Everett Rogers’ Innovation Diffusion Theory (IDT) (1962) explains technology dissemination through five key attributes: relative advantage (perceived superiority), complexity (perceived difficulty), compatibility (alignment with values), trialability (ability to experiment), and observability (visibility of outcomes) (Jiang et al., 2025).

In higher education, these attributes shape ICT adoption: relative advantage (e.g., enhanced efficiency) and observability (peer/instructor use) drive uptake, while complexity (e.g., non-localized interfaces) hinders it—especially for non-technical or low-English-proficiency students (Gosavi et al., 2024; Wu, 2024). Institutional factors (e.g., mandatory platform use) and cultural context (e.g., Hong Kong’s bilingual needs) further influence diffusion, with students favoring tools accommodating Chinese-English switching (correlating with higher adoption).

IDT categorizes adopters into innovators (tech-savvy, experimental), early adopters, early/late majority, and laggards (hindered by outdated devices/skills), aiding targeted support (Jiang et al., 2025). This study extends IDT by integrating two contextual factors—cultural compatibility and institutional mandates—into the traditional framework, aiming to explore how these extended dimensions, alongside classic IDT attributes (relative advantage, complexity, observability), shape ICT acceptance in Hong Kong’s multicultural professional education context. The core variables in this study are defined as follows: independent variables (relative advantage, complexity, observability), moderators (cultural compatibility, institutional mandates), dependent variable (ICT adoption intention), and mediator (adopter categories).

Relative advantage, complexity, and observability are key to ICT adoption: perceived advantages (e.g., better collaboration) and observable peer/instructor use boost uptake, while complexity (e.g., English-only interfaces) acts as a barrier (Gosavi et al., 2024; Nascimento et al., 2024).

Cultural differences impact adoption: non-local students (especially from Mainland China) prefer bilingual (Chinese-English) tools, with language compatibility correlating strongly with use (Dahri et al., 2025; Zou et al., 2025). Institutional mandates (e.g., required Canvas use) motivate adoption, particularly for hesitant learners (Steyn et al., 2024).

Cultural compatibility extends IDT’s compatibility construct by enhancing the relevance of innovation attributes through linguistic and cultural localization, which significantly impacts non-local students’ technology adoption (Dahri et al., 2025; Huang et al., 2024). This relationship aligns with IDT’s focus on the importance of perceived compatibility in fostering adoption, and in bilingual settings, cultural and linguistic fit is a key determinant of technology acceptance (Dalle et al., 2024; Zhao et al., 2025). On the other hand, institutional mandates accelerate ICT diffusion by strengthening perceived advantages and reducing complexity barriers, largely due to policy support and institutional pressure (Bhattacherjee and Sanford, 2006). In this context, institutional mandates provide an infrastructure that simplifies and supports the integration of ICT tools into students’ academic routines, especially for those who may be resistant to new technologies (Alamri, 2025). Both factors amplify the positive effects of relative advantage and observability, while also mitigating the negative impacts of complexity on adoption intention. These moderators serve to enhance the adoption process by aligning the innovation with student needs and institutional goals. Hence, it is proposed the following hypothesis:

Adopters fall into five groups: innovators (tech-savvy, experimental), early adopters, early/late majority, and laggards (hindered by outdated devices/skills). Recognizing these categories helps tailor support (Mishra and Shrawankar, 2025). These categories help determine how individual students evaluate the relative advantage, observability, and complexity of innovations, which directly impact their adoption intention (Almaiah and Al Mulhem, 2019; Vagnani and Volpe, 2017). For example, in Hong Kong’s bilingual and multicultural context, the mediation effect of adopter categories is particularly pronounced, as these categories serve as filters through which students process innovation attributes, aligning them with their personal identity and cultural context. Specifically, adopter categories transmit a significant proportion (34%) of the effect of relative advantage through identity confirmation, which shapes how students recognize and integrate ICT tools into their learning practices. As such, the following hypothesis is proposed:

ICT integration aims to enhance learning: tools offer flexibility, support diverse styles, and improve access to development. Perceived usefulness/ease of use drives engagement, leading to better outcomes. As students develop higher levels of ICT adoption intention, this intention translates into increased usage, deeper engagement, and more effective interaction with ICT tools. This process is captured in the “acceptance-participation-performance” mechanism, where students’ intention to adopt ICT directly influences their participation and performance (Shareef et al., 2009). The greater the intention to adopt, the more likely students are to engage in sustained use of ICT tools, which enhances knowledge acquisition and improves cross-cultural collaboration outcomes.

In Hong Kong’s educational context, a positive relationship has been observed between ICT adoption intention and academic performance, where improvements in ICT adoption intention are correlated with a 15–20% better performance in both individual and collaborative learning tasks (Cheng and Tsang, 2021). This supports the theoretical connection within IDT, highlighting that ICT adoption intention, by fostering greater engagement, directly influences digital learning outcomes, enhancing overall academic achievement.

Therefore, this study introduces the following hypothesis:

Innovation Diffusion Theory (IDT) posits that perceived attributes of innovation, such as relative advantage, complexity, and observability, significantly influence adoption behavior (Rogers, 2003). In the context of ICT adoption, cultural compatibility and institutional mandates are key contextual factors that shape adoption intention (Dahri et al., 2025; Bhattacherjee and Sanford, 2006). Additionally, adopter categories play a mediating role in the relationship between innovation attributes and adoption behavior (Vagnani and Volpe, 2017). The hypothesized research model based on IDT is shown in Figure 1.

Figure 1 illustrates the hypothesized relationships among classic IDT attributes (Relative Advantage, Complexity, Observability) and extended contextual variables (Cultural Compatibility, Institutional Mandates) in predicting ICT adoption intention and digital learning outcomes. Solid lines represent significant positive paths, whereas dashed lines indicate negative paths.

Questionnaires were designed with items and scales meticulously derived from validated organizational behavior research to ensure construct validity (Tahat et al., 2023; Zhu and Ren, 2025). The selected measures, having been widely employed in previous studies, showcase their robustness in measuring key constructs. Notably, the original research underpinning these measures underwent strict validation processes—including factor analysis, reliability assessments, and expert panel reviews—thereby reinforcing the reliability and validity of the adopted items.

The research population comprised undergraduate students in professional and continuing education programs at Hong Kong’s post-secondary institutions. Two public institutions—CPCE at Hong Kong Polytechnic University and Tung Wah College—were selected, with a combined enrollment of 19,279 students. Using Krejci and Morgan’s sample size formula, a target sample of 384 was determined. After obtaining formal approval from department administrators, survey questionnaires were randomly distributed to students, with researchers present to address queries. Following data collection, 6 missing and 19 invalid questionnaires were excluded, yielding 359 valid responses (93.4% response rate) for analysis (as shown in Table 1).

The researchers calculated the descriptive statistics of the respondents’ demographic data. Descriptive analysis of the gathered data revealed that most of the participants were female, and a significant proportion were between 18 to 20 years old. Further, study respondents indicated diverse academic majors, with a notable percentage from design and IT-related fields. Table 1 shows the descriptives of respondents’ demographics.

Figure 2 illustrates the proportional distribution of the 359 valid respondents across institutions (CPCE vs. Tung Wah College), academic majors (Design/IT vs. other fields), and student origin (local vs. non-local).

The sampling frame comprised all full-time undergraduate students enrolled in professional and continuing education programs at CPCE and Tung Wah College. A simple random sampling technique was applied (Wang et al., 2025) by assigning each student a unique numerical code based on institutional enrolment lists and generating random numbers via Microsoft Excel’s RAND() function. Questionnaires were distributed both in paper form during compulsory tutorial sessions and electronically via institutional email, with unique access codes to prevent duplicate submissions.

Assessing the normality of continuous data is a critical step in statistical analysis, as it informs the selection of central tendency measures and guides appropriate statistical methods. When data exhibit a normal distribution, parametric tests are typically used for group comparisons. Using the Kolmogorov–Smirnov test, all p-values exceeded 0.7, leading to the non-rejection of the null hypothesis and confirming that the data followed a normal distribution in this study.

The research tool’s internal consistency and convergent validity were verified using Cronbach’s alpha and composite reliability (CR) assessments. All constructs demonstrated strong reliability (α > 0.7), with average variance extracted (AVE) values exceeding 0.60 (Diao et al., 2025) (0.615, 0.605, and 0.636 respectively) (Table 2), indicating adequate convergent validity. VIF (Variance Inflation Factor) values were also evaluated for multicollinearity, with all values falling within an acceptable range (1.2, 1.3, and 1.1 respectively), supporting the validity of the model. Furthermore, F-squared values confirmed that all constructs made significant contributions to the model (0.34, 0.15, and 0.20 respectively). Model fit indices via SmartPLS further confirmed satisfactory fit: Tucker-Lewis Index (TLI) = 0.926, Normed Fit Index (NFI) = 0.87, chi-square = 2.647, and Standardized Root Mean Square Residual (SRMR) = 0.149.

Discriminant validity was assessed using two criteria: (1) Fornell-Larcker criterion, where the square root of the average variance extracted (AVE) for each construct exceeded its correlation with other constructs (e.g., relative advantage = 0.829, complexity = 0.798), confirming no significant overlap among latent variables; (2) Heterotrait-Monotrait (HTMT) ratio, with all values below the recommended threshold of 0.90 (as shown in Table 3). Both criteria confirm the distinctiveness of the research model’s constructs.

Figure 3 illustrates bivariate correlations between latent variables, with diagonal values representing the square root of average variance extracted (AVE) to support discriminant validity. All correlation coefficients are below 0.7, indicating no significant multicollinearity.

To assess the predictive capability of the research model, the team computed R² coefficients (Coefficients of Determination). The obtained R² values, ranging from 0.604 to 0.717, indicate a strong predictive power for the proposed model (as shown in Table 4). After assessing the predictive power of the measurement model, the researcher conducted the path analysis. The researcher took 1,000 resamples and bootstrapped them to determine the path values, t-values, and significance values. The analysis revealed that Relative Advantage positively influences ICT Adoption Intention, Complexity exerts a negative effect on it, and Observability significantly promotes such intention. Additionally, Cultural Compatibility and Institutional Mandates were found to positively impact ICT Adoption Intention. Undergraduate students’ adoption behavior is also shaped by Adopter Categories, which in turn affect ICT Acceptance. Thus, path analysis results indicated that the collected data aligned with most of the proposed hypotheses (as shown in Table 5).

Sample restriction: the sample is limited to two institutions (CPCE and Tung Wah College), primarily focusing on design and IT majors (75.1%), which may limit the generalizability of findings to students in humanities or social sciences. Future studies should include a broader sample across different disciplines and institutions throughout Hong Kong to capture a more representative view of ICT adoption patterns.

Cross-sectional design: this study utilizes a cross-sectional design, which precludes the ability to make causal inferences about long-term adoption dynamics (Kim and Kim, 2025). Longitudinal research tracking ICT usage across multiple semesters could provide a clearer understanding of how institutional mandates and students’ ICT skills evolve over time.

Unmeasured variables: while the study addresses several key factors, cultural factors, such as learning styles (e.g., collectivism vs. individualism) and prior digital literacy, were not explicitly measured but could play a significant role in ICT adoption. Future studies might benefit from integrating constructs like the Technology Acceptance Model (TAM) to explore potential mediating effects of these variables.

Emerging technologies: this study does not address recent advancements, such as generative AI tools like ChatGPT, which may significantly influence adoption patterns. Future research could investigate how emerging technologies are diffused among different adopter categories and how these tools are integrated into educational settings (Jin et al., 2025).

Comparative analysis: a cross-cultural comparison between Hong Kong, Mainland China, and other multilingual regions (e.g., Singapore) could offer valuable insights into how language policies and institutional structures influence ICT integration. This would help identify universal trends as well as contextual differences in technology adoption, providing a more global understanding of these dynamics.