Examining public acceptance intentions for government digital human through expectation confirmation and technology acceptance models Authors: Jing-hao Chen, Yi-de Chen, Feng Jia, Lian Gu, Xiao-yu Qiu Published: 2026 ABSTRACT AbstractGovernment digital humans represent a significant application of the combination of generative artificial intelligence and digital governance. This study aims to explore the factors influencing public acceptance of government digital humans, providing insights for enhancing their services and user experience. By integrating expectation confirmation theory (ECT) and the technology acceptance model (TAM), this study expands upon ECT by incorporating additional user perception factors. A structural equation model is employed to explain the influencing factors and their interrelationships, while a mediation effect model tests the mechanisms of these factors. The findings indicate that public expectation confirmation enhances satisfaction with government digital humans, which positively impacts acceptance intention. Perceived information quality, perceived intelligence, perceived convenience, perceived attractiveness, perceived usefulness, and artificial intelligence trust collectively mediate the relationship between public expectation confirmation, satisfaction, and acceptance intention. The findings highlight the importance of enhancing the perceived usefulness, perceived intelligence, perceived information quality, and perceived attractiveness of government digital humans. Relevant departments should focus on improving these attributes to optimize digital government services and enhance user experience and utilization rates.Keywords: Government digital human, Digital government, Public acceptance intention, ECT, TAMSubject terms: Psychology, Human behaviour Introduction Enhancing the digitalization and intelligence of government operations has consistently been a core objective in the development of digital governments worldwide. In 2021, China’s “14th Five-Year Plan and the 2035 Long-Range Objectives Outline” explicitly called for the comprehensive promotion of digital and intelligent transformation of government operations, business processes, and service models. This plan emphasized deepening the “Internet + Government Services” initiative and enhancing the functionality of integrated online service platforms1. In 2023, the Central Committee of the Communist Party of China and the State Council released the “Overall Layout Plan for Digital China Construction,” which stressed the development of highly efficient and collaborative digital governance, improving digital service levels, accelerating the implementation of “one-stop services,” and promoting the integration of online and offline services2. In the global context of digital and intelligent integration, the rise of generative artificial intelligence (AI) presents a significant technological opportunity for the digital and intelligent upgrading of public governance. This advancement is pivotal in enhancing government service levels and modernizing national governance systems and capacities. Government digital humans represent a significant application of generative AI in the domain of digital governance. In 2023, China introduced several government digital humans, including Guangzhou’s tax digital human “Yue Yue,” Nanchang’s “Xiao Gan Shi” self-service machine, and Kunshan’s “Xiao Hua Xiao Qiao.” These government digital humans leverage generative AI technology, combined with advanced technologies such as cloud computing, big data, and the metaverse. They present themselves as virtual avatars (e.g., three-dimensional cartoon images or intelligent self-service machines) and provide public services such as administrative approval, intelligent Q&A, and business processing through dialogue and human-computer interaction. Government digital humans not only offer higher intelligence and flexibility but also enhance the public experience through anthropomorphic interactions, covering all business scenarios and delivering 24/7 intelligent services. Compared to traditional government machines, government digital humans address the limitations of fixed programs, offering smarter and more efficient public services. Despite the potential application prospects of technologies such as the metaverse in the field of government digital humans, their widespread adoption still faces numerous challenges, requiring comprehensive consideration of technical, policy, security, and privacy issues3. Therefore, this study focuses on government digital humans based on generative AI and natural language processing technologies, which provide convenient business consultation and administrative services through anthropomorphic virtual avatars. Academic research on the integration of generative AI into digital governance has primarily focused on two aspects. At the macro level, it addresses the “theoretical foundations” of technological integration, including theoretical logic4, value goals5, risk prevention6, and feasible paths7. At the micro level, it examines the “practical being” of technological implementation, encompassing case studies8,9, reflective evaluations10, and areas for improvement10,11. Existing research on integrating generative AI into digital governance, while advancing at both macro-theoretical and micro-practical levels, often adopts a “supply-side” logic that emphasizes how technological upgrades enhance government services through standardization and efficiency gains. However, a critical yet often overlooked dimension is the core driver of intelligent public services: responsiveness to public demand and the public’s willingness to embrace emerging technologies11. If individuals remain disengaged owing to concerns or negative experiences, even the most sophisticated technologies may fail to generate meaningful public value. This issue is especially relevant to the emerging domain of government digital humans, which serve as the “digital face” of public services and interact directly with citizens. Public acceptance is therefore essential, shaping not only the diffusion of such technologies but also the fabric of government–citizen relationships in the digital age. Accordingly, this study adopts a public-centered perspective to address the research gap concerning the underexplored “demand side.” We argue that understanding how and why individuals accept government digital humans is critical for supporting their responsible development and improving digital governance outcomes. Although expectation confirmation theory (ECT) and the TAM are established frameworks for explaining user behavior, we posit that the unique role of government digital humans as AI-enabled public service agents requires a more nuanced approach than simply applying or merging existing models. To this end, this study presents a theoretical extension by integrating ECT and TAM while incorporating attributes specific to government digital humans (e.g., perceived intelligence, perceived information quality, perceived anthropomorphism, perceived attractiveness, perceived convenience, and AI trust). These dimensions are closely associated with public perceptions of the digital human’s “image appeal,” “service capability,” and “human-computer relationship,” and their relevance exceeds the traditional scope of standard models. By constructing and testing this extended model, the study identified the key factors and mechanisms that shape public acceptance of government digital humans. The findings provide actionable insights for optimizing digital human design, improving public service experience, and strengthening human-computer trust. Literature review Digital humans and application fields A digital human is a computer-generated, virtual, human-like figure capable of simulating human behavior, engaging in language interactions, and expressing emotions. These entities interact with users through advanced AI technologies, such as natural language processing and machine learning. With the continuous advancement of AI technology and computational power, digital humans have evolved from simple virtual response assistants to highly intelligent and interactive agents12,13. Currently, digital humans are widely used in various fields. In customer service, Apple’s Siri and Amazon’s Alexa provide portable voice interaction functions, forming intelligent ecosystems within their respective product ranges14. In the healthcare sector, digital humans offer one-on-one virtual medical services and mental health consultations, providing patients with convenient and reliable assistance11. In e-commerce, virtual customer service agents enhance the online shopping experience by assisting users15. Additionally, digital humans are applied in movies, games, and online educational courses, offering users human-computer interaction and immersive experiences16. Government service machines, government intelligent answering robots, and government digital humans Government service kiosks and intelligent response robots are modern technological devices used by government departments to enhance service efficiency and operational effectiveness. Each method has specific application scenarios and functions. Government service kiosks, as self-service terminals integrating multiple services, provide convenient self-service channels for the public and are generally installed in government offices, public places, or community centers. These kiosks operate around the clock, allowing citizens to independently handle various government services, such as ID card applications, traffic violation inquiries, and social security information queries, thereby achieving instant and convenient government services. However, their interactivity and ability to handle complex tasks are limited, making them incapable of addressing complex or unstructured public demands17. For example, when handling complex administrative approvals that require cross-departmental coordination, these government service kiosks often fail to provide effective solutions. Contrastingly, intelligent response robots collect information and perform business transactions through preset programs and rules, primarily assisting with fixed administrative tasks. These robots interact with the public through voice and text, presenting predefined response templates, but they still fall short when dealing with diverse and complex tasks7. The emergence of government digital humans has transformed the provision of intelligent government services across multiple dimensions. In terms of communication upgrades, government digital humans enable interactive human-computer feedback18. Utilizing advanced natural language understanding and generation capabilities, generative AI allows the public to not only inquire about information from digital humans but also to complete business transactions through interactive dialogues, reducing human intervention and overcoming the limitations of traditional, “unintelligent” AI19,20. This advancement ensures that users receive immediate and effective service feedback through their interactions with digital humans. In terms of reshaping service scenarios, government digital humans integrate data from multiple departments through a single interface, enabling cross-departmental communication and sharing. This integration facilitates synergy between physical spaces and virtual environments, advancing the “one-stop service” reform, streamlining processes, and reducing service costs18. By minimizing the need for users to navigate between different departments, government digital humans enhance overall service efficiency and user satisfaction21. Regarding the enhancement of the public’s primary status, traditional government administration often exhibits an asymmetric power dynamic between administrative entities and the public, leading to issues such as indifferent attitudes and poor service quality7. Government digital humans provide equal communication and services that transcend temporal and spatial limitations, catering to user needs and significantly improving the user experience. Moreover, as virtual entities capable of simulating human appearances and voices, government digital humans eliminate the drawbacks of human services, promote equal opportunities for disadvantaged groups, and enhance the warmth of services20. Factors influencing public acceptance intention of government digital humans The core of intelligent governance lies in enhancing responsiveness to public needs21. As emerging public service agents that simulate interpersonal interaction, government digital humans require public acceptance to ensure their technological effectiveness and to convey warmth in service delivery. Therefore, identifying the various factors that influence public acceptance of government digital humans is a necessary focus of this research. Core constructs from traditional technology acceptance theories, such as TAM, particularly perceived usefulness (PU) (i.e., the ability to solve problems) and perceived ease of use (PE) (i.e., operational convenience), remain validated drivers in recent AI-focused public service research22,23. This suggests that fulfilling basic utility expectations and reducing usage barriers continue to underpin technology acceptance, regardless of technological advancement. However, public expectations for government digital humans go beyond mere usefulness and ease of use. In managing complex public affairs, perceived intelligence (PI) (i.e., the capacity to understand and respond to multifaceted issues) and perceived IQ (i.e., accuracy and authoritativeness) represent critical indicators of service “hard power.” Recent studies have examined how individuals assess AI’s intelligent performance and information reliability in government contexts, which directly shape perceptions of service efficiency and institutional capability24,25. Simultaneously, PC, such as the ability to deliver 24/7, one-stop services, reflects the principle of technology serving the people and continues to garner scholarly attention26,27. The “warmth” of human-computer interaction should not be overlooked. In government services, perceived attractiveness (PA) (i.e., an affable and aesthetically pleasing image) and perceived anthropomorphism (ANT) (i.e., natural and empathetic interaction) are becoming key focal points. Research suggests that good design can enhance emotional experience and bridge the digital divide17,28, indicating these features may hold new value in building harmonious digital government–citizen relationships. Trust is the cornerstone of digital interaction. For government digital humans handling sensitive information and exercising public power, AI trust (TR) is particularly vital. The public is concerned not only with technical ability but also with benevolence and adherence to data regulation (benevolence and integrity trust). Recent research has highlighted public concern with privacy protection, algorithmic fairness, and perceived risks (functional, privacy, ethical, etc.) in government settings, forming key acceptance thresholds15,29–31. In summary, factors influencing public acceptance intention for government digital humans are multidimensional, involving technical functions, user experience, trust, ethics, and more. Although existing studies have identified many possible factors, their relative weight, interaction effects, and integrated theoretical interpretation in the context of government digital humans require further deepening. Specifically, how these elements (including intelligence, attractiveness, and trust, though “mature” elsewhere) carry new meaning in the government setting and how they collectively influence public expectations, satisfaction, and acceptance intention calls for more empirical study. Results Participant demographic The final sample included 801 participants. Demographic details are provided in Table 1. In terms of gender, 37.1% (n = 297) identified as male and 62.9% (n = 504) as female. In terms of age distribution, the following age groups accounted for majority of the sample population: 18–25 years (36.0%), 26–35 years (20.3%), and 36–45 years (25.7%). Participants aged 46–55 and over 55 accounted for 11.6% and 6.4%, respectively. Regarding educational attainment, 17.2% had a high school education or lower, 27.6% held an associate degree, 41.2% held a bachelor’s degree, 13.1% held a master’s degree, and 0.9% held a doctorate. Familiarity with government digital humans varied: 10.1% had never heard of them, and 17.9% had heard but were unfamiliar. Most participants reported some degree of awareness, with 34.7% somewhat familiar, 25.1% quite familiar, and 12.2% very familiar with the technology. Table 1Demographic characteristics of the sample.VariableCategoryFrequency/NumberPercentageGenderMale29737.1%Female50462.9%Age18–2528836.0%26–3516320.3%36–4520625.7%46–559311.6%Above 55526.4%Highest education levelHigh school or below13817.2%Associate degree22127.6%Bachelor’s degree33041.2%Master’s degree10513.1%Doctoral degree70.9%Level of Understanding of Government Digital HumansNever heard of it8110.1%Heard of it, but not familiar14317.9%Somewhat familiar27834.7%Quite familiar20125.1%Very familiar9812.2% Demographic characteristics of the sample. Reliability and validity analysis As shown in Table 2, Cronbach’s α values for all variables exceeded 0.7, and composite reliability (CR) for each variable also exceeded 0.7, indicating strong internal consistency of the research scale. Validity refers to how well a measurement instrument captures the intended constructs. In EFA, all standardized factor loadings were above 0.7, reflecting good construct validity (Table 2). The average variance extracted (AVE) for all variables was greater than or equal to 0.5, supporting convergent validity (Table 2). Table 2Reliability and convergent validity test results.VariableItemStandardized Factor LoadingCronbach’s αCRAVEPerceived IQIQ10.8060.8290.8350.559IQ20.720IQ30.734IQ40.727PIPI10.7640.8130.8160.526PI20.735PI30.716PI40.683Perceived Attractiveness (PA)PA10.7950.8530.8580.602PA20.677PA30.827PA40.797Perceived Anthropomorphism (ANT)ANT10.6550.8100.8130.522ANT20.747ANT30.712ANT40.771PCPC10.7230.7720.7730.532PC20.751PC30.714AI Trust (TR)TR10.7380.8630.8670.520TR20.684TR30.768TR40.704TR50.704TR60.726PUPU10.7400.8420.8420.517PU20.720PU30.712PU40.712PU50.710Perceived Ease of Use (PE)PE10.7430.8330.8340.558PE20.763PE30.697PE40.782Expectation Confirmation (CE)CE10.8070.8280.8360.561CE20.735CE30.709CE40.741Satisfaction (SAT)SAT10.7440.8740.8840.604SAT20.756SAT30.798SAT40.757SAT50.827Intention to Use (CI)CI10.7160.7750.7790.541CI20.757CI30.733 Reliability and convergent validity test results. Correlation analysis A correlation analysis was conducted to examine relationships among latent variables and assess discriminant validity. As shown in Table 3, all pairwise Pearson correlations were significantly positive (P < 0.01), consistent with the hypothesized path directions in the structural model. Discriminant validity was confirmed using the Fornell and Larcker32 criterion: the square root of the AVE for each construct exceeded its correlations with any other construct (Table 3). These results confirm the distinctiveness of the latent variables, supporting the structural model analysis. Table 3Discriminant validity test results.IQMSDIQPIPAANTPCTRSATCEPUPECI5.860.750.748PI5.800.810.478** 0.725 PA5.470.950.328**0.306** 0.776 ANT5.380.960.338**0.385**0.428** 0.722 PC6.080.740.441**0.425**0.273**0.333** 0.729 TR5.910.720.460**0.501**0.320**0.382**0.466** 0.721 SAT5.890.780.562**0.615**0.420**0.515**0.587**0.596** 0.719 CE5.740.820.396**0.385**0.269**0.401**0.304**0.399**0.559** 0.747 PU6.100.670.319**0.328**0.123**0.284**0.461**0.368**0.520**0.308** 0.749 PE5.990.730.323**0.334**0.168**0.296**0.419**0.363**0.509**0.307**0.370** 0.777 CI6.030.730.523**0.559**0.342**0.491**0.559**0.575**0.689**0.490**0.500**0.413** 0.736 **, and * represent P < 0.01, and P < 0.05, respectively. The bold numbers on the diagonal are the square roots of AVE. Discriminant validity test results. **, and * represent P < 0.01, and P < 0.05, respectively. The bold numbers on the diagonal are the square roots of AVE. Common method bias To evaluate the impact of common method bias (CMB) on the study’s findings, we performed the tests described in the methodology section, yielding the following results: Harman’s Single-Factor Test: EFA results (Table 4) showed that the first unrotated principal component explained 31.701% of the total variance, below the recommended threshold of 40%. In addition, the single-factor model estimated using CFA demonstrated poor fit indices (Table 5; χ²/df = 7.561 > 5, RMSEA = 0.091 > 0.08, SRMR = 0.081 > 0.08, CFI = 0.634 < 0.9, TLI = 0.617 < 0.9). Together, these findings suggest that no single-factor dominates the variance, indicating that CMB is not a serious concern. Harman’s Single-Factor Test: EFA results (Table 4) showed that the first unrotated principal component explained 31.701% of the total variance, below the recommended threshold of 40%. In addition, the single-factor model estimated using CFA demonstrated poor fit indices (Table 5; χ²/df = 7.561 > 5, RMSEA = 0.091 > 0.08, SRMR = 0.081 > 0.08, CFI = 0.634 < 0.9, TLI = 0.617 < 0.9). Together, these findings suggest that no single-factor dominates the variance, indicating that CMB is not a serious concern. ULMC Test: We compared the model incorporating a common method variance (CMV) factor (11 factors + CMV) with the baseline measurement model (11 factors), with fit indices shown in Table 5. The results indicated that adding the CMV factor did not substantially improve model fit compared to the baseline. Changes in the fit indices did not exceed the recommended validation criteria33, suggesting that the effects of the unmeasured method have limited explanatory power in this study. ULMC Test: We compared the model incorporating a common method variance (CMV) factor (11 factors + CMV) with the baseline measurement model (11 factors), with fit indices shown in Table 5. The results indicated that adding the CMV factor did not substantially improve model fit compared to the baseline. Changes in the fit indices did not exceed the recommended validation criteria33, suggesting that the effects of the unmeasured method have limited explanatory power in this study. Full Collinearity Assessment: We calculated the VIFs for all latent variables serving as predictors in the structural model (Table 6). The results showed that all the VIF values were well below the conservative threshold of 3.3, with the highest value being 2.606 (for Satisfaction). Therefore, no serious collinearity issues exist at the structural model level. Full Collinearity Assessment: We calculated the VIFs for all latent variables serving as predictors in the structural model (Table 6). The results showed that all the VIF values were well below the conservative threshold of 3.3, with the highest value being 2.606 (for Satisfaction). Therefore, no serious collinearity issues exist at the structural model level. Table 4Harman’s single-factor test.ComponentInitial EigenvaluesTotal% of VarianceCumulative %114.58331.70131.70122.9406.39238.09332.0884.53842.63141.9004.13146.76251.8143.94350.70561.7003.69554.40071.4783.21457.61481.3302.89160.50591.1412.48062.984 Harman’s single-factor test. Table 5Model fit comparison.Fitχ2dfχ2/dfRMSEANFIRFIIFISRMRCFITLIBaseline Model1427.0129341.5280.0260.9240.9160.9720.0310.9720.969Single-Factor Model7478.0049897.5610.0910.6010.5830.6350.0810.6340.617Baseline Model + CMV Factor1307.8418881.4730.0240.9300.9190.9770.0250.9760.972Criteria< 5< 0.08< 0.08> 0.9> 0.9 Model fit comparison. Table 6Full collinearity test.VariableVIFPerceived Anthropomorphism (ANT)1.265Expectation Confirmation (CE)1.497Intention to Use (CI)2.210Perceived IQ1.588Perceived Attractiveness (PA)1.244PC1.233Perceived Ease of Use (PE)1.190PI1.553PU1.412Satisfaction (SAT)2.606AI Trust (TR)1.113 Full collinearity test. Based on converging results from these methods, we conclude that CMB is not a significant concern in this study. Hypothesis testing The hypothesis test results are shown in Table 7. The path coefficient from expectation confirmation to satisfaction was 0.254 (P < 0.01), and the path coefficient from satisfaction to acceptance intention was 0.740 (P < 0.001), indicating significant positive relationships between expectation confirmation and satisfaction and between satisfaction and acceptance intention, thus supporting hypotheses H1a and H1b. The path coefficients from PU, perceived attractiveness, perceived anthropomorphism, PI, perceived information quality (IQ), PC, AI trust, and perceived ease of use to satisfaction were 0.137, 0.065, 0.078, 0.136, 0.086, 0.162, 0.101, and 0.105, respectively, with P values less than 0.05, indicating a significant positive impact on satisfaction and supporting hypotheses H2a, H3a, H3c, H4a, H4c, H4e, H5a, and H6b. The path coefficients from expectation confirmation to PU, perceived attractiveness, perceived anthropomorphism, PI, perceived IQ, PC, and AI trust were 0.424, 0.596, 0.732, 0.953, 0.694, 0.635, and 0.703, respectively. All associated P values were less than 0.05, indicating a significant positive impact of expectation confirmation on these variables, supporting hypotheses H2c, H3b, H3d, H4b, H4d, H4f, and H5b. The path coefficients from perceived ease of use to PU and from PU to acceptance intention were 0.212 and 0.103, respectively, with P values less than 0.01, supporting hypotheses H6a and H2b. The results are shown in Fig. 1. Table 7Structural equation model test results.HypothesisHypothesized PathEstimateStandard ErrorSignificanceTest ResultH1aCE → SAT0.2540.084**SupportedH1bSAT → CI0.7400.049***SupportedH2aPU → SAT0.1370.027***SupportedH2bPU → CI0.1030.034**SupportedH2cCE → PU0.4240.051***SupportedH3aPA → SAT0.0650.018***SupportedH3bCE → PA0.5960.058***SupportedH3cANT → SAT0.0780.026**SupportedH3dCE → ANT0.7320.061***SupportedH4aPI → SAT0.1360.025***SupportedH4bCE → PI0.9530.068***SupportedH4cIQ → SAT0.0860.031**SupportedH4dCE → IQ0.6940.049***SupportedH4ePC → SAT0.1620.034***SupportedH4fCE → PC0.6350.050***SupportedH5aTR → SAT0.1010.032**SupportedH5bCE → TR0.7030.050***SupportedH6aPE → PU0.2120.045***SupportedH6bPE → SAT0.1050.026***Supported***, **, and * represent P < 0.001, P < 0.01, and P < 0.05, respectively. Structural equation model test results. ***, **, and * represent P < 0.001, P < 0.01, and P < 0.05, respectively. Theoretical framework. Note: “+” indicates a positive correlation. Alt text: Diagram showing the theoretical framework that integrates the technology acceptance model (TAM) and expectation confirmation theory (ECT), including constructs such as perceived usefulness, perceived ease of use, satisfaction, and intention to use, as well as their interrelationships. Path analysis: chain mediation effect As shown in Table 8, the chain mediation test results, conducted using PROCESS Model 6 with 5000 bootstrap samples and controlling for demographics and familiarity, revealed significant mediation effects for seven of the eight tested pathways. Specifically, the pathway involving perceived attractiveness as the primary mediator (Pathway 4: CE→PA→CI) did not show a significant overall indirect effect from expectation confirmation (CE) to acceptance intention (CI), as the 95% confidence interval contained zero. Table 8Serial mediation effects.No.Mediating variablePathwayIndirect effectBoot S.E.95%CIMediation effectLower limitUpper limit1Perceived IQ, Public SatisfactionCE→IQ→CI0.0600.0160.0300.095YesCE→SAT→CI0.1750.0240.1290.224YesCE→IQ→SAT→CI0.0630.0140.0380.093Yes2Perceived Intelligence, Public SatisfactionCE→PI→CI0.0630.0180.0320.100YesCE→SAT→CI0.1600.0240.1150.208YesCE→PI→SAT→CI0.0630.0140.0380.092Yes3PC, Public SatisfactionCE→PC→CI0.0590.0190.0270.100YesCE→SAT→CI0.1600.0210.1190.204YesCE→PC→SAT→CI0.0510.0150.0240.081Yes4Perceived Attractiveness, Public SatisfactionCE→PA→CI0.0110.009−0.0040.031NoCE→SAT→CI0.2380.0350.1740.311YesCE→PA→SAT→CI0.0300.0090.0150.050Yes5Perceived Anthropomorphism, Public SatisfactionCE→ANT→CI0.0540.0140.0280.083YesCE→SAT→CI0.1930.0260.1440.247YesCE→ANT→SAT→CI0.0540.0120.0330.080Yes6AI Trust, Public SatisfactionCE→TR→CI0.0780.0190.0430.119YesCE→SAT→CI0.1530.0210.1130.195YesCE→TR→SAT→CI0.0660.0160.0380.099Yes7PU, Public SatisfactionCE→PU→CI0.0490.0180.0180.090YesCE→SAT→CI0.1820.0230.1390.228YesCE→PU→SAT→CI0.0500.0150.0230.082Yes8PU, Public SatisfactionPE→PU→CI0.0660.0220.0280.114YesPE→SAT→CI0.1880.0250.1420.242YesPE→PU→SAT→CI0.0740.0180.0420.112Yes Serial mediation effects. All other tested pathways demonstrated significant mediation effects. Therefore, expectation confirmation influences acceptance intention indirectly by enhancing perceived IQ, PI, PC, perceived anthropomorphism (ANT), AI Trust (TR), and PU, which in turn increase satisfaction (SAT) and/or directly affect acceptance intention (CI). Thus, perceived service capabilities (IQ, PI, perceived convenience (PC)), appearance-related human-likeness (ANT), the human-machine relationship (TR), and PU all function as significant mediators in the relationship between expectation confirmation and acceptance intention. Additionally, the analysis confirmed the core mediation mechanism proposed by ECT (“expectation confirmation → satisfaction → acceptance intention”) in the significant pathways. Pathway 8 also demonstrated a significant mediation effect, aligning with the perspective of TAM. Endogeneity test To further assess the robustness of the model estimation results, we conducted an endogeneity test on key model paths using the Gaussian Copula method, as outlined in the methodology section. The results showed that the coefficients of all constructed Gaussian Copula paths were not statistically significant. According to the criteria proposed by Hult et al.34 and Hair et al.35, the general non-significance of the Gaussian Copula paths suggests no statistical evidence of substantial endogeneity bias in the primary path estimations of this model. While endogeneity can never be completely ruled out in studies using observational data, the results of this rigorous test increase our confidence in the model’s reliability. These findings suggest that endogeneity did not exert a substantive influence on the study’s core conclusions, including hypothesis testing and mediation analysis. Therefore, the theoretical inferences and managerial implications derived from the model demonstrate good robustness. Discussion This study integrates ECT and TAM based on government digital human technology to analyze user perceptions and satisfaction. It introduces core variables such as perceived IQ, PI, perceived attractiveness, perceived anthropomorphism, PC, and AI trust, and constructs a structural equation model to deeply explore the factors and pathways influencing public acceptance of government digital humans. First, expectation confirmation, PU, and user satisfaction are important factors influencing the public’s acceptance intention of government digital humans. Expectation confirmation had the most significant positive impact on satisfaction, indicating that when public expectations aligned with actual experiences, satisfaction with government digital humans was higher. Previous studies have emphasized the importance of expectation confirmation in user acceptance behavior17,36,37, and this study provides specific support in the field of government digital humans. This phenomenon can be intuitively understood as follows: when individuals perceive that their expectations of a digital service are either met or exceeded, they naturally experience greater satisfaction and are consequently more likely to continue using or adopting the service. Second, there is a significant positive impact between satisfaction and acceptance intention, indicating that as long as experience and satisfaction with government digital humans are good, public acceptance intention will be higher, which is consistent with existing research findings38,39. Additionally, PU has a significantly positive impact on satisfaction and acceptance intention, with a stronger effect on satisfaction. When users perceive a digital service as useful and believe it improves efficiency and quality, they feel more satisfied. This increased satisfaction then enhances their willingness to accept and continue using the technology. When the public believes that AI can effectively improve service efficiency and quality during the actual use of government digital humans, their satisfaction increases, making them more willing to accept emerging technology22. Moreover, expectation confirmation also has a significant positive impact on PU, further emphasizing the importance of public expectations in forming perceptions of government digital humans. It is noteworthy that PU also serves as a mediator in the impact of expectation confirmation on satisfaction. Gupta et al.39 confirmed this mechanism in their study of smart fitness wearables. Third, the appearance of government digital humans significantly enhances public satisfaction, thereby stimulating public acceptance intention. In terms of appearance characteristics, expectation confirmation significantly influences perceived anthropomorphism and attractiveness. Both perceived anthropomorphism and perceived attractiveness have significant positive impacts on satisfaction. Huang and Yu17 emphasized in their study on virtual news anchors that the more similar virtual images are to humans in appearance and personality, the higher their public acceptance. The more anthropomorphic the traits of government digital humans meet public expectations, the better the user experience. Therefore, in the future, government digital humans should incorporate personalized and anthropomorphic design elements to better meet public emotional needs. Furthermore, perceived attractiveness determines the first impression of whether the public is willing to learn about and use the technology. Beyond appearance, future efforts should focus on precisely meeting public needs to enhance internal attractiveness. Park et al.40 categorized the attractiveness of service robots into relational, appearance, and functional types, all of which can elicit positive emotions from users. Furthermore, perceived attractiveness significantly influences satisfaction (H3a) and shapes the public’s first impression. Our mediation analysis shows that although the direct path from expectation confirmation through perceived attractiveness to acceptance intention (CE→PA→CI) was not significant, a significant indirect effect was observed through the sequential path via satisfaction (CE→PA→SAT→CI). Therefore, enhanced attractiveness, driven by confirmed expectations, contributes to acceptance intention mainly by increasing user satisfaction, emphasizing the role of attractiveness in the design of government digital humans40. Fourth, the service capabilities of government digital humans are directly related to the enhancement of public satisfaction and influencing public acceptance intentions. In terms of service capabilities, expectation confirmation significantly influences PI, perceived IQ, and PC. All three perceptions have a significant positive impact on satisfaction. As an intelligent service method, the intelligence level of government digital humans indicates the quality of service. It is crucial to enhance their intelligent interaction capabilities, which involves not only technical innovation but also comprehensive and systematic considerations in user interface design and service process optimization24. Additionally, the accuracy, timeliness, and comprehensiveness of the information provided are key factors affecting public expectations and satisfaction. IQ is directly related to public satisfaction with the output results41. When information provided by government digital humans meets public expectations, satisfaction levels increase. Convenience is also an important measure of public service; the greater the convenience in the operation of government digital humans, the smoother the public experience. Therefore, during the digital transformation process, government departments should adhere to public needs by providing convenient, simple, efficient, and intelligent digital services to truly implement the “one-stop service” concept. Furthermore, PI, IQ, and convenience all serve as mediating variables in the relationship chain, with expectation confirmation significantly positively influencing satisfaction through these factors. Fifth, public trust in AI is crucial for enhancing satisfaction. In terms of human-machine relationships, expectation confirmation significantly influences AI trust, which in turn significantly promotes public satisfaction. Currently, privacy and security issues have a major negative impact on public trust in AI. This phenomenon indicates that addressing privacy and security issues is crucial for enhancing public trust in AI. Yigitcanlar et al.15 found that, compared to concerns about AI being more intelligent than humans, the public is more worried about AI infringing on privacy. Bedué and Fritzsche29 discovered that when users perceive risks associated with AI, such as data breaches, they are less likely to use it. Public trust in government digital humans essentially reflects trust in AI. Developers need to invest more in human-computer interaction, system transparency, credibility, and security to enhance public trust in AI technology, thereby promoting increased satisfaction. Trust in AI also serves as a mediating variable for the impact of expectation confirmation on acceptance intention. Sixth, perceived ease of use can enhance perceptions of usefulness and improve public satisfaction. In line with classic theories in the field of human-computer interaction, ease of use is often considered the cornerstone of the user experience22. Human-computer interaction systems with high ease of use are more easily accepted and mastered by the public, resulting in more positive experiences22,42. In the design and development of government digital humans, emphasizing ease of use not only helps improve public operational efficiency but also directly influences overall public recognition. Therefore, in the design process of government digital humans, optimizing aspects such as interface design, operational processes, and user guidance is beneficial for comprehensively enhancing the user experience and advancing the digitalization of government services. This study addresses why and how the public accepts digital humans from the perspective of public demand, thereby enriching and improving the research framework for digital and intelligent government services. The theoretical framework integrates ECT and TAM and extends existing theories by introducing three key dimensions: appearance characteristics (e.g., perceived attractiveness and perceived anthropomorphism), service capabilities (including perceived IQ, PI, and PC), and TR (AI trust). This comprehensive theoretical model offers a richer set of dimensions and indicators for understanding the public acceptance of government digital humans, providing valuable insights into their future upgrades and promotion. As AI continues to be embedded in digital governance, government agencies can enhance public acceptance of government digital humans, prevent mismatches between public service supply and demand, reduce digital divides, and avoid ineffective service deployments through the following three strategies: First, we should optimize the supply strategy for government digital humans from both functional enhancement and appearance design perspectives. The primary purpose of the public use of government digital humans is to obtain high-quality information. Enhancing problem recognition capabilities, transaction processing speed, and efficiency is crucial for maximizing public satisfaction. Additionally, it is important to avoid traditional, clunky, and rigid robot images. Government digital humans should increasingly integrate metaverse technologies to make their appearance, voice, tone, and actions as anthropomorphic as possible while ensuring that their overall demeanor is friendly and approachable. This enhances the service’s attractiveness and warmth. Second, the operational difficulties of government digital humans should be reduced. Operational complexity is a major barrier to the user adoption of new technologies. Creating simpler and more intuitive service interfaces, adding voice and text prompts for related operations, and reducing the learning curve can help alleviate public concerns regarding the use of this technology. Additionally, the operational design of government digital humans should include considerations for vulnerable groups, addressing their specific needs to ensure the interface is accessible and easy to use. Ensuring user privacy is also essential for enhancing public trust. Privacy breach risks significantly reduce public trust in AI services. Maintaining data security, clearly defining departmental responsibilities, and providing a secure environment for public use are crucial for alleviating privacy concerns and increasing acceptance intentions. Transparent policies and practices are essential for building and maintaining public trust. Despite offering valuable insights, the study limitations must be acknowledged. First, although the sample size was large (N = 801) and demographically balanced, the reliance on an online platform (Credamo) may limit representativeness, requiring caution in generalizing the findings. Second, while the use of video stimuli was necessary for standardization and feasibility, it differs from actual user interaction. This method cannot fully capture the complexity of real-world experience or predict long-term adoption behavior, thus limiting external validity. Third, the cross-sectional design prevents analysis of changes over time. Fourth, the reliance on self-reported data may introduce reporting bias. Future studies should address these limitations. Research employing more representative sampling, especially targeting underrepresented groups or regions, is needed. Studies involving direct interaction with government digital humans—through laboratory experiments or field studies—are crucial for capturing authentic user experience. Longitudinal designs are also essential to examine the evolution of public attitudes and trust over time. Finally, integrating objective behavioral data (e.g., usage logs, task completion rates) with self-reported responses will yield a more comprehensive understanding of acceptance. Addressing these areas will advance knowledge of effective and responsible implementation of government digital humans in public services. Materials and methods Research hypotheses Prior studies have increasingly adopted integrated theoretical frameworks to explain technology acceptance and continuance intention, particularly in contexts where user behavior is shaped by multiple cognitive and experiential mechanisms. For example, Soren and Chakraborty integrated the theory of planned behavior, flow theory, and habit to explain consumers’ continuance intention toward over-the-top platforms, showing that complementary theories can jointly capture different dimensions of post-adoption behavior43. In a similar vein, the acceptance of government digital humans involves both post-use evaluations, such as expectation confirmation and satisfaction emphasized by ECT, and technology-related beliefs, such as perceived usefulness and perceived ease of use emphasized by TAM. Therefore, integrating ECT and TAM is theoretically appropriate for this study, as it provides a more comprehensive framework for explaining public acceptance intention toward government digital humans. Extended expectation confirmation theory Expectation confirmation, satisfaction, and intention to use According to ECT, users develop initial expectations about the objects they will use based on their cognition and experience. After usage, they evaluate the performance of the product based on their actual experiences, leading to a comparison between initial expectations and actual experiences, resulting in either confirmation or disconfirmation17,45. These users typically refer to the general public who actually use these products or services. Expectation confirmation occurs when users’ expectations align with their actual experiences, whereas disconfirmation occurs when a discrepancy exists between them46. Similarly, when using government digital humans, the public compares their pre-use expectations with their post-use experiences, resulting in either confirmation or disconfirmation. Relevant research indicates that when users’ initial expectations match their actual usage levels, they experience higher levels of satisfaction, reflecting a psychological and emotional comparison between expectations and reality47. When actual usage meets or exceeds initial expectations, user satisfaction manifests as pleasant and positive feelings. Conversely, when actual usage falls short of expectations, satisfaction is expressed as negative and unpleasant emotions36. Public acceptance intention is the level of recognition and acceptance formed by the public based on their value judgments and perceptions of the objects they interact with48. ECT posits that user satisfaction positively influences public acceptance intentions. The prerequisite for public acceptance of government digital humans is that the services provided must satisfy the public. If the public is dissatisfied with government digital humans, this psychological state drives them to discontinue the use of the technology. For instance, Gan found that user dissatisfaction increases the intention to close a video in his study on discontinuance behavior on short-video platforms49. Similarly, Tam et al.50 confirmed this relationship in their study on user attitudes toward mobile apps. Therefore, this paper proposes the following hypotheses: H1a: Public expectation confirmation positively influences satisfaction with government digital humans. H1b: Public satisfaction positively influences the intention to use government digital humans. Perceived usefulness PU is a key user perception variable in classic ECT. This variable refers to the extent to which users believe that using a particular system enhances their job performance51. When users realize that using government digital humans facilitates their administrative processes and reduces operational costs, their satisfaction with digital humans increases, leading to a positive attitude toward them. PU has a positive relationship with user satisfaction. For instance, Lee38 found that when users perceived an e-learning system as beneficial for enhancing their knowledge, their satisfaction with the system increased, thereby boosting their acceptance intentions. Numerous studies have shown that PU plays a crucial role in user satisfaction and continuous usage intentions22,52,53. Additionally, according to Cognitive Dissonance Theory, if users’ prior expectations or cognitive experiences are inconsistent with the actual results, they may experience cognitive dissonance45. To reduce this dissonance, users of government digital humans attempt to adjust their perception of usefulness to align more closely with reality, which implies that expectation confirmation increases PU, whereas expectation disconfirmation reduces it39. Similar cognitive dissonance may apply to other beliefs related to the use of government digital humans, such as perceived attractiveness, perceived anthropomorphism, PI, perceived IQ, PC, and AI trust. This paper hypothesizes the following: H2a: PU positively influences public satisfaction with government digital humans. H2b: PU positively influences the public’s intention to use government digital devices. H2c: Expectation confirmation positively influences PU. Appearance characteristics: perceived attractiveness and perceived anthropomorphism Currently, scholars have explored different influencing factors based on ECT according to the differences in research objects and contexts49. For example, Oghuma et al.36 added user interface, perceived enjoyment, and perceived security to the basis of PU to examine user acceptance of mobile instant messaging. Cheng37, in studying user acceptance of hospital electronic systems, incorporated Flow theory to complement ECT. In this study, we also expanded user perception factors based on the application context of government digital humans. In terms of appearance, government digital humans serve the public through digital human avatars, typically possessing human-like physical and personality traits that mimic human expressions, actions, and voices, thus enhancing the realism of interaction with the public16. With the rise of metaverse technology, digital human avatars are being used in a wide range of service scenarios. The anthropomorphism of digital human avatars enhances their functional use, simplifying and promoting human-computer interactions14. People tend to project social patterns and expectations of human interaction onto human-computer interactions. When they find communication with robots akin to real social interactions, human-computer interaction is more readily accepted40. Additionally, the appearance, functional, and relational characteristics of AI attractiveness bring positive emotional value to the public, further stimulating user acceptance of AI services24. Huang and Yu17 found that the higher the perceived attractiveness and anthropomorphism of virtual avatars, the greater the public’s satisfaction and acceptance intention. Therefore, this study selected perceived attractiveness and perceived anthropomorphism as variables representing the public perception of the appearance characteristics of government digital humans. The former represents the attractiveness of a digital human’s appearance and demeanor to the public, whereas the latter indicates the extent to which the digital human resembles a real person. The study hypothesizes the following: H3a: Perceived attractiveness positively influences public satisfaction with government digital humans. H3b: Expectation confirmation positively influences the perceived attractiveness of government digital humans. H3c: Perceived anthropomorphism positively influences public satisfaction with government digital humans. H3d: Expectation confirmation positively influences the perceived anthropomorphism of government digital humans. Service capabilities: perceived intelligence, perceived information quality, and perceived convenience In terms of service capabilities, the fundamental goal of government digital humans as providers of public services is to deliver services that satisfy the public and respond to their needs, which is a major source of competitive advantage. PI mainly depends on individuals’ perception of the performance and capabilities of AI applications. Generally, when intelligent entities exhibit characteristics such as problem-solving, goal achievement, reasoning, and learning, people perceive them as intelligent49,53. When users perceive an interacting entity as highly intelligent, they tend to have a positive attitude toward its service capabilities, thereby enhancing their acceptance intentions. PI represents the ability and professionalism of government digital humans to meet unstructured public service needs53. IQ reflects the excellence of the information output by the service provider and directly influences the public’s evaluation of digital human services. Accurate and relevant information output that meets public needs is crucial for effective user communication because information that directly addresses user needs fosters higher levels of satisfaction41. Although the content of government digital human services may shift according to the service context, high-quality information output is key to maintaining user engagement. PC is an important consideration when choosing government digital humans over other service providers based on opportunity costs. On one hand, convenience directly reduces the time, money, and effort costs for users in seeking and utilizing services, thereby directly enhancing public satisfaction with the service54. On the other hand, convenience signals that public service design is centered on public needs, embodying the principle of serving the people: H4a: PI positively influences public satisfaction with government digital humans. H4b: Expectation confirmation positively influences the PI of government digital humans. H4c: Perceived IQ positively influences public satisfaction with government digital humans. H4d: Expectation confirmation positively influences the perceived IQ of government digital humans. H4e: PC positively influences public satisfaction with government digital humans. H4f: Expectation confirmation positively influences the PC of government digital humans. Human-machine relationship: AI trust In terms of human-machine relationships, the experiential aspect of human-machine interaction has begun to exhibit social attributes and significance similar to real interpersonal interactions. The presence of social cues makes trust in AI an increasingly important factor influencing human-machine interaction in the public use and selection of AI29. Transparency, explainability, certification, public knowledge, and experience are crucial for building trust15,55. Public trust in AI enhances engagement with the targeted service and increases user loyalty to the service56. For instance, research indicates that AI can significantly improve public satisfaction with the service by cultivating positive psychological impressions. These positive psychological impressions may include acknowledgement of AI’s capabilities, trust in AI’s behavior, and satisfaction with the quality of AI services49,57. Additionally, for anthropomorphic technology applications, Lankton et al.57 proposed measuring user trust in three dimensions: ability, benevolence, and integrity. This multidimensional trust model aids in gaining a more comprehensive understanding and improving user trust in AI. Therefore, this study hypothesizes the following: H5a: AI trust positively influences public satisfaction with government digital humans. H5b: Expectation confirmation positively influences AI trust in government digital humans. Technology acceptance model TAM posits that PU and user attitudes jointly determine the usage of technology. User attitudes refer to the psychological evaluation users have toward the services they receive. In this study, user satisfaction was used to represent user attitudes toward government digital humans because public satisfaction with the services provided by government digital humans often determines their attitude toward the technology58. Satisfaction is a dimension of attitude, a treatment also employed by Na et al.59. According to the TAM, all else being equal, people are more inclined to engage in behaviors toward which they hold a positive attitude22. Thus, in the application of government digital humans, public satisfaction influences acceptance intentions. As a belief or perception, PU can directly influence the public’s acceptance intention when emotions do not fully activate the attitude toward triggering usage intentions, as the potential performance improvement brought about by this usefulness serves as an incentive55. Perceived ease of use refers to the degree to which a user expects the target product or service to be easy to use50. Perceived ease of use influences user attitudes through two fundamental mechanisms: self-efficacy and instrumentality. As a tool, the easier a technology is to interact with, the greater the user’s sense of efficiency and personal control, both of which directly induce a positive attitude22. Additionally, improving perceived ease of use can directly affect PU. Enhanced ease of use can save effort by allowing individuals to accomplish more work with the same amount of effort. If the public perceives this change, their perception of its usefulness will also increase61. On one hand, PU and ease of use can directly enhance public attitudes and intentions. Cheon et al.50 found that PU and perceived ease of use have a significant positive impact on university faculty’s intention to adopt mobile information platforms. On the other hand, perceived ease of use indirectly affects public attitudes and behavior by enhancing PU. Venkatesh and Davis62 noted this indirect effect in the extended TAM. This study posits that when the public realizes that government digital humans are easy to operate, they will not only be satisfied with the technology itself but also recognize that ease of operation can enhance work efficiency, indicating that perceived ease of use can enhance PU. This study hypothesizes the following: H6a: Perceived ease of use positively influences PU. H6b: Perceived ease of use positively influences public satisfaction with government digital humans. The research hypotheses and theoretical framework of this study are presented in Fig. 2. Structural equation model hypothesis testing. Alt text: Diagram showing the path coefficients and significance levels of the structural equation model, illustrating the relationships and positive correlations among variables such as perceived usefulness, perceived ease of use, satisfaction, and intention to use. Research design Questionnaire design Data were collected using a survey questionnaire, which was designed to include the following four parts: (1) Questionnaire instructions detailing the purpose of the survey, basic requirements, and guidelines. (2) Collection of demographic information of the participants, including gender, age, education level, and their understanding of government digital humans. (3) To ensure participants fully understood the service process of government digital humans, they were required to watch two introductory videos: one on the news report of the Jiangxi Nanchang “Xiao Gan Shi” self-service machine for drainage permit applications, and another on consulting a highly anthropomorphic digital virtual police officer in terms of voice, appearance, and actions. (4) The participants completed a questionnaire to measure the observed variables. Variable sources This study included 11 latent variables. The independent variables were expectation confirmation and perceived ease of use. The mediating variables were perceived IQ, PI, perceived attractiveness, perceived anthropomorphism, PC, AI trust, PU, and satisfaction. The dependent variable was the intention to use. This study adopted mature scales and used a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree). Items with factor loadings less than 0.6 were removed, resulting in 46 retained items (Table 9). Table 9Measurement items and sources of variables.VariableItemQuestionSourcePerceived Information Quality (IQ)IQ1I find the information provided by the government digital human to be highly credible.Ghasemaghaei and Hassanein63IQ2I find the information provided by the government digital human to be very accurate.IQ3I find the information provided by the government digital human to be rich, complete, and comprehensive.IQ4I find the information provided by the government digital human to be very clear and understandable.Perceived Intelligence (PI)PI1I believe the government digital human has strong professional capabilities and is competent in providing government consultation services.Balakrishnan et al24PI2I believe the government digital human can provide government consultation services in a more efficient and intelligent manner.PI3I believe the government digital human possesses good government service capabilities.PI4I believe the government digital human has the professional skills to provide government consultation services.Perceived Attractiveness (PA)PA1I find the government digital human’s facial appearance to be very attractive.Filieri et al64.PA2I find the government digital human to be well-groomed and elegant.PA3I like the appearance of the government digital human.PA4I find the image of the government digital human to be attractive to me.Perceived Anthropomorphism (ANT)ANT1I find the appearance of the government digital human to be very human-like.Huang and Yu17ANT2I find the voice of the government digital human to be very natural.ANT3I find the government digital human’s gestures to be graceful and smooth when providing consultation services.ANT4I find the facial expressions of the government digital human to be very vivid and natural when answering questions.Perceived Convenience (PC)PC1The use process of the government digital human kiosk is very convenient.Lin26PC2The government digital human kiosk is very patient and makes the service process clearer and easier to understand.PC3The government digital human kiosk helps me quickly obtain the information I need.AI Trust (TR)TR1The interactions with the government digital human are effective and competent.Bedué and Fritzsche29; Lankton et al57.TR2The government digital human can fulfill its role well.TR3The government digital human is capable and proficient in its tasks.TR4The government digital human’s responses are honest.TR5The government digital human’s service is sincere and genuine.TR6The government digital human acts in my best interest.Perceived Usefulness (PU)PU1Using the government digital human kiosk helps improve government service performance.Davis et al60; Bhattacherjee45PU2Using the government digital human kiosk helps improve the efficiency of government services.PU3The government digital human kiosk makes handling tasks faster and more efficient.PU4The government digital human kiosk meets my service needs and is very helpful to me.PU5Overall, the government digital human kiosk is very useful for handling resident affairs.Perceived Ease of Use (PE)PE1I find the operation of the government digital human kiosk to be very simple.Davis et al60. ; Venkatesh and Davis62PE2Learning to use the government digital human kiosk is very easy for me.PE3I am confident that I can quickly learn and proficiently use the government digital human kiosk.PE4Overall, I find the government digital human kiosk easy to use.Expectation Confirmation (CE)CE1I find the government digital human to be more natural and fluent than I expected.Bhattacherjee45;Huang and Yu17CE2I find the appearance of the government digital human to be more realistic than I expected.CE3I find the performance of the government digital human to be better than I expected.CE4Overall, most of my expectations for the government digital human have been confirmed.Satisfaction (SAT)SAT1I believe using the government digital human in government service halls is a wise choice.Li et al.61SAT2Communicating with the government digital human makes me feel satisfied.SAT3I am satisfied with the performance of the government digital human.SAT4I prefer the government digital human to provide consultation services over human services.SAT5Overall, the service of the government digital human is satisfactory.Intention to Use (CI)CI1I am highly likely to continue using the government digital human for government consultation services in the future.Ashfaq et al65CI2If possible, I will recommend the government digital human to others.CI3I will try to use the government digital human to obtain more government-related information in the future. Measurement items and sources of variables. Bhattacherjee45; Huang and Yu17 Data collection Data for this study were collected using Credamo, a professional online survey platform. Credamo was chosen for its large user base and, importantly, its ability to set specific screening criteria for targeted participant recruitment. This function allowed better control over the sample composition and helped achieve a more balanced demographic profile for the study. Each participant who passed the quality check received a reward of 3 yuan (CNY), resulting in 1000 questionnaires. For data quality control, the research team excluded questionnaires with the following issues: (1) response time that was too long (> 2100 s) or too short (< 300 s); (2) multiple consecutive items with the same choice; and (3) incorrect answers to quality check questions based on the introductory videos. In the end, 801 questionnaires were retained, yielding an effective response rate of 80.1%. This study was conducted in accordance with the ethical principles of the World Medical Association Declaration of Helsinki and relevant institutional guidelines and regulations. The study protocol, including the online questionnaire survey and data collection procedures, was reviewed and approved by the Medical Ethics Committee of Guangxi University (Approval No. GXU-2025-003). The study was a non-clinical questionnaire-based study and involved no physical intervention, disease diagnosis, treatment, or prevention procedures. Before participating in the survey, all participants were provided with information about the purpose of the study, the voluntary nature of participation, the confidentiality and anonymity of their responses, and their right to withdraw. Informed consent was obtained from all participants before they entered the questionnaire. The survey was conducted anonymously, and no personally identifiable information was collected. Empirical testing Reliability and validity analysis The collected questionnaires were subjected to reliability and validity tests using SPSS 27.0. Reliability refers to the consistency, stability, and dependability of measurement results. Common method bias Considering the potential risk of CMB, as the data in this study were mainly collected via self-administered questionnaires at a single point in time, we adopted a strategy combining procedural controls and statistical tests to mitigate and assess its impact. For procedural controls, during questionnaire development, we conducted a pre-survey to refine item wording and reduce ambiguity. Respondents were also assured of anonymity to minimize social desirability bias. For statistical tests, given the limitations of any single diagnostic method, we employed multiple complementary techniques for systematic assessment, following methodological recommendations66: Harman’s Single-Factor Test: As a preliminary diagnostic, this included exploratory factor analysis (EFA) to examine the variance explained by the first unrotated principal component and confirmatory factor analysis (CFA) to assess the goodness-of-fit of a single-factor model. Harman’s Single-Factor Test: As a preliminary diagnostic, this included exploratory factor analysis (EFA) to examine the variance explained by the first unrotated principal component and confirmatory factor analysis (CFA) to assess the goodness-of-fit of a single-factor model. Unmeasured Latent Method Construct (ULMC) Technique: This involved adding a CMV factor to the measurement model and comparing the model’s fit to that of the baseline model to evaluate the explanatory power of the method factor. Unmeasured Latent Method Construct (ULMC) Technique: This involved adding a CMV factor to the measurement model and comparing the model’s fit to that of the baseline model to evaluate the explanatory power of the method factor. Full Collinearity Assessment: This method used variance inflation factors (VIFs) to test collinearity among latent variables in the structural model, helping determine whether CMB caused collinearity problems33. Full Collinearity Assessment: This method used variance inflation factors (VIFs) to test collinearity among latent variables in the structural model, helping determine whether CMB caused collinearity problems33. These multifaceted tests were used to more comprehensively evaluate the potential impact of CMB on this study’s results. The specific findings are presented in the Results section. Hypothesis testing The structural equation model was constructed using Amos Graphics 26.0. The model fit results were as follows: χ²/df = 2.192 (< 3), RMSEA = 0.039 (< 0.08), SRMR = 0.054 (< 0.08), CFI = 0.935 (≥ 0.9), NFI = 0.887 (≥ 0.8), RFI = 0.879 (≥ 0.8), IFI = 0.935 (≥ 0.8), and TLI = 0.930 (≥ 0.9). All indices met their respective threshold criteria, indicating a good model fit67. Path analysis: chain mediation effect This study employed the non-parametric bootstrap method with repeated sampling using PROCESS Module 6 in SPSS to test the chain mediation effects of eight pathways, with gender, age, education level, and familiarity with government digital humans as control variables. After 5000 repeated samples, if the 95% confidence interval did not contain 0, the mediation effect was considered significant68. Endogeneity test Recognizing the limitations of traditional methods and the characteristics of our survey data, this study adopted a more advanced and suitable approach recommended in the structural equation modeling field to systematically test for potential endogeneity within the model. Specifically, we followed the procedural guidance of Hult et al.34 and applied the Gaussian Copula method34,35. A key advantage of this method is its ability to assess endogeneity without relying on external instrumental variables. It does so by directly constructing and testing “Gaussian Copula paths” that model the correlation between antecedent constructs and the error terms of their corresponding endogenous constructs, thus quantifying the presence and extent of endogeneity. The specific results of this test are reported in the Results section. Author contributions Conceptualization, J.H.C. and F.J.; methodology, Y.D.C.; software, F.J.; validation, J.H.C., F.J. and Y.D.C.; formal analysis, F.J.; investigation, Y.D.C.; resources, J.H.C.; data curation, F.J.; writing—original draft preparation, F.J., L.G; writing—review and editing, J.H.C., F.J., Y.D.C., X.Y.Q.; visualization, X.Y.Q.; supervision, J.H.C.; project administration, J.H.C.; funding acquisition, J.H.C. All authors have read and agreed to the published version of the manuscript. Funding This study was supported by the Major Project of the National Social Science Foundation of China, entitled “Innovative and Applied Research on Information Analysis Methods Based on the Integration of Digital and Intelligence” (Grant No. 22&ZD326). Data availability The data used to support the findings of this study are included within the article. Code availability No custom code, bespoke computational tool, new algorithm, new software, or bespoke analytical pipeline was developed or used in this study. The statistical analyses were conducted using standard statistical software, including SPSS 27.0, Amos Graphics 26.0, and the PROCESS macro for SPSS. Declarations Competing interests The authors declare no competing interests. Ethical approval Ethical approval and informed consent are described in the Materials and methods section. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data used to support the findings of this study are included within the article. No custom code, bespoke computational tool, new algorithm, new software, or bespoke analytical pipeline was developed or used in this study. The statistical analyses were conducted using standard statistical software, including SPSS 27.0, Amos Graphics 26.0, and the PROCESS macro for SPSS.