The specific analysis results are as follows.

As shown in

Table 4. Discriminant validity (Fornell–Larcker criterion).

	AINT	ATT	OPEN	PA	PC	PCOM	PI	PMR	PRN	PSQ
AINT	0.737
ATT	0.710	0.742
OPEN	0.368	0.372	0.726
PA	0.602	0.468	0.344	0.742
PC	0.572	0.537	0.411	0.551	0.732
PCOM	−0.503	−0.490	−0.382	−0.408	−0.408	0.725
PI	0.553	0.501	0.449	0.637	0.597	−0.419	0.740
PMR	0.539	0.539	0.325	0.507	0.530	−0.394	0.502	0.785
PRN	0.597	0.520	0.465	0.619	0.595	−0.428	0.646	0.504	0.715
PSQ	0.591	0.553	0.299	0.507	0.514	−0.422	0.436	0.582	0.448	0.712

, the square root of each construct’s AVE is greater than its highest correlation with other constructs, indicating that the discriminant validity of the scale is guaranteed. As presented in

Table 5. Results of reliability and validity.

		Indicator Reliability	Convergent Validity	Consistency Reliability
Construct	Item	Factor Loading	AVE	Cronbach’s Alpha	CR
PC	PC1	0.760	0.536	0.570	0.776
	PC2	0.750
	PC4	0.684
PA	PA1	0.775	0.550	0.594	0.786
	PA2	0.705
	PA4	0.745
PRN	PRN1	0.664	0.511	0.679	0.806
	PRN2	0.753
	PRN3	0.709
	PRN4	0.729
PMR	PMR4	0.802	0.616	0.377	0.762
	PMR5	0.767
PI	PI1	0.806	0.547	0.722	0.828
	PI2	0.751
	PI3	0.674
	PI4	0.722
PSQ	PSQ1	0.649	0.507	0.513	0.755
	PSQ2	0.721
	PSQ5	0.762
OPEN	OPEN1	0.765	0.527	0.850	0.886
	OPEN3	0.749
	OPEN4	0.688
	OPEN5	0.748
	OPEN6	0.669
	OPEN9	0.743
	OPEN10	0.715
PCOM	PCOM1	0.675	0.526	0.782	0.847
	PCOM2	0.749
	PCOM3	0.706
	PCOM4	0.783
	PCOM5	0.710
ATT	ATT1	0.751	0.551	0.728	0.830
	ATT2	0.699
	ATT3	0.721
	ATT4	0.795
AINT	AINT1	0.737	0.543	0.719	0.826
	AINT3	0.715
	AINT4	0.789
	AINT5	0.703

, the factor loading of each item among all constructs is above 0.6, varying between 0.649 and 0.806. The remaining items are higher than 0.7, except for ATT2, OPEN4, OPEN6, PC4, PCOM1, PI3, PRN1, and PSQ1. Overall, each construct has a good factor loading and the indicator reliability satisfied the requirements. Although the Cronbach’s $\alpha$ coefficients of four constructs (PC, PA, PMR, PSQ) are less than 0.6, the composite reliability of all constructs is greater than 0.7, which is more important [111]. Therefore, in general, the consistency reliability of each construct is acceptable. In addition, the AVE values of each construct are higher than 0.5. All in all, the reliability and validity of the scale are relatively satisfactory, reflecting the accuracy and reliability of this study to a certain extent.

5.2. Structural Equation Model

To examine the results of the hypothesis, we perform bootstrap and two-tail tests (significance level = 5%) using 5000 samples to obtain the significance level of the path coefficients. As shown in

Table 6. The summary results of hypothesis testing.

Hypotheses	Path	Path Coefficient (β)	T-Statistics	${\mathit{f}}^2$	p-Value	Result
H1	ATT -> AINT	0.388	6.321	0.217	0.000	Supported
H2a	PC -> ATT	0.158	2.423	0.023	0.015	Supported
H2b	PC -> AINT	0.063	1.254	0.005	0.210	Not supported
H3a	PA -> ATT	−0.003	0.037	0.000	0.970	Not supported
H3b	PA -> AINT	0.180	2.867	0.042	0.004	Supported
H4a	PRN -> ATT	0.155	2.506	0.021	0.012	Supported
H4b	PRN -> AINT	0.129	2.211	0.020	0.027	Supported
H5a	PMR -> ATT	0.176	3.392	0.031	0.001	Supported
H5b	PMR -> AINT	0.015	0.345	0.000	0.730	Not supported
H6a	PI -> ATT	0.109	1.744	0.010	0.081	Not supported
H6b	PI -> AINT	0.023	0.423	0.001	0.672	Not supported
H7a	PSQ -> ATT	0.253	4.763	0.067	0.000	Supported
H7b	PSQ -> AINT	0.144	3.799	0.031	0.000	Supported
H8	OPEN -> AINT	−0.016	0.387	0.001	0.699	Not supported
H9	PCOM -> AINT	−0.089	2.183	0.015	0.029	Supported
Indirect Path		Path Coefficient (β)	Bca [2.5%, 97.5%]		T-Statistics	p-Value
PRN -> ATT -> AINT		0.060	[0.012, 0.121]		2.150	0.032
PSQ -> ATT -> AINT		0.098	[0.048, 0.148]		3.854	0.000
PC -> ATT -> AINT		0.061	[0.012, 0.116]		2.317	0.021
PMR -> ATT -> AINT		0.068	[0.023, 0.113]		3.007	0.003

, in addition to H6 and H8, hypotheses H1, H4, H7, and H9 are fully supported, and H2, H3, and H5 are partially supported. Out of 15 hypotheses, a total of 9 hypotheses are established. According to the path coefficient and significance level, it is not difficult to find that ATT ($\beta$ = 0.388, p < 0.001), PA ($\beta$ = 0.180, p < 0.01), PRN ($\beta$ = 0.129, p < 0.05), and PSQ ($\beta$ = 0.144, p < 0.001) have a positive impact on AINT, while PCOM ($\beta$ = −0.089, p < 0.05) negatively and significantly affects AINT. In addition, PC ($\beta$ = 0.158, p < 0.05), PRN ($\beta$ = 0.155, p < 0.05), PMR ($\beta$ = 0.176, p < 0.001), and PSQ ($\beta$ = 0.253, p < 0.001) positively and significantly affect ATT.

Notably, PC ($\beta$ = 0.061, p < 0.05) and PMR ($\beta$ = 0.068, p < 0.01) have an indirect effect on AINT via ATT, though they not directly affect AINT. PRN ($\beta$ = 0.060, p < 0.05) and PSQ ($\beta$ = 0.098, p < 0.001) also have an indirect effect on AINT through ATT, not only have a direct impact on AINT. The overview of the hypothesis structural models is shown in Figure 2.

The effect size $f^2$ is used to measure the effect of the corresponding path independent variable on the dependent variable. According to the threshold values of $f^2$ from Chin [120], we find that ATT has a medium effect on AINT (0.15 < $f^2$ < 0.35), while PA, PRN, and PSQ have weak effect on AINT (0.02 < $f^2$ < 0.15), and PCOM has an even weaker effect on AINT ($f^2$ < 0.02). In addition, PC, PRN, PMR and PSQ have weak effect on ATT (0.02 < $f^2$ < 0.15).

From

Table 7. R² and Q².

Variables	R2	Adjusted R2	Q2
AINT	0.641	0.635	0.335
ATT	0.454	0.448	0.238

, we assess the quality of the proposed model by R² and Q², which show the explanatory and predictive effects. Overall, our model predicts 63.5% of the variance of AINT, which is between 0.333 and 0.670, indicating a near substantial explanatory power [121]. In addition, the Q² values need to be higher than 0, which indicates that the exogenous constructs have predictive relevance for the endogenous construct under consideration [119,120]. Obviously, Q² values of the two endogenous latent variables satisfy the requirements. The parameters of model fit such as normed fit index (NFI), root mean square residual covariance ($\mathrm{RMS}\_\mathrm{theta}$), d_G and standardized root mean square residual (SRMR) as presented in

Table 8. Model fit summary.

Index	NFI	RMS_Theta	d_G	SRMR
Value	0.655	0.117	0.93	0.069
Threshold	0.9 < NFI < 1	<0.95	<0.95	<0.08

, which showed good fitness of the model.

6. Discussion

In general, it can be seen that the proposed model has good explanatory power, except for hypotheses H6 and H8. Hypotheses H1, H4, H7, and H9 are completely supported. Hypotheses H2, H3, and H5 are partially supported. For RQ1, our results show that ATT, PA, PRN, and PSQ have significant positive impacts on AINT, while PCOM has a negative direct effect on AINT. Although the effect of PC and PMR on AINT is not significant, they have indirect effects on AINT through PC/PMR -> ATT -> AINT. Notably, PI and OPEN have no effect on AINT. In a word, ATT, PC, PA, PRN, PMR, PSQ, and PCOM all affect AINT directly or indirectly.

From the perspective of extrinsic motivation, PSQ is the most significant extrinsic motivation factor affecting AINT, which both directly and positively influences AINT and indirectly affects AINT through ATT. PMR also has no direct effect on adoption intention but indirectly affects AINT through ATT. The result that PI has no direct or indirect effect on AINT is somewhat surprising to us. The reasons can be explained as follows: while intelligent messaging is an upgrade from traditional SMS, its interactive method is similar to OTT produces (e.g., instant messaging apps, chatbot-enabled Internet platform accounts). The interactive speed of intelligent messaging depends on the quality of the mobile network. Considering functional maturity, some pre-commercialized intelligent messaging accounts have a longer response time and less content compared with OTT products, and it is confirmed by Kim et al. that timeliness is an important factor in PI [81]. In addition, the performance of a chatbot function is homogeneous with other customer service bots in OTT products. Users may not perceive intelligent messaging as significantly different from existing OTT services. Therefore, participants’ AINT and ATT of intelligent messaging are not affected by PI. In the future, the system and service quality of intelligent messaging service should be improved to facilitate PI and then an innovative interaction method needs to be adopted (e.g., voice/gesture-control) in this all-in-one platform.

As for the perspective of intrinsic motivation, compared with the factors of extrinsic motivation, the hypothesis of intrinsic motivation is better supported. Similar to the previous study [62], PA and PRN have direct positive impacts on AINT. PRN has an indirect effect on AINT through ATT, which is similar to [63]. PC also has an indirect effect on AINT through ATT. In conclusion, all factors of intrinsic motivation have some effects on AINT. However, in extrinsic motivation, only PMR and PSQ have a direct or indirect positive influence on AINT, but PI has no effect on AINT. Therefore, we have reason to believe that the effect of intrinsic motivation on AINT is more significant, which is satisfied with both the expectation of self-determination theory and the orientation of audience-oriented research.

Similar to previous studies in the field of education [122], openness has no effect on AINT. The reason why OPEN is not significant is similar to the explanation for PI. Except for the acquisition channel, the service type and content provided by the intelligent messaging itself are less different from OTT products and mobile Internet platform service, the curiosity of mobile users may not be inspired before or after experiencing intelligent messaging services, which results in insufficient novelty perception by mobile users. Although the intelligent messaging service is the emerging form of SMS with attractive features, the users would not like to seek new experiences. Therefore, OPEN has no significant effect on AINT in this study. In the future, if intelligent messaging can break through the current service model and allow a user experience different from other OTT products, OPEN may have a positive impact on AINT.

Notably, although the impact of PCOM on AINT is negative, the effect is minimal ($f^2$< 0.02), which is also related to the characteristics of intelligent messaging itself. As an upgrade service for traditional SMS, intelligent messaging inherits the advantages of simple operation and ubiquitous access to traditional SMS. At the same time, differing from other online services [105,106], this app-free application greatly reduces the threshold and operation difficulty for mobile users. However, it cannot be ignored that, in the initial stage of user experience, due to unfamiliar operations and network delays, mobile users feel that there is a certain degree of complexity in intelligent messaging. In general, PCOM negatively affects the AINT of intelligent messaging, but its impact is relatively weak.

In addition to promoting the quality of services, the internal needs of mobile users need to be focused on to further improve the autonomy and relatedness in the usage of intelligent messaging. Letting mobile users perceive the ability to use intelligent messaging is also important. It also requires researchers to combine technical means with the inherent needs of mobile users. By improving extrinsic motivations such as technology, product form, and software quality, the intrinsic motivations of mobile users will be gradually stimulated, so that mobile users would like to adopt intelligent messaging.

6.1. Theoretical Implications

First, instead of traditional theoretical models (e.g., TAM and UTAUT), this study adopts SDT from a psychological perspective to study the adoption intentions for new applications. In addition, our study also supplements the empirical research of SDT on the intention to adopt new technologies and applications. Through empirical research, we found that the number of supported hypotheses in intrinsic motivation (H2a, H3b, H4a, and H4b) is larger than that in extrinsic motivation (H5a, H7a, H7b), i.e., the intrinsic motivation has a relatively larger impact on mobile users’ adoption intentions than the extrinsic motivation, which indicates the effectiveness of the self-determination theory. This indicates that intrinsic motivation, perceived competence, perceived autonomy, and perceived relatedness play a necessary role in optimal development [38]. Second, based on the characteristics of intelligent messaging and interviews, we summarize the proper constructs to concretize the extrinsic motivations in SDT, including perceived media richness, perceived interactivity, and perceived system quality, which contribute to making the SDT explicit. This study found that, in addition to perceived interactivity, the other two factors of extrinsic motivations have a direct or indirect impact on AINT, which can provide a reference for similar research. Finally, this study further complements and perfects the connotation of self-determination theory by introducing two extended variables: openness to experience in personality trait theory and perceived complexity in innovation diffusion theory. This study found that perceived complexity negatively affected AINT, but its effect was small or even negligible. In addition, although openness did not affect AINT in this study, similar studies in the future can examine the influence of other personality traits on adoption intention on the basis of this study in order to continuously enrich the connotation of self-determination theory.

6.2. Practical Implications

Firstly, this study revealed that intrinsic motivation has a greater impact on mobile users’ adoption intentions than extrinsic motivation. As a new media application, the service provider needs to focus on the inherent interest and satisfaction of mobile users rather than mandated learning and instruction. It would be better to integrate more personalized AI-driven services (e.g., multi-modal interaction, AI assistants, and large language model applications) in this platform, which could stimulate users’ intrinsic motivation and then let them spontaneously use intelligent messaging. Secondly, extrinsic motivation such as system quality should be considered to better depend on the advantages of the technology characteristic. In addition, the system quality needs to be further improved to achieve high media richness. The media ecosystem of an intelligent messaging service could be enriched with sufficient content providers and institutional users, which aim to improve the perceived attitudes among mobile users. Thirdly, service providers also need to create product differentiation compared with the existing mobile applications to achieve a higher level of interactive experience. The operational complexity of intelligent messaging should be reduced as much as possible to complete the last mile in usage. Finally, this research helps to enhance the “user-centered" awareness for the technical developers and service providers, which can further accelerate the large-scale popularization and development of intelligent messaging. Moreover, it also contributes to the construction of an omni-channel media environment.

6.3. Limitations

First, the age of the participants was mainly concentrated between 18 and 45, while participants over the age of 45 years only accounted for 4.93%. No participants came from the over-60-year-old group. Although intelligent messaging has been upgraded by adopting new technologies, it still uses the entrance of traditional SMS; as such, its attributes of convenience and benefit to the people have not undergone essential changes. Especially for people over the age of 45, SMS has been a familiar communication tool since they were young. Therefore, we believe that people over the age of 45 are more likely to psychologically accept the upgraded form of traditional SMS, which does not require the installation of mobile apps and familiarization with their operation steps. Therefore, it is more representative to investigate this group’s adoption intention. Second, the participants were mainly from economically developed regions, such as Guangdong in Southern China and Shandong and Jiangsu in Eastern China. However, economically underdeveloped areas were less covered, such as Northwest China and Northeast China. Third, as intelligent messaging is still in the process of commercialization and industrialization, its user interface or presentation form will be updated. In addition, the stimulus materials (text and video) from tele-operator advertising (or other popular science articles/videos) displayed in the survey may not have completely simulated the mobile users’ experience. For participants who have not used 5G RCS, their feelings may only be superficial and lack profound practical experience. In future work, we plan to conduct long-term participatory observation among intelligent messaging users to explore the impact of external environmental factors (e.g., personal economic status) on the intention to adopt intelligent messaging.

7. Conclusions

In this study, based on self-determination theory, we have investigated the psychological factors and their relationships in the context of mobile users’ intelligent messaging service adoption intention in China. Except for PI and OPEN, other proposed constructs were found to directly or indirectly affect the adoption intention of intelligent messaging. ATT, PA, PRN, and PSQ have significant positive impacts on AINT of intelligent messaging, while PCOM has a negative direct impact on AINT. Among them, ATT has the greatest effect on AINT, followed by PA. In addition, although PC and PMR have insignificant effects on AINT, there exist indirect paths of PC -> ATT -> AINT and PMR -> ATT -> AINT. Overall, the effect of intrinsic motivation is greater than that of extrinsic motivation, which is consistent with self-determination theory. According to the self-determination theory, it is not difficult to conclude that mobile users can continuously convert external conditions into extrinsic motivations and then internalize extrinsic motivations as intrinsic motivations when using intelligent messaging because they will endorse its value or worth [32]. This will further satisfy their basic psychological needs (perceived competence, perceived autonomy, and perceived relatedness). During this internalization process, mobile users acquire intrinsic motivations, which drive them to adopt intelligent messaging actively. Moreover, it is necessary to improve external technical conditions to meet the intrinsic motivation of mobile users, thereby accelerating the widespread popularization of intelligent messaging.

In terms of future work, the investigation of specific AI-empowered scenarios and AI messaging platforms can be considered after full commercialization with a larger penetration rate and active users. For example, the mobile users’ attitudes and intentions towards governmental and financial services could be assessed, which could provide relevant feedback and promote concrete industrial applications. For elderly users, a friendly user interface and multi-modal interactive means should be provided in order to improve the usability of intelligent messaging services by service providers (e.g., easier access through a simplified interface hierarchy and a large print version). Furthermore, an age-appropriate service must be considered by industrial/governmental institutions in order to obtain higher coverage among mobile users after full commercialization. In addition, we plan to assess long-term participatory observations among intelligent messaging users (especially elderly users) to reveal the impact of external environmental factors (e.g., personal economic status) and explore other personality traits (e.g., neuroticism, extraversion, agreeableness, and conscientiousness). More importantly, attention should be paid to the cross-cultural differences and usage behavior among mobile users in the North American, Asia-Pacific, and European regions.