The unified theory of acceptance and use of DingTalk for educational purposes in China: an extended structural equation model

The paper addresses the rapid expansion of online learning platforms and the need to understand factors influencing their acceptance and use, focusing on DingTalk, a widely used synchronous online communication platform in China. Although prior research has applied statistical models to various educational technologies, little rigorous, UTAUT-based evidence exists for DingTalk, especially in the post-pandemic era. The study’s purpose is to empirically investigate acceptance and use of DingTalk for educational purposes by extending the UTAUT framework with additional constructs and testing 17 hypotheses. The extended model evaluates how classic UTAUT variables—effort expectancy, performance expectancy, social influence, and facilitating conditions—together with new variables—attitude toward behavior, self-efficacy, and received feedback—affect behavioral intention and actual use. The importance lies in informing E-learning platform integration, improving user experience, and enhancing learning outcomes in China’s educational context.

The study adopts UTAUT (Venkatesh et al., 2003) as the core framework due to its higher explanatory power and notes its extension (UTAUT2) with hedonic motivation, price value, and habit. Prior educational applications of UTAUT show varied effects of performance expectancy (PE), effort expectancy (EE), social influence (SI), and facilitating conditions (FC) on behavioral intention (BI) and use behavior (UB), with demographic moderators (gender, age, experience, voluntariness) influencing relationships. Mixed findings in education motivate re-examination. The authors define traditional constructs for the DingTalk context (EE, PE, SI, FC, BI, UB) and review evidence on their roles, noting inconsistencies especially for FC and SI across studies. To extend UTAUT, the authors add: Attitude toward behavior (ATB), frequently linked to BI in MALL, gamified vocabulary, and digital reading contexts and posited as a mediator; Self-efficacy (SE), users’ belief in their capability to use DingTalk, previously impacting BI and perceived ease/usefulness; Received feedback (RF), the extent of feedback during online lessons, supported by literature on peer/teacher and automated feedback improving motivation and outcomes. The study proposes 17 hypotheses (H1–H17) linking these constructs to ATB, BI, and UB, including mediating roles of ATB and BI.

Design: Cross-sectional survey with structural equation modeling (SEM). Participants: 930 initial responses; 856 valid (92.0% valid rate) from teachers and students in China who had used DingTalk for online courses. Inclusion criteria: stakeholders (teachers/students), consented, and had conducted/attended DingTalk online courses. Exclusion: non-teachers/students, unwilling, unfamiliar with online/synchronous tools, straight-lining responses, completion time <40 seconds. Demographics included gender (21.03% male, 78.97% female), age distribution primarily 22–25 (61.92%), occupations (undergraduate 43.34%, masters/doctor 37.15%, teachers 16.82%), and experience with DingTalk (>2 years 40.07%). Instruments: A 46-item questionnaire (5 demographic, remaining on 5-point Likert scale) measuring UTAUT core and extended constructs: EE, PE, SI, FC, BI (from Venkatesh & Davis, 2000; Venkatesh et al., 2003); SE (Schwarzer & Jerusalem, 1995; Wang et al., 2022); ATB (Zhang & Yu, 2022); RF (Akkuzu & Uyulgan, 2014); UB (Waheed et al., 2016). Back-translation used; English and Chinese versions provided; ethical approval and consent obtained. Procedure: Questionnaire designed in Questionnaire Star, expert-reviewed, distributed via online/offline channels using convenience sampling; incentives provided. Data were downloaded, cleaned in SPSS 23.0, and analyzed in Amos 24.0 and SPSS 23.0. Analysis: Two-step SEM (Anderson & Gerbing, 1988). Measurement model evaluation via CFA: items removed by modification indices for better fit (deleted: PE1, SI2, SI4, SE2, PI items 1–4, BI4, UB3; retained FC4 for fit). Reliability and validity assessed: standardized loadings, composite reliability (CR), average variance extracted (AVE), Cronbach’s alpha; discriminant validity via Fornell-Larcker. Model fit indices reported (CMIN/DF, CFI, GFI, NFI, TLI, SRMR, RMSEA). Structural model evaluation: R² for ATB, BI, UB; path coefficients with significance tests; mediation tested via bootstrapping (5,000 samples, 95% bias-corrected CI).

- Measurement model reliability/validity: Construct reliability and validity met accepted thresholds (typical CR > 0.7; AVE mostly > 0.4; Cronbach’s alpha adequate), and discriminant validity satisfied (Fornell-Larcker). Model fit was satisfactory to excellent: CMIN/DF=2.432; CFI=0.959; GFI=0.929; NFI=0.932; TLI=0.952; SRMR=0.038; RMSEA=0.041. - Explained variance (R²): ATB=73.5%; BI=60.9%; UB=46.0% (moderate to strong explanatory power). - Significant predictors of Attitude toward behavior (ATB): • EE → ATB: β=−0.104, p<0.05 (negative) • PE → ATB: β=0.451, p<0.001 (positive) • FC → ATB: β=0.111, p<0.05 (positive) • SE → ATB: β=0.189, p<0.001 (positive) • RF → ATB: β=0.377, p<0.001 (positive) • SI → ATB: ns (β=−0.029, p=0.572) - Significant predictors of Behavioral intention (BI): • SI → BI: β=0.232, p<0.001 (positive) • FC → BI: β=0.111, p<0.05 (positive) • ATB → BI: β=0.301, p<0.001 (positive) • RF → BI: β=0.189, p<0.001 (positive) • EE, PE, SE → BI: not significant - Significant predictors of Use behavior (UB): • BI → UB: β=0.554, p<0.001 (positive) • FC → UB: β=0.28, p<0.001 (positive) • RF → UB: β=−0.163, p<0.05 (negative) • SI → UB: not significant - Supported hypotheses: H3, H5, H6, H7, H8, H9, H11, H12, H13, H15, H16, H17 (12 of 17). Rejected: H1, H2, H4, H10, H14. - Mediation (bootstrap 5,000; 95% CI): • PE → ATB → BI: B=0.159, 95% CI [0.064, 0.306] • RF → ATB → BI: B=0.142, 95% CI [0.055, 0.275] • FC → ATB → BI: B=0.037, 95% CI [0.003, 0.104] • SI → BI → UB: B=0.19, 95% CI [0.069, 0.347] • FC → ATB → BI → UB: B=0.023, 95% CI [0.002, 0.067] - Overall: The extended UTAUT model effectively explains intention and use of DingTalk; ATB and BI act as important mediators.

The findings clarify determinants of DingTalk acceptance and use in China. Contrary to many UTAUT-based studies, EE and PE did not directly predict BI, possibly due to mandatory use policies and increasing digital literacy diminishing perceived effort/utility relevance. EE slightly reduced ATB, suggesting that perceived effort erodes favorable attitudes. PE strongly improved ATB, indicating that perceived usefulness shapes positive attitudes even if it does not directly drive intention under mandated contexts. SI significantly increased BI but not ATB, implying social/organizational pressures can drive intention without altering personal attitudes, and SI’s effect on UB was indirect via BI. FC consistently supported ATB, BI, and UB, highlighting the importance of resources, technical support, and infrastructure (e.g., network quality) for synchronous platforms. SE enhanced ATB but not BI, suggesting confidence boosts positive perceptions but may not translate to intention when external constraints dominate. RF emerged as a novel, impactful construct: it improved ATB and BI but negatively affected UB, potentially reflecting complexities of feedback processes or constraints (e.g., bandwidth) in synchronous settings; nonetheless, RF’s positive influence on attitudinal and intentional outcomes underscores the centrality of interaction quality. Mediation results emphasize ATB and BI as mechanisms linking antecedents (PE, FC, RF, SI) to BI and UB, respectively. Collectively, the results address the research question by demonstrating that an extended UTAUT with ATB, SE, and RF explains DingTalk acceptance and use better in China’s educational context and indicates where platform design and institutional support can improve adoption and sustained use.

This study extends UTAUT by incorporating attitude toward behavior, self-efficacy, and received feedback to explain acceptance and use of DingTalk for education in China. The model shows strong/satisfactory fit and explains substantial variance in ATB (73.5%), BI (60.9%), and UB (46.0%). Key contributions include identifying determinants of ATB (EE−, PE+, FC+, SE+, RF+), BI (SI+, FC+, ATB+, RF+), and UB (BI+, FC+, RF−), and demonstrating mediating roles of ATB and BI along several paths. The work offers theoretical insights into technology acceptance under policy-driven use and highlights the importance of feedback and facilitating conditions in synchronous learning platforms. Future research should extend acceptance models with additional psychological/behavioral constructs and moderators, examine policy and organizational factors, and explore feedback-related mechanisms (including AI chatbots) to enhance interactivity and outcomes in E-learning.

- Cross-sectional data collected over four months during easing of pandemic restrictions; platform changes in this period may have influenced responses. - Potential response bias due to participants’ emotions and pandemic-related impacts affecting objectivity. - Heterogeneous participant groups (students and teachers) may vary across dimensions; failure to detect moderating effects reduced the extended model’s reliability for subgroup differences.