Abstract
This study, positioned at the critical juncture of China’s transition from digitalization to intelligence, explores how Internet use is associated with social trust and perceptions of social fairness, as well as how these relationships vary across different social contexts. Based on nationally representative data from the Chinese General Social Survey (CGSS) 2021, regression analysis and Bootstrap methods were employed to systematically examine the relationship between internet usage frequency and perceptions of social fairness as well as the mediating role of social trust. Urban-rural heterogeneity was identified as the primary source of variation through within-sample comparison, while Henan Province was examined as a validating case to test mechanism robustness. The research found that in the national sample, internet usage frequency was significantly negatively associated with perceptions of social fairness, and this relationship was partially mediated by the decrease in social trust, with the mediating effect accounting for 24% to 37% of the total effect depending on model specification. Notably, urban residents experience twice the negative impact compared to rural residents (β_urban = −.080 vs. β_rural = −.040, p < .01), revealing significant digital divide effects. The Henan case (n = 398) demonstrates consistent mechanism direction with the national sample, with effect sizes even larger (β = −.089) though marginally significant due to smaller sample size, validating the theoretical robustness of the identified pathways. These findings provide important implications for digital governance practice: digital inclusion policies should shift from “universality-oriented” to “quality-oriented”; mechanisms should be constructed to protect social trust in the digital era; and differentiated urban-rural digital development strategies should be implemented to promote positive interaction between digitalization and social harmony.
Plain Language Summary
This study examined how internet use affects people’s feelings about fairness in Chinese society. With over 1.1 billion internet users in China, understanding these effects is crucial for social harmony. We analyzed survey data from 8,000 people across China. We found that people who use the internet more frequently tend to feel society is less fair. This happens because internet use reduces their trust in others—when people trust others less, they also see society as less fair. Trust explains about one-quarter to one-third of this relationship. Importantly, this effect is much stronger for city dwellers than for rural residents. Urban internet users experience twice the negative impact compared to rural users, likely because they encounter more algorithm-driven content showing social inequalities and have weaker face-to- face community connections. These findings suggest that digital policies should go beyond simply providing internet access. Instead, they should focus on helping people use the internet wisely, protecting social trust online, and recognizing that cities and rural areas need different approaches to digital development.
Keywords
Introduction
With the rapid development of new-generation information technologies such as artificial intelligence, big data, and the Internet of Things, the world is experiencing an accelerated transition from digitalization to intelligence. 2024 marks the 30th anniversary of China’s full-functional access to the international Internet. Over these 30 years, China’s Internet has achieved leapfrog development from nothing to something, from small to large, and from large to strong, building the world’s largest and technologically advanced internet infrastructure and establishing the world’s largest online retail market and internet user base. Internet users have grown from 620,000 in 1997 to 1.108 billion in 2024, with the Internet penetration rate rising to 78.6% (CNNIC, 2025).
At this critical juncture of the transition from digitalization to intelligence, understanding how Internet use is associated with social trust and perceptions of social fairness and how these relationships differ across various levels of regional development is not only a key perspective for understanding China’s social transformation in the digital era but also a fundamental theoretical issue for intelligent social governance (Ahmed, 2021).
The advent of the intelligent era has brought new characteristics to the relationship between the Internet and society. On the one hand, technological applications such as algorithmic recommendations, information cocoons, and deepfakes are profoundly reshaping citizens’ ways of obtaining information and forming social cognition. Unlike traditional media environments where information exposure was relatively random and diverse, algorithmic curation creates personalized information ecosystems that may amplify certain content while filtering out others. This algorithmic differentiation effect can exacerbate social cognitive polarization—users are increasingly exposed to information that confirms their existing beliefs while being shielded from contradictory perspectives (Ceron & Memoli, 2015). Information cocoons, reinforced by recommendation algorithms, may intensify users’ perception of social inequality by repeatedly exposing them to similar narratives about unfairness, while deepfake technologies erode the credibility of online information, further undermining the foundation of social trust. On the other hand, the penetration of intelligent technology has intensified the diversification and complexity of the digital divide, evolving from an “access divide” to a “capability divide” and “cognitive divide” (Van Deursen & Van Dijk, 2018). Against this background, existing theories present multiple perspectives on the relationship between internet use and perceptions of social fairness. One view holds that the Internet promotes social equity by lowering barriers to information access, expanding social networks, and increasing channels for participation (Warschauer & Matuchniak, 2010). Conversely, another view asserts that the Internet may exacerbate social inequality because the distribution of digital dividends is constrained by original resource endowments and usage capabilities (Ragnedda, 2018). This theoretical divergence becomes more complex in intelligent transformation, as intelligent technologies can narrow social gaps through inclusive services or widen cognitive divisions due to complex algorithmic biases and technical barriers.
Empirical research also presents contradictory results. Zhu et al. (2020), through research on rural households in China, found that internet use has a significant negative impact on perceptions of social fairness. D. Zhou et al. (2022) further confirmed that internet use reduces individuals’ perception of social fairness by an average of 5%. However, some studies suggest that internet use may produce positive effects under specific conditions. Q. Zhou et al. (2021) showed that digital financial inclusion promotes social integration by providing equal opportunities and rights, enabling low-income and vulnerable groups to enjoy modern financial services. J. Chen and Huang (2021) confirmed that digital government transformation, as a technological driving force, can effectively promote the economic participation of vulnerable groups. These seemingly contradictory research results suggest that internet use and social fairness may not be a simple linear relationship but moderated by a series of mediating variables. Social trust is likely the key mechanism connecting the two.
As a core element of social capital, Social trust has a fundamental role in maintaining social stability (Putnam, 2000; Fukuyama, 2001). In the transition to an intelligent society, the formation mechanism of social trust is undergoing profound changes, shifting from traditional trust based on face-to-face interaction to network trust mediated by technology (Hampton & Wellman, 2018). Research shows that internet use may influence social trust levels by changing social interaction patterns and channels for obtaining information (Friemel, 2016). The way information flows on the Internet differs significantly from traditional media environments; users can access more diverse viewpoints and participate in broader social discussions, and these changes profoundly impact the construction of social trust (Gong et al., 2020). Meanwhile, social trust directly affects residents’ subjective well-being, including perceptions of social fairness (Huhe et al., 2015). Synthesizing these studies reveals our perspective: There should be a connecting relationship between internet use, social trust, and perceptions of social fairness.
As digital economies restructure traditional economies and artificial intelligence reshapes the social division of labor, understanding how internet use is associated with perceptions of social fairness through social trust provides important implications for building an inclusive, intelligent society. However, existing research still has three key gaps: first, there is a lack of systematic verification of the complete mediating pathway of internet use→social trust→perceptions of social fairness; Second, most studies focus on national-level patterns while overlooking internal heterogeneity within societies. China’s rapid urbanization and persistent urban-rural divide create distinct digital usage contexts and social structures, yet few studies have systematically examined how internet-society relationships vary between urban and rural populations. Third, methodologically, there is a need for research designs that combine large-scale representative validation with context-specific case studies to test both the universality and boundary conditions of theoretical mechanisms.
To fill these research gaps, this study, based on nationally representative data from the Chinese General Social Survey (CGSS) 2021, adopts a multi-stage research design: (1) validating the mediating mechanism using the full national sample; (2) identifying sources of heterogeneity through urban-rural and age-cohort comparisons within the national sample; and (3) testing mechanism robustness through a case study of Henan Province, a representative central region province. The study incorporates multidimensional control variables, including age, gender, ethnicity, religious belief, education level, and personal income, to enhance the reliability of results. Specifically, this study aims to answer the following research questions: How does internet usage frequency relate to individuals’ perceptions of social fairness? Does social trust mediate the relationship between internet use and perceptions of social fairness? Do these relationships vary across different social contexts, particularly between urban and rural populations and across age cohorts? Are the mechanisms identified in national data robust when tested in specific regional contexts, such as Henan Province?
Why Henan Province? The selection of Henan Province as a validating case is grounded in both theoretical considerations and empirical characteristics. As a core province in China’s “Rise of Central China” strategy, Henan represents the transitional zone between developed eastern regions and less developed western areas. With an urbanization rate of 55.43% (compared to the national average of 64.72%), GDP per capita of ¥59,410 (ranked 18th among 31 provinces), and internet penetration of approximately 75% (close to the national 78%), Henan occupies a middle-tier position in China’s digital development spectrum. As China’s most populous province with over 42 million rural residents (ranked 1st nationally), Henan exhibits pronounced urban-rural duality characteristic of China’s broader modernization challenges.
In summary, the theoretical contributions of this study are mainly reflected in four aspects: First, by verifying the mediating Model of Internet use→social trust→social fairness perception, it achieves innovative integration of digital divide theory and social capital theory, providing a new perspective for understanding the “digital psychological divide” in the intelligent era —moving beyond material access gaps to cognitive and emotional consequences of digital engagement. Second, identifying urban-rural heterogeneity as a primary source of variation provides critical insights into how digital impacts are structured by pre-existing social stratification. This finding bridges digital divide research with urbanization studies, revealing that digital tools do not operate in a social vacuum but interact with existing inequalities. Third, the case validation approach—testing whether mechanisms identified in large-scale national data hold in specific contexts—provides a methodological template for studying context-dependent digital effects. This design balances generalizability (national validation) with contextual sensitivity (case study), addressing a persistent tension in digital society research. Finally, revealing the micro-psychological mechanisms of internet social effects provides a theoretical foundation for bridging social cognitive divisions in the transition from digitalization to intelligence. At the practical level, the research findings will provide empirical evidence for formulating targeted digital inclusion policies and narrowing urban-rural digital divide, promoting positive interaction between digital intelligent development and social harmony.
Literature Review
Internet Use and Perceptions of Social Fairness
Perceptions of social fairness refer to individuals’ subjective evaluations of the justness of social resource distribution, an important psychological foundation for social harmony and stability (Tyler & Smith, 1995). As the Internet deeply integrates into social life, its relationship with perceptions of social fairness has become a focus of academic attention. Existing views show significant divergence regarding the relationship between internet use and perceptions of social fairness.
On the one hand, the technological empowerment perspective suggests that the Internet can promote social equity by lowering barriers to information access, expanding social networks, and increasing channels for participation. The Internet, as an “information equalizer,” can break traditional information monopolies and help disadvantaged groups access knowledge resources and social opportunities. Norris (2003) points out that the Internet may bridge participation gaps between different groups, enhance citizen empowerment, and increase perceptions of social fairness. This view is supported by some empirical research, such as Robinson et al. (2015), who found that digital technology can promote social inclusion and reduce inequality in specific contexts.
On the other hand, the digital divide perspective argues that the Internet may exacerbate rather than alleviate social inequality. Hargittai’s (2008) concept of the “second-level digital divide” points out that even with equal access opportunities, differences in digital literacy, usage patterns, and ability to benefit will still lead to uneven distribution of technological dividends. Internet use may also reduce perceptions of social fairness through the following mechanisms: first, the information exposure effect, where the Internet increases the frequency of exposure to social inequality phenomena; second, the social comparison effect, which expands the range and frequency of social comparisons, potentially strengthening feelings of relative deprivation (Verduyn et al., 2017); third, the algorithmic differentiation effect, where recommendation systems and information cocoons may exacerbate social cognitive polarization and weaken the basis of consensus.
In the Chinese context, the digital divide perspective has received some support in research by Chinese scholars. Zhang and Liu’s (2021) analysis shows that internet usage frequency negatively correlates with perceptions of social fairness; Li et al. (2023) found a similar pattern in their longitudinal comparison using multiple years of CGSS data.
Based on the above analysis, considering the Chinese context, this study proposes:
Internet Use and Social Trust
Social trust refers to individuals’ general trust expectations toward others and social institutions. Academics have formed three main perspectives on how internet use is associated with social trust.
The time substitution perspective argues that internet use occupies time for face-to-face social interaction, potentially leading to increased social isolation and decreased levels of trust (Nie & Erbring, 2002). Virtual interactions lack non-verbal cues and contextual feedback compared to face-to-face communication, making it challenging to establish deep trust. Stepanikova et al. (2010) found that internet usage time is significantly associated with social suspicion. The anonymity and low-threshold expression in the internet environment may increase conflict and polarization, further weakening the foundation of trust.
Conversely, the network supplementation perspective posits that the Internet may promote social trust formation by expanding social networks and enhancing the ability to maintain existing relationships (Boulianne, 2015). The Internet does not replace but supplements face-to-face interaction, providing new channels for maintaining community connections. Ellison et al. (2007) found that in terms of bridging social capital, social media use is significantly positively correlated with social capital formation.
The third perspective emphasizes that different types of internet activities have different social effects (Shah et al., 2005). This view distinguishes between social, informational, and entertainment-oriented internet use, believing that their direction and strength of influence on social trust differ. Gil de Zúñiga et al. (2012) found that information-oriented social media use positively correlates with social trust, while entertainment-oriented use is negatively correlated. This differentiated effect suggests that the key to how the Internet influences social trust lies not in the amount of use but in the mode of use and content quality (Brandtzæg, 2012).
In the Chinese context, internet use and social trust show complex association patterns. Specifically, there may be a non-linear relationship between internet use and social trust, with moderate use having a positive impact but excessive use potentially producing adverse effects. Han et al.’s (2017) research on forms of social trust in China further indicates that in Chinese society, there are significant differences in the formation mechanisms and manifestations of particularized trust (trust in acquaintances) and generalized trust (trust in strangers). This finding provides an important reference framework for understanding how internet use is associated with different types of social trust. Considering all these viewpoints and conducting this research against the backdrop of China’s social transformation, where traditional relationship culture intertwines with modern digital networks, we believe the time substitution perspective may be more applicable to the Chinese context, namely:
Social Trust and Perceptions of Social Fairness
Social trust and perceptions of social fairness are two key psychological indicators of social integration. Procedural justice theory contends that social trust indirectly shapes perceptions of distributive fairness by influencing cognition of the procedural justice of social institutions (Tyler & Blader, 2003). When individuals have higher trust in social institutions and decision-makers, they are more inclined to believe that resource allocation processes follow fair principles and, thus, are more likely to accept existing distribution results, even if there is a certain degree of inequality. Zmerli and Newton’s (2008) cross-national research shows that after controlling for factors such as socioeconomic status, general social trust significantly correlates with institutional performance evaluation, enhancing perceptions of social fairness.
Social identity theory emphasizes that social trust reduces individuals’ sensitivity to inequality by strengthening social cohesion and collective identity (Hogg, 2016). When individuals have high trust in the social community, they form stronger feelings of belonging and identification, can understand social phenomena from the perspective of collective interests, and consequently hold more tolerant attitudes toward social distribution conditions. Gao and Zhao’s (2022) research found that high-trust individuals demonstrate stronger social inclusiveness, higher acceptance of income disparities, and more positive perceptions of fairness.
The information processing perspective suggests that social trust influences how individuals acquire and interpret social information, adjusting their evaluation of social fairness (van den Bos et al., 1998). High-trust individuals tend to access and believe official information, make positive attributions when facing social uncertainty, reduce attention to unfair phenomena, and enhance the overall evaluation of social fairness. Bianchi et al.’s (2014) experimental research found that participants with high trust in social institutions are more inclined to make fair judgments when facing ambiguous information.
In the context of China’s social transformation, Yang and Tang (2010) found that in Chinese society, interpersonal trust and institutional trust affect perceptions of social fairness through different pathways: institutional trust directly promotes the formation of perceptions of social fairness, while interpersonal trust mainly indirectly affects perceptions of fairness by strengthening social norm identification. Guo and Zhang’s (2024) analysis based on CGSS2021 data also found that social trust has a significant positive impact on perceptions of social fairness, and this relationship remains stable after controlling for demographic characteristics and socioeconomic status.
Existing research consistently indicates that social trust positively is associated with perceptions of social fairness. Based on the above analysis, this study proposes:
The Mediating Role of Social Trust
Integrating the theoretical analysis and empirical evidence from the previous three sections, this part explores social trust as a mediating mechanism through which Internet use is associated with perceptions of social fairness.
From the perspective of information exposure, internet use increases exposure to social inequality and adverse events (Ceron & Memoli, 2015). This information may be negatively associated with trust in social institutions, and related to lower perceptions of social fairness. Under algorithmic recommendation mechanisms, negative information and polarized views are more likely to be disseminated, reinforcing the sense of social division and being associated with lower social trust. From the perspective of social relationship reconstruction, Internet use changes traditional social relationship networks, weakening strong connections based on geographic communities and increasing weak virtual connections based on interests. This change in social relationship structure may be associated with the foundation of trust formation, thereby relating to the construction process of perceptions of social fairness. From the cognitive processing perspective, internet use is associated with how individuals acquire and process social information, changing attribution patterns (Valenzuela et al., 2009) and indirectly shaping evaluations of social fairness by its association with social trust.
In recent years, Chinese scholars have begun to pay attention to the association between internet use, social trust, and perceptions of social fairness. Xu et al.’s (2024) research based on CGSS2017 data found that internet use indirectly affects elderly people’s perceptions of social fairness by influencing social trust, and this mediating effect is particularly significant among elderly people with lower digital literacy. X. Chen and Yang’s (2025) latest research shows that, after controlling for factors such as education, the relationship between internet usage frequency and perceptions of social fairness is partially mediated by social trust. These findings provide important evidence for understanding the internal mechanisms of how the Internet influences perceptions of social fairness.
However, against the background of unbalanced regional development, the mediating mechanism of internet use being associated with perceptions of social fairness through social trust may show significant differences. This heterogeneity may stem from the following aspects: First, differences in internet development level and usage patterns. In economically developed regions, the Internet has deeply integrated into daily life and social interaction, and its influence may be more systematic and profound, while in less developed regions, internet use may still be a supplementary activity with relatively limited social and psychological associations. Second, the moderating role of traditional social structures. Less developed regions typically retain more traditional social relationship networks, and these existing social structures may buffer the associations of internet use with social trust. Third, there are differences in diversity in information acquisition channels. Residents in developed regions usually have more diverse sources of information, and the associations of Internet use with their social cognition is relatively dispersed. In contrast, residents in less developed regions may have a higher dependency on the Internet as an information source but with limited depth and breadth of use.
In summary, this study proposes:
In addition to the hypotheses formed through the literature review above, another contribution of this study is to summarize and illustrate the logical relationships among variables included in existing models. In recent years, Chinese scholars have discussed the relationships between internet use, social trust, and perceptions of social fairness from different perspectives, as shown in Table 1. These studies provide an important historical foundation for this paper. Building on these historical foundations, we have completed the “last mile” of integration and constructed a research model, as shown in Figure 1, which we will test for applicability in the national sample and validate through case study.
Overview of the Relationship between Internet Use, Social trust, and Social Fairness.
Theoretical model framework.
Research Method and Data
Data Source and Sample Characteristics
This study is based on nationally representative data from the latest Chinese General Social Survey (CGSS) 2021 version. CGSS is China’s most authoritative nationwide, comprehensive, and continuous large-scale social survey project, employing multi-stage stratified probability sampling methods with high sample representativeness. The survey design employs three-stage sampling: (1) Primary Sampling Units (PSUs): Counties/districts selected with probability proportional to size; (2) Secondary Units: Neighborhoods/villages randomly selected within PSUs; (3) Households: Systematic sampling within each secondary unit, with one adult (18+) randomly selected per household using Kish selection grid. The sample frame covers all 31 provinces/municipalities/autonomous regions in mainland China, with a completed sample of 8,148 interviews and a response rate of 72.3%. The data are publicly available at the CGSS official website (http://cgss.ruc.edu.cn).
This study uses 8,148 valid samples from the CGSS 2021 data. To ensure the reliability of the analysis results, this study excluded observations with missing values for key variables. The sample distribution is consistent with the initial sample, indicating no obvious systematic bias in the sample screening process.
Variable Measurement
Dependent Variable: Perceptions of Social Fairness
This study uses item A35 in the CGSS questionnaire, “Overall, do you think today’s society is fair or unfair?” to measure perceptions of social fairness. This item uses a 5-point scale, from “1 = completely unfair” to “5=completely fair,” with higher scores indicating stronger perceptions of social fairness. In the data processing, “98 = don’t know” and “99 = refuse to answer” were coded as missing values. Descriptive statistics show that among valid samples, 4.25% of respondents consider society “completely unfair,” 13.88% consider it “relatively unfair,” 21.29% consider it “neither fair nor unfair,” 51.95% consider it “relatively fair,” and 7.92% consider it “completely fair.” This distribution reflects that the overall fairness evaluation of Chinese society shows a relatively positive structure with some variation.
Independent Variable: Internet Use
Internet use is measured through item A28_5 in the CGSS questionnaire: “In the past year, your usage of the internet (including mobile internet).” This item uses a 5-point scale, from “1 = never” to “5 = very frequently,” with higher scores indicating higher frequency of internet use. In the data processing, “98 = don’t know” and “99 = refuse to answer” were coded as missing values. Descriptive statistics show that 28.17% of respondents “never” use the Internet, 6.21% use it “rarely,” 7.58% use it “sometimes,” 20.30% use it “frequently,” and 37.65% use it “very frequently.” This distribution reflects the digital divide phenomenon in China’s digital transformation process. Although most respondents (57.95%) frequently use the Internet, more than a quarter still never use the Internet, showing obvious digital exclusion. This heterogeneity in digital usage provides a good variance foundation for this study to explore the social associations of internet use.
Mediating Variable: Social Trust
Social trust is measured through item A33 in the CGSS questionnaire: “Overall, do you agree or disagree that most people in this society are trustworthy?” This item uses a 5-point scale, from “1 = strongly disagree” to “5 = strongly agree,” with higher scores indicating higher levels of social trust. In the data processing, “98 = don’t know” and “99 = refuse to answer” were coded as missing values. Descriptive statistics show that 3.29% of respondents “strongly disagree,” 13.32% “somewhat disagree,” 13.66% “neither agree nor disagree,” 54.06% “somewhat agree,” and 14.83% “strongly agree.” Chinese citizens have relatively high levels of social trust (nearly 70% of respondents expressed positive trust attitudes), but there are still some differences.
Control Variables
This study includes the following control variables to control for the potential influence of individual characteristics on research results. Age: calculated by subtracting birth year from 2021. The sample has an average age of 51.53 years (SD = 17.56), with a minimum of 18 years and a maximum of 94 years. Gender: a binary variable (0 = female, 1 = male), with females accounting for 54.85% and males accounting for 45.15% of the sample. Ethnicity: a binary variable (0 = non-Han, 1 = Han), with Han accounting for 92.64% and non-Han accounting for 7.36% of the sample. Religious belief: a binary variable (0 = no religious belief, 1 = has religious belief), with those with no religious belief accounting for 92.50% and those with religious belief accounting for 7.50% of the sample. Education level: a four-category variable, 1 = junior high school and below (32.60%), 2 = high school/technical secondary school (28.36%), 3 = junior college/bachelor’s degree (18.27%), 4 = graduate degree and above (20.51%). Personal income: logarithmically transformed personal annual income to reduce the impact of skewed data distribution on estimation results.
The selection of these control variables is based on the consensus in the existing literature on research regarding perceptions of social fairness and social trust (Ma et al., 2024; Knight & Gunatilaka, 2024), which helps eliminate the interference of confounding factors and improve the reliability of research results. Including education level and income, variables can control for the potential influence of socioeconomic status on internet use and social evaluation, while controlling for demographic variables such as age, gender, ethnicity, and religious belief helps reduce the interference of individual heterogeneity on estimation results.
Analytical Strategy
This study adopts a three-stage research design to systematically examine the relationship between internet use and perceptions of social fairness, the mediating role of social trust, and the heterogeneity of these relationships across different social contexts.
Stage 1: Mechanism Validation Using National Sample
In this stage, we validate the core theoretical mechanisms using the full national sample, which includes all 31 provinces including Henan. The following regression equations are constructed:
(1) Perceptions of Social Fairness =β1Internet Use +γX +ε1
(2) Social trust =β2Internet Use +γX +ε2
(3) Perceptions of Social Fairness =β3Internet Use +β4Social trust +γX +ε3
Where X represents a vector of control variables, including age, gender, ethnicity, religious belief, education level, and personal income, ε1, ε2, and ε3 are random error terms. We emphasize that our cross-sectional design can identify associations but cannot definitively establish causality. The proposed pathway (Internet Use → Social Trust → Fairness Perception) remains theoretically driven and requires validation through longitudinal or experimental designs in future research.
This stage adopts a two-phase analysis strategy: First, simplified models (without control variables) and complete models (with control variables) are constructed separately to evaluate model robustness; second, the Bootstrap method (with 1,000 repetitions) is used to calculate standard errors and confidence intervals for precise estimation of the mediating effect. Compared with the traditional Sobel test, the Bootstrap method does not require the sampling distribution to satisfy the normality assumption and can provide more accurate parameter estimation and significance testing (Preacher & Hayes, 2008).
Stage 2: Heterogeneity Exploration Within National Sample
To identify sources of variation in the mechanisms identified in Stage 1, we conduct heterogeneity analysis within the national sample by examining whether relationships differ across theoretically relevant subgroups:
(1) Urban-Rural Heterogeneity (Primary Analysis): China’s persistent urban-rural divide creates distinct contexts for digital engagement and social structure. We estimate separate models for rural residents and urban residents, comparing path coefficients across groups using likelihood ratio tests. This analysis addresses whether digital-society relationships are structured by pre-existing social stratification.
(2) Age-Cohort Heterogeneity (Exploratory Analysis): Given cohort differences in digital nativity and usage patterns, we explore whether mechanisms vary across age groups: Youth (≤35), Middle-aged (36–50), and Older adults (>50). This exploratory analysis helps identify age-related boundary conditions.
This within-sample comparison approach eliminates the statistical power concerns inherent in comparing large and small subsets, providing robust tests of heterogeneity.
Stage 3: Case Validation Using Henan Province
Rather than treating Henan as a “contrast region” separate from the national sample, we examine it as a validating case to test whether mechanisms identified in Stage 1 remain robust in a specific regional context. Henan Province was selected based on its middle-tier development characteristics. We estimate the same model structure as Stage 1 for the Henan subsample and evaluate:
(1) Direction Consistency: Whether all paths show the same directional relationships as in the national sample, regardless of statistical significance.
(2) Effect Size Comparison: Whether the magnitude of coefficients (β values) is comparable, accounting for the smaller sample’s wider confidence intervals.
(3) Mechanism Robustness: Whether Bootstrap confidence intervals overlap between national and Henan samples, indicating that non-significance in Henan reflects statistical power limitations rather than mechanism failure.
Data processing and statistical analysis were completed using Stata 18.0 software. To enhance the reliability of results, all models use robust standard errors for estimation to address potential heteroscedasticity issues. Through this rigorous analytical strategy, this study can systematically evaluate the associations of internet usage frequency with perceptions of social fairness and the mediating role of social trust and explore the heterogeneity of these association patterns against the background of diverse social contexts.
Research Results
Descriptive Statistics and Correlation Analysis
Table 2 presents descriptive statistics of the main research variables for both the full national sample and the Henan subsample. In terms of means, the average level of perceptions of social fairness is 3.457 (SD = 0.970) in the national sample and 3.528 (SD = 0.922) in Henan, which are at a moderately high level on the 5-point scale, indicating that Chinese citizens hold relatively positive evaluations of social fairness overall. The mean internet usage frequency is 3.343 (SD = 1.668) nationally and 3.319 (SD = 1.793) in Henan, reflecting relatively frequent internet usage habits, but the significant standard deviation indicates an evident digital divide phenomenon in the sample. The mean of social trust is 3.645 (SD = 0.997) nationally and 3.719 (SD = 0.982) in Henan, showing that Chinese citizens generally have high levels of social trust. Regarding control variables, the sample has an average age of 51.527 years, a male proportion of 45.3%, a Han ethnicity proportion of 92.7%, a religious belief proportion of 7.5%. The average education level is 2.277 (SD = 1.123), indicating that the typical respondent has completed high school or technical secondary education. The mean log-transformed income is 8.895 (SD = 4.666). Notably, the Henan sample exhibits a significantly higher rural population proportion (70.35%) compared to the national average (59.95%), reflecting its characteristics as a major agricultural province in central China.
Descriptive Statistics and Sample Characteristics.
Henan’s rural population proportion is 10 percentage points higher than the national average, reflecting its characteristics as a major agricultural province in central China.
The correlation analysis results among the main variables show that internet usage frequency is significantly negatively correlated with perceptions of social fairness (r = −0.084, p < .001), providing preliminary support for Hypothesis 1; internet usage frequency is also significantly negatively correlated with social trust (r = −0.099, p < .001), providing preliminary support for Hypothesis 2; social trust is significantly positively correlated with perceptions of social fairness (r = 0.326, p < .001), consistent with the expectation of Hypothesis 3. These correlation results suggest that higher internet usage frequency is associated with lower social trust and perceptions of social fairness. In contrast, while higher levels of social trust are associated with higher perceptions of social fairness. However, correlation analysis can only reflect simple associations between variables and cannot reveal directional pathways or mediating mechanisms, thus requiring further exploration through regression analysis.
Stage 1: Mechanism Validation Using National Sample
This section presents the core mechanism validation using the full national sample. Table 3 displays the regression results for both baseline models (without control variables) and full models (with control variables) to systematically assess robustness.
Mechanism Validation—National Sample and Henan Case.
Note. Standard errors in parentheses. Control variables include age, gender, ethnicity, religious belief, education, and log income. n.s. = not significant.
p < .001. *p < .05. +p < .10.
In the baseline Model, the analysis results of the national sample show that all three main paths reach statistical significance: internet usage frequency is significantly negatively associated with perceptions of social fairness (β = −.052, p < .001), supporting Hypothesis 1; This suggests that individuals who use the internet more frequently tend to report lower perceptions of social fairness. Second, internet usage frequency is significantly negatively associated with social trust (β = −.099, p < .001), supporting Hypothesis 2; This indicates that higher internet usage is associated with lower levels of generalized trust in others. Third, social trust is significantly positively associated with perceptions of social fairness (β = .321, p < .001), supporting Hypothesis 3. These results suggest that higher levels of social trust are associated with more positive evaluations of social fairness, consistent with social capital theory.
After including control variables, in the full model, the direction and significance of the main path coefficients remain stable: the negative association of internet usage frequency with perceptions of social fairness remains significant (β = −.054, p < .001); the negative association of internet usage frequency with social trust, though slightly reduced in coefficient magnitude, remains significant (β = −0.055, p < .001); and the positive association of social trust with perceptions of social fairness also remains stable (β = .316, p < .001). The consistency between baseline and full models demonstrates that, even after controlling for individual characteristics, the negative associations of internet use with perceptions of social fairness and the mediating role of social trust robustly persist in the national sample.
To precisely quantify the mediating mechanism, we employed the Bootstrap method (with 1,000 repetitions) to estimate standard errors and confidence intervals for the indirect effect. As shown in Table 3, that internet usage frequency produces a significant negative indirect effect on perceptions of social fairness through social trust. In the baseline model, the indirect effect is β = −.031 (p < .001, 95% CI [−0.035, −0.027]), accounting for approximately 37% of the total effect (total effect β = −.084, p < .001). In the full model with control variables, the indirect effect is β = −.017 (p < .001, 95% CI [−0.023, −0.011]), accounting for approximately 24% of the total effect (total effect β = −.072, p < .001). The consistency of significant mediation across both specifications—with the mediation proportion ranging from 24% to 37%—provides strong support for Hypothesis 4 in the national context, social trust plays a significant partial mediating role in the relationship between internet use, and perceptions of social fairness, with the mediating effect remaining substantial regardless of model specification. The slight reduction in mediation proportion from 37% (baseline) to 24% (full model) suggests that part of the relationship is attributable to compositional differences in internet users, but a significant mediation pathway persists even after accounting for these factors.
Table 3 also presents results from the Henan subsample, which demonstrates consistent directional relationships across all paths. Notably, the magnitude of Henan’s baseline direct effect (β = −.089) is comparable to or larger than the national average (β = −.052), with marginal significance (p = .063) primarily reflecting smaller sample size rather than mechanism failure, as evidenced by overlapping Bootstrap confidence intervals. These results validate mechanism robustness across different contexts.
Stage 2: Heterogeneity Exploration Within National Sample
Having established the mediating mechanism in the full national sample, this section examines whether these relationships vary across subgroups within the national sample. We conduct two types of analysis: (1) urban-rural heterogeneity as the primary analysis, and (2) age-cohort heterogeneity as exploratory analysis. Table 4 presents standardized path coefficients for each subgroup along with likelihood ratio tests.
Heterogeneity Analysis—National Sample.
Note. Older adults show significant negative paths while younger groups do not, suggesting age-differentiated digital impacts requiring further validation. Note for both panels: Standard errors in parentheses. All coefficients are standardized.
p < .10, indicating exploratory findings.
p < .001. **p < .01. *p < .05. +p < .10.
Table 4, Panel A compares rural residents and urban residents. The likelihood ratio test reveals significant heterogeneity (LR χ2(3) = 12.78, p = .005), indicating that relationships differ substantially across urban and rural contexts.
The key finding is that urban residents experience approximately twice the negative association between internet use and fairness perceptions compared to rural residents (β_urban = −.080, p < .001 vs. β rural = −.040, p < .01). This substantial difference suggests internet use has more pronounced negative relationships with fairness evaluations in urban contexts.
The mediating pathway shows interesting patterns. Rural residents show stronger trust erosion (β_rural = −.125, p < .001 vs. β_urban = −.060, p < .001), yet their overall fairness perceptions are less affected. This paradox likely reflects that rural residents retain stronger offline social networks that buffer against internet-induced trust decline. The Trust→Fairness path remains consistent across both groups (β_rural = .330, p < .001; β_urban = .304, p < .001).
These findings reveal that digital impacts are structured by pre-existing social stratification. Urban residents’ higher exposure to algorithmically curated content and greater social comparison opportunities may amplify awareness of inequality. In contrast, rural residents’ more basic internet usage and stronger offline community ties provide resilience against negative digital effects.
Table 4, Panel B examines three age groups: Youth, Middle-aged, and Older adults. The likelihood ratio test shows marginally significant heterogeneity (LR χ2(6) = 10.73, p = .097), warranting cautious interpretation.
The pattern reveals that older adults show significant negative associations across all paths (Internet→Fairness: β = −.074, p < .001; Internet→Trust: β = −.062, p < .001), while younger cohorts show non-significant or weaker associations. Youth (≤35) show no significant relationships between internet use and either fairness perceptions (β = .010, p = .665) or social trust (β = .008, p = .722).
This age gradient suggests digital natives may have developed cognitive frameworks that buffer against negative effects observed in older cohorts. However, these findings should be interpreted cautiously given the exploratory nature and marginal significance. Age differences may also proxy for different internet activities—younger users engage more in social networking while older users focus more on news consumption.
The heterogeneity analyses demonstrate that internet-fairness relationships vary significantly across social contexts. Urban-rural heterogeneity emerges as the most robust finding, with urban residents experiencing substantially stronger negative associations. Age-cohort analysis provides exploratory evidence of differential vulnerability, requiring further validation. These findings advance beyond aggregate patterns to reveal how digital impacts are shaped by pre-existing social structures.
Stage 3: Case Validation Using Henan Province
This section examines whether the mechanisms identified in the national sample remain robust in a specific regional context. Henan Province was selected as a validating case based on its middle-tier development characteristics. Rather than treating Henan as a “contrast region” separate from the national sample, we assess whether the pathway Internet Use→Social Trust→Fairness Perception operates consistently in this context. Results are presented in Table 3 alongside national sample results to facilitate direct comparison.
The Henan baseline model demonstrates consistent directional relationships across all three paths: Internet→Fairness (β = −.089, p = .063, marginally significant), Trust→Fairness (β = .359, p < .001, significant), and Internet→Trust (β = −.101, p < .05, significant). The indirect effect is also significant (β = −.036, p < .05, 95% CI [−0.071, −0.001]), accounting for approximately 29% of the total effect.
Notably, the magnitude of the direct effect in Henan’s baseline model (β = −.089) is larger than the national average (β = −.052), suggesting the mechanism operates with comparable or stronger intensity in this middle-tier development context. The marginal significance (p = .063) rather than conventional significance primarily reflects smaller sample size (n = 398 vs. N = 8,005), as evidenced by wider confidence intervals. The Bootstrap confidence interval for Henan’s indirect effect ([−0.071, −0.001]) overlaps with the national interval ([−0.035, −0.027]), indicating point estimates fall within expected sampling variation rather than representing a fundamentally different mechanism.
In the Henan full model, the total effect becomes non-significant (β = −.085, p = .262) due to reduced statistical power with multiple control variables and limited sample size. However, the direction of all paths remains consistent with the national pattern: Internet→Fairness (β = −.049), Trust→Fairness (β = 0.352, p < .001, still significant), and Internet→Trust (β = −.104). The indirect effect magnitude (β = −.037) is comparable to the national full model (β = −.017), though the Bootstrap confidence interval ([−0.094, 0.020]) includes zero, reflecting insufficient power rather than mechanism absence.
The fact that the Trust→Fairness path remains robustly significant across both Henan models (β = .359*** in baseline, β = .352*** in full model) provides strong evidence that this core component of the theoretical framework operates consistently. The weakening of other paths in the full model is attributable to the statistical challenge of estimating multiple parameters with fewer than 400 observations, not to substantive differences in underlying mechanisms.
These findings validate the robustness of mechanisms identified in the national sample. All paths show consistent directions, effect sizes are comparable or larger in magnitude, and Bootstrap confidence intervals overlap with national estimates. The pattern of significance differences between Henan and national samples aligns with expectations based on statistical power rather than indicating regional mechanism failure. This case validation demonstrates that the pathway Internet Use→Social Trust→Fairness Perception operates across different developmental contexts, supporting the theoretical framework’s generalizability within China’s diverse regional landscape.
Discussion
Theoretical Significance of Research Findings
In the national sample, we find that internet usage frequency is significantly negatively associated with perceptions of social fairness. This result supports the view that the Internet may strengthen rather than weaken social inequality. As an information amplifier, the Internet increases the frequency with which individuals are exposed to social inequality phenomena, being associated with their evaluation of social fairness through a relative deprivation mechanism. Even after controlling for individual socioeconomic characteristics, this relationship remains robust, indicating that the associations of internet use on social evaluation stem not only from the digital divide but represent a more complex social psychological process.
The negative association of internet usage frequency with social trust is also validated in the national sample. This indicates that in the Chinese social context, internet use is associated with reduced time for face-to-face social interaction, and related to weaker traditional social relationship networks, increasing the uncertainty of social interaction, and being associated with weaker the foundation of social trust.
Additionally, this study confirms that social trust has a significant positive association with perceptions of social fairness. In the national sample, social trust plays an important partial mediating role in the relationship between internet use and perceptions of social fairness, with the mediation proportion ranging from 24% to 37% depending on model specification, demonstrating robustness across baseline and full models. This finding reveals the internal mechanism by which the Internet is associated with social evaluation: Internet use reshapes social relationship structures and trust foundations, being associated with individuals’ judgments of the fairness of social distribution. This mediating relationship expands our understanding of the Internet’s social effects, shifting from a simple information acquisition perspective to a more comprehensive analysis of social psychological mechanisms.
Importantly, this study makes a key methodological and substantive contribution by systematically examining heterogeneity within the national sample. The urban-rural heterogeneity analysis provides important evidence for understanding the context-dependency of Internet social effects. Urban residents experience approximately twice the negative association between internet use and fairness perceptions compared to rural residents (β_urban = −.080 vs. β_rural = −.040, p < .01), revealing that digital impacts are structured by pre-existing social stratification. This heterogeneity reflects how traditional offline social structures—stronger in rural areas—buffer against digital impacts, while urban contexts characterized by weaker community ties and greater exposure to algorithmic content amplify negative associations. This finding suggests that internet effects should not be simply generalized; we pay attention to the key role of the social environment and development stage in shaping internet social influence.
The Henan case validation demonstrates that the theoretical mechanisms identified in the national sample operate consistently across different regional contexts. All paths show consistent directional relationships, with effect sizes in Henan’s baseline model (β = −.089) comparable to or larger than national averages (β = −.052). The pattern of significance differences aligns with expectations based on statistical power rather than substantive mechanism failure, as evidenced by overlapping Bootstrap confidence intervals. This case validation approach—testing whether mechanisms identified in large-scale data hold in specific contexts—provides a methodological template for studying context-dependent digital effects, balancing generalizability with contextual sensitivity.
Against the backdrop of China’s social transformation, these findings collectively constitute a more complex and dynamic analytical framework for Internet social effects. At the theoretical level, this framework helps deepen understanding of the internal logic of social psychological changes in the process of digital transformation. This framework integrates digital divide theory with social capital theory, revealing the micro-psychological pathways through which internet use is associated with social evaluation, providing a new perspective for understanding the “digital psychological divide” in the intelligent era.
Practical Significance of Research Findings
This study’s empirical findings have important implications for digital governance, social harmony, and balanced urban-rural development.
The negative relationship between internet usage frequency and perceptions of social fairness reminds policymakers to pay attention to social cognitive division in digital transformation. Internet use indirectly weakens perceptions of social fairness by being associated with lower trust. This finding suggests that digital governance should focus not only on access but also, on enhancing digital literacy, and promoting healthy use. In practice, we can consider conducting hierarchical digital literacy education targeted at different groups’ characteristics, improving citizens’ ability to discern online information, while optimizing the online information ecology and reducing the spread of polarized and harmful content.
The mediating role of social trust emphasizes the special importance of building social capital in the digital age. As the Internet deeply integrates into social life, traditional social relationship structures face reconstruction, and the foundation of social trust is weakened. While promoting digital development, maintaining social trust networks requires targeted interventions. At the policy level, specific measures include: (1) Promote the integration of online and offline community development, encourage the construction of geography-based social networking platforms, and combine virtual interaction with physical contact to strengthen social bonds. (2) Improve the transparency and accountability of the internet environment. (3) Enhance citizens’ institutional trust in cyberspace to alleviate the negative associations of digital interaction with social trust. (4) Design platform features that facilitate authentic social connections, such as verified identity systems and community-based forums tied to geographic localities.
The urban-rural heterogeneity finding—that urban residents experience twice the negative association between internet use and fairness perceptions—has critical policy implications. Digital inclusion policies cannot simply aim for equal access but must account for how digital tools interact with existing social structures. For urban areas, where residents face greater algorithmic exposure and weaker offline social ties, interventions should focus on: (1) Information literacy programs emphasizing critical evaluation of online content and managing social comparison effects; (2) Promoting community-building initiatives that strengthen local social bonds; (3) Regulating algorithmic recommendation systems to reduce polarization and inequality awareness amplification.
For rural areas, where stronger traditional networks buffer digital impacts but basic connectivity remains limited, strategies should prioritize: (1) Infrastructure development ensuring reliable internet access; (2) Digital skill training tailored to practical needs (e.g., agricultural e-commerce, rural finance); (3) Leveraging existing offline social capital while introducing digital tools—’digital + traditional’ integration rather than replacement; (4) Developing platforms that connect local products to national markets, ensuring digital dividends reach rural communities.
The Henan case validation demonstrates that theoretical mechanisms operate consistently across contexts, suggesting that targeted interventions addressing trust erosion and information ecology apply broadly, though implementation strategies must adapt to local social structures and development levels.
Overall, the social effects of China’s digital transformation process show complex and diverse characteristics, and their associations vary across multiple factors such as urban-rural context. When formulating digital development strategies, pay attention to urban-rural differentiation and adopt context-sensitive policy measures that both fully leverage the positive role of the Internet in promoting social inclusion and equal opportunities and vigilantly guard against the risks of social cognitive division.
Conclusion and Implications
Based on data from the Chinese General Social Survey (CGSS) 2021, this study reveals the relationship between internet usage frequency, Social trust, and perceptions of social fairness and through a three-stage research design encompassing national validation, heterogeneity exploration, and case validation. The results show that internet usage frequency is significantly negatively associated with perceptions of social fairness (β = −.054, p < .001 in full model), and this relationship is partially mediated by social trust. Specifically, higher internet usage frequency is associated with lower individual social trust levels, which relates to decreased perceptions of social fairness, with the mediating effect of social trust accounting for 24% to 37% of the total effect,depending on model specification, demonstrating robustness.
Critically, heterogeneity analysis within the national sample reveals that urban residents experience approximately twice the negative association between internet use and fairness perceptions compared to rural residents (β_urban = −.080 vs. β_rural = −.040, p < .01), indicating that digital impacts are structured by pre-existing social stratification. This urban-rural divide emerges as the study’s primary substantive contribution. Additionally, the Henan case validates mechanism robustness across contexts: all paths show consistent directions, with effect sizes comparable to national averages and overlapping Bootstrap confidence intervals, demonstrating that the pathway Internet Use→Social Trust→Fairness Perception operates consistently despite regional developmental differences.
These findings provide important implications for digital governance practice: First, digital inclusion policies should shift from “universality-oriented” to “quality-oriented.” The negative association between internet use and perceptions of social fairness suggests that increasing usage frequency does not automatically enhance perceptions of social fairness. Policy focus should shift from internet infrastructure construction to improving usage quality, enhancing users’ critical information filtering ability, reducing information cocoon effects and confirmation bias, and promoting healthy and positive internet usage patterns.
Second, mechanisms to protect social trust in the digital era should be constructed. The discovery of social trust as a mediating mechanism highlights that the negative associations of internet use with perceptions of social fairness partially stem from the associations with lower social trust. We should promote the integration of online and offline community development, encourage the construction of geography-based social networking platforms, and combine virtual interactions with physical contacts to strengthen social bonds; at the same time, improve the transparency and accountability of the network environment and enhance citizens’ institutional trust in cyberspace, which may alleviate the negative associations of digital interaction with social trust.
Finally, differentiated urban-rural digital development strategies should be implemented. The urban-rural heterogeneity finding indicates that interventions must be context-sensitive: urban areas require information literacy programs and algorithmic regulation to manage social comparison effects, while rural areas need infrastructure development and strategies that preserve beneficial offline social capital while expanding digital capabilities. This differentiated approach ensures digital tools interact productively with existing social structures rather than uniformly imposing standardized solutions.
Limitations and Future Prospects
Although this study has achieved important results in exploring the relationship between internet use, social trust, and perceptions of social fairness, some limitations exist.
As a cross-sectional study, the causal order among variables cannot be definitively established. While we frame the pathway as Internet Use→Social Trust→Fairness Perception based on theoretical reasoning, reverse causality remains possible—individuals with lower fairness perceptions may use the internet differently, or those with lower trust may engage more intensively online. Longitudinal tracking designs or quasi-experimental methods are needed to more accurately reveal causal relationships. Future research should employ panel data or natural experiments to establish temporal precedence and rule out alternative explanations. The measurement of internet use is relatively crude, using only frequency as the indicator, ignoring differences in usage content, purpose, and mode. Information acquisition, social, and entertainment uses may have different or even opposing effects on social trust and perceptions of fairness. Future research should employ detailed measures of internet usage types to explore their differentiated social effects. Despite the comparison between Henan Province and the national sample, systematic comparison across multiple regions is still lacking. The social effects of the Internet on developed eastern coastal regions, less developed western mountainous regions, and border ethnic regions may each have their characteristics. Future research could extend to more diverse regional backgrounds to further test the context-dependency of Internet social effects. Additionally, the reliance on single-item measures for complex constructs like social trust and perceptions of social fairness, while common in large-scale surveys, limits reliability and validity. Multi-item validated scales would provide more robust measurement in future studies.
While Henan Province provides valuable insights as a typical central region case, the generalizability of case-specific findings requires caution. The case validation approach—testing whether mechanisms identified in large-scale national data hold in a specific context—is methodologically appropriate, but we do not claim Henan’s patterns are representative of all Chinese provinces. Future research should extend case validation to provinces with different developmental characteristics (e.g., highly developed eastern provinces like Zhejiang, less developed western provinces like Guizhou) to further test boundary conditions and mechanism robustness across diverse contexts.
With the rapid development of artificial intelligence, algorithmic recommendations, and deepfake technologies, the mechanism of Internet continues to evolve. Future research should shift focus from digitalization to intelligence, exploring how emerging technologies affect social trust and perceptions of fairness through mechanisms like algorithmic polarization, information cocoons, and AI-generated content.
Footnotes
Ethical Considerations
This study was based on secondary analysis of publicly available data from the Chinese General Social Survey (CGSS) 2021. As the research involved no direct interaction with human participants and all data were fully anonymized, this study met the criteria for exemption from institutional ethics review per research ethics guidelines at Henan Forestry Vocational College. The original CGSS data collection received ethical approval from the Ethics Committee at Renmin University of China. (
).
Consent to Participate
The CGSS research team obtained informed consent from all survey participants prior to data collection. Participants were informed about the survey purpose, voluntary nature of participation, data confidentiality, and right to withdraw. No additional consent was required for this secondary analysis of de-identified data.
Author Contributions
The author confirms being the sole contributor of this work and has approved it for publication.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The author declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Data Availability Statement
The data that support the findings of this study are publicly available from the Chinese General Social Survey (CGSS) database.