Influence on Educators’ Decisions Regarding Continued Use of the Virtual Learning Environment Blackboard in Public School Systems

Abstract

The purpose of this study was to analyse educators’ decisions on the continued use of the virtual learning environment (VLE) Blackboard and its associated e-learning technologies in the classroom within the public school system. This cross-sectional descriptive quantitative research collected 306 responses from educators in 30 public schools in Gauteng Province, South Africa. The results revealed that the empirical data’s mean performance expectancy (PEY) was lower than the ‘agree’ range of the hypothesised population, implying that the educators’ assumption is that the deployed technology does not improve their work performance. Furthermore, the results showed that learning tradition (LTD) has a complementary partial mediation effect on the relationship between PEY and continued use intention (CUI). Additionally, facilitating conditions (FCCs) also have a complementary partial mediation effect on the relationship between PEY and CUI. Conditional mediation (CoMe) from the path SOI x PEY -> LTD -> CUI was statistically significant. In probing the conditional indirect effect, the results showed that, if the social influence (SOI) increased, the mediation effect of LTD decreases. On the contrary, if it decreased, the mediation effect of LTD increased. This was also evident in the Johnson-Neyman plot. SOI did not moderate the mediation effect of FCC on the relationship between PEY and CUI. This study concludes that social and operational factors highly influence the dynamics of continued use of VLE and its associated e-learning technologies and cannot be discounted by practitioners and policy-makers in their quest to increase technology use in the school system. This study contributes to the unified technology acceptance and use theory model (UTAUT), advancing the idea that facilitating conditions and learning traditions can be mediators and social influence moderators within certain contexts and research settings.

1. Introduction

Fourth industrial revolution technologies (Eleyyan, 2021; Mtotywa et al., 2024), the coronavirus (COVID-19) pandemic that caused social distancing and lockdowns (De et al., 2020; Renu, 2021), and the shift to customised learning have increased technology use in the school system (Schmid et al., 2022). Haleem et al. (2022) argued that digital technology has become crucial in achieving the United Nations’ Sustainable Development Goal for 2030 of providing high-quality education for all. The acceleration of technology in the education system is regarded as having a critical, positive role in improving the education system (Jobirovich, 2021; Mtotywa et al., 2024). This is because it offers a productive and interactive environment and reaches a large population (Ali et al., 2018; Samsudeen & Mohamed, 2019). Haleem et al. (2022) posit that this transition to high use of digital technology in learning is necessary, as the traditional classroom has fallen short of creating an instant learning environment, and can expedite evaluations and foster higher levels of participation. The efficiencies offered by technologies are regarded as unparalleled compared to traditional learning approaches (Major et al., 2021; Wai Yan & Chao Wei, 2022), through the introduction of technology-assisted learning with smartboards, virtual laboratories, simulations, and other related technologies (Major et al., 2021; Haleem et al., 2022). Despite this, integrating digital technology into education systems is complex and challenging (Brown, 2015; Salavati, 2017).

Blackboard is a type of virtual learning environment (VLE), which are artificial environments built with information technology to support student learning. It includes features such as course management, content delivery, and communication tools that facilitate the educational process (Albakri & Abdulkhaleq, 2021; Hossain et al., 2017; Y. Wang, 2018). The virtual learning environment Blackboard is a widely used platform in educational institutions, particularly for facilitating online learning and teaching. It provides a comprehensive suite of tools that support various educational activities, including content delivery, communication, and assessment (Goeser & Williams, 2021). Blackboard serves as a digital solution that enhances the learning experience by providing a platform for teachers and students to interact and communicate effectively (Lai, 2019). It supports a variety of educational activities, from content delivery to evaluations and feedback, making it a versatile tool in education institutions (Albakri & Abdulkhaleq, 2021). The platform is part of a broader category of virtual learning environments that extend learning beyond physical spaces, allowing students to access educational resources and engage in learning activities remotely (Goeser & Williams, 2021). Usability evaluations suggest that, while Blackboard is effective in supporting educational activities, improvements are needed to enhance its user friendliness and accessibility (Asim et al., 2023). Its features contribute to improved learning quality, suggesting that its broader adoption could enhance educational experiences (Hossain et al., 2017; Lai, 2019). Despite its benefits, Blackboard faces usability challenges (Asim et al., 2023). A study comparing Blackboard with another learning management system, Canvas, shows that both platforms have usability scales and other statistical tools, indicating similar levels of user satisfaction and acceptability scores (Gumasing et al., 2022). Both platforms require strategic implementation and support, including instructor training and technical support, to maximise their educational benefits and mitigate limitations (Oudat & Othman, 2024).

The effective use of technologies in schools on a larger scale is frequently lacking. Several aspects influence the effectiveness of using e-learning technology for teaching and learning, such as facilitating conditions, social influence, and learning tradition (Al-Adwan et al., 2022). There are connections among behavioural intentions, social influence, enabling conditions, performance expectancy, and ease of use (Szajna, 2016). Social influence, including institutional support, affects willingness to adopt new technologies. Facilitating conditions, such as the availability of resources and support, also play a crucial role. These factors are particularly important in contexts where traditional learning methods are prevalent, as they can either support or hinder the transition to digital learning environments (Awotunde et al., 2021; Tewari et al., 2023). Learning tradition can negatively impact the continued usage intentions of technologies. This is because established educational practices may resist the integration of new technologies, leading to a preference for traditional methods over digital ones (Al-Adwan et al., 2022). Research studies have used technology-driven models to forecast technology-related intentions and behaviours in various contexts (Amadin et al., 2018). The probability of adopting a technology is contingent on the direct impact of crucial factors such as performance expectation, learning tradition, social influence, and facilitating conditions.

Within this complex environment, this study aimed to understand the influences on educators’ decisions regarding the continued use of technology in the public school system. This study was investigated using the following objectives. The first objective determined the levels of performance expectancy from the use of a virtual learning environment with its associated e-learning technology in the classroom to understand the educators’ assumptions regarding whether the particular technology improves their work performance (L.-Y.-K. Wang et al., 2019). The second objective determined the influences of learning tradition, facilitating conditions, and social influences on the relationship between performance expectancy and continued use of virtual learning environment with its associated e-learning technology.

The rest of the paper has the following sections. First, we present the grounding theory, followed by the development of the hypotheses and conceptual model. The next section presents the research methodology used in this study. The following sections present the results and analysis of the research, followed by a discussion that includes the implications for educators, school management, and policy-makers. The final section highlights the conclusions, limitations, and proposed future areas of study.

2. Theory and Literature

The Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT) are two prominent frameworks used to understand technology adoption (O’Dea, 2025; Safitri & Sari, 2024). Both models aim to predict and explain user behaviour towards technology, but they differ in their complexity and the factors they consider (Safitri & Sari, 2024). TAM focuses on perceived usefulness and perceived ease of use as primary determinants of technology acceptance (Or, 2024), while UTAUT incorporates additional factors such as social influence and facilitating conditions (Hakimi et al., 2024). UTAUT extends TAM by including additional constructs such as social influence and facilitating conditions (Safitri & Sari, 2024). These elements provide a more comprehensive view of technology adoption (Nguyen & Nguyen, 2024).

The unified technology acceptance and use theory model (UTAUT) is the theoretical basis for this study (Da Silva Soares et al., 2025; Dalogdog et al., 2024). The seminal work underlying this theory comes from Venkatesh et al. (2003). It was developed to explain why technology is accepted in various situations. This theory is in line with the theory of rational action (TRA), theory of planned behaviour (Fishbein & Ajzen, 1975), the technological acceptance model (TAM) (Davis, 1986), and TAM 2 (Venkatesh & Davis, 2000). Some elements of early technology theories were used to develop the UTAUT model. The need for a theory that could better explain the adoption and acceptance of technology than previous models signalled the formation of the theory. Venkatesh et al. (2003) posited that the UTAUT model can account for 50% of actual use of technology and 70% of aspirations for the use of technology. UTAUT is the most successful model for predicting both intended and actual uses of technology in its effectiveness. Recently, improved unified technology acceptance and use theory models have been studied (Al-Adwan et al., 2022). The UTAUT model adds the characteristics most relevant to learning management systems (LMS) to its components, since online learning depends on context.

The UTAUT model contains the following structures: traditional learning, facilitating conditions, and social influences (Al-Adwan et al., 2022). UTAUT has been critiqued for its limited applicability in developing countries, where cultural and infrastructural factors may influence technology acceptance differently than in developed contexts (Bayadea & Du Plessis, 2024). This supports the assertions of Amadin et al. (2018), who posit that a technology model needs to be context-specific. Sultana (2020) argued with different lenses and considered the UTAUT model to be in favour of this idea because it has wide applicability and the ability to explain the use and intention of the technology. Although the UTAUT model has been criticised, it remains the most effective and practical model for predicting the intentions and behaviours of technology use. Several studies have used the UTAUT model as a theoretical basis for research (Abbad, 2021). However, minimal prior research has looked at the mediating role that learning traditions play in facilitating conditions and the moderating influence of social impact on the relationship between performance expectancy and the continuous use of virtual learning environment with its associated e-learning technology.

3. Conceptual Model and Hypotheses

3.1. Performance Expectancy

Performance expectancy refers to the users’ assumption that a particular technology will improve their work performance (Venkatesh et al., 2003; L.-Y.-K. Wang et al., 2019). This view is also shared by Gunasinghe et al. (2020) and Abbad (2021), in arguing the effectiveness of a certain technology in helping them achieve their work-related goals. Based on the perspectives of learners and educators, the use of virtual learning environment with its associated e-learning is believed to improve their performance in assigned tasks. This is consistent with the performance goals outlined in the study. Educators and learners who have access to a certain e-learning technology typically develop the assumption that it will either improve their performance or not. Al-Adwan et al. (2022) assert that there is a similarity between perceived usefulness in the TAM and performance expectancy. Therefore, as suggested in the literature, we proposed the following hypothesis.

H1. 

The mean performance expectancy for the sample is similar to the ‘agree’ range of the hypothesised population.

3.2. Influences of Use of Virtual Learning Environment with Its Associated e-Learning Technology

Al-Adwan et al. (2022) argued that the intention to persist in using the technology refers to the desire and goal to continue doing so. Maintaining consistent use of e-learning for educational reasons is essential, as this can guarantee that students fully utilise e-learning technology. Employing e-learning allows education to be accessible to a wide range of individuals, while also being cost-effective, innovative, and imaginative. Additionally, it allows students to acquire knowledge from any geographical location (Ali et al., 2018; Samsudeen & Mohamed, 2019). Educators are responsible for teaching, while students are expected to use the e-learning platform to enhance their knowledge. Lwoga and Komba (2015) found that performance expectancy is a stronger predictor of both technology use intentions and actual behaviour. Cheah et al. (2021) also support this view and posit that performance expectancy is more effective than perceived usefulness (Cheah et al., 2021).

3.2.1. Learning Tradition

The learning tradition refers to the established and known status quo framework. Kleijnen et al. (2009) suggested that individuals typically adhere to established customs and conventions inside an institution, making them resistant to any efforts to deviate from the familiar. The impact of learning tradition on technology acceptance can vary significantly in different cultural and educational contexts. Some studies have shown that cultural factors can influence the acceptance of educational technologies, suggesting that learning traditions rooted in specific cultural contexts may mediate technology adoption differently (Göğüş et al., 2012; Z. Zhang et al., 2022). The tradition can also affect the flow experience in e-learning systems. Integrating job requirements and skill certificates into e-learning design can enhance the flow experience, suggesting that traditional educational goals can be aligned with digital learning environments to improve acceptance (Y. Zhang et al., 2020). When the current status quo is in conflict with the proposed technology, it becomes challenging for the user to embrace and incorporate the suggested innovation (Ma & Lee, 2018). Several authors have utilised the learning tradition to assess its effectiveness in predicting individuals’ intentions and behaviours around the use of technology (Mpungose, 2020). The use of learning tradition was employed in this study due to its ability to predict the use and rationale for adopting e-learning technology.

H2. 

Learning tradition has a mediation effect on the relationship between performance expectancy and the continued use of VLE and associated e-learning technology.

3.2.2. Facilitating Conditions

Facilitating conditions encompass the degree to which users of technology perceive that sufficient infrastructure has been established to facilitate the use of technology in a specific environment. Venkatesh et al. (2003) explained facilitating conditions as the user’s recognition of the organisation’s infrastructure, essential for ensuring a positive user experience. Kamaghe et al. (2020) argued that consumers’ capacity to have a positive experience is hindered by a lack of knowledge, inadequate help, and insufficient resources. In this study, the facilitating conditions include factors such as strong connectivity, a well-designed website, and the availability of information technology support. Every educational institution must ensure that it has established the required circumstances to guarantee a positive user experience for technology users. Multiple researchers have used favourable situations in various studies to forecast intentions and behaviour regarding the use of technology (Venkatesh et al., 2012; Ambarwati et al., 2020). These conditions act as mediators by providing the necessary support and resources, enhancing the technology adoption likelihood. They are crucial in contexts where users may face barriers to technology use, such as outdated resources or lack of training (Adli, 2023). This is particularly relevant within South Africa, where there are barriers to effectively leveraging technology (Mtotywa et al., 2022a).

H3. 

Facilitating conditions have a mediation effect on the relationship between performance expectancy and continued use of VLE with its associated e-learning technology.

3.2.3. Social Influence

Social influence refers to the degree to which a user believes that important individuals should adopt a particular technology (Da Silva Soares et al., 2025). Social impact is the extent to which a user believes that important individuals approve of using technology (Venkatesh et al., 2003). Abbad (2021) expanded on this and posited that the social effect is the extent to which individuals perceive that important persons support their use of technology. Social influence can be connected to social norms, one of TAM’s components. Social impact strongly predicts technological intention and use in different settings (Lwoga & Komba, 2015; Gunasinghe et al., 2020).

Social influence significantly impacts intention and use behaviour, as users are influenced by recommendations and social norms, suggesting that social influence can moderate the relationship between other UTAUT constructs and technology adoption, emphasising the importance of social norms in technology acceptance (Istiqomah et al., 2024). Social influence is an important factor that influences the use behaviour of new technologies (Muangmee et al., 2021), and was shown to have a positive impact on users’ intentions to use new technologies (Sair & Danish, 2018).

H4. 

Social influence moderates the mediation effect of learning tradition on the relationship between performance expectancy and the continued use of VLE with its associated e-learning technology.

H5. 

Social influence moderates the mediation effect of facilitating conditions on the relationship between performance expectancy and continued use of VLE with its associated e-learning technology.

Therefore, the conceptual model can be visualised as shown in Figure 1. This model incorporates the latent variables: performance expectancy (predictor), continued use intention (outcome), traditional learning (mediator), facilitating conditions (mediator), and social influences (moderator) (Al-Adwan et al., 2022). Age, gender, and education level are frequently considered in UTAUT studies to understand their impact on technology acceptance (Tahir, 2023). In this study, demographic variables were investigated as control variables to account for their potential influence on technology acceptance. This approach helps isolate the effects of UTAUT’s core constructs. The integration of demographic variables provides a more nuanced understanding of how different groups may respond to technology adoption in educational contexts. Several studies have investigated the relationships between UTAUT constructs and technology acceptance in educational settings (Chu & Dai, 2021; Sarkam, 2019). Others have conducted moderated mediation on the utilisation of online educational resources in teaching (Kio & Lau, 2017).

4. Methodology

4.1. Research Design

The study had ethical clearance [Ref #: FCRE2023/FR/08/009-MS (2)]. It was based on a cross-sectional descriptive quantitative design (Seoke et al., 2023). The advent of educational tools, such as VLEs with their associated e-learning technology, including Blackboard and myLMS, has influenced the education system to use e-learning (Palvia et al., 2018). The technology evaluated in this study was the Blackboard Collaborate technology platform, an internet-based educational platform that allows virtual teaching and learning.

4.2. Population and Sample

The population of this research was educators who understood the factors that affect continued use of the Blackboard platform. There were 93,453 basic education educators in Gauteng, South Africa, spread across 2991 schools in 2022 (Statista, 2024). This population resulted in a sample size of 383 based on Yamane (1967) for a finite population. The study obtained 306 randomly selected participants from 30 schools, with 10–12 educator representatives per school. The 306 responses equate to a response rate of 79.9%. Within the responses, 44.4% were in the 36 to 45 years age group, 33.3% were 46 years and older, 11.1% were between the ages of 18 and 25 years, and another 11.1% were in the 26 to 35 years age group. Among these respondents, 55.6% were male, 33.3% were female, and 11.1% were other or preferred not to say. Regarding education, 33.3% had a Bachelor’s degree, 55.6% held an Honours degree, and 11.1% had a Master’s degree.

4.3. Instrument and Data Collection

The instrument collected empirical data related to the latent variables of interest in using collaborative Blackboard technology: performance expectancy, continued use intention, learning tradition, facilitation conditions, and social influence. This instrument was adapted from Davis (1986) and Al-Adwan et al. (2022). Four statements related to each latent variable were assessed using a five-point Likert scale (strongly disagree, disagree, neither disagree nor agree, agree, and strongly agree). The instrument is provided in Appendix A. This was a self-administered survey, with respondents completing and returning it to the central collection point.

4.4. Data Analysis

The data collected were analysed using IBM SPSS version 29 and SmartPLS version 4. The empirical data did not have problems with missing values, as the missing values were under the threshold of 5% (Schafer, 1999). There were also no extreme outliers based on the interquartile range (Dash et al., 2023), with all data being within three times the interquartile range. For sample sizes larger than 300, skewness and kurtosis are preferred instead of z-values to determine the normality distribution (Sovey et al., 2022). The range of skewness was −0.673 to 1.390, and there was a kurtosis of −1.000 to 1.528 for all variables. These were within the ±2 and ±7 range, indicating normally distributed data (Hair et al., 2010). Harman’s one-factor test was used to test common-method bias, and the result of 35.78%, which is less than the 50.00% threshold, indicated no common-method bias issues (Podsakoff et al., 2003). For objective one, one-sample t-test was used to test H1 by comparing the mean of a sample with the population mean (Gerald, 2018; Al-kassab, 2022). This was achieved by assessing whether the sample mean differed substantially from the hypothesised population mean, using ${\mu }_0$ = 3.40. The value was based on the range > 3.40, which is in the ‘agree’ range (Sözen & Güven, 2019). Partial Least Square Structural Equation Modelling (PLS-SEM) (Hair et al., 2019) and PROCESS Models (Sarstedt et al., 2020) were used to investigate objective two testing mediation and moderated mediation analysis, H2–H5. This was based on bootstrapping using 10,000 sub-samples using bias-corrected and accelerated bootstrap (BCa) and considering the control variables—age, gender, and education—in the calculation.

5. Results

5.1. Level of Performance Expectancy

Four variables (PE1, PE2, PE3, and PE4) which were part of the instrument were used to determine the latent variable, performance expectancy. The mean value (M) ranged from M = 1.80 to 2.21 (

Table 1. Inter-item correlations and Cronbach alpha coefficient.

			Inter-Item Correlation Matrix
	Mean	SD	PE1	PE2	PE3	PE4
PE1	2.21	1.182	1.000	0.541	0.321	0.352
PE2	1.80	1.022	0.541	1.000	0.517	0.427
PE3	2.15	1.123	0.321	0.517	1.000	0.365
PE4	1.83	1.012	0.352	0.427	0.365	1.000
Number of elements = 4 Cronbach alpha (α) = 0.74

). The inter-item correlation matrix of these variables showed values of 0.321 to 0.541, which is within the acceptable range. In addition, the Cronbach alpha value, α = 0.74, indicated inter-item reliability, as it was higher than 0.7 (George & Mallery, 2019), implying that these four items measure the same latent variable and are appropriately assigned to the scale.

The results of the analysis using the one-sample t-test showed a statistically significant negative difference between the sample mean and the hypothetical mean, ${\mu }_0$ = 3.40 with t(305) = −30.12, p < 0.001, with Cohen’s d = 1.722 (

Table 2. Results from the one-sample t-test.

	Test Value = 3.40
	t	df	Significance		Mean Difference	95% CI of the Difference		Cohen’s d $\left|\mathit{d}\right|$
			One-Sided p	Two-Sided p		Lower	Upper
PEY	−30.117	305	0.000	0.000	−1.40327	−1.4950	−1.3116	−1.722

CI = confidence interval.

). This implies that educators viewed the performance expectancy as 1.722 standard deviations lower than the ${\mu }_0$ = 3.40. As such, H1 was not supported, as the mean performance expectancy for the sample was lower than the ‘agree’ range of the hypothesised population.

5.2. Measurement Model

The confirmatory factor analysis was determined using the measurement model. Hair et al. (2022) cautioned against reporting the model’s fit in PLS-SEM. Among these indices, the standardised root mean square residual (SRMR) is the most developed, with a threshold of 0.10 for an acceptable model fit. Based on this threshold, the model was considered a good fit, with SRMR = 0.093 for the estimated model and SRMR = 0.087 for the saturated model (

Table 3. Factor loadings, composite reliability, and convergence validity.

	Factor Loadings (λ)	Cronbach Alpha (α)	Composite Reliability (rho_a)	Composite Reliability (rho_c)	Average Variance Extracted (AVE)
Performance expectancy		0.744	0.752	0.839	0.568
PEY1	0.734
PEY2	0.844
PEY3	0.732
PEY4	0.695
Learning tradition		0.708	0.733	0.805	0.510
LTD1	0.789
LTD2	0.640
LTD3	0.667
LTD4	0.749
Facilitating conditions		0.876	0.877	0.915	0.728
FCC1	0.850
FCC2	0.867
FCC3	0.830
FCC4	0.867
Social influence		0.832	0.842	0.888	0.667
SOI1	0.842
SOI2	0.832
SOI3	0.718
SOI4	0.866
Continued use intention		0.793	0.795	0.879	0.707
CUI2	0.828
CUI3	0.840
CUI4	0.855
SOI X PEY	1.000
Age	1.000
Gender	1.000
Education	1.000

Model fit: SRMR = 0.093 (estimated model) SRMR = 0.087 (saturated model).

). Henseler et al. (2014) had previously suggested that SRMR is useful in detecting misspecification of the PLS-SEM model. Factor loadings were greater than 0.70, except for LTD2 (λ = 0.640) and (λ = 0.667), which were greater than 0.6, with the average of LTD1–LTD4 being acceptable, as it was more than 0.70. Cheung et al. (2024) highlight that, in addition to factor loading, the extracted average variance (AVE) and composite reliability should be analysed (Table 3).

To establish an appropriate level of convergence validity, the AVE must be greater than or equal to 0.5. This indicates that the underlying latent variables explain at least 50% of the variance in the indicators (Fornell & Larcker, 1981). All AVE values in this model were greater than 0.5, ranging from 0.568 to 0.728, implying acceptable convergence validity. The model shows composite reliability, with rho_c = 0.805–0.915 and rho_a = 0.733–0.877. The model also had Cronbach alpha values higher than 0.7, ranging from α = 0.708 to α = 0.832.

Table 4. Discriminant validity analysis with HTMT.

	Age	Education	FCC	Gender	LTD	PEY	SOI	CUI	SOI X PEY
Age
Education	0.375
FCC	0.045	0.033
Gender	0.516	0.029	0.028
LTD	0.101	0.055	0.856	0.061
PEY	0.165	0.042	0.609	0.093	0.680
SOI	0.124	0.041	0.499	0.055	0.783	0.633
CUI	0.062	0.008	0.719	0.018	0.841	0.656	0.836
SOI X PEY	0.009	0.047	0.174	0.058	0.299	0.373	0.297	0.344

provides the discriminant validity using a heterotrait–monotrait ratio matrix (HTMT). The results confirmed discriminant validity, with all values less than 0.90 (HTMT₉₀) for the measured latent variables (Hair et al., 2022).

Correlation Matrix

Table 5. Means, standard deviations, and correlations among factors.

	Mean	SD	1	2	3	4	5	6	7	8
CUI	2.38	0.789
PE	2.00	0.815	0.509 **
FCC	2.24	0.882	0.530 **	0.492 **
LTD	2.80	0.831	0.681 **	0.478 **	0.656 **
SOI	2.23	0.853	0.703 **	0.494 **	0.428 **	0.562 **
Age	__	__	−0.044	−0.140 *	−0.042	−0.045	−0.112
Gender	__	__	0.010	0.078	−0.012	0.020	0.048	−0.516 **
Education	__	__	0.013	0.031	0.031	−0.019	−0.010	−0.375 **	0.029

N = 306; ** p < 0.01, * p < 0.05.

presents the correlation matrix, highlighting the interrelationships among the different factors: PEY, FCC, LTD, SOI, CUI, socio-demographic variables, age, gender, and education.

PEY had a statistically significant positive and strong correlation with CUI (r = 0.509, p < 0.01), FCC with CUI (r = 0.530, p < 0.01), LTD with CUI (r = 0.681, p < 0.01), and SOI with CUI (r = 0.509, p < 0.01). Demographic variables, age, gender, and education did not have a statistically significant correlation with CUI. There was also a statistically significant correlation between FCC and LTD (r = 0.656, p < 0.01), FCC and SOI (r = 0.428, p < 0.01), and LTD and SOI (r = 0.562, p < 0.01). The substantial correlations among these different factors provided initial support for research hypotheses and a baseline for potential mediation and moderation effects.

5.3. Structural Model

The hypotheses of the study were tested using the structural model. The study first considered the standard assessment criteria, which are the explanatory power (R²), also known as the predictive power (Rigdon, 2012), and the effect size (f²), assessing the change in R² of the endogenous factor when a predictor is removed from the model. It also assessed Stone-Geisser’s Q² for the predictive relevance of the endogenous variables and the model’s multicollinearity to confirm no bias in the regression results (

Table 6. Explanatory power (R²), effect size (f²), and predictive relevance (Q²) of the model.

	R2	Q2 Predict	RMSE	MAE	f2 FCC	f2 LTD	f2 CUI
CUI	0.519	0.413	0.773	0.586
LTD	0.546	0.524	0.694	0.558			0.207
FCC	0.291	0.250	0.872	0.586			0.040
PEY					0.148	0.114	0.037
SOI					0.062	0.574
SOI x PEY					0.001	0.015
Age					0.001	0.002	0.000
Education					0.001	0.000	0.000
Gender					0.001	0.000	0.002

). There was substantial explanatory power based on R² > 0.26 (Cohen, 1988), with R² = 0.519 for CUI, R² = 0.546 for LTD, and R² = 0.291 for FCC.

Generally, there was a negligible to small effect when removing the different predictors from the model, with a large effect size (Cohen, 1988) for SOI on LTD (f² = 0.574), medium for LTD and CUI (f² = 0.207) and PEY on FCC (f² = 0.148), and small for PEY on LTD (f² = 0.114). The model also had good predictive relevance based on Q² > 0 (Hair et al., 2022), with CUI and LTD having a strong degree of predictive relevance (Q² = 0.413 and Q² = 0.524, respectively) and FCC having a moderate degree of predictive relevance (Q² = 0.250).

Mediation Analysis

Mediation analysis was performed to determine the mediation effect of LTD on PEY and CUI and the mediation effect of FCC on PEY and CUI (

Table 7. Results of the mediation and moderated mediation analyses.

Effects	Paths	β	t-Statistic	p-Value
Total effect	FCC -> CUI	0.254	3.795	0.000
	LTD -> CUI	0.303	4.479	0.000
	PEY -> CUI	0.528	7.146	0.000
	PEY -> FCC	0.455	2.500	0.012
	PEY -> LTD	0.695	5.279	0.000
	SOI -> CUI	0.315	3.977	0.000
	SOI -> FCC	0.314	2.096	0.036
	SOI -> LTD	0.782	6.988	0.000
	SOI x PEY -> CUI	−0.055	1.912	0.056
	SOI x PEY -> FCC	−0.025	0.352	0.725
	SOI x PEY -> LTD	−0.167	3.655	0.000
Direct effect	FCC -> CUI	0.254	3.795	0.000
	LTD -> CUI	0.303	4.479	0.000
	PEY -> CUI	0.211	2.728	0.006
	PEY -> FCC	0.455	2.500	0.012
	PEY -> LTD	0.695	5.279	0.000
	SOI -> FCC	0.314	2.096	0.036
	SOI -> LTD	0.782	6.988	0.000
	SOI x PEY -> FCC	−0.025	0.352	0.725
	SOI x PEY -> LTD	−0.167	3.655	0.000
Specific indirect effect	PEY -> LTD -> CUI	0.208	3.874	0.000
	PEY -> FCC -> CUI	0.109	2.671	0.008
	SOI -> LTD -> CUI	0.237	3.738	0.000
	SOI -> FCC -> CUI	0.078	1.934	0.053
	SOI x PEY -> LTD -> CUI	−0.050	3.019	0.003
	SOI x PEY -> FCC -> CUI	−0.005	0.359	0.720
Conditional indirect effect	PEY -> LTD -> CUI conditional on SOI at +1 SD	0.054	2.800	0.005
	PEY -> FCC -> CUI conditional on SOI at +1 SD	0.095	3.287	0.001
	PEY -> LTD -> CUI conditional on SOI at −1 SD	0.139	4.094	0.000
	PEY -> FCC -> CUI conditional on SOI at −1 SD	0.103	3.746	0.000
	PEY -> LTD -> CUI conditional on SOI at mean	0.096	4.023	0.000
	PEY -> FCC -> CUI conditional on SOI at mean	0.099	3.969	0.000

).

The results showed a statistically significant indirect effect of PEY on CUI through LTD (H2: β = 0.203, t = 3.912, p < 0.001). The total effect of PEY on CUI was statistically significant (β = 0.522, t = 7.095, p < 0.001). With the inclusion of the mediator, LTD or FCC, the effect of PEY on CUI was still statistically significant (β = 0.211, t = 2.728, p < 0.01), based on the direct effect results. The results also showed a complementary partial mediation effect (Hair et al., 2022) of LTD on the relationship between PEY and CUI.

Therefore, H2 was supported. FCC had a statistically significant indirect effect on the relationship between PEY and CUI (H3: β = 0.108, t = 2.643, p < 0.01). The results also showed that FCC had a complementary partial mediation effect on the relationship between PEY and CUI. Thus, H3 was supported.

5.4. Moderated Mediation Analysis

Conditional mediation (CoMe) analysis combines mediation and moderation analysis to investigate and test hypotheses regarding how mediated interactions change due to environment, boundaries, or individual variations (Cheah et al., 2021). Hayes (2015) posited that the CoMe index is used to determine the effect of the moderator on the mediated relationship with the CoMe index, which is statistically significantly different from zero, indicating a conditional mediation. The CoMe index = −0.049 from the path, and SOI x PEY -> LTD -> CUI was statistically significant (p < 0.01). When probing the indirect conditional effect, the results showed that PEY -> LTD -> CUI was conditional on SOI at mean, β = 0.094: if the SOI was increased (+1 SD), the mediation effect decreased, β = 0.053, while if it decreased, it increased, β = 0.136. At all three levels (Mean, −1 SD and +1 SD), the paths were statistically significant, meaning the changes between the levels were significant. This was also evident from the Johnson-Neyman plot (Figure 2).

As such, H4 was supported. For the path SOI x PEY -> FCC -> CUI, the CoMe index = −0.005, and was not statistically significant, at p > 0.05, and as such, H5 was not supported.

6. Discussion

We live in a complex world with an environment that involves multiple factors; therefore, investigating the effects of only one or two variables in isolation would be artificial and insignificant (Sarstedt et al., 2020). This is particularly true for emerging economies such as South Africa, where the adoption and use of technology is highly influenced by prevailing socio-economic factors (Tambotoh et al., 2015; Mtotywa et al., 2022b) and barriers to leveraging technologies, such as poor digital culture and the digital divide (Mphahlele et al., 2021; Mtotywa et al., 2022a). The unified technology acceptance and use theory model (UTAUT) has been used to analyse the dynamics of technology in education (Da Silva Soares et al., 2025; Dalogdog et al., 2024). This study’s results showed that the mean performance expectancy (PEY) was lower than the ‘agree’ range in support of hypothesis one. This implies that the educators’ assumption is that the deployed technology does not improve their work performance. This is plausible, as Olofsson et al. (2019) argued that digital educator competence is difficult and remains elusive because the necessary conditions, possibilities, challenges, and contextual and societal situations are always changing. Research also indicates that educators experience stress and sometimes anxiety when acquiring technological skills in an educational setting (Sharaievska et al., 2022).

The results also showed that learning tradition (LTD) has a complementary partial mediation effect on the relationship between PEY and CUI, supporting hypothesis two. Additionally, FCC also has a complementary partial mediation effect on the relationship between PEY and CUI, supporting hypothesis three, in that facilitating conditions had a mediation effect on the relationship between performance expectancy and continued use of VLE with its associated e-learning technology. Ji et al. (2019) highlighted that the facilitating conditions are the extent to which stakeholders perceive that they have received adequate support to ensure a positive experience when using technology within education. The absence of essential assistance hinders the user’s ability to have a satisfactory technological experience.

Learners and educators will consistently evaluate the current learning and teaching environment to determine its effectiveness. When VLE with its associated e-learning technology does not perform effectively, persuading learners and educators to embrace subsequent e-learning technology becomes challenging. This study not only investigated the mediation effect of LTD and FCC, but also tried to understand whether the moderator, SOI, changed the strength of indirect effects. We also performed a more comprehensive analysis using CoMe, as Jacoby (1978) had previously noted and as was recently supported by Sarstedt et al. (2020). The SOI was critical to strengthening the mediation effect of the LTD, but not of the FCC. These results support hypothesis four but not hypothesis five.

This study has implications for educators and school management. Educators must acknowledge that technology has revolutionised our daily lives and has become necessary in learning, so digital competency and use by educators are imperative. Educators need the support of school management to achieve this competency and use technology effectively for instruction and learning. Furthermore, it is imperative for educators’ education programmes to prioritise the integration of technical, pedagogical, and topic knowledge in the training of educators (Ley et al., 2021) and to incorporate technology in day-to-day education and learning (Major et al., 2021; Haleem et al., 2022). This study also has implications for policy-makers. Policy adoption and directives in countries around the world promote the use of technology (Law & Lee, 2023). Policy-makers must create an enabling environment that ensures that facilitating conditions are good and that learning traditions change with time and strengthen the realities of technological advancements introduced by the fourth (Haleem et al., 2022; Mtotywa et al., 2024) and fifth industrial revolutions (Noble et al., 2022). This study contributes both theoretically and methodologically. Cheah et al. (2021) posit that, although CoMe analysis can be an important component of empirical investigations aimed at advancing theory, there is a dearth of studies on its applicability in general, and particularly in PLS-SEM.

7. Conclusions

Advances in technology, the coronavirus outbreak, and customised learning have resulted in a noticeable increase in the number of schools that use digital technology. The use of technology has become a modern paradigm within the world’s education system. This study concludes that social and operational factors greatly influence the dynamics of continued use of technology and cannot be discounted by practitioners and policy-makers in their quest to increase technology use in the school system. It can be concluded that educators view the technology deployed as not improving their work performance. It can also be concluded that learning tradition and facilitating conditions mediate the relationship between performance expectancy and continued use of VLE with its associated e-learning technology. Social influence moderates the mediating of the learning tradition in the relationship between performance expectancy and continued use intention. In probing the conditional indirect effect, the results showed that, if the social influence increased, the mediation effect of learning tradition decreased. On the contrary, if it decreased, the mediation effect of learning tradition increased.

The study’s limitations are that the sample was recruited from 30 schools and was not generalised across South Africa. Furthermore, the study only covered public schools and did not include private schools, which now make up about 10% of South African schools (Codera, 2024). Future studies should also be conducted to understand the influence of working conditions and unionised environments (Mafisa, 2017; Han & Maloney, 2021) on performance expectancy and continued use of technology in the public school system. Additionally, a similar study with a comparative analysis of public vs. private schools is proposed as a future study to understand how socio-economic dynamics influence educators’ decisions to continue the use of VLE with its associated e-learning technology.