Perceived usefulness (PU) refers to the extent to which individuals believe that using a particular technology will enhance their performance (Davis, 1989). In the context of Learning Management Systems (LMS), PU reflects students’ perceptions of how digital platforms contribute to their academic success—whether through improved access to course materials, streamlined communication, or greater flexibility in learning. A growing body of research confirms that PU is a strong predictor of behavioural intention to use LMS platforms. For example, Barz et al. (2024) found that PU significantly influenced students’ willingness to adopt LMS tools in higher education settings. Similarly, Zulherman and Zaelani (2023) demonstrated that PU plays a central role in shaping LMS engagement among university students, validating its relevance in post-pandemic digital learning environments.

Hypothesis 1 (H1): Perceived usefulness positively influences students’ behavioural intention to use Learning Management Systems.

Perceived ease of use (PEOU) refers to the extent to which a user believes that interacting with a system will be free of effort (Davis, 1989). In the context of Learning Management Systems (LMS), this construct captures students’ perceptions of how intuitively and efficiently they can navigate digital platforms for academic purposes. PEOU has proven particularly critical in developing contexts, where digital literacy and prior exposure to educational technologies may be limited (Rhema and Miliszewska, 2014). Systems perceived as easy to operate are more likely to be embraced, especially when students face infrastructural constraints or lack formal ICT training. Alharbi and Drew (2014) further confirms that usability plays a decisive role in shaping students’ intention to adopt LMS tools, reinforcing the importance of interface simplicity and user support in the acceptance of technology.

Hypothesis 2 (H2): Perceived ease of use positively influences students’ behavioural intention to use Learning Management Systems.

Subjective norms—defined as the perceived social pressure to perform or not perform a certain behaviour—introduce a socio-psychological dimension to the Technology Acceptance Model (TAM) (Ajzen, 1991). In educational environments, students’ behavioural intentions are often shaped by peer influence, instructor expectations, and institutional advocacy. These social cues can reinforce or discourage engagement with digital platforms, especially in contexts where technology adoption is still emerging. Chu and Chen (2016) found that subjective norms significantly influenced learners’ attitudes toward e-learning systems, particularly when institutional support was visible. Similarly Hussein (n.d.), demonstrated that peer and faculty endorsement of LMS platforms positively affected students’ intention to use them, underscoring the importance of social validation in technology uptake.

Hypothesis 3 (H3): Subjective norms positively influence students’ behavioural intention to use Learning Management Systems.

While behavioural intention does not guarantee action, it is widely acknowledged as its strongest predictor of actual technology use (Abbad, 2021). In educational settings, intention often aligns with opportunity and access; students who intend to use LMS platforms are more likely to translate that intention into actual usage—particularly when system availability, institutional support, and digital infrastructure are sustained. This relationship has been validated across multiple TAM-based studies, where behavioural intention consistently mediates the link between attitudinal constructs and observable engagement with digital (Zulherman and Zaelani, 2023).

Hypothesis 4 (H4): Behavioural intention positively influences the actual usage of Learning Management Systems by students.

The theoretical model (Figure 1) for this study builds on these four hypotheses, positing that perceived usefulness, perceived ease of use, and social norms influence behavioural intention, which in turn affects actual system usage. This model will be tested using structural equation modelling (PLS-SEM), a rigorous method for assessing both direct and indirect effects among variables.

This study adopted a quantitative research design and employed the Extended Technology Acceptance Model (TAM) to investigate the factors influencing students’ adoption and usage of Learning Management Systems (LMS). The model integrated three key constructs—perceived usefulness, perceived ease of use, and subjective norms—as predictors of behavioural intention. Behavioural intention was then modelled as a mediating variable influencing actual system usage. This structure reflects established TAM extensions validated in educational technology research, particularly in developing contexts where socio-psychological and infrastructural factors shape digital engagement (Chu and Chen, 2016; Zulherman and Zaelani, 2023).

The target population for this study comprised university students enrolled in ten higher education institutions across three provinces of the Democratic Republic of Congo: North Kivu, Ituri, and Kasai-Central. All participants had prior experience using at least one Learning Management System (LMS), most commonly Google Classroom, as part of their academic coursework. This criterion ensured that respondents could provide meaningful reflections on LMS adoption and usage.

Participants were recruited using purposive sampling from six Congolese universities that had prior experience with Learning Management Systems (LMS). These universities were selected according to four criteria: (1) prior institutional LMS deployment (evidence of at least one semester of LMS use), (2) geographic representation across provinces affected by both conflict and policy change (North Kivu, Ituri, Kasai-Central), (3) institutional size and diversity of disciplines (to capture undergraduate students across STEM and social sciences), and (4) accessibility and willingness of local administrators to facilitate data collection. Within each selected university, students were sampled through convenience/purposive approaches targeting classes that used LMS in the preceding academic terms. A total of 400 valid responses were collected, exceeding the minimum sample size recommended for structural equation modelling (SEM) using SmartPLS. According to (Hair et al., 2021), a sample of 200 or more is generally sufficient to achieve statistical power and model stability when estimating complex path relationships.

Data were collected using a structured questionnaire developed in French, the primary language of instruction across the participating universities. The instrument was designed to measure five core constructs derived from the Extended Technology Acceptance Model (TAM): perceived usefulness, perceived ease of use, subjective norms, behavioural intention, and actual system use. Each construct was operationalised using a minimum of three items, adapted from validated TAM instruments. Responses were recorded on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree), ensuring consistency in measurement and facilitating structural equation modelling (SEM) analysis.

The questionnaire was administered using KoboToolbox, a platform specifically designed for data collection in low-resource and offline environments. This choice ensured accessibility across diverse institutional settings, with data gathered both online and offline depending on campus infrastructure and connectivity. Before participation, students provided informed consent, and the study adhered to established ethical protocols, including confidentiality safeguards and assurances of voluntary participation.

Structural Equation Modelling (SEM) was conducted using SmartPLS version 4, a variance-based approach well-suited for exploratory research and complex models with latent constructs (Hair et al., 2021). The measurement model was assessed for internal consistency reliability using Cronbach’s Alpha and Composite Reliability (CR), both of which are standard indicators for evaluating construct stability (Sarstedt et al., 2017). Convergent validity was examined through the Average Variance Extracted (AVE), while discriminant validity was confirmed using the Fornell-Larcker criterion and the Heterotrait-Monotrait (HTMT) ratio, as recommended for PLS-SEM applications (Henseler et al., 2015).

The structural model was evaluated using path coefficients, coefficient of determination (R²), and predictive relevance (Q²), which collectively assess the explanatory and predictive power of the model. Bootstrapping with 5000 subsamples was employed to test the statistical significance of all hypothesised relationships, ensuring robustness in parameter estimation

Discriminant validity was evaluated using both the Fornell–Larcker criterion and the Heterotrait–Monotrait (HTMT) ratio, as recommended for reflective measurement models in PLS-SEM (Henseler et al., 2015; Hair et al., 2021). As shown in Table 3, the square root of the Average Variance Extracted (AVE) for each construct (diagonal values) was consistently greater than its correlations with other constructs (off-diagonal values), thereby satisfying the Fornell–Larcker criterion (Fornell and Larcker, 1981). This indicates that each construct shares more variance with its own indicators than with those of other constructs, supporting discriminant validity.

Complementarily, Table 4 reports the Heterotrait–Monotrait (HTMT) ratios, all of which were below the conservative threshold of 0.85, thereby confirming discriminant validity (Henseler et al., 2015). The HTMT criterion is considered a more robust and reliable alternative to traditional methods such as the Fornell–Larcker criterion, particularly in detecting discriminant validity issues in variance-based structural models (Franke and Sarstedt, 2019). Values below 0.85 suggest that the constructs are empirically distinct and not excessively correlated, reinforcing the conceptual separation of latent variables within the model (Hair et al., 2021).

Collectively, these results confirm that each latent construct is empirically distinct from the others, satisfying the requirements for discriminant validity (Fornell and Larcker, 1981; Henseler et al., 2015). Taken together, Tables 1–4 provide robust evidence of measurement model adequacy, demonstrating high internal consistency reliability, strong indicator loadings, satisfactory convergent validity, and clear discriminant validity. These findings affirm the psychometric soundness of the instrument and its suitability for structural equation modelling using PLS (Hair et al., 2019).

As reported in Table 6, the structural model accounts for 42.3% of the variance in behavioural intention (INT) and 34.3% in actual usage (USAGE). These R² values reflect moderate explanatory power, as per (Chin, 1998) classification, which considers values around 0.33 as indicative of acceptable model strength in behavioural and social sciences. The findings affirm that perceived usefulness (PU), perceived ease of use (PEOU), and social norms (SN) are salient predictors of intention and, indirectly, of usage. However, the residual unexplained variance suggests that the model may benefit from the inclusion of additional constructs to capture the multifaceted nature of technology adoption.

Specifically, variables such as habit, facilitating conditions, and perceived behavioural control have been shown to enhance predictive accuracy in extended frameworks like UTAUT and the Theory of Planned Behaviour (Ajzen, 1991; Venkatesh et al., 2003). Habit, for instance, reflects the automaticity of repeated behaviour and has demonstrated incremental validity in explaining continued usage. Facilitating conditions (Limayem et al., 2007)—such as access to infrastructure, technical support, and organisational readiness—may also moderate the intention–usage link, particularly in resource-constrained environments. Thus, while the current model offers a parsimonious and theoretically grounded explanation, future research should consider integrating these complementary dimensions to improve its explanatory scope and contextual relevance.

As reported in Table 7, the Q² values obtained via cross-validated redundancy were 0.406 for behavioural intention (INT) and 0.339 for actual usage (USAGE), both substantially above zero. These values indicate that the model possesses meaningful predictive relevance for its endogenous constructs, as per the guidelines established by Stone (1974) and Geisser (1974). In the context of PLS-SEM, Q² values above 0.25 are generally interpreted as indicative of strong predictive accuracy (Hair et al., 2019), particularly when assessing behavioural outcomes in technology adoption research.

The use of blindfolding procedures to generate Q² statistics provides a robust test of the model’s out-of-sample predictive capability, complementing the explanatory power indicated by R² values. The results suggest that the latent predictors—namely perceived usefulness, perceived ease of use, and social norms—offer not only theoretical relevance but also practical utility in forecasting user intentions and behaviours. This reinforces the model’s applicability in real-world settings and supports its deployment in future empirical studies seeking to understand technology acceptance dynamics

The findings of this study offer both theoretical and practical implications. Theoretically, the validated model reinforces the enduring relevance of TAM constructs—particularly Perceived Usefulness (PU) and Perceived Ease of Use (PEOU)—in predicting behavioural intention, while demonstrating the added value of integrating Social Norms (SN) to capture socio-contextual influences. This aligns with calls for culturally sensitive extensions of TAM and UTAUT, especially in non-Western and resource-constrained environments (Straub et al., 1997; Srite and Karahanna, 2006). The results confirm that technology adoption in such contexts is shaped not only by individual perceptions but also by social expectations and collective behavioural patterns.

Practically, the study provides specific, actionable strategies for policymakers, system designers, and institutional leaders seeking to foster effective and inclusive digital adoption:

Collectively, these practical insights demonstrate that digital transformation in emerging economies requires addressing not only technological factors but also cultural expectations, resource limitations, and the broader social environment. By aligning system design and institutional practices with these behavioural drivers, universities can meaningfully enhance digital adoption and improve learning outcome

Despite its contributions, the study is subject to several limitations. First, the model excluded potentially influential constructs such as habit, facilitating conditions, and trust, which may further explain variance in usage behaviour. Second, the use of cross-sectional data limits the ability to capture temporal dynamics or infer causality. Third, although the study aimed to validate the extended TAM in a Congolese context, demographic data (e.g., age, gender, year of study) were not collected, which prevented the inclusion of demographic characteristics as control variables or the examination of subgroup differences. This absence restricts the ability to assess how demographic variations might influence the studied relationships. Fourth, the generalisability of the findings may be constrained by sample characteristics and institutional context. Finally, while subjective norm (SN) was included, other cultural dimensions—such as power distance or uncertainty avoidance—were not explicitly modelled, which may influence behavioural outcomes in culturally diverse settings.

Building on these findings, future research should consider several avenues for extension. First, incorporating additional constructs such as habit, perceived behavioural control, and facilitating conditions may enhance the model’s explanatory and predictive power (Ajzen, 1991; Venkatesh et al., 2003). Second, longitudinal designs would allow for the examination of behavioural evolution over time, capturing sustained usage patterns and post-adoption dynamics. Third, exploring moderating effects—such as age, gender, digital literacy, or institutional support—could reveal subgroup-specific adoption mechanisms and inform targeted interventions.

Moreover, applying the model across diverse technological domains (e.g., e-health, mobile finance, AI-supported learning) and regional contexts would test its robustness and adaptability. Finally, integrating qualitative approaches—such as interviews or ethnographic methods—could provide deeper insights into the socio-cultural and emotional dimensions of technology acceptance, complementing the quantitative findings and enriching theory development.