Mind the Gap, Not the Interface: Gendered Expectation–Perception Gaps in Educational Technology Services Among International Students in China

1. Introduction

The digital transformation of higher education has moved from a peripheral innovation agenda to the core infrastructure through which universities deliver, administer, and legitimize their services. In China, this transformation is policy-driven and nationally coordinated: the Education Informatization 2.0 Action Plan (2018) set the goal of an internet-based, smart education system encompassing digital campuses, smart classrooms, and cloud-based resource platforms in all schools, a trajectory consolidated by the National Education Digitalization Strategic Action (2022) and the Smart Education of China platform [1,2,3]. Because educational technology is now the primary interface between institutions and their students, the quality and equity of that interface are directly relevant to the sustainable development agenda: Sustainable Development Goal (SDG) 4 (quality education, notably Targets 4.3 and 4.5), SDG 5 (gender equality), and SDG 10 (reduced inequalities) all presuppose that digitalization narrows, rather than reproduces social disparities in educational opportunity [4,5]. Understanding who expects what from, and who is satisfied with, digital campus services is therefore not merely a marketing question but a diagnostic of sustainable educational transformation.

China constitutes a critical case for such an inquiry because it combines world-scale campus digitalization with one of the world’s largest and most structurally distinctive international student populations. According to official statistics, China hosted 492,185 international students from 196 countries in 2018, of whom 59.95% originated from Asia and 16.57% from Africa, with approximately 52.95% drawn from Belt and Road Initiative (BRI) partner countries [6,7]. This inbound mobility is predominantly a Global-South, scholarship-engineered phenomenon, sustained by the BRI Education Action Plan (2016) and associated scholarship instruments [8,9]. After two decades of enrollment-first expansion, policy discourse has pivoted “from quantity to quality,” as service and management capacity demonstrably lagged recruitment growth [10]. The COVID-19 pandemic further elevated digital service quality from convenience to a retention-critical condition for international students [11]. For this population, often arriving with high expectations of a technologically advanced destination, the digital service environment is encountered first (online admission), daily thereafter (websites, social media, smart classrooms, digital libraries), and judgment of it shapes satisfaction, recommendation, and the sustainability of the entire inbound-mobility model [12,13]. Marketing research on this population has further shown that incoming students most value the campus environment, professional teachers, student life, and cultural diversity, while the least satisfactory aspects cluster in the promotion of international studies and in the partnerships between universities [14].

Against this backdrop, the gender equity of digital education remains incompletely theorized. Two decades of research on the technology acceptance model (TAM) and its successors have established that gender moderates how technology is perceived: males report higher technology self-efficacy and more strongly weight performance expectancy, whereas females weight ease of use and social influence [15,16,17,18]. Males also systematically overestimate their own digital skills at equal measured proficiency, suggesting that male-favoring differences operate at the level of beliefs and priors rather than behavior or outcomes [19]. Crucially, meta-analytic evidence indicates that gender differences in actual e-learning satisfaction are small or absent globally and in China specifically [20], a pattern captured by Gefen and Straub’s [21] classic finding that women and men “perceive differently but use alike.” Yet this literature examines gender almost exclusively as a moderator of acceptance and intention; within expectation–confirmation theory (ECT), where satisfaction is produced by comparing experienced performance against prior expectations [22], the expectation stage itself has rarely been gender decomposed. International student studies in China report either main effects or female disfavoring satisfaction differences [12,23], and recent modeling work explicitly calls for gender moderation tests [24], but no study has separately measured how male and female international students rate the importance of educational technology services versus their satisfaction with those same services, nor quantified the resulting expectation–perception gap. This omission matters because the gender digital divide in students’ home regions is conventionally framed as a divide of access and skills [5,25]; whether a subtler divide persists at the expectation stage, even where access and experienced service are equal, is unknown.

The main goal of this study is to identify whether, and at which stage of the satisfaction-formation process, male and female international students differ in their evaluations of educational technology services. To achieve this goal, the study analyzes survey data from 329 international students (135 males, 194 females) in China,the same survey program whose comparative analysis of incoming and outgoing students’ perceptions is reported in [26],who rated both the perceived importance and the experienced satisfaction of six educational technology services, online admission, diversified information access, timely website information, social media information, smart classrooms, and digital libraries, on five-point Likert scales. Four research questions guide the analysis. RQ1: Do male and female international students differ in the importance they attach to educational technology services? (H1: males rate importance higher than females.) RQ2: Do they differ in experienced satisfaction with the same services? (H2: no significant gender difference in satisfaction.) RQ3: Do they differ in the expectation–perception (importance–satisfaction) gap? (H3: males exhibit a larger negative gap.) RQ4 (exploratory): Is the gendered gap robust to region of origin, urban–rural provenience, and family income? The results support the hypothesized asymmetry: males rate technology importance significantly higher (composite M = 4.46, SD = 0.53 vs. M = 4.29, SD = 0.61; Welch’s t = 2.55, p = 0.011, Cohen’s d = 0.278), satisfaction is gender-neutral (p = 0.614), and males consequently show a significantly larger negative expectation–perception gap (−0.48 vs. −0.27; p = 0.016, d = −0.269). The gap is partly compositional: after controlling for region of origin and family income, the gender coefficient attenuates to non-significance (β = −0.105, p = 0.286), although it persists within the Asian subsample (p = 0.012). Because importance and satisfaction were measured concurrently, importance ratings are treated throughout this article as retrospective evaluative anchors rather than directly observed pre-arrival expectations, and the causal language of Expectation-Confirmation Theory is used interpretively rather than as a strictly tested causal chain.

This study makes four contributions. First, it reframes gender equity in digital education from the extensively studied stages of access and acceptance to the expectation stage of satisfaction formation, thereby extending expectation–confirmation theory with a gendered account of how reference standards are formed [15,22]. Second, it provides, to the authors’ knowledge, the first two-dimensional (importance × satisfaction) gender decomposition of educational technology service evaluations among international students in China, complementing prior single-dimensional satisfaction studies in this population [12]. Third, it documents and quantifies a pattern, expectation inflation without experience discrimination,that is consistent with stereotype and self-efficacy driven differences in expectation formation: shared institutional delivery is accompanied by gender-equal reported satisfaction, while stereotype and self-efficacy-driven priors inflate male expectations, producing a measurable disconfirmation penalty concentrated among male students [19,20]. Fourth, it demonstrates that the gender gap is structurally embedded, intersecting with region of origin and family income, advancing an intersectional rather than essentialist reading of gender inequity in digital education that speaks directly to SDGs 4, 5, and 10 [4].

The remainder of this paper is organized as follows. Section 2 reviews the literature on gender and technology acceptance, expectation–confirmation theory, and importance–performance analysis, and develops the hypotheses. Section 3 describes the research design, instrument, sample, and analytical strategy. Section 4 reports the results for RQ1–RQ4, including importance–performance analysis by gender. Section 5 discusses the theoretical, practical, and sustainability implications, together with limitations. Section 6 concludes.

2. Literature Review and Hypothesis Development

2.1. Educational Technology Services and Sustainable University Quality

The quality of a contemporary university is increasingly experienced through its digital interface. Smart classrooms, digital libraries, online admission systems, institutional websites, social-media information channels, and online learning platforms no longer merely support academic services; for many students they are the services through which the institution is encountered, evaluated, and remembered [1,3]. This is especially true in China, which has pursued an aggressive, policy-driven digital transformation of higher education. The Education Informatization 2.0 Action Plan (2018) set the goal of an internet-based smart education system with digital campus construction in all schools, marking a shift from technology-driven to innovation-driven development [1,2]. Under the subsequent National Education Digitalization Strategic Action, According to the Ministry of Education, the Smart Education of China platform, officially launched on March 28, 2022, has attracted a great number of users from more than 200 countries/areas [27]. Empirical evidence confirms that this infrastructure is consequential for how students judge their institutions: students’ satisfaction with smart classrooms depends on usability and instructional integration rather than equipment alone [28], digital-library satisfaction is shaped by system, information, and service quality [29], and information technology infrastructure ranks among the most influential determinants of international student satisfaction in China [30].

These digital services acquire heightened significance for international students, who often interact with the host university online, through admission portals, websites, and social media, before ever setting foot on campus, and whose daily academic lives are thereafter mediated by the same channels. China hosts the largest international student population in Asia and one of the largest worldwide, enrolling 492,185 students from 196 countries in 2018, of whom 59.95% were Asian, 16.57% African, and 52.95% from Belt and Road Initiative (BRI) partner countries [6,7]. This inflow is substantially engineered by the BRI Education Action Plan (2016), which expanded government scholarships to partner countries as an instrument of South–South cooperation [31]. Within the United Nations 2030 Agenda, such scholarship-driven mobility operationalizes Sustainable Development Goal (SDG) Target 4.b (scholarships for developing countries), while the quality and equity of the educational experience received bear directly on SDG 4 (quality education), SDG 5 (gender equality), and SDG 10 (reduced inequalities) [4]. The COVID-19 pandemic stress-tested this model: when delivery moved fully online, the quality of digital services became decisive for international student retention, consolidating educational technology as the core service interface of Chinese international education [11]. Understanding how international students evaluate these services, and whether their evaluations are equitable, is therefore a question about the sustainability of the entire inbound-mobility model, not merely about user experience design.

2.2. Gender and Technology Perception: From Acceptance to Expectation

A robust literature in the technology acceptance tradition demonstrates that gender systematically shapes how individuals perceive, evaluate, and form beliefs about technology. In the seminal gender-moderation study of the Technology Acceptance Model (TAM), Venkatesh and Morris [15] showed that men’s technology use decisions are driven more strongly by perceived usefulness, whereas women’s are driven more strongly by perceived ease of use and subjective norm, with these effects attenuating with experience. The Unified Theory of Acceptance and Use of Technology (UTAUT) likewise positions gender as a moderator of the performance expectancy, effort expectancy, and social influence paths, with performance expectancy particularly salient for men [16]. Gefen and Straub [21] anticipated the pattern most relevant to the present study: women and men perceive information technology differently but do not use it differently, implying that gender operates at the level of perception and belief rather than behavior. Studies in educational and Chinese heritage contexts replicate this asymmetry: among learners in Taiwan, China, men reported higher computer self efficacy and stronger usefulness-driven acceptance of e-learning [32], and gender differences in Internet use, confidence, and experience were larger in China than in the United Kingdom, indicating that cultural context amplifies rather than erases gendered technology beliefs [33].

Meta-analytic evidence establishes the magnitude and persistence of these belief-level gaps. Across 82 studies and approximately 40,000 participants, Whitley [17] found that males exhibited stronger sex-role stereotyping of computers (d = 0.541), higher computer self-efficacy (d = 0.406), and more positive affect toward computers (d = 0.259), with smaller differences in actual usage behavior (d = 0.326),the foundational demonstration that belief-level gaps exceed behavior-level gaps. Seventeen years later, a meta-analysis of studies from 1997–2014 confirmed that males still held more positive attitudes toward technology use, a persistence attributed to societal norms framing technology as a male domain [18]. Critically, these attitudinal gaps are partly artifacts of calibration rather than competence: men systematically overestimate their online skills relative to measured performance, while women underestimate theirs at equal actual skill levels [19], and gender differences persist in motivation and self-reported use but vanish in actual performance measures [34]. Gender-role socialization and the enduring “technology is masculine” stereotype [17,35] thus produce a male-salient orientation toward technology that is largely independent of objective capability or experience.

We argue that these perception and belief-level differences should manifest most strongly at the expectation stage of service evaluation. Importance ratings of educational technology services, smart classrooms, digital libraries, online admission, websites, social media, online learning, tap precisely the performance-expectancy construct that men weight most heavily [15,16], and they function as priors informed by self-efficacy and stereotype-consistent identification [18,19]. Because international students in China have been positively selected into a technologically saturated environment, access constraints do not suppress these priors. Accordingly:

H1. 

Male international students attribute higher importance to educational technology services than female international students.

2.3. Expectation–Confirmation Theory and the Gendered Expectation–Perception Gap

Expectation–Confirmation Theory (ECT) provides the canonical account of how satisfaction is produced. Oliver [22] modeled satisfaction as the outcome of a comparison between pre-consumption expectations and perceived performance: the resulting confirmation, positive disconfirmation, or negative disconfirmation determines satisfaction and downstream intentions, with expectations serving as the cognitive anchor or reference standard [22,36]. Bhattacherjee [37] extended this logic to information systems in the Expectation-Confirmation Model (ECM), in which confirmation of expectations shapes perceived usefulness and satisfaction, which in turn drive continuance intention, now the dominant framework for explaining continued use of e-learning platforms and educational technology services. In service-quality research, the same comparison logic underlies the gaps model, in which service quality is conceptualized as the discrepancy between expectations and perceptions (SQ = P − E) [38], and Importance–Performance Analysis (IPA), which locates attributes on an importance × performance grid to derive improvement priorities [39,40]. In higher education, personal factors including gender jointly shape expectations and satisfaction, and managing student expectations is recognized as a lever of satisfaction management [41].

Integrating ECT with the gendered-perception literature of Section 2.2 yields a specific, testable implication. ECT distinguishes two inputs to satisfaction, expectations and perceived performance, that the gender literature suggests are differentially gendered. Gender-role socialization, self-efficacy, and overconfidence operate on the formation of expectations (priors), whereas perceived performance is anchored in the experienced service itself [19,22]. Because male and female students on the same campus consume the similar smart classrooms, digital libraries, websites, and admission systems, their experienced performance should converge; stereotype-driven expectation priors are updated toward a shared reality [21]. Gender should therefore differentiate the expectation stage but not the experienced-performance stage of the ECT chain.

ECT and its information-systems descendant have been extensively validated in educational contexts. The ECM explains continuance intention toward online learning, massive open online courses, and learning management systems [37], and ECT has been applied directly to international students’ university satisfaction, showing that initial expectations moderate the impact of post-arrival challenges on satisfaction and recommendation [41]. In China-focused research, expectation-disconfirmation logic underlies both service-quality audits of smart public services and satisfaction models of international students [12]. Importantly, despite the widely acknowledged evidence that gendered beliefs precede and shape technology evaluation, gender has largely been confined to a control or moderating role at the acceptance stage; its involvement at the expectation-formation stage, the inception of the ECT chain, remains theoretically nascent [15,21].

The satisfaction-parity side of this prediction is well supported. A meta-analysis of 20 high-quality studies across world regions found generally no significant gender differences in e-learning satisfaction, self-efficacy, motivation, attitude, or performance, and specifically no gender difference in attitudes in China (d = 0.09, p = 0.292) [20]. Multi-group analyses in Chile and Spain found no significant gender difference in overall e-learning acceptance despite path-level variation [42], and in the exact population of the present study, international students in China, a recent small-scale survey found no significant gender difference in satisfaction (T = 1.815, p = 0.070) [43], a pattern echoed by larger national samples in which gender effects on satisfaction were weak or absent once other factors were considered [24,44]. This parity evidence must, however, be set against a conflict zone in the literature: some studies report female-favoring satisfaction differences, attributed to better planning and more instructor interaction [45], while others report male-favoring differences, including a Sustainability-published study of 618 international students in Wuhan in which female students were significantly less satisfied than males with teaching, advisory, and overall services [12], and a national survey in which male international students in China reported higher satisfaction [44]. Given that the modal finding globally, in China generally, and in the target population specifically,is parity, and that the present study examines technology-mediated rather than interpersonal services, we hypothesize:

H2. 

Male and female international students do not differ in satisfaction with educational technology services.

If H1 and H2 hold jointly higher male expectations but equal experienced satisfaction, then a third implication follows arithmetically and theoretically: the expectation–perception gap (satisfaction minus importance), the within-person analogue of the disconfirmation construct in ECT [22,38], must be more negative for males. Gendered expectation formation without gendered service delivery mechanically produces a gendered gap; the interesting empirical question is whether the predicted asymmetry materializes. IPA research in other service domains confirms that gender differentiates both the level of importance and the size of expectation–perception discrepancies [46]. Hence:

H3. 

Male international students exhibit a larger negative expectation–perception gap (satisfaction − importance) with educational technology services than female international students.

2.4. The Gender Digital Divide Meets Global-South Mobility

The gender digital divide remains a structural feature of the countries from which most international students in China originate. Women in low and middle income countries are 14% less likely than men to use mobile internet, approximately 235 million fewer women online, with the widest gaps in South Asia (32%) and Sub-Saharan Africa (29%), driven by affordability, digital skills, and safety barriers [25]. UNESCO [5] further estimates that women and girls are 25% less likely than men to know how to leverage digital technology for basic purposes and four times less likely to know how to programme, positioning digital-skills equality as a prerequisite for SDG 4 and SDG 5 attainment. For students who cross these divided digital environments into China’s technologically saturated campuses, the divide does not disappear so much as change form: access is equalized by the host institution, but the confidence, self-efficacy, and expectation structures acquired under conditions of unequal access travel with the student. The relevant divide in a mobile student population is thus plausibly one of expectation calibration rather than access.

This reframing matters because inbound mobility to China is predominantly a Global-South phenomenon with a distinctive gender composition. African and Asian students are pulled by scholarships, low cost, BRI-linked career prospects, and China’s technological image [9], and scholarship recruitment channels,central to China, Africa educational exchange, have been shown to produce mismatches between policy goals and students’ lived experiences, including uneven program quality and constrained post-graduation pathways [47,48]. Recruitment expansion has historically outpaced service capacity, motivating a national policy pivot “from quantity to quality” in international student education [10]. Moreover, expectations formed pre-arrival, through scholarship narratives, recruitment materials, and social media,strongly shape later satisfaction judgments [13], and these pre-arrival information environments may themselves be gendered. If male-dominated mobility channels (e.g., scholarship pipelines from African source countries) coincide with the male overconfidence mechanisms of Section 2.2, the aggregate gender gap in expectations may be partly compositional: embedded in region-of-origin, provenience, and family-income structures rather than reducible to individual gender. Because the literature offers no direct evidence on this intersection, we treat it as an exploratory research question:

RQ4 (exploratory): Is the gendered expectation–perception gap robust to region of origin, urban/rural provenience, and family income?

2.5. Conceptual Model and Positioning

Gender shapes the expectation stage of the ECT chain through self-efficacy, stereotype-driven identification, and overconfidence (path a, Section 2.2); both genders’ expectations are then confronted with a common institutional service reality, producing perceived performance that does not differ detectably by gender (path b, Section 2.3); the within-person discrepancy between expectations and perceived performance, negative disconfirmation, yields satisfaction and the gap score G = S − I, which is predicted to be more negative for males (H3). Region of origin, provenience, and family income enter as structural covariates that may carry part of the gender effect (RQ4), reflecting the compositional embedding of gender in Global-South mobility structures (Section 2.4).

The present study is positioned against four closest works, each of which it extends on specific dimensions. First, Yasmin et al. [12], published in this journal, applied ECT to 618 international students in Wuhan and found female-disfavoring satisfaction differences, but examined satisfaction alone without an importance–satisfaction decomposition, and studied predominantly interpersonal services rather than educational technology. Second, León-Quismondo et al. [46] demonstrated the value of gender-separated IPA matrices in service settings, but in fitness centers rather than education, and without a theoretical account of why gender should differentiate expectations versus experiences. Third, Ding [23] documented expectation–experience gaps among international students in China qualitatively, but did not measure them quantitatively or test gender moderation. Fourth, Yu and Deng [20] established satisfaction parity meta-analytically but treated satisfaction as an outcome rather than as one stage of an expectation–disconfirmation chain, leaving the expectation stage unexamined. No existing study decomposes international students’ evaluations of educational technology services into gendered expectations versus gendered experiences within a single ECT-based framework, and none tests whether the resulting gap is compositional in region and income. The present study addresses this configuration of gaps using a survey of 329 international students in China.

3. Materials and Methods

3.1. Research Design

This study adopts a quantitative, cross-sectional survey design to examine gender differences in international students’ evaluations of educational technology services at their host universities in China. The design follows the expectation–perception gap paradigm that originates in expectation-confirmation theory (ECT), in which satisfaction results from a comparison between prior expectations and experienced performance [22], and in the service-quality gap tradition, where quality is operationalized as the discrepancy between expectations and perceptions [49]. In higher education and adjacent service settings, this paradigm is commonly implemented as an importance–performance design: respondents rate both the importance of a set of service attributes and their satisfaction with (or perceived performance of) the same attributes, and the attribute-level difference is used diagnostically [39]. This paired importance–satisfaction format is particularly appropriate for international student samples, where expectations are formed through pre-arrival information channels and satisfaction is evaluated against the delivered campus environment [12,50].

Three features of the design warrant explicit justification. First, the cross-sectional, single-wave administration entails that importance ratings are retrospective rather than pre-consumption expectations; they may therefore be contaminated by experienced performance, a recognized weakness of single-shot expectation–perception surveys [51]. We mitigate this concern by anchoring the importance items conceptually (“how important is this component to you”) rather than as recalled pre-arrival expectations, and we acknowledge the residual limitation in Section 5. Accordingly, throughout this article the term “expectation” refers to this concurrently measured importance anchor, a proxy for, rather than a direct observation of, pre-consumption expectations, and Expectation–Confirmation Theory is used as an interpretive framework for the importance–satisfaction discrepancy, not as a strictly tested causal chain. Second, because the same respondents rate both batteries, gap scores can be computed and tested at the individual level using paired procedures, which removes between-person heterogeneity in scale use and substantially increases statistical power relative to independent comparisons [52]. Third, the study focuses on gender as the grouping variable of interest while collecting region of origin, urban/rural provenience, and family income as covariates, enabling a compositional robustness check: if a bivariate gender difference merely reflects the different national and socioeconomic composition of the male and female subsamples, it should attenuate once composition is controlled.

3.2. Instrument and Measures

Data were collected with an online questionnaire titled “Marketing Survey about Studying Abroad”. The core of the instrument consists of two mirrored batteries covering 32 components of the study-abroad experience. The first battery measures perceived importance of each component on a five-point Likert scale (1 = not important at all to 5 = very important); the second battery measures satisfaction with the same components as delivered by the host university (1 = very dissatisfied to 5 = very satisfied). The present study analyzes the six components that constitute the educational technology service environment: the online admission process; diversified access to university information; timely information on the university website; university information in social media; smart classrooms; and the digital library. In addition, a single-item attitude statement, “Online classes will be indispensable in future abroad study” (five-point agreement scale), captures respondents’ forward-looking orientation toward online learning, and the face-to-face teaching item from the two main batteries serves as a non-technology control, allowing us to test whether any observed gender pattern is specific to technology services or reflects a general response-style difference.

Composite scores were computed as the unweighted mean of the six technology items for importance and satisfaction separately. Mean composites were preferred to sum scores because they preserve the original 1–5 metric, facilitate comparison across respondents with occasional missing items, and are the conventional aggregation in importance–performance applications [39,40]. Internal consistency was satisfactory for both composites: Cronbach’s α = 0.830 for the technology importance scale and α = 0.885 for the technology satisfaction scale, exceeding the conventional 0.70 threshold [12]. Internal consistency was comparable across gender groups (males: α = 0.809 importance, 0.889 satisfaction; females: α = 0.837, 0.884). Following the service-quality gap operationalization [49], an individual-level gap score was computed as G = Satisfaction − Importance for the composite and for each item, such that negative values indicate that delivered service falls short of the importance attached to it (an expectation shortfall) and values near zero indicate calibrated expectations. We acknowledge the well-known criticism that difference scores can have lower reliability and may not mirror the cognitive process of satisfaction formation, and that performance-only measures sometimes perform better [53,54]. Two design features mitigate these concerns: the gap is computed from paired within-person ratings rather than independent samples, and all gap-based conclusions are cross-checked against the separate importance and satisfaction analyses, so that no conclusion rests on the difference score alone. Relatedly, because stated importance is known to correlate with performance in importance–performance designs [51,55], we interpret the gap primarily as an expectation–perception discrepancy within the ECT framework rather than as a stand-alone managerial priority index.

3.3. Sample and Data Collection

The questionnaire was administered in English between 2023 and 2024 through Wenjuanxing, a widely used Chinese online survey platform. The survey link was distributed through online, international student offices, international student WeChat groups, and mailing lists, etc, and respondents were encouraged to forward the link to fellow international students. Eligibility was restricted to respondents who self-reported that they were studying, or had studied, at a Chinese university; a screening question at the start of the questionnaire terminated the survey for all others. Participation was voluntary and anonymous; no personally identifiable information was collected, and completion of the questionnaire was taken to imply informed consent. Data cleaning proceeded in three steps: header and invalid rows were removed; incomplete responses and responses with missing values were excluded pairwise per analysis (listwise for the composite scales). The final analytic sample comprises N = 329 valid responses (135 male, 194 female). Given the lack of a comprehensive sampling frame, a convenience sampling method was employed; the generalizability constraints this entails are addressed in the Limitations (Section 5.6). Table 1 summarizes the demographic composition of the sample.

The sample exhibits several features that are consequential for the analysis. First, women are over-represented (59.0%), producing unequal group sizes that motivate the use of Welch’s rather than Student’s t-tests (Section 3.4). Second, the sample is heavily concentrated in Asia, and Vietnam alone accounts for just over half of all respondents; this reflects the actual composition of international student flows into China, which are dominated by Asian and, increasingly, African source countries, but it limits generalizability and motivates the region subgroup analyses reported in Section 4. Third, the sample skews toward lower-income households, with 78.7% reporting family annual income below US$30,000, consistent with a population substantially supported by scholarships; income is therefore retained as an ordinal control (bands coded 1–4) in the robustness regressions. Finally, and most importantly, preliminary inspection revealed that gender and region of origin are strongly associated in this sample (45.2% of male respondents, but only 6.7% of female respondents, originate from African countries), so any bivariate gender difference must be re-examined after compositional controls before being attributed to gender per se.

3.4. Analytical Strategy

Analyses proceeded in four stages. First, descriptive statistics and within-gender paired-samples t-tests compared composite importance and satisfaction to establish whether an expectation shortfall exists for each gender separately; paired tests are appropriate because the two batteries are completed by the same respondents [52]. Second, between-gender comparisons of the importance, satisfaction, and gap composites, and of each individual item, were conducted with Welch’s two-sample t-tests. Welch’s procedure was preferred over Student’s t-test because the gender groups are unequal in size (135 vs. 194) and exhibited unequal variances on several measures; Welch’s test does not assume variance homogeneity and remains well calibrated under unequal n. To guard against violations of normality inherent in ordinal Likert data, all primary Welch tests were replicated with Mann–Whitney U tests; the two procedures agreed for all three composite comparisons, and the item-level Mann–Whitney results are reported alongside the Welch tests in Section 4. Because six items per family are tested, Benjamini–Hochberg false-discovery-rate adjustment is applied within each item family (importance, satisfaction, gap); confirmatory claims are restricted to the three composite comparisons, and item-level results are interpreted as exploratory. Effect sizes are reported as Cohen’s d computed with the pooled standard deviation, following recommendations that gender comparisons report magnitude, not only significance [42,46].

Third, to assess whether the gender difference in the gap score is robust to sample composition, we estimated ordinary least squares (OLS) regressions of the composite gap score on gender. Model 1 includes gender only (male = 1); Model 2 adds urban provenience (urban = 1), the ordinal family income variable (1–4), and region-of-origin dummy variables (Africa as the reference category). Because the dependent variable is a bounded difference score and residual variance may differ across groups, all regressions use HC3 heteroskedasticity-robust standard errors; HC3 is recommended over HC0–HC2 in moderate samples because it applies a degrees-of-freedom correction that performs well when high-leverage observations are present [56]. Fourth, subgroup analyses stratified by region (Asia vs. Africa) tested whether the gender pattern replicates within more compositionally homogeneous subsamples, addressing the intersectional concern that gender effects may be confounded with origin-region effects in mobility populations.

The significance threshold was set at p < 0.05, with results in the range .05 ≤ p < 0.10 flagged as marginal rather than interpreted as confirmatory. All analyses were conducted in Python 3 using the pandas, scipy, and statsmodels libraries. Given the single-source, single-wave design, common-method bias cannot be ruled out; procedural remedies were applied at the design stage (respondent anonymity, separation of the importance and satisfaction batteries, and varied scale anchors), and the residual risk is acknowledged as a limitation in Section 5.

3.5. Ethical Considerations

The study was conducted in accordance with the Declaration of Helsinki. The survey was administered anonymously to adult respondents, participation was entirely voluntary, and no personally identifiable or sensitive information was collected; completion of the questionnaire was taken to imply informed consent. Ethical review and approval were waived for this study due to its anonymous, non-interventional survey design involving adult participants.

4. Results

This section reports the empirical results in six steps. Section 4.1 establishes the descriptive pattern common to both genders; Section 4.2, Section 4.3 and Section 4.4 test H1–H3 using Welch’s t-tests on composite and item-level scores; Section 4.5 presents a gender-disaggregated Importance–Performance Analysis (IPA); Section 4.6 addresses RQ4 through robustness regressions and subgroup analyses; and Section 4.7 reports supplementary checks. All tests are two-tailed with α = 0.05; effect sizes are Cohen’s d. Full descriptive and inferential statistics are consolidated in Table 2 (Section 4.2).

4.1. A Universal Expectation Shortfall

Before examining gender differences, we assessed whether the expectation–perception structure posited by Expectation–Confirmation Theory (ECT) and the service-quality gaps model [22,38] holds in this sample at all. It does, and strongly. Male respondents rated the composite importance of the six educational technology services at M = 4.46 (SD = 0.53) against a satisfaction mean of M = 3.98 (SD = 0.75), a significant shortfall under a paired-samples t-test, t(134) = 7.48, p < 0.001. Female respondents showed the same configuration in attenuated form: importance M = 4.29 (SD = 0.61) versus satisfaction M = 4.02 (SD = 0.70), t(193) = 5.01, p < 0.001. The shortfall is therefore not a male idiosyncrasy: every one of the twelve item-level gap cells in Table 2 is negative, ranging from −0.14 (female, online admission) to −0.70 (male, social media information), consistent with disconfirmation-based accounts in which perceptions fall short of high prior expectations [22,38]. The theoretically interesting question, addressed in Section 4.2, Section 4.3 and Section 4.4, is whether the magnitude of this shortfall , and of its two components, expectations and experienced performance , differs by gender.

4.2. H1: Gendered Technology Expectations

H1 predicted that male students attach greater importance to educational technology services than female students. The composite test supports the hypothesis: male importance scores (M = 4.46, SD = 0.53) significantly exceed female scores (M = 4.29, SD = 0.61), Welch’s t(311) = 2.55, p = 0.011, d = 0.278 (Mann–Whitney p = 0.012), a small-to-medium effect. As shown in Table 2, the item-level pattern is broadly consistent. The largest difference occurs for diversified access to university information (male M = 4.55 vs. female M = 4.34, t = 2.68, p = 0.008, d = 0.291, Mann–Whitney p = 0.012), followed by digital library (t = 2.19, p = 0.029, d = 0.238, Mann–Whitney p = 0.028) and social media information (t = 2.21, p = 0.028, d = 0.235, Mann–Whitney p = 0.137). The online admission process differs only marginally (t = 1.89, p = 0.060, d = 0.212, Mann–Whitney p = 0.021), and two services , timely website information (p = 0.249) and smart classroom (p = 0.266) , show no significant difference, although male means are nominally higher for all six items without exception. One item is sensitive to test choice: the social-media importance difference is significant under Welch’s test (p = 0.028) but not under the Mann–Whitney test (p = 0.137), reinforcing the exploratory status of the item-level results.

Table 2 contains the study’s entire argument in miniature. The importance columns show a consistent directional pattern, male means higher for six of six services, with significance concentrated where effect sizes approach d ≈ 0.24–0.29. The satisfaction columns, by contrast, show no significant gender difference on any item or on the composite, and the direction of the trivial differences is mixed. The gap columns combine these two asymmetries: because male expectations are higher while experienced performance shows no statistically detectable gender difference, the male shortfall is mechanically and significantly larger for four of six services and for the composite. Notably, the two services without significant gap differences are not those with the smallest male shortfalls , the male gap for timely website information (−0.56) is the second largest in the table , but those with the largest within-group variances, indicating that statistical power, rather than the absence of a directional pattern, drives the non-significance there. H1 is therefore supported at the composite level and for three of six items.

Figure 1(a) visualizes these importance differences; Figure 1(b) previews the satisfaction results discussed in Section 4.3.

Two features of Figure 1 deserve emphasis. First, the vertical ordering of services is nearly identical across genders within each panel: both male and female students rank diversified information access and timely website information among the most important services and social media information and the smart classroom as (relatively) less important. The gender difference in expectations is thus a matter of level, not of structure , men and women want the same things, but men want them slightly more intensely. Second, the contrast between the two panels is visually stark: the male bar exceeds the female bar on every item in panel (a), whereas in panel (b) the bars are essentially interleaved, with female means nominally higher for online admission, diversified information access, social media, and the digital library. This visual dissociation between uniformly higher male expectations and undifferentiated experienced performance foreshadows the gap analysis of Section 4.4 and illustrates why pooling importance and satisfaction into a single “evaluation” score would obscure the mechanism at work.

4.3. H2: No Detectable Gender Difference in Experienced Satisfaction

H2 predicted no gender difference in experienced satisfaction with the delivered technology services. The results are consistent with H2 at every level of analysis. The composite satisfaction scores are statistically indistinguishable , male M = 3.98 (SD = 0.75), female M = 4.02 (SD = 0.70), Welch’s t(276) = −0.51, p = 0.614, d = −0.057 (Mann–Whitney p = 0.553) , and the effect size is near zero. Item-level tests (Table 2) yield the same conclusion for all six services, with p-values ranging from 0.082 (social media information) to 0.971 (timely website information). As Figure 1(b) shows, the largest nominal difference runs in the female direction: female students report somewhat higher satisfaction with university information in social media (M = 3.90, SD = 0.97) than male students (M = 3.70, SD = 1.07), although this difference does not reach significance (t = −1.74, p = 0.082, d = −0.199).

We interpret this pattern as consistent with, though not proof of, gender-neutral service delivery, with three caveats. First and foremost, our data measure students’ reported satisfaction, not the delivery process itself: the absence of a significant gender difference in satisfaction cannot demonstrate that institutional service delivery is gender neutral or identical for men and women, only that no gender difference in perceived satisfaction was detected at the measurement occasion. Second, absence of significance is not proof of absence of effect; however, the composite effect size (d = −0.057) is so small that any true difference is likely practically negligible. A two one-sided tests (TOST) equivalence test on the composite satisfaction difference (observed Δ = −0.041, 90% CI [−0.174, 0.092]) established equivalence at a moderate smallest effect size of interest (d = ±0.3) but not at a small one (d = ±0.2); satisfaction is therefore best described as statistically indistinguishable across gender with a negligible observed effect (d = −0.057), although very small differences cannot be excluded. Third, the marginal female advantage for social media satisfaction (p = 0.082) coincides with the service for which the male expectation premium is also pronounced , a combination that produces the single largest gendered gap in Table 2 (Section 4.4). Overall, the results are consistent with H2: whatever drives the gendered evaluations documented in this study, it does not appear to operate at the stage of experienced service performance, at least as perceived and reported by students.

4.4. H3: The Gendered Expectation–Perception Gap

H3 predicted that male students exhibit a larger negative expectation–perception gap (satisfaction−importance) than female students. The composite test supports the hypothesis: the male gap (M = −0.48, SD = 0.74) is significantly more negative than the female gap (M = −0.27, SD = 0.76), Welch’s t(293) = −2.41, p = 0.016, d = −0.269 (Mann–Whitney p = 0.027). Item-level analyses (Table 2; Figure 2) show that the difference is significant for four of six services: social media information (t = −2.91, p = 0.004, Mann–Whitney p = 0.023), online admission (t = −2.27, p = 0.024, Mann–Whitney p = 0.017), diversified information access (t = −2.09, p = 0.038, Mann–Whitney p = 0.027), and digital library (t = −2.06, p = 0.040, Mann–Whitney p = 0.086). It is not significant for timely website information (t = −0.79, p = 0.433) or the smart classroom (t = −0.25, p = 0.804). The digital-library gap is sensitive to test choice (Welch p = 0.040 vs. Mann–Whitney p = 0.086), reinforcing its exploratory status.

These item-level results are exploratory. After Benjamini–Hochberg false-discovery-rate adjustment within the gap family, the social-media gap remains significant (adjusted p = 0.023), while the online-admission, diversified-access and digital-library gaps are marginal (adjusted p = 0.060); within the importance family, diversified access remains significant (adjusted p = 0.047) with social media and digital library marginal (adjusted p = 0.059). The composite conclusions are unchanged: both key composites survive adjustment within the three-test composite family (adjusted p = 0.025 for both importance and gap).

The selectivity of the significant gaps is theoretically informative. Three of the four significant items, online admission, diversified information access, and social media information, are information-intermediated services whose quality is judged largely against expectations formed before or outside direct campus experience, through websites, agents, and social-media channels; the fourth, the digital library, is likewise an information service whose performance is experienced as content adequacy. In contrast, the smart classroom, a physical-infrastructure service encountered in situ, shows essentially identical gaps for men (M = −0.27) and women (M = −0.25). We caution, however, that the dichotomy is not clean: timely website information is also information-intermediated yet shows no significant gender difference, plausibly because its large within-gender variances (SD = 1.10 and 1.17) inflate the standard error despite a nominally substantial male shortfall (−0.56 vs. −0.46). The safest conclusion is that the gendered gap is real at the composite level, is driven jointly by the male expectation premium (Section 4.2) and the near-zero satisfaction difference (Section 4.3), and is most clearly detectable for services whose evaluation depends on institutionally managed information channels. H3 is supported in the bivariate analysis; its robustness is examined in Section 4.6.

4.5. Importance–Performance Analysis by Gender

To translate the group comparisons into managerially interpretable diagnostics, we plotted the six services on a gender-disaggregated IPA matrix [39,40], with satisfaction as the performance axis and the data-centered crosshair at the grand means (importance 4.36, satisfaction 4.00). Figure 3 shows each service’s position for male and female respondents, with arrows indicating the male → female shift.

Three observations emerge. First, most services occupy the high-importance/high-performance region (“keep up the good work”) for both genders, indicating that the host institution’s digital service portfolio is broadly aligned with what both groups value. Second, the male → female arrows are almost uniformly directed downward (lower female importance) and slightly rightward (equal or higher female satisfaction), a graphical restatement of H1 and H2. Third, the consistent exception is social media information. For males it combines the lowest satisfaction of all six services (3.70) with above-mean importance (4.40) , the upper-left “concentrate here” quadrant, precisely the configuration that IPA logic flags for corrective attention [39]. The female point, in contrast, shifts down and right into the lower-left “low priority” region in Figure 3, reflecting below-mean importance (4.21, below the 4.36 crosshair) paired with below-mean satisfaction (3.90, below the 4.00 crosshair). In practical terms, social-media-based information provision is the weakest link in the service chain for both genders, but for different reasons: for men it underperforms against high expectations; for women it is simply a low-salience, low-performing channel.

4.6. RQ4: Robustness and the Compositional Nature of the Gap

RQ4 asked whether the gendered gap survives controls for students’ structural locations. Table 3 reports two OLS models of the composite gap score with HC3 heteroskedasticity-consistent standard errors. Model 1 reproduces the bivariate result: the male coefficient is β = −0.202 (p = 0.016). Model 2 adds urban provenience, family income, and region-of-origin dummies (Asia and Other, with Africa as reference). The male coefficient attenuates by roughly half to β = −0.105 and loses significance (p = 0.286), while Asian origin (β = +0.293, p = 0.022) and family income (β = +0.114, p = 0.025) emerge as significant predictors of smaller (less negative) gaps.

The attenuation pattern in Table 3 indicates that the bivariate gender gap is partly compositional. Gender and region of origin are strongly confounded in this sample: 45.2% of male respondents originate from African countries, compared with only 6.7% of female respondents, and African-origin students report the most negative gaps at the region level, and the African female cell (−0.79) is the single most negative gender × region cell, although it rests on only 13 respondents; the Other-region female cell (−0.60) is likewise strongly negative (Figure 4). Once this unequal regional composition is held constant, the independent association between gender and the gap is small and statistically unreliable. Two further results qualify a purely compositional reading. First, income independently predicts better-calibrated expectations, suggesting that socioeconomic resources, not gender per se , shape how demanding students’ reference standards are. Second, the gender difference does not vanish everywhere: within the Asian subsample, the male gap (M = −0.44) remains significantly more negative than the female gap (M = −0.19, p = 0.012), whereas the African subsample shows no significant gender contrast, although with only thirteen African female respondents this test is severely underpowered. The gendered gap is thus neither a universal main effect nor a pure artifact: it is detectable within the largest regional subgroup but is amplified in the pooled sample by the over-representation of high-gap African students among men.

Figure 4 makes the intersectional structure visible. The male-larger-gap ordering holds only in the Asian subsample , the only subgroup with adequate cell sizes in both genders. In the African subsample the female mean (−0.79) is nominally more negative than the male mean (−0.55), and in the residual Other category the female mean (−0.60) likewise exceeds the male mean (+0.28) in magnitude; neither contrast is statistically reliable given the tiny cells involved (only 13 female respondents in the African subsample, and only 3 male respondents in the Other category). We therefore refrain from interpreting these reversals substantively, but they reinforce the central caution of this section: a binary gender contrast estimated on a regionally unbalanced sample aggregates heterogeneous subgroup patterns. In sum, the gendered expectation–perception gap is robust within the Asian majority subsample, attenuated to non-significance under controls, and structurally entangled with region of origin and family income, a pattern we interpret in the Discussion as evidence that gender equity in digital education is embedded in mobility structures rather than reducible to individual dispositions.

4.7. Supplementary Checks

Two supplementary analyses bound the interpretation of the main findings. First, gender differences do not extend to attitudes toward online learning per se: agreement with the statement that online classes will be indispensable in future study abroad was statistically indistinguishable between male (M = 3.64, SD = 1.12) and female (M = 3.75, SD = 0.99) students, t = −0.90, p = 0.368, d = −0.103. The gendered patterns documented above therefore concern the evaluation of specific institutional technology services, not a generalized male enthusiasm for, or female skepticism toward, technology-mediated education. Second, the face-to-face (F2F) teaching control item replicates the expectation–satisfaction dissociation outside the technology domain: male students rated F2F teaching as significantly more important than female students did (M = 4.58 vs. M = 4.39, p = 0.020), while satisfaction with F2F teaching did not differ (p = 0.624). This parallel suggests that the mechanism underlying H1–H3, inflated male expectations against experienced performance that shows no detectable gender difference, is not specific to educational technology but reflects a more general calibration difference in how male and female international students set evaluative reference points, a point taken up in the Discussion.

5. Discussion

5.1. Summary of Findings

Taking into account the results presented in Section 4, the evidence converges on a consistent pattern: gender differentiates the importance students attach to technology services but not their experienced satisfaction. The resulting expectation–perception gap falls disproportionately on male students and on information-intermediated services whose quality is judged against remotely formed priors. This gap is structurally embedded in region of origin and family income rather than purely individual. The Discussion therefore focuses on the mechanism, boundary conditions, and implications of this pattern rather than restating the statistical results reported in Section 4.

5.2. Expectation Inflation Without Experience Discrimination

The central empirical pattern of this study , higher male expectations, identical experienced performance, and a consequently larger male shortfall , can be summarized as expectation inflation without experience discrimination. This phrase names a mechanism that the gender-and-technology literature predicts but had not previously demonstrated in international education. Meta-analytic evidence shows that belief-level gender gaps exceed behavior-level gaps: males report higher computer self-efficacy and more positive technology attitudes while actual performance differences are small or absent [17,18]. Critically, part of the belief gap is calibration error rather than competence: men overestimate their digital skills at equal measured proficiency, whereas women underestimate theirs [19,34]. Our importance ratings plausibly tap exactly this overconfident prior. The technology acceptance tradition locates gender at the perception-formation stage, men weight performance expectancy more heavily [15,16], and women and men “perceive differently but use alike” [21] , and the present results are consistent with extending this logic one step upstream in the ECT chain, gender may operate at the expectation-formation stage rather than the perceived-performance stage. Because importance and satisfaction were measured concurrently in our design, we advance this mechanism as an interpretation consistent with the data, not as a demonstrated causal sequence. Because both genders on the same campus consume the same smart classrooms, portals, and repositories, their experienced performance converges on a shared reality, while stereotype and self-efficacy-driven priors remain differentiated. Comparing our results with the reviewed literature, it can be seen that the satisfaction parity we observe (H2) is consistent with the modal finding globally and in China specifically [20], and the gender-neutral attitude toward online classes confirms that the mechanism concerns evaluative reference points rather than technology orientation per se.

This interpretation must, however, be positioned honestly against conflicting evidence. Yasmin et al. [12], studying 618 international students in Wuhan and publishing in this journal, found female students less satisfied than males with teaching, advisory, and overall services; Zhao and Ma [44] similarly report male-favoring satisfaction differences, while González-Gómez et al. [45] report the reverse. We do not read our parity result as refuting these studies but as delineating a boundary condition: service-domain specificity. The female-disfavoring differences in Yasmin et al. [12] were concentrated in interpersonal and administrative services, where gendered interaction dynamics and differential treatment are plausible; our instrument covers institutionally standardized technology services, whose delivery is largely impersonal and therefore structurally less susceptible to gendered discrimination. The two literatures are thus compatible: gender differences in experienced satisfaction may reside in non-technology, human-mediated domains, while technology services are evaluated uniformly in reported satisfaction. This reconciliation carries an important qualification for our own mechanism: expectation inflation, not service discrimination, is what produces gendered dissatisfaction with digital services, but it cannot rule out discrimination elsewhere in the service bundle.

The selectivity of the gap across services further refines the mechanism. The gendered shortfall is significant precisely for information-intermediated services , social media information, online admission, diversified information access, and the digital library, whose quality is judged against expectations formed before or outside direct campus experience, through recruitment narratives, agents, and online channels [13]. These are services with high expectation elasticity: their reference standards are constructed remotely and can drift upward under the influence of male-salient performance expectancy and overconfidence priors. The smart classroom, by contrast, is a physical-infrastructure service encountered in situ; its evaluation is anchored in immediate, shared sensory experience, leaving little room for divergent priors to operate, and its gaps are indeed nearly identical for men and women. The parallel pattern for the face-to-face teaching control item, higher male importance, equal satisfaction, corroborates this reading: where prior expectations can be inflated, gender differentiates the resulting disconfirmation; where experience anchors judgment directly, it does not. The replication of the dissociation outside the technology domain additionally suggests a general male calibration tendency rather than a technology-specific one, consistent with the overconfidence account [19] and with expectation management as a general determinant of student satisfaction [41].

5.3. The Compositional and Intersectional Nature of the Gender Gap

The RQ4 results complicate any essentialist reading of the bivariate gender gap. Gender and region of origin are strongly confounded in this sample: male respondents are drawn disproportionately from African source countries, African-origin students report the most negative regional gaps, and the most extreme gender × region cells involve female students from underrepresented regions, albeit on very small subsamples. Once this composition is controlled, the independent male coefficient attenuates substantially and loses significance, while Asian origin and family income emerge as significant predictors of better-calibrated expectations. We read this attenuation cautiously: the aggregate gender gap is partly compositional, aggregating the expectation structures of differently composed mobility streams rather than expressing a uniform individual level gender disposition.

A plausible structural interpretation, offered as interpretation, not established fact, is that scholarship-engineered mobility channels select and shape students differently across sending regions. China–Africa educational exchange rests substantially on government scholarship pipelines, which can generate mismatches between policy promises and lived experiences [47,48], and African students are pulled by scholarship availability and China’s technological image in ways that may inflate pre-arrival expectations [8,9]. Where such channels are male-dominated, reflecting gendered access to scholarships in source countries [25], the pooled sample assigns African students’ larger shortfalls disproportionately to the male group. The income association supports a related interpretation: socioeconomic resources may function as a calibration resource by improving access to realistic pre-arrival information. Crucially, the gender difference does not reduce entirely to composition: within the compositionally homogeneous Asian subsample, the male gap remains significantly more negative. The gender effect is therefore real but context-bound, and equity diagnostics must disaggregate by the intersection of gender with origin and class, lest a binary contrast on an unbalanced mobility sample conflate individual dispositions with the political economy of who gets to move [47].

5.4. Managerial Implications for Universities

For university administrators, the findings redirect attention from service delivery to expectation calibration management. If, as the pattern suggests, male dissatisfaction reflects inflated priors rather than deficient delivery, an account requiring longitudinal confirmation (see Section 5.6) , the cost-effective lever is not upgrading infrastructure for one gender but managing the information environment in which expectations are formed [41]. In this sense, it is recommended to the management of the universities to treat expectation calibration as an integral part of education service quality management. Three implications follow. First, pre-arrival information quality matters disproportionately: admission portals, virtual campus tours, and recruitment materials should present realistic, not aspirational, depictions of digital service capacity, since expectations formed through these channels are precisely where the gendered shortfall originates. Second, social media deserves particular attention: it is simultaneously the weakest-performing service for men and the channel through which pre-arrival promises are most aggressively amplified. Universities should audit the gap between what their official and affiliated social-media channels promise and what the campus delivers, treating social-media promise management as a satisfaction-management instrument rather than a marketing exercise. Third, expectation-setting induction programs targeted at high-expectation segments, which this study suggests will skew male and African-origin, could recalibrate reference standards early, before disconfirmation consolidates into dissatisfaction.

Equally important is what universities should not change. The satisfaction parity documented under H2, whatever it reflects about the underlying delivery process, is an achievement consistent with SDG 5, and interventions must preserve it: expectation calibration should lower inflated priors toward reality, never differentiate the delivered service. Finally, segment-specific gap scores can serve as early-warning indicators. Because the expectation–perception gap is computable from routine paired surveys and is sensitive to composition shifts in incoming cohorts, monitoring it by gender × region cells would allow institutions to detect emerging mismatch populations, for instance, cohorts recruited through new scholarship channels , before dissatisfaction manifests in attrition or reputational damage.

5.5. Implications for Sustainable International Education

These findings bear directly on the sustainability of China’s inbound-mobility model. China’s internationalization has pivoted “from quantity to quality” precisely because service capacity lagged recruitment growth [10], and satisfaction is the proximal mechanism linking service quality to the model’s viability: dissatisfied students attrit, discourage compatriots through word-of-mouth, and erode the reputational capital on which scholarship-driven recruitment depends [11,12]. Expectation mismanagement is a quiet threat to this system. Our results suggest that a substantial share of measured dissatisfaction, concentrated among male and African-origin students, reflects not deficient services but uncalibrated expectations engineered upstream by scholarship narratives and recruitment channels [9,47]. Sustainable internationalization therefore requires expectation governance across the whole mobility chain, from sending-country recruitment to on-campus delivery, not merely campus-level quality improvement. For the policy makers at the national level, in line with the national strategy of China, it is recommended to integrate expectation calibration into the management of the “Study in China” brand and into scholarship recruitment guidelines, so that pre-arrival promises made through official channels match the digital services actually delivered.

At the same time, the study documents a genuine equity achievement with monitoring value. Digital services, standardized and impersonally delivered, appear to function as perceptual equalizers: reported satisfaction with them did not differ detectably between men and women (H2) even in a population drawn from countries with wide gender digital divides [5,25], although satisfaction parity cannot by itself demonstrate that the underlying delivery is identical. This supports the SDG 10 promise of digitalization as inequality-reducing, provided access is institutionally guaranteed. We propose that the two-dimensional (importance × satisfaction) gender decomposition used here be adopted as an equity diagnostic for SDG 4/5 monitoring in international education: satisfaction parity alone can mask expectation-stage inequities, while importance differences alone are harmless unless they generate disconfirmation; only the joint decomposition reveals whether a system is equitable in delivery, in expectation formation, or in both.

5.6. Limitations and Future Research

Several limitations bound these conclusions. First, the design is cross-sectional and self-reported; importance ratings were collected concurrently with satisfaction and are therefore retrospective priors that may be contaminated by experienced performance [51]. Although the face-to-face control item and the service-selectivity pattern argue against pure contamination, a contaminated measure should not differentiate information-intermediated from in-situ services, longitudinal designs measuring expectations before arrival and satisfaction after are needed to establish causal ordering. Second, the convenience sample is regionally unbalanced, the generalizability of our findings to all international students should be made with caution: Vietnam alone contributes 52% of respondents, the African female cell contains only thirteen participants, and all respondents studied in a single host country. The Asian-subsample persistence result mitigates but does not eliminate the resulting constraints; multi-host-country and multi-institution comparisons are required to test generalizability. Third, gender was operationalized binarily as recorded in the source instrument; future designs should include non-binary and self-described categories, both for inclusivity and because the essentialism critique of Section 5.3 applies to the analyst’s categories as well. Fourth, we relied on stated rather than derived importance; stated importance is known to correlate with performance and to distort IPA priorities [51,55], and our data-centered IPA crosshair is one of several defensible placements whose sensitivity we did not exhaustively test. Fifth, difference scores have documented reliability and validity criticisms [53], which we mitigated by cross-checking all gap conclusions against the separate importance and satisfaction analyses, but derived-importance and performance-only specifications would strengthen replication. Sixth, the single-source, single-wave instrument leaves common-method bias possible despite procedural remedies; and the regression models are parsimonious, omitted variables such as discipline, program level, scholarship status, and length of stay may carry part of the compositional effect attributed to region and income. Seventh, item-level comparisons are exploratory: under Benjamini–Hochberg false-discovery-rate adjustment within each item family, only the social-media gap and diversified-access importance survive adjustment, with several further items marginal. Eighth, measurement invariance across gender was not formally tested, although internal consistency was comparable across groups, so mean-level comparisons assume at least approximate metric equivalence.

These limitations define a research agenda. Longitudinal expectation tracking across the pre-arrival–arrival–graduation arc would test whether the male expectation premium is formed pre-departure, as our mechanism implies, and how it updates with experience , the acceptance literature predicts attenuation with experience [15], a testable implication our cross-section cannot address. Qualitative mechanism studies could examine how scholarship channels, agents, and social media construct the expectations that male students in particular carry into Chinese campuses. Comparative designs across host countries would establish whether expectation inflation without experience discrimination is specific to China’s scholarship-engineered mobility or a general feature of Global-South student flows. Finally, as institutions increasingly deliver services algorithmically, extending the two-dimensional gender decomposition to AI-mediated advising and learning analytics would test whether the equalizing property of standardized digital delivery survives the personalization turn, or whether personalization reintroduces the gendered differentiation that impersonal delivery currently avoids.

6. Conclusions

As educational technology becomes the primary interface between universities and their students, the equity of that interface becomes a diagnostic of sustainable educational transformation. This study examined whether gender differentiates international students’ evaluations of educational technology services in China, separately measuring the importance and satisfaction ratings of six services among 329 international students and quantifying the resulting gendered expectation–perception gaps. The evidence yields a coherent answer: gender enters satisfaction formation at the expectation stage rather than the experience stage. Male students hold systematically higher expectations, experienced satisfaction is indistinguishable between genders, and the resulting shortfall falls disproportionately on male students and on information-intermediated services, while remaining partly rooted in the region and income composition of mobility streams.

Theoretically, these findings extend expectation–confirmation theory one step upstream, locating gender at the expectation-formation rather than the perceived-performance stage, and name the underlying pattern expectation inflation without experience discrimination. The two-dimensional (importance × satisfaction) decomposition further shows that satisfaction parity alone is an incomplete equity metric, since it can mask expectation-stage inequities. Because importance and satisfaction were measured concurrently, this mechanism is advanced as an interpretation consistent with the data, not a demonstrated causal sequence. Practically, the findings redirect institutional attention from service delivery to expectation calibration: pre-arrival information, recruitment narratives, and social-media promises function as satisfaction-management instruments, and calibration should lower inflated priors toward reality without differentiating the delivered service.

For sustainable international education, the standardized, impersonally delivered digital services examined here appear to function as perceptual equalizers, supporting—though not proving—the inequality-reducing promise of digitalization, while the gendered expectation shortfall warns that equity in expectation formation must be governed across the whole mobility chain, from sending-country recruitment to on-campus delivery, if digitalization is to narrow rather than silently reproduce social disparities in educational experience. Future research should track expectations longitudinally across the pre-arrival, arrival, and graduation stages, examine qualitatively how scholarship channels, agents, and social media construct gendered priors, and test the pattern comparatively across host countries and beyond binary gender categories.