A uses and gratifications approach to examining users’ continuance intention towards smart mobile learning

The paper examines why users continue to use smart mobile learning (SML) systems that integrate AI technologies (e.g., NLP, speech recognition, machine learning, recommendation systems) to deliver personalised language learning. Using Liulishuo as the focal SML app, the study addresses gaps in prior uses and gratifications (U&G) research, which largely considered broad gratification categories without specifying constructs for the intelligent dimension of SML. The research questions center on which specific gratifications drive continuance intention in SML and the relative importance of these gratifications. The study proposes introducing perceived intelligence as a novel technology gratification construct and aims to empirically determine a hierarchy of gratification influences on continuance intention using PLS-SEM on survey data from Liulishuo users.

The study adopts Uses and Gratifications (U&G) theory (Katz et al., 1973) to explain post-acceptance technology use by focusing on user needs, expectations, and gratifications derived from media/system usage. The literature recognises U&G’s applicability to computer-mediated media (internet, smartphones, SNS, online video, online games, instant messaging, and mobile learning) and highlights its strength in capturing emotional and social factors often omitted by technology diffusion models like TAM. Satisfaction in U&G encompasses both need fulfilment and pleasurable elements. Prior typologies in U&G for new media identify gratifications including technology/process, content, social, utilitarian/personal integrative, hedonic/affective. Building on this, the study derives five gratification types—technology, hedonic, social, utilitarian, and content—and specifies constructs for each (perceived intelligence, convenience; perceived enjoyment, concentration; status; achievement; education), grounded in incentive theory of motivation, flow theory, diffusion of innovations, self-determination theory, and learning theory. Table-based prior work shows analogous gratifications across media, underscoring the relevance of these categories to SML.

Design: Quantitative survey study targeting users of the Liulishuo SML app. Measurement development: Scales adapted from prior literature to the SML context, assessed on seven-point Likert items. Constructs measured: continuance intention (Bhattacherjee, 2001), perceived intelligence (Bartneck et al., 2009), convenience (Ko et al., 2005), perceived enjoyment (Moon and Kim, 2001; van der Heijden, 2003), concentration (Koufaris, 2002), status (Venkatesh and Brown, 2001), achievement (Wu et al., 2010), education (Stafford et al., 2004). Translation: Items originally in English underwent forward translation, back translation, comparison and revision, and pilot testing to ensure linguistic/conceptual equivalence. Pilot: 107 pilot responses (excluded from main analysis) provided preliminary reliability/validity evidence. Sampling and data collection: Conditional random sampling via Baidu sample service (China); only prior Liulishuo users proceeded past the screening question. The app skews to younger adults; the sample reflects this user base. Data collection spanned ~15 days, yielding 495 valid responses. Descriptives: 51.9% female; 95.8% with college degree or above; 98% aged 19–45. Analysis approach: Partial least-squares structural equation modelling (PLS-SEM) using SmartPLS 3.0. Measurement model: Internal consistency and convergent validity supported (CR > 0.80; Cronbach’s alpha > 0.70; AVE > 0.50; all item loadings > 0.70 and higher on their own constructs). Discriminant validity: Fornell–Larcker criteria satisfied; cross-loadings supported discriminant validity; multicollinearity not a concern (squared correlations < 0.80). Structural model: Goodness of Fit (GoF) = 0.64 (above large benchmark 0.36); R² for continuance intention = 0.631. Bootstrapping: 5000 resamples to assess significance of paths with bias-corrected 95% CIs.

• The model explains 63.1% of the variance in continuance intention (R² = 0.631). Goodness of fit GoF = 0.64. - Significant positive predictors of continuance intention (standardised path coefficients, significance): • Perceived intelligence (technology gratification): β = 0.171, p < 0.01 (strongest effect). • Education (content gratification): β = 0.150, p < 0.05. • Status (social gratification): β = 0.147, p < 0.05. • Achievement (utilitarian gratification): β = 0.131, p < 0.05. • Convenience (technology gratification): β = 0.125, p < 0.05. • Perceived enjoyment (hedonic gratification): β = 0.120, p < 0.05. - Non-significant predictor: Concentration (hedonic gratification) was not significantly related to continuance intention (H4 not supported). - Hierarchy of influence indicated by effect sizes: intelligence (highest), followed by education, status, achievement, convenience, and perceived enjoyment. - Empirical evidence introduces and validates perceived intelligence as a technology gratification construct uniquely salient to AI-enabled SML.

The findings address the core research question of which gratifications drive continued SML use and their relative importance. Users’ expectations about intelligent system capabilities dominate continuance intention, indicating that in AI-enabled learning contexts, technology gratification—particularly perceived intelligence—outweighs intrinsic hedonic states. Content gratification via educational value, as well as social status and utilitarian achievement, substantially reinforce continued use, aligning with the multifaceted motivations captured by U&G theory. Convenience further supports continuance, reflecting the mobility and ease central to SML. Perceived enjoyment contributes modestly, whereas concentration does not, suggesting that flow-related focused immersion may be less critical than perceived intelligent guidance and tangible learning value in this context. These results underscore that SML users prioritise smart, personalised tutoring features and educational outcomes over purely affective immersion, extending U&G understanding to AI-driven learning systems.

This study extends U&G theory to the SML context by specifying and empirically validating gratification constructs that shape users’ continuance intention. It pioneers perceived intelligence as a technology gratification construct and shows it is the most potent driver of continued use. Content (education), social (status), utilitarian (achievement), technology (convenience), and hedonic (perceived enjoyment) gratifications also positively influence continuance intention, while concentration does not. The work clarifies the hierarchy of gratifications for AI-enabled mobile learning and highlights that user expectations regarding intelligent functionality outweigh intrinsic hedonic needs. Future research can further develop U&G theory in smart communication technologies and examine generalisability across platforms and cultures, exploring additional intelligent features and interactions that may shape post-acceptance behaviour.

• Geographic and cultural scope: Data were collected in China, potentially introducing cultural bias; findings from a developing-country context may not generalise to developed economies. - Demographic skew: The sample reflects Liulishuo’s primarily young adult user base, which may limit generalisability to other age groups. - Platform specificity: Results are centred on Liulishuo and may vary with other SML platforms. Future studies should include cross-country comparisons (including Western developed countries) and multiple platforms to enhance generalisability.