Students' online learning adaptability and their continuous usage intention across different disciplines

With the rapid development of information technology, online learning has become integral to modern education. Despite widespread adoption in China’s large higher education system, controversies remain around student engagement, satisfaction, and willingness to continue using online platforms. Understanding how to enhance students’ continuance intention is crucial. Prior studies suggest overall online learning adaptability among college students is relatively low and is a critical factor for learning quality and assessment in online environments, but mechanisms linking adaptability to continuance intention are underexplored. Given online technologies’ extensive use across disciplines and observed disparities in outcomes, this study investigates how students’ online learning adaptability affects their continuance intention across 12 disciplines in China and examines the mediating role of satisfaction towards online teaching. Using a random sample of 11,832 students from 334 universities and structural equation modeling, the study analyzes the comprehensive impact of online learning adaptability on continuance intention and potential mediation by satisfaction across disciplines.

The study is grounded in the Technology Acceptance Model (TAM) and Adaptive Learning Systems (ALS) theory. Continuance intention refers to learners’ intention to keep using online learning as a primary method. TAM posits perceived usefulness (PU) and perceived ease of use (PEU) as core determinants of technology acceptance; evidence suggests both positively influence intention to use online learning. ALS emphasizes human–machine interaction adaptability, including learners’ adaptation to technology and systems’ adaptation to learners via learner models. The authors conceptualize online learning adaptability as a multidimensional, two-way process including PEU, PU, and system environment adaptation (SEA, i.e., functional adaptability of software to diverse learner styles). Prior research links satisfaction towards online teaching with course quality and performance, with PU, PEU, and system adaptability influencing satisfaction, which in turn predicts continuance intention, potentially mediating effects of adaptability on intention. Hypotheses: H1a–H1c: PU, PEU, and SEA positively affect continuance intention. H2a–H2c: PU, PEU, and SEA positively affect satisfaction towards online teaching. H3: Satisfaction positively affects continuance intention. H4a–H4c: Satisfaction mediates relationships between PU/PEU/SEA and continuance intention. To address limitations of traditional TAM in complex environments, the authors integrate TAM and ALS into the ASL-TAM framework, modeling adaptability (PU, PEU, SEA) as predictors of satisfaction and continuance intention with satisfaction as mediator.

Data source: Survey data were collected online by a Teacher Development Centre of a public university in mainland China (IRB No. NB-HEC-20200328L) during 2020–2021. The survey was distributed via academic affairs offices, included two lie-detection questions, and limited one submission per account (latest response overwrote prior). From 256,504 student responses across 334 universities, a stratified random sample of 1,000 students per each of 12 disciplines (12,000 total) was drawn. Data cleaning removed 162 cases based on lie checks, atypical completion times (<5 or >20 min, 3σ rule), implausible ages (<15 or >25, 3σ), invalid school names, and non-users of online learning, yielding 11,832 valid cases (986 per discipline). Disciplines: philosophy, economics, law, education, literature, history, natural science, engineering, agriculture, medicine, management, arts. Instrumentation: A 33-item questionnaire measured five constructs: PU (11 items), PEU (3), SEA (10), satisfaction towards online teaching (ST; 7), and continuance intention (CIN; 2). Overall reliability: Cronbach’s alpha 0.924; KMO 0.937; Bartlett’s test p<0.001. Exploratory factor analysis extracted subdimensions: PU—teaching resources (PU_TR), classroom teaching (PU_CT), teaching evaluation (PU_TE); PEU—technical training (PEU_TT), pedagogical training (PEU_PT), proficiency levels (PEU_PL); SEA—technical service (SEA_TSER), teaching support (SEA_TSUP), policy support (SEA_PS); ST—teaching effectiveness (ST_TE), teaching experience (ST_TEXP), learning outcomes (ST_LO); CIN—online mode (CIN_ON), blended mode (CIN_BL). Scales were adapted from Davis (1993) for PU/PEU, Igbaria (1990) for SEA, Ajzen & Fishbein (1980) for ST, and Chen & Tseng (2012) for CIN. Analysis: Descriptive statistics and reliability analyses were conducted in SPSS 25.0. Common method bias was evaluated via Harman’s single-factor test. One-way ANOVA tested disciplinary differences in observed variables. Structural equation modeling (SEM) using AMOS 24.0 estimated the ASL-TAM for each discipline with maximum likelihood. Model modification indices guided adding a residual covariance (example: e2<->e3) to improve fit. Goodness-of-fit indices were examined (GFI, AGFI, RMR, RMSEA, NFI, CFI, TLI, PNFI, PGFI, CMIN/DF).

• Reliability: Across 12 disciplines, Cronbach’s alpha for observed-variable composites was ≥0.90, indicating high reliability. Data were suitable for SEM. - Common method bias: Harman’s single-factor test showed the first factor explained 29.21% (<40%), suggesting CMB was not a major concern. - One-way ANOVA across disciplines: On average across disciplines, PEU (mean 3.62) > SEA (3.60) > PU (3.47). ST (3.47) slightly exceeded CIN (3.44). PU was the main weak link of adaptability, particularly teaching evaluation (PU_TE) with the lowest PU subscore (3.26). Within PEU, technical training (PEU_TT) was lowest (3.58); within SEA, technical service (SEA_TSER) was lowest (3.53); within ST, teaching effectiveness (ST_TE) was lowest (3.28). Philosophy showed the lowest overall evaluation (approx. 3.41). All 14 observed variables exhibited significant between-discipline differences (p<0.001). - Correlation analysis: PU, PEU, and SEA were significantly and positively correlated with each other and with ST and CIN (p<0.001). ST and CIN were also significantly positively correlated (p<0.001). - Model fit: The ASL-TAM achieved acceptable fit in all 12 disciplines. Representative ranges: GFI ~0.936–0.953; AGFI ~0.904–0.924; RMSEA ~0.066–0.079; NFI ~0.952–0.971; CFI ~0.958–0.976; TLI ~0.945–0.967; PNFI and PGFI >0.5; CMIN/DF ~5.10–6.56. - Structural paths: In all 12 models, ST → CIN was significant and positive, supporting H3. Direct effects of PU, PEU, and SEA on CIN were significant and positive across disciplines, supporting H1a–H1c. PU, PEU, and SEA each significantly and positively predicted ST, supporting H2a–H2c. Mediation analyses indicated ST partially mediated the relationships of PU, PEU, and SEA with CIN, supporting H4a–H4c. - Disciplinary differences: While the ASL-TAM fit all disciplines, path coefficients varied significantly by discipline. Compared to six STEM disciplines, six humanities disciplines exhibited more pronounced internal differences, especially in PU-related observed variables and in paths involving ST. Overall, humanities showed greater variability and lower evaluations in perceived usefulness than STEM.

Findings demonstrate that online learning adaptability robustly drives students’ continuance intention, both directly (via PU, PEU, SEA) and indirectly through satisfaction towards online teaching. This indicates that both student-side perceived adaptation (ease and usefulness) and system-side adaptability to learner needs are critical. The applicability of the ASL-TAM across 12 disciplines validates the TAM core structure for online learning acceptance and extends it by incorporating system environment adaptation from ALS theory. The partial mediation by satisfaction clarifies the mechanism: adaptability enhances satisfaction with online teaching quality and experience, which in turn strengthens intentions to persist with online learning. Disciplinary heterogeneity matters: humanities fields show larger disparities, particularly in perceived usefulness and satisfaction pathways, suggesting pedagogical and technological designs must be tailored to disciplinary characteristics. Enhancing human-computer interaction intelligence and teacher–student adaptive teaching strategies can elevate satisfaction and continuance intention.

This study integrates TAM and ALS into the ASL-TAM framework and empirically validates it across 12 disciplines using a large, multi-institution sample. Main contributions: (1) Identifies weak links in online learning adaptability in China—lower perceived usefulness (notably teaching evaluation), limited technical service, and insufficient technical training—coupled with generally modest continuance intention. (2) Demonstrates that satisfaction towards online teaching is a central antecedent partially mediating the effects of PU, PEU, and SEA on continuance intention. (3) Reveals significant disciplinary differences between humanities and STEM in perceived usefulness and satisfaction pathways, underscoring the need for discipline-sensitive online learning designs. Implications include building widely applicable platforms while accommodating discipline-specific pedagogy, enhancing technical support and human-centered interactivity (especially for humanities), improving performance evaluation and industry–education integration (especially for STEM), and accelerating adaptive learning environments responsive to diverse learner profiles. Future research should refine adaptive systems and instructional models tailored to disciplinary needs and further probe peer effects and cultural factors shaping perceived usefulness and satisfaction.

Two key limitations: (1) The large number of participating universities may differ in disciplinary settings and standards, potentially introducing heterogeneity not fully controlled in analysis. (2) Participant location, university level, and academic year—factors known to affect satisfaction—could not be fully controlled or eliminated. Future work should address these sources of variation and explore additional covariates.