Introduction

The integration of innovative technologies in everyday life is becoming more prevalent, with rapid technological advancements affecting various aspects of people’s lives. With the advent of new technologies, the world is transitioning into an era characterized by technology, intelligence, and automation. Particularly in the context of artificial intelligence within the automotive industry, autonomous vehicles (AVs) may not only result in changes in traffic regulations1 but also reshape relationships with other sectors, including insurance, healthcare, construction, transportation, and energy2.

An evident illustration of this is the substantial investment allocated to AVs. In 2019, AVs attracted the most significant investment3, with market projections predicting a valuation of $235.393 billion by 20324. Autonomous driving technology is classified into six levels, which range from level 0 (no automation) to level 5 (full automation)5. Although the transfer of driving functions from human drivers to vehicles6 has generated significant attention, the reception of this technology remains mixed.

Notably, John Krafcik has contended that attaining complete autonomy in driving is unlikely7 and perceives this technological shift as a series of incremental improvements rather than a transformative leap8. Additionally, public perception varies, as individuals in developed countries and higher-income groups tend to articulate more pessimistic and critical views regarding the adoption of AVs9,10,11.

Research has demonstrated that more than 50% of surveyed individuals prefer AVs to those driven by humans12. When deployed in shared usage models, AVs are anticipated to reduce private vehicle ownership by up to 20%13, with each AV potentially replacing over ten conventional vehicles14,15,16.

Prior studies emphasize that public acceptance and attitudes are essential for the extensive adoption of AVs17, with individuals who have prior experience with this technology exhibiting more positive attitudes12,18. As research on AVs advances, a growing number of factors influencing their success has been identified. These factors include message framing19, perceived safety20, environmental issues21, ethical framework22, personality traits23, media environment24,25, and both positive and negative information26.

Because AVs can enter urban road testing, the United States and China completed the commercial operation of autonomous car taxis in 2020 and 2021, respectively27,28. This indicates the potential for the complete integration of AVs into the market, supplanting traditional vehicles. Therefore, exploring potential measures to encourage consumers to purchase AVs is crucial.

Previous research on consumer adoption intentions has widely utilized the Technology Acceptance Model (TAM). However, the TAM primarily assesses the impact of a product’s utility and convenience on adoption intentions. To address TAM’s limitations, research has evolved to include models that analyze consumer states, such as self-efficacy and subjective norms. However, these models frequently neglect the environmental factors, such as electronic word-of-mouth (eWOM), that influence the consumer’s state.

Social cognitive theory and the stimulus-organism-response (SOR) theory elucidate that the environment can trigger actions. Social cognitive theory posits that the environment can induce people’s behavioral outcomes. Meanwhile, SOR theory suggests that external environments (stimuli) affect an individual’s psychological state, triggering specific responses that result in changes in resultant behaviors29,30. An individual’s internal emotions or behaviors can be perceived as reactions to external environmental factors31,32. In other words, SOR framework emphasized the influence of external environmental stimuli on individuals, while social cognitive theory paid more attention to individuals’ internal cognition, attitudes, and beliefs.

The SOR theory is extensively used in marketing and consumer behavior research to explain how environmental factors influence the internal state of consumers and elicit specific behavioral responses29,30. In the context of AVs, external stimuli such as media coverage, news of technological advancements, or eWOM) from other users can affect consumers’ emotional and cognitive responses to autonomous technology, thereby influencing their purchase decisions33.

Social cognitive theory underscores the significance of individuals’ beliefs, attitudes, and self-efficacy in influencing behaviors. Within the AV domain, a consumer’s belief in their ability to effectively utilize this technology and their perceptions of its safety crucially shape their acceptance levels34.

However, outside of the AVs space35, reported in 2017 that how theoretical structures based on social cognitive theory and stimulus-response (S-R) explain students’ engagement, prerequisites, and learning performance. In addition36, integrated social cognitive theory (SCT), stimulation-organ-response model (SOR), and planned behavior theory (TPB) together and applied them to ecotourism research36. ’s study validated the validity of these three important integrated forms of social psychological behavioral theories and provided a new perspective for understanding behavioral motivations in ecotourism. Furthermore37, innovatively proposed a conceptual model that includes psychosocial cognition and mental activity processes by combining SCCT, which is an extended model of social cognitive theory (SCT), with the SOR model. They found that adding the SOR model can more rigorously explain the evolution process of students’ internal psychological cognition in the learning environment when receiving external stimulus and revealed its role in improving learning satisfaction.

Although the SOR theory has been acknowledged by scholars such as Dzandu, Pathak and Gulliver33, Cho, Jeong and Bae38, and Shuo39 in the context of consumer acceptance and willingness to utilize Avs. However, there are still few researches on AVs adoption intention based on SOR theory and social cognition theory. If the SOR framework explains how external circumstances trigger behavior by influencing an individual’s emotional and cognitive states, then social cognitive theory complements how individuals form beliefs about behavior through observation, learning, and self-assessment. Therefore, combining environmental stimuli with personal cognition can better explain the influence of the external environment on an individual’s mental state, as well as the interaction between an individual’s internal beliefs, attitudes, and behaviors. Consequently, this study aims to integrate the SOR theory with social cognitive theory to assess the direct impact of external environments on behavior and evaluate the influence of consumers’ internal belief systems on consumer behavior. This approach aims to comprehensively elucidate the interplay between environmental stimuli and personal cognitive responses in shaping consumer actions.

This research aims to empirically analyze the factors affecting consumers intentions to adopt AVs through a three-tier theoretical model that encompasses external stimulus perceptual (human and product) and behavioral (adoption intentions) factors in China. Figure 1 is a conceptual model based on stimulus-organism response and social cognitive theory. Specifically, it aims to empirically examine how environmental factors (eWOM and Mass Media), human factors (Self-efficacy and Subjective Norm), and product factors (Trust) indirectly or directly influence consumer willingness to adopt AVs. Consequently, it provides a more comprehensive understanding of the dynamics of adopting AVs. This approach offers valuable insights into the complex relationship between media influence and consumer behavior in the context of AV adoption.

Fig. 1
figure 1

Conceptual model based on Stimulus-Organism Response (SOR) and social cognitive theory.

Full size image

Literature review

As the commercial operation of AVs taxis becomes normalized and the potential for full marketization approaches, it is increasingly crucial to determine the pre- preconditions influencing consumers’ decisions to purchase AVs. Environmental issues have been confirmed to stimulate the adoption of autonomous electric vehicles21. Additionally, ethical frameworks, such as utilitarianism, deontology, relativism, absolutism (monism), and pluralism, are critical to the acceptance of AVs22. Several studies have examined the influence of both positive and negative information on consumers’ willingness to adopt AVs26.

Moreover, personality traits—such as openness, neuroticism, conscientiousness, agreeableness, and extraversion—and trait anxiety have been shown to affect consumer perceptions of AVs23. When focusing on millennials, a demographic emblematic of youth, perceived personal and social benefits, as well as perceived safety, emerge as core determinants of whether they adopt or reject AVs20.

Table 1 presents a comprehensive summary of research on AVs conducted from 2014 to the present. The table outlines each study’s theoretical foundations, research methodologies, and key variables, offering an overview of the core themes and findings within the field of AV research.

Table 1 Related research.
Full size table

The full commercialization of AV taxi services in China and the United States has enabled consumers to access information about AVs not only through traditional mass media but also from online sources, such as shared experiences and reviews from active users, as well as content from social media influencers and bloggers. Considering the significant influence of both mass media and eWOM in shaping consumer perceptions, this study proposes a media-based model to comprehend the adoption of AVs.

Social cognitive theory highlights the critical role of individual and product-related cognition in shaping behavior. It suggests that human behavior is influenced by the dynamic interaction between the individual, the environment, and the product34. This theory is widely used to analyze and predict human behaviors and decision-making patterns, positioning it as a key theoretical framework in academic research. In the field of AVs24, employed social cognitive theory as the foundational framework to examine the interaction between environmental influences, personal cognition, and behaviors. Building upon this approach54, added media exposure and self-efficacy elements to the social cognitive theory framework to assess what drives consumers to adopt autonomous vehicles (AVs).

55 observed that individuals exposed to extensive information on AVs from both social and traditional media are likely to cultivate a heightened belief in their ability to effectively utilize these technologies. In a different context25, distinguished between the impacts of social and traditional media by creating a media-based perceptions and adoption model, with the aim of deepening the understanding of how various media types affect consumer perceptions and the adoption process of AVs.

Both the social learning theory (SLT)56 and the theory of planned behavior (TPB)57 emphasize the significance of self-efficacy and subjective norms. Self-efficacy is an individual’s confidence in their ability to execute a specific task58, whereas subjective norms pertain to an individual’s social pressures that influence their behavior59. Moreover, trust is a critical element, especially in the acceptance of innovative technologies, as a lack of trust can significantly impede their widespread adoption60,61. Given these considerations, self-efficacy, subjective norms, and trust are key components in the proposed model for predicting consumer readiness to adopt AVs.

Building on the SOR framework and social cognitive theory, this study proposes the following research model (as shown in Fig. 2):

Fig. 2
figure 2

Model’s constructs and hypotheses.

Full size image

Research hypotheses

Environmental factors: mass media and eWOM

Mass media, which encompass mediums such as film, radio, recorded music, and television, disseminates information to extensive audiences through written, broadcast, or spoken channels62,63. eWOM) refers to the sharing of information and experiences regarding products or services through electronic means, such as social media, forums, and blogs. This form of word-of-mouth extends beyond formal review systems to include all forms of electronic recommendations, discussions, communications, sharing, and dissemination activities64.

Consumers frequently use social media and online reviews to guide purchasing decisions65. Studies have indicated that mass media significantly enhances self-efficacy24,25, with both social and conventional media positively influencing this psychological trait55.

Mass media has demonstrated efficacy in enhancing the self-efficacy of individuals operating AVs25. Albayrak and Ceylan66 conducted a meta-analysis to confirm that established 16 of the 21 online word-of-mouth factors as significantly influential on purchase intentions, among which self-efficacy is included.

Hypothesis H1: Mass media positively affect self-efficacy.

Hypothesis H1*: eWOM positively affect self-efficacy.

Subjective norms significantly influence public behavior and attitudes regarding emerging technologies, such as AVs24,25. Research by Du et al.24 and Zhu et al.25 confirms mass media’s and social media’s significant impact on subjective norms. Further, Lee et al.55 investigated how traditional media and social media influence the willingness to adopt AVs. Consequently, they discovered that both traditional media and social media positively influence subjective norms.​ ​ In addition, Mao & Lyu67 confirmed the relationship between subjective norms and online word-of-mouth. They indicated a positive correlation among the two, a finding that reinforces the earlier findings of Gao et al.68.

Additionally, research indicates that eWOM significantly enhances subjective norms67, while the influence of mass media on these norms has been consistently affirmed24,25. These insights collectively highlight the critical influence of media in modulating subjective norms; the subsequent assumptions pertain to their interrelationship:

Hypothesis H2: Mass media positively impact subjective norms.

Hypothesis H2*: eWOM positively impacts subjective norms.

Media reports, especially when unfavorable, significantly influence public trust and perceptions of safety concerning AVs69. Trust issues in AVs can be effectively addressed by disseminating detailed, concrete information70. Research demonstrates that various media channels exert distinct effects on trust; social media tends to diminish trust, whereas traditional media generally enhances it55. Additionally, mass media significantly influences consumer trust71. The credibility of eWOM is also crucial in determining the public’s willingness to adopt new technologies, including AVs72. Notably, mass media significantly enhances trust concerning technology and car usage71, underscoring the essential role of trust in the context of AVs42,44,46,48. This observation forms the basis for the following hypothesis:

Hypothesis H3: Mass media positively affect trust.

Hypothesis H3* eWOM positively affect trust.

The integration of traditional media and new social media marketing has become a fundamental course for enterprises72. The importance of word-of-mouth in consumer behavior is evident73,74. Dichter75 described it as an informal mode of communication. The proliferation of the internet has led to the emergence of online word-of-mouth76. Moreover, researchers have noted that online word-of-mouth is more effective than its traditional counterpart77. Komthong et al.78 advocate for prioritizing online word-of-mouth as a principal marketing strategy.

The positive or negative bias propagated by mass media significantly influences public perception and behavior regarding AVs71. Studies have established that various communication channels influence the intention to adopt technologies such as mobile banking in unique ways62. Mass media frequently exerts a greater influence than traditional media54.

Alnoor et al.79 expanded the scope of eWOM within the e-commerce sector by integrating perceptions of chaos from both positive and negative reviews. This integration enhances the understanding of eWOM dynamics. Similarly, Shankar et al.80 discovered that positive eWOM, characterized by assertions of argument quality, valence, and consistency, significantly boosts consumer willingness to utilize mobile banking services.

Further research by Du et al.54 highlighted that both social and traditional media influence the adoption intentions of AVs via perceived value, with social media having a more substantial impact. Additional studies have demonstrated that both mass media and word-of-mouth effectively promote travel intentions81,55. The survey revealed that both social and traditional media positively influenced people’s willingness to adopt. Based on this, their relationship can be interpreted as follows:

Hypothesis H4: Mass media positively affects adoption intention.

Hypothesis H4*: eWOM positively affects adoption intention.

Personal factors: Self-efficacy and subjective norms

Subjective norms, as defined by Ajzen82 and Ajzen & Fishbein83, reflect an individual’s perception of the societal pressures related to performing specific actions, encapsulating the impact of widely held opinions on personal perceptions59. Consumer decisions are frequently influenced by group opinions and suggestions, although the information guiding these decisions may carry biases based on diverse requirements84. Furthermore, the extent of an individual’s behavior and trust can be influenced by the impact of these subjective norms61.

Prior research robustly indicates that subjective norms substantially influence various dependent variables85,86. Self-efficacy emerges as a pivotal third predictor of intentions, enhancing the predictive power of attitudes and subjective norms87. Self-efficacy and subjective norms are essential factors influencing public acceptance of AVs54,88. Furthermore, the interplay between entrepreneurship education and subjective norms significantly enhances students’ self-efficacy89.

A study on mobile learning discovered that self-efficacy accounts for 39.7% of the variation in subjective norms, with a significantly robust relationship evidenced by a path coefficient (β = 0.630) and high statistical significance (t = 13.751) and effect size (f2 = 0.658)90. These findings indicate a profound interconnectedness between self-efficacy and subjective norms, reinforcing their role in behavioral predictions. Therefore, the following hypotheses may be proposed concerning the relationship:

Hypothesis H5, H5*: Self-efficacy positively affects subjective norms.

Product factor: trust

The opinions and recommendations of social circles significantly influence consumer decision-making84. Individuals with high self-efficacy are more likely to trust AVs because they believe they can safely utilize such technologies91. Moreover, significant gender differences exist in the levels of self-efficacy and the impact of subjective norms; males typically exhibit higher self-efficacy92, whereas females are more influenced by subjective norms93. The interplay between self-efficacy, subjective norms, and trust positively correlates, thus jointly impacting trust24,55,71.

According to research, self-efficacy has been established as a critical determinant in the relationship between driver-assistive technology usage and trust in such systems94. Increased usage and experiential interactions with AVs enhance public trust71,95. Consequently, a nuanced understanding of these dynamics is crucial. Consequently, their relationship may be characterized as follows:

Hypothesis H6, H6*: Self-efficacy positively affects trust.

Hypothesis H7, H7*: Subjective norms positively affect trust.

Behavioral factors: adoption intention

Self-efficacy in the AV sector is conceptualized as individuals’ confidence and capability to address unforeseen challenges encountered during the adoption and operation of AV technology96. This construct is operationalized according to the definition provided by96, highlighting the critical role of self-efficacy in navigating the complexities of AV technology utilization.

Empirical evidence indicates that self-efficacy significantly influences public adoption intentions toward AVs. A notable study by Zhu et al.25 quantifies this impact, demonstrating a significant effect size (β = 0.31, t = 5.58) directly linked to self-efficacy. Supporting this finding, research by Lee et al.55 and Du et al.24 further corroborates the positive correlation between self-efficacy and the willingness to adopt AV technology.

Prior research has extensively examined relationship between self-efficacy and various forms of behavioral intention24,55,97,98. For instance90, reports that the influence of self-efficacy on mobile learning behavior intention accounts for a substantial 55.9% of the behavioral outcome. This statistic underscores the pervasive impact of self-efficacy across various technological domains.

To extend this line of inquiry, Seuwou99 hypothesizes and subsequently validates a positive linkage between self-efficacy and the intention to use AVs. The data from Seuwou’s study reinforce the hypothesis, confirming a robust positive correlation between self-efficacy and usage intentions.

Subjective norms, particularly those stemming from family, friends, and colleagues, are crucial in shaping adoption intentions toward AVs100. Research by Yuen et al.88 highlights the pivotal role of these social pressures, whereas Wishart et al.100 assert that the influence exerted of family and friends is not significantly different from of colleagues; both significantly influence adoption intentions.

Direct impacts of subjective norms on the adoption intentions of AV technology have been documented25. Studies such as those by24,25,55 provide robust evidence that subjective norms directly and positively influence intentions to adopt AVs. Moreover, the relationship between subjective norms and the intention to use AVs extends beyond mere adoption. Seuwou99 and Herrenkind et al.101 have documented a consistent positive correlation between subjective norms and usage intentions. In addition, in cultural contexts such as China and Southeast Asia, familial and peer influences are pivotal in millennials’ automotive decisions, particularly regarding environmentally friendly vehicles102,103,104.

Trust is undeniably significant48. Trust in technology, particularly in AVs, emerges as a critical determinant of behavioral intentions71. Increased trust correlates with a higher likelihood of technology adoption47,105. This pattern applies to specific applications, such as autonomous buses, where increased trust correlates with greater willingness to utilize the technology101.

Given these insights, it is posited that self-efficacy and subjective norms significantly interact with trust to influence the adoption and use of AV technologies. Therefore, we have justifiable reasons to assume that their relationship are as follows:

Hypothesis H8, H8*: Self-efficacy positively affects adoption intention.

Hypothesis H9, H9*: Subjective norms positively affect adoption intention.

Hypothesis H10, 10*: Trust positively affects adoption intention.

Methodology

All experiments in this study strictly follow ethical and scientific norms to ensure that the rights and privacy of the subjects are fully protected. The methodology of this study was developed in accordance with the Declaration of Helsinki and all relevant international, national, and institutional requirements and was approved by the Ethics Committee of Dongguk University (approval number: [2024-04-24]). All participants signed a detailed informed consent form online. Participants voluntarily participate in this study after being fully aware of the study purpose, procedure, possible risks. All data will be kept strictly confidential, only used for scientific research analysis and will never be disclosed to third parties.

This study employs a structured questionnaire to facilitate large-scale data collection on individual attitudes and reported intentions, capitalizing on the advantages of questionnaire surveys in gathering a broad spectrum of consumer opinions and providing quantitative data. The primary research subjects in this study are consumers, who are pivotal for validating the research hypotheses. Consequently, the study adopts a quantitative research methodology, employing a structured questionnaire to systematically gather data from Chinese consumers.

Each item within the questionnaire was evaluated using a Likert scale, enabling a quantitative analysis of consumer attitudes and perceptions. The sample includes residents of various ages and diverse educational backgrounds from multiple Chinese cities.

Regarding analytical methods, the data requirements for Covariance-Based Structural Equation Modeling (CB-SEM) are more stringent than those of Partial Least Squares Structural Equation Modeling (PLS-SEM), which is more accommodating; however, both methods tend to yield comparable results106. Factor-based models are more apt for CB-SEM, whereas composite-based models are preferable for PLS-SEM106.

The empirical analysis through PLS-SEM aims to optimally delineate and predict the structural relationships among dependent variables. This method aims to minimize the dispersion of explanations for dependent variables by adjusting model parameters.

Empirical analysis using PLS-SEM can explore, optimally explain, and predict the structural relationships of dependent variables. PLS-SEM minimizes the explanatory dispersion of dependent variables by adjusting model parameters. It possesses the benefits of handling small samples, complex models, and non-conventional data distributions107. The PLS-SEM approach focuses on explaining and predicting endogenous latent variables. It estimates path coefficients to maximize explanatory power (R2) by minimizing the error terms of the endogenous latent variables107.

Simultaneously, the research question focuses on identifying the effect of different types of media directly and indirectly on the adoption of AVs, a multifaceted issue involving psychological states, products, and intent. PLS-SEM has strong predictive power and can maximize the variance of endogenous underlying variables. This is crucial for comprehending the mechanisms by which mass media and eWOM influence consumer psychological states (self-efficacy and subjective norms), products (trust), and adoption intentions. Therefore, considering the above factors, this study uses PLS-SEM for empirical analysis.

Measurements

To ensure the precision and relevance of the research, the investigators meticulously developed a theoretical model underpinned by prior studies. They crafted a questionnaire informed by previous empirical findings and the perceived gaps in existing knowledge. The questionnaire used in the study included 37 questions: there are 9 demographic questions and 6 variables (28 rating scales) using a 7-point Likert scale, where 1 = “strongly disagree” to 7 = “strongly agree.”The higher the number chosen by the respondents, the higher their level of approval.

Survey items were meticulously developed using validated scales from prior research to ensure both reliability and validity in measuring constructs such as trust, self-efficacy, and subjective norms. Specifically, the items measuring mass media influences were derived from the mature scales mentioned in24,25, thereby ensuring consistency with established research. eWOM factors, particularly those related to AV features, were adapted from blending industry-specific attributes with consumer perceptions81,108. Based on the references in24,25, subjective norms have been appropriately worded and modified. Self-efficacy items were adapted from established scales in24,25, thus ensuring the measurement’s conceptual integrity. Constructs measuring trust were based on scales modified from prior research references50,109. Lastly, the scales of adoption intention were adapted from validated constructs in24,25.

Data collection

To ensure the diversity and representativeness of the sample, this study adopted a variety of measures to optimize the survey distribution. First of all, Credamo was applied, a national online research platform that covers different regions and social groups across the country which able to avoid having a single sample source. Secondly, the study adopted a random sampling method to ensure that the data collection is not limited to specific income, age, occupational educational background, etc., which can improve the representativeness of the data. In addition, in the process of data collection, we eliminated invalid samples in the first stage by setting trap questions, by using quick questions and other forms. Finally, the participants included Chinese people from different regions, which covering the whole of China from first-tier cities to fifth-tier cities. Compared with the sample from a single designated area, we ensure the diversity of the sample and enhance the universality of the study of diversity.

Despite the measures above, there may still be some biases in this study that need to be considered when interpreting the results. Firstly, because the survey was conducted entirely online, participants who are not accustomed to use Internet (such as the elderly) may be underestimated, and future studies can be combined with offline interviews or telephone surveys to obtain a more balanced sample. Secondly, the respondents had a higher level of education (68.9% had a bachelor’s degree or above), which may mean that the results of the study are more inclined to the views of the highly educated group, and future studies may consider including more groups with lower educational background to improve the socioeconomic diversity of the sample. Thirdly, although the sample covered First-tier cities to Fifth-tier cities, users of the online questionnaire platform are mainly concentrated in economically developed areas, and the representativeness of samples in some remote areas may be relatively low. Therefore, future research can be combined with offline research to improve regional coverage.

Data collection for this study was conducted for three days, from October 10 to October 12, 2023, via the professional online platform Credamo (www.credamo.com). Out of 385 collected questionnaires, 315 were deemed valid, resulting in an effectiveness rate of 81.6%.

The sample’s demographic scope encompassed a diverse cross-section of the Chinese population residing in China. In terms of demographic information, the gender distribution within the sample was predominantly female, representing 58.4% of respondents. While 131 male respondents were accounted for 41.6%. Most participants (70.8%) were categorized as young adults aged between 18 and 34. Among the respondents, the 25–34 age group constituted the largest proportion, with 135 individuals and accounting for 42.9% of the total. Which was followed by the 18–24 age group, representing 27.9% of total. The proportion of respondents aged 35–44 was 19.7%, while those aged 45–54 accounted for 5.7%. The smallest proportion was observed among individuals aged 55 and above, comprising only 3.8% of the sample. Educational attainment among respondents was notably high, with 68.9% holding at least a bachelor’s degree. Respondents with a high school education or below comprised the smallest proportion, representing only 5.1% of the total sample (16 individuals). Those with a junior college (three-year program) background accounted for 10.8%. Meanwhile, individuals holding a master’s degree or higher constituted 15.2% of the respondents, totaling 48 participants. The predominant employment status was company employees, comprising 60% of the sample. A total of 66 students participated in the study, representing 21.0% of the sample. Civil servants comprised 5.7% of the respondents, totaling 18 individuals. The self-employed category had the lowest representation, accounting for only 3.8%. Meanwhile, respondents in other occupations made up approximately 9.5% of the total sample.

Regarding economy and income, only 20% of the respondents reported an annual post-tax income that exceeded RMB 150,000. Approximately 80% of respondents reported an annual income of RMB 150,000 or below. Among them, 85 individuals (27.0%) earned RMB 50,000 or less, while 29.8% had an income between RMB 50,001 and 100,000. Additionally, 23.2% fell within the RMB 100,001–150,000 range. Those earning between RMB 150,001 and 200,000 accounted for 9.5% of the sample, whereas 10.5% reported an income exceeding RMB 200,000. In terms of geographical distribution, respondents were predominantly from second-tier cities, comprising 51.1% of the sample and forming the primary interviewee group. Third-tier cities followed, representing 21.9% of the respondents. The proportions of individuals from first tier and fourth-tier cities were relatively similar, at 12.1% (38 individuals) and 10.8% (34 individuals), respectively. The lowest representation came from fifth-tier cities, which accounted for only 4.1% of the total sample.

Experience with AVs was relatively prevalent among the respondents, with 64.4% reporting familiarity. The remaining 35.6% of respondents indicated that they had never experienced autonomous vehicles. The respondents’ vehicular preferences were almost evenly split between fuel-powered vehicles (42.5%) and new energy vehicles (46.3%). A minority of 35 people (11.1%) reported frequent use of autonomous vehicles. Furthermore, 60.6% of the respondents had experienced a motor or non-motor vehicle-related traffic incident. 39.4% (124 people) had never experienced a traffic accident. Table 2 shows the descriptive statistical characteristics of the respondents.

Table 2 Demographic characteristics.
Full size table

Data validation

The questionnaire’s internal consistency, assessed via Cronbach’s alpha coefficient, revealed a robust overall value of 0.94 for the 37 items, surpassing the accepted threshold of 0.7, which indicated reliable scale reliability110,111,112 Utilizing SPSS 26.0, specific constructs within the questionnaire demonstrated commendable consistency, with Cronbach’s alpha for mass media at 0.766, eWOM at 0.788, self-efficacy at 0.889, subjective norms at 0.945, trust at 0.927, and adoption intention at 0.909. Each variable’s Cronbach’s alpha value reached the standard evaluation range of greater than 0.7113.

Table 3 shows the correlation analysis further elucidated the relationships among variables. Correlations were classified as either negative or positive, with values ranging from 0.1 to 0.29 indicating a weak correlation, 0.30–0.49 suggesting a moderate correlation, and 0.5-1.0 signifying a strong correlation114,115. Empirical results indicated positive correlations among variables, thus aligning with theoretical expectations. Mass media exhibited a moderate positive correlation with adoption intention (r = 0.483), whereas eWOM demonstrated a weaker linkage (r = 0.219). Furthermore, self-efficacy (r = 0.626), subjective norms (r = 0.609), and trust (r = 0.698) demonstrated strong positive associations to adopt AVs, underscoring their pivotal roles in influencing adoption behaviors.

Table 3 Means, standard deviations, reliabilities, and correlations among study constructs.
Full size table

Results

Convergent validity validation

The convergent validity of the research instruments was ascertained through assessments of factor loadings, average variance extraction (AVE), and composite reliability (CR), based on the criteria established by116,117. CR assesses the questionnaire’s internal consistency, with a requisite value exceeding 0.7 as stipulated by118, which indicates robust internal consistency. AVE evaluates the explanatory power of latent variables over the variance of the measurement variables, necessitating a benchmark AVE value exceeding 0.5 to confirm sufficient explanatory capacity117.

In this study, as observed from Table 4, all factors exhibited factor loadings greater than 0.5, AVE values above the threshold of 0.5, and CR values exceeding 0.7, thereby affirming the strong convergent validity and reliability of the measurement model employed116,119.

In addition, the Kaiser-Meyer-Olkin (KMO) measure was used to evaluate the data’s suitability for factor analysis. When the KMO value is closer to 1, the correlation between variables is higher, and the dataset is more suitable for factor analysis119. The statistical software SPSS 26.0 was used to obtain a high value of 0.950, indicating a significant correlation among the variables, which is suitable for factor analysis.

Furthermore, the results of Bartlett’s test of sphericity were statistically significant, with a p-value of 0.000. This test verifies that the correlation matrix of the measured variables is not an identity matrix (variables are not independent), confirming that the variables are interrelated.

Factor extraction was conducted using an analytical method, and as a result, all factors demonstrated extraction values above 0.5. Additionally, factor rotation was performed via the principal component analysis method, which converged after six iterations. This process extracted six factors with eigenvalues exceeding 1, accounting for a cumulative variance of 73.954%. These results suggest that a substantial portion of the main components within the analyzed dataset can be explained, highlighting the effectiveness of the factor analysis in identifying the data’s key dimensions.

Table 4 Measurement items and convergent validity analysis.
Full size table

Convergent validity validation

The Fornell-Larcker criterion has traditionally evaluated discriminant validity117; however, it has faced criticism for its limited sensitivity and specificity120,121. In response58, introduced the heterotrait-monotrait ratio (HTMT) as a more robust method for assessing discriminant validity. This method stipulates that an HTMT value exceeding 0.85 implies a significant overlap among constructs, thereby indicating weak discriminant validity. High HTMT ratios may imply a complete lack of discriminant validity122. Thus, employing the HTMT ratio is crucial for affirming the independence among variables and preserving research integrity.

Discriminant validity requires a distinct separation among scale items that measure different constructs123. Specifically, it requires that the square root of the AVE for each factor should exceed the correlation coefficients involving that factor in the corresponding matrix117. Strict HTMT values should generally adhere to a standard of less than 0.8558,121. Strict adherence to HTMT thresholds below 0.85 guarantees minimal overlap among constructs and upholds the validity of discriminative assessments124.

This study rigorously assessed discriminant validity using the HTMT ratio criterion, which was set below 0.85124. It utilized SmartPLS 4.0.9.2 to calculate the HTMT values (heterotrait-monotrait ratio of correlations). Table 5 results confirm that all HTMT ratios between constructs fall below the threshold, and each variable’s square root of AVE exceeds its associated correlation coefficients117. These findings substantiate the discriminant validity of the variables examined, ensuring that each construct uniquely contributes to the model without undue overlap.

Table 5 Discriminant validity analysis.
Full size table

Structural model results

The Partial Least Squares Structural Equation Modeling (PLS-SEM) approach, which aims to explain and predict endogenous latent variables, utilizes path coefficient estimation to minimize error terms and maximize explanatory power (R²)107. This study employed SmartPLS 4.0.9.2 for model testing, incorporating a bootstrapping procedure with 5,000 subsamples to delineate the distinct contributions of each pathway within the model. Figure 3 indicates that the proposed factors account for 57.5% and 56.4% of the variance in adoption intention. Hypotheses H2, H3, H4*, H9, and H9* were not supported, whereas the remaining hypotheses were confirmed.

Analysis revealed the influence of mass media on the structure of human factors is not significant. Namely, the communication content or influence of mass media played a role in shaping consumers’ perception of social pressure (subjective norm (β = -0.114, t = 1.432, p = 0.152 > 0.05) and trust in autonomous vehicles (H3: β = -0.083, t = 1.198, p = 0.231 > 0.05) did not show a statistically significant effect. In other words, the sample data in this study did not demonstrate that mass media significantly increased people’s acceptance of autonomous vehicles due to social expectations or effectively increased their trust in the technology. This may mean that the influence of mass media is more indirect or that the audience may be more influenced by other factors (such as personal experience, word of mouth in social networks, industry expert opinions, etc.) in forming subjective norms and trust, rather than relying solely on the transmission of information by mass media.

Additionally, the statistical test results failed to support the H4* hypothesis (β = -0.001, t = 0.027, p = 0.978 > 0.05), indicating that the influence path of electronic word-of-mouth on the adoption intention of autonomous vehicles is not significant. This suggested that, within the study sample, consumers’ adoption intention was not significantly driven by electronic word-of-mouth. In other words, information from social media, online reviews, or user recommendations may not have effectively enhanced their acceptance of autonomous vehicles.

Finally, the statistical test results indicate that hypothesis H9 (β = -0.111, t = 1.574, p = 0.115 > 0.05) and H9* (β = -0.119, t = 1.692, p = 0.091 > 0.05) were not supported, demonstrating that the influence of subjective norms on the adoption intention of autonomous vehicles is not statistically significant. This suggests that, within this sample, individuals’ intention to adopt autonomous vehicles was not substantially driven by social pressure or the expectations of others.

In this study, the mass media model 1 and the electronic word-of-mouth model 2, as two comparative models, aim to explore the mechanism and differences of the two different information dissemination channels in influencing consumers’ attitudes and behavioral intentions towards autonomous vehicles.

Model 1, the mass media model, mainly examined the influence paths of traditional mass media on consumers’ cognition and behavior, including the effects of media content on consumers’ perceived usefulness, subjective norms, trust, and final purchase intention. In Model 1, the study found that mass media significantly affects purchase intention but has no significant impact on subjective norms and trust. This indicates that mass media has an influence in enhancing the public’s attention to and overall acceptance of AVs, but its effect in shaping social identity (subjective norms) and consumer trust is limited. This might be due to the one-way information sources and lack of interactivity of mass media. It is difficult to establish deep trust and social identity.

Model 2, the Electronic Word-of-Mouth Model, explored the influence paths of electronic word-of-mouth (eWOM), such as non-official information sources like social platforms and forum comments, on consumers’ psychology and behavior. The results showed that electronic word-of-mouth has a significant positive impact on subjective norms and trust, indicating that eWOM, which was highly interactive and derived from “others’ experiences,”is more capable of establishing consumers’ trust and herd mentality. However, the influence of electronic word-of-mouth on the final purchase intention was not significant. This might be because although people trust the content of electronic word-of-mouth, such information often contains a lot of negative or neutral comments, which can easily lead to a wait-and-see mood and thereby reduce the tendency of direct purchase actions.

Model 1 showed that mass media can directly influence purchase intentions and was applicable to enhancing consumers’ initial interest in new technologies and brand recognition.

Model 2 indicated that electronic word-of-mouth has a more effective influence on consumers’ trust and social cognition and is suitable for strengthening positive identification and confidence among consumer groups in product promotion.

The differences in the results between the two revealed that mass media have a greater “wide publicity” effect, while electronic word-of-mouth has a greater “deep penetration” effect. When promoting autonomous vehicles, enterprises should combine two communication methods: using mass media to enhance exposure and using electronic word-of-mouth to strengthen trust building and community atmosphere.

Fig. 3
figure 3

The PLS-SEM results of model.

Full size image

Discussion

The rapid advancement of automation in the transportation sector indicates that the commercialization of AVs is imminent26. Therefore, understanding the key factors that influence AVs’ consumer adoption is critical. With AVs entering the urban road-testing phase and approaching commercialization, exploring potential measures to promote their widespread acceptance is especially necessary. This study aims to identify the pre-factors that affect consumers’ purchase of AVs and to analyze current media channels’ direct and indirect role to provide a theoretical basis for promoting the market popularization of AVs.

Such as 1) How do environmental factors (mass media and eWOM), perceptual factors (self-efficacy and subjective norms), and product factors (trust) influence consumers’ willingness to purchase AVs?

2) How can the power of media channels be maximized to enhance consumer willingness to adopt AVs?

eWOM significantly influences consumer subjective norms and trust, whereas mass media insignificantly impact these factors. This difference underscores that, while information can be obtained through various media channels, its effect on consumer cognition varies depending on the source. Notably, the more information consumers obtain through eWOM, the stronger the impact on subjective norms. Additionally, consumer trust in AVs is significantly influenced by eWOM. In other words, when online evaluations of AVs are predominantly positive, it significantly enhances consumer trust.

As indicated by prior findings, both mass media and eWOM positively influence consumers’ self-efficacy. Specifically, mass media accounts for 29.3% of the variance in self-efficacy (R² = 0.293), whereas eWOM accounts for only 8.3% (R² = 0.083). These findings indicate that mass media significantly enhances self-efficacy. This result aligns with the findings of125, reinforcing that mass media serves as a highly effective tool for enhancing awareness and fostering consumer confidence.

Mass media significantly and positively affected adoption intentions, whereas the hypothesis regarding eWOM was rejected. This contradicts the findings of81, which identified word-of-mouth and mass media as key predictive factors. The research results indicate that eWOM does not directly influence consumers’ willingness to adopt AVs despite its importance.

The main reason eWOM does not directly affect consumers’ willingness to adopt AVs is because mass media, including TV, radio, and movies, generally have higher credibility and trust among a wider audience than eWOM. Public media, such as mass media, possess greater credibility when adopting new technologies such as autonomous vehicles. In addition, mass media tends to deliver controllable, similar, and well-crafted messages, rendering it more persuasive. In contrast, eWOM frequently presents different opinions from numerous different sources, some of which may be questionable, thus reducing the public’s intention to adopt AVs.

Furthermore, self-efficacy was positively correlated with subjective norms. Notably, mass media and self-efficacy together explained 37.8% of the variance in subjective norms (R² = 0.378), whereas eWOM and self-efficacy accounted for 39.9% (R² = 0.399). This implies that although self-efficacy positively impacts the development of subjective norms, its influence varies depending on the media channel. Specifically, the media source from which information is acquired influences the strength and formation of subjective norms. This result is supported by90. Individuals with high self-efficacy believe they can effectively utilize emerging technologies such as AVs. As a result, they may be more inclined to conform to or agree with their peers’ social norms or opinions, especially if those norms favor the adoption of the technology. Individuals with greater confidence in their technological proficiency exhibit increased receptiveness to external social influences, which may further validate their belief in the autonomous driving technology’s utility.

Similarly, the influence of self-efficacy and subjective norms on trust exhibited a comparable trend. Mass media, self-efficacy, and subjective norms collectively accounted for 56.6% of the variance in trust (R² = 0.566), whereas eWOM, self-efficacy, and subjective norms accounted for 57.8% of trust (R² = 0.578). Specifically, the mass media indirectly influences consumers’ trust in autonomous vehicles through self-efficacy. eWOM can influence consumers’ trust in autonomous vehicles both directly and indirectly through their self-efficacy and subjective norms. These findings indicated that while both media channels can increase consumer trust in AVs but depending on the path it affects. Overall, consumers were more likely to rely on information obtained from eWOM. This highlighted the significance of peer reviews and user-generated contents in shaping consumer trust, as70 pointed out, trust issues in autonomous vehicles can be effectively resolved by disseminating detailed and specific information.

Trust is a key prerequisite for consumer adoption of new technologies61, particularly in complex and potentially risky emerging technologies such as autonomous driving. Consumers with a high sense of self-efficacy possess confidence in their capacity to address contingencies and potential issues associated with autonomous driving technology. This ability enables them to reduce their anxiety regarding the unknown and uncertainty. Similarly, subjective norms, through other people and social support, can alleviate consumer uncertainty regarding emerging technologies, thereby increasing their propensity to perceive autonomous driving technology as safe and reliable. In other words, self-efficacy and supervisory norms reflect the state of consumer psychological activity and enhance the willingness to accept autonomous driving technology by shaping psychological comfort.

The hypothesis that subjective norms influence the willingness to adopt AVs was overturned and rejected, contrary to the findings of24,25. This means that the opinions and expectations of significant people have no material impact on consumers’ willingness to adopt AVs. In this case, consumers rely more on their insights, self-confidence (self-efficacy), and trust in technology than on others’ opinions and expectations. Furthermore, the perceived value of individual assessment and effective use of technology outweighs the influence of social pressures or norms. In other words, individuals prioritize their assessment of risks and benefits over societal expectations when adopting disruptive emerging technologies such as autonomous vehicles.

In addition, autonomous vehicles remain an emerging technology with varying levels of public acceptance. In contrast to established technologies, there exists strong social norms or expectations for the adoption of autonomous vehicles. In contrast, autonomous vehicles have not yet developed a definitive or broad consensus in society. The lack of a cohesive societal stance diminishes the impact of subjective norms, resulting in individuals being less inclined to perceive pressure from others or society to encourage the adoption of the technology.

Finally, self-efficacy and trust positively and significantly impact adoption intentions. Previous research has confirmed that trust in new things or technologies affects consumers’ willingness59 and that trust helps consumers purchase AVs126. In other words, when media channels report positively on AVs, individuals’ recognition and trust in their abilities will increase, making them more likely to adopt AVs56,67.

In other words, people with high self-efficacy exhibit a greater propensity to engage with and adopt emerging technologies. Their confidence in overcoming potential risks increases the probability of adopting emerging technologies. In addition, trust enhances consumer confidence in the reliability of self-driving technology, thereby increasing their willingness to adopt it. Consumers will adopt autonomous driving technology when they perceive it as reliable and safe, and when it is endorsed by reliable public sources such as regulators or trusted media.

The study highlights the complex relationship between various media channels and consumers’ willingness to adopt AVs, highlighting the necessity of understanding and strategically utilizing media channels, including eWOM and mass media. Specifically, electronic word-of-mouth indirectly affected consumers’ adoption intention of autonomous vehicles through self-efficacy and trust. On the other hand, mass media can both indirectly influence adoption intentions through self-efficacy and directly influence consumer adoption intentions for autonomous vehicles. This finding is consistent with the conclusion of81, It shows that different media channels have different effects on consumers.

eWOM helps create subjective norms and trust by providing truthful, transparent, and diverse information, as well as facilitating interaction with other users. This aligns with the view of127 that people are willing to grasp the experiences of existing consumers76 and provide information72. The information posted by experienced users bolster consumer confidence in emerging technologies127.

Related stakeholders, government departments, and manufacturers should fully leverage the advantages of eWOM and promote the posting of authentic and positive reviews. This is attributable not only to consumers’ increased willingness to trust information from their peers72 but also suggested that eWOM indirectly influences adoption via trust and subjective norms but not directly. However, the fact that the mass media directly impacts consumers’ willingness to adopt AVs cannot be ignored. Therefore, targeted, reasonable, and wise planning and usage of media channels will achieve a much greater effect.

Theoretical and practical significance

Based on TAM, this study makes a theoretical contribution by introducing external environmental factors (mass media and eWOM), psychological factors (self-efficacy, subjective norms), and product factors (trust) to establish a three-stage model. Another key theoretical contribution is the distinction among the functions of various media channels (mass media and eWOM). Mass media is more direct and effective in enhancing consumers’ self-efficacy and adoption intention, whereas eWOM is superior to mass media in influencing consumers’ trust and subjective norms.

Traditional SCT emphasizes the interaction among individuals, behaviors, and the environment and mostly focuses on individual cognitive variables such as self-efficacy. This study introduced two types of external stimuli, namely mass media and electronic word-of-mouth, to reveal how they respectively activate the three psychological paths of self-efficacy, subjective norms, and trust, thereby influencing behavioral intentions. In particular, the discovery of the role of electronic word-of-mouth in shaping subjective norms and trust has broken through the traditional assumption in SCT that “subjective norms indirectly affect intentions.” Mass media can more directly enhance self-efficacy and purchase intention, providing a new digital context for the observational learning mechanism. Furthermore, through the complementary between SOR and SCT, the relatively broad framework of “behavior acquisition and motivation” in SCT can be supplemented. After combination, it is possible to specifically analyze how the information source activates the internal psychological mechanism and guides the behavioral outcome. Specifically, the SOR model provides a clear path of “stimulus → organism → response” but lacks refinement of the internal mechanisms of the “organism.”SCT focuses on cognitive and social factors but seldom distinguishes different stimulus sources. After integration, we embed the core variables of SCT, such as trust, self-efficacy, and subjective norms, into the “organism” level of SOR. Clearly distinguishing the differentiated effects of the two stimuli, mass media and electronic word-of-mouth, on psychological variables and behavioral intentions.

This study expanded the application paths of Stimulus-Organism Response (SOR) Theory and Social Cognitive Theory (SCT) in the field of emerging technology adoption at the theoretical level. Compared with previous studies that emphasized technical characteristics (such as ease of use and usefulness) or individual cognitive variables (such as perceived usefulness and adoption intention) (such as the TAM model), this study emphasized revealing through “organism” variables such as “subjective norms”, “trust”, and “self-efficacy”. How external information stimuli are transformed into specific behavioral intentions through an individual’s internal psychological processes. It further enriched the mechanism construction of the “organism” link in the SOR theory.

Furthermore, in combination with the empirical results of the latest research in recent years (2024–2025) on the influence mechanism of the media environment. For example128, a systematic analysis was conducted on how negative media coverage (stimulus) inhibits consumers’ purchase intention (response) for self-driving cars by enhancing perceived risks and weakening trust (organism)129. started from the risk perception paradigm and the theory of conditional value, combined with variables such as trust, self-efficacy, and subjective norm,.This paper further explored the influence of media public opinion (including reports from mainstream media and comments on social platforms) on subjective norms and trust perception and realized the final influence path through behavioral intentions.

In conclusion, this study has constructed a theoretical bridge between SOR theory and SCT, deepened the causal chain among external information stimuli, psychological mediating variables, and the intention of technology adoption, and provided a new perspective and theoretical support for understanding the decision-making mechanism of consumers in a highly uncertain technological environment.

Contrary to numerous previous studies24,25, this study established that subjective norms have no direct impact on consumers’ willingness to adopt AVs. This finding has prompted stakeholders, governments, and manufacturers to re-evaluate the role of social impact in the adoption of autonomous driving technologies. Individual factors (self-efficacy) and product factors (trust) outweighed social pressures in influencing adoption decisions.

In terms of building consumer trust, relevant stakeholders, such as government departments and manufacturers, can enhance consumers’ self-efficacy through rational use and management of media channels, actively use the network positive evaluation, reduce negative evaluation, and provide accurate and reliable positive evaluation. In addition, they can focus on providing accurate and balanced information, using online word-of-mouth management tools or services to monitor and manage online comments and published content, and dealing with negative information promptly.

In contrast, stakeholders such as manufacturers, policymakers, and others should focus on measures to enhance consumer reliability and confidence in AV technology, build consumer trust, emphasize safety and reliability to ease consumer anxiety and highlight the utility of AVs. For example, they should enhance the transparency of AVs and provide detailed product information and clear usage guidelines to enable consumers to better understand the product’s functionality and operability. In addition, giving test-drive opportunities or test-drive experiences enables consumers to directly perceive the safety and convenience of autonomous driving technology. Moreover, knowledge of safety standards and technical reliability can further foster trust.

Based on mass media has a positive impact on adoption intentions of autonomous vehicles. Therefore, it is essential to establish credible mass media channels that facilitate targeted transmission of information pertaining to autonomous driving technology to promote its benign development. Channel communication should focus on educating the public about AVs’ safety, reliability, and sustainability. Invite credible news media and field experts to cooperate to enhance credibility and persuasiveness.

Finally, government agencies should actively promote reliable information about AVs through mass media while supporting eWOM by encouraging experienced users to generate authentic review content, thereby gaining wider support for AV adoption.

Limitations and directions for future research

Indeed, certain aspects of the study remain inadequately addressed, constituting a limitation of the study. First, the sample respondents originate from specific groups (Chinese), regions (China), and cultures. Subsequently, the influence of different groups, regions, and social cultures on the adoption intention of AVs is not fully considered. Therefore, the research results are not broadly representative, and their universality in different cultural settings will be limited. Secondly, the information transmission method is incomplete. The research focuses on eWOM and mass media; it fails to fully explore the potential impact of environmental factors on the adoption intention of AVs by considering other environmental factors, such as traditional media and social media. Thirdly, this study adopted a cross-sectional study. The data was only collected at a specific time point, and the evolution process of target variables at various time points could not be obtained. As trust in media continues to evolve, especially with the decline in mass media trust and the rise in the influence of social media influencers (SMIs) led by electronic word of mouth (eWOM), eWOM’s role is likely to continue to grow, changing consumers’ decision-making processes, brand trust, and willingness to buy130. Consumers are increasingly relying on personalized and interactive content sources, such as social media influencers, rather than traditional mass media channels130. This would also mean that the decline in trust in mass media and the rising trend of eWOM led by opinion leaders could affect the application of future studies. Therefore, future research should employ longitudinal research design to monitor the change process of target variables over time and comprehensively grasp the information of the time dimension.