Abstract
Short-video platforms increasingly support informal learning, yet the determinants of students’ voluntary educational use remain under-specified. This study models Chinese undergraduates’ educational use of Douyin (TikTok’s China-market version) by integrating the Technology Acceptance Model (TAM) and Social Cognitive Theory (SCT), specifying two complementary routes: a platform-facing Adoption route—operationalized as a reflective second-order construct indicated by perceived usefulness, perceived ease of use, and social influence—and a person-centered Personal Factor route. A cross-sectional survey of N = 486 undergraduates from a public university in Henan, China, was analyzed using confirmatory factor analysis and structural equation modeling. Results show that both routes significantly and positively predict educational usage (Adoption: β = .54, p < .001; Personal Factor: β = .41, p < .001). These findings indicate that repurposing an entertainment-first platform for learning depends jointly on platform appraisals and learner dispositions. The study contributes an integrated TAM–SCT specification of voluntary educational usage on short-video platforms and offers practical implications for instructional design and platform features/policies.
Plain Language Summary
Once viewed purely as an entertainment app, Douyin—the Chinese version of TikTok—is rapidly emerging as an informal learning platform among university students. This study sheds light on a timely and underexplored phenomenon: why students are choosing to use Douyin voluntarily for educational purposes. With over 250 million users engaging with knowledge-related content, Douyin is quietly reshaping how young people in China access academic support in their daily digital routines. To explore this shift, the researchers surveyed 486 undergraduate students at a public university in Henan Province. Using a statistical model that combines two well-established theories—Technology Acceptance Model (TAM) and Social Cognitive Theory (SCT)—the study examined how students’ perceptions of Douyin’s usefulness, ease of use, and social influence, along with their own confidence and learning goals, predict educational engagement on the platform. The findings reveal a powerful combination of factors driving this behavior. Students are more likely to use Douyin for learning when they believe it is easy to use, socially supported, and academically helpful. More importantly, students who feel confident navigating the platform and expect meaningful learning outcomes are even more motivated to use it as a self-directed educational tool. These insights show that Douyin is not merely a distraction but a flexible learning space shaped by both technology and personal agency. This research offers fresh perspectives on how social media can be reimagined for education. It provides practical guidance for educators, content creators, and platform designers who wish to reach students where they already are—on their phones. By understanding what motivates students to engage with platforms like Douyin for learning, stakeholders can harness this momentum to make education more accessible, relevant, and aligned with the habits of a digital-native generation.
Introductıon
With advancements in broadband and mobile network technology, new media platforms have become deeply integrated into university students’ daily lives. Among these platforms, Douyin (the Chinese counterpart to TikTok) has rapidly gained prominence among undergraduates, who increasingly report using it not only for leisure but also for study-related purposes. Recent statistics indicate remarkable global growth of TikTok, with monthly active users increasing from approximately 507 million in December 2019 to over 1 billion by September 2021 (Kemp, 2023; Radin & Light, 2022). Notably, the largest user group comprises young adults aged 18 to 24, accounting for 38.5% of total users. This underscores the critical role TikTok plays in the daily routines of contemporary undergraduates (Sun, 2021). Yet the very features that optimize entertainment—algorithmic curation and habitual short-form consumptionalgorithmic curation and habitual short-form consumption— do not automatically explain students’ voluntary shift toward education-oriented use, raising the question of when and why such repurposing occurs.
A similar phenomenon is evident within China, where Douyin serves a parallel role tailored specifically to domestic users. According to the 2023 report by the China Internet Network Information Center (CNNIC), China has 1.067 billion Internet users, of whom 1.012 billion are short-video consumers, representing 94.8% of total Internet users. College students, as the most active demographic, spend over 40% of their smartphone usage time engaging with short videos, predominantly on platforms like Douyin (Feng & Fu, 2022). Although TikTok and Douyin share core features, such as short video content creation and personalized recommendation algorithms, Douyin uniquely incorporates culturally specific elements aligned with local norms, exemplified by initiatives such as its “Positive Energy” content section (Kaye et al., 2021). From a practical standpoint, even small shifts in this entrenched usage can influence course design, academic support, and platform governance in higher education.
The educational affordances of Douyin have become increasingly recognized by educators and students alike. Compared to traditional online educational platforms, social media platforms like Douyin offer diverse educational possibilities and an inherently familiar interface that facilitates seamless adoption in educational contexts (Joosten, 2012). Undergraduate students, characterized as digital natives immersed in digital technology from an early age (Turner, 2015; Twenge, 2017), are adept at leveraging such platforms for autonomous and creative learning (Galanek et al., 2018; Arteaga Sánchez et al., 2014). They frequently utilize Douyin independently, actively seeking educational content to address academic challenges (Oriji & Torunarigha, 2019). Nevertheless, recognizing these affordances is analytically distinct from explaining sustained voluntary educational usage; the latter requires specifying the determinants that move students beyond casual consumption toward education-oriented engagement.
The prevalence of educational usage on Douyin is supported by significant empirical evidence. For instance, knowledge-related content on Douyin grew by 35.4% from January to October 2022, with over 250 million users specifically engaging with educational content—a 44.1% year-on-year increase (Oceanengine, 2023a). Remarkably, more than 95 million Douyin users participated in live university lectures in 2022, equivalent to approximately one in every fourteen people in China attending virtual college courses via the platform (Oceanengine, 2023b). Such evidence demonstrates the platform's burgeoning role as a supplemental educational resource for undergraduates, highlighting the need to explore the underlying factors driving this phenomenon.
However, despite this promising trend, research exploring the determinants of students’ voluntary educational usage of Douyin remains limited. Existing studies on Douyin’s educational applications have predominantly focused on teachers’ instructional methods rather than students' self-directed learning behaviors (Gao, 2022; Yang, 2022; Zeng, 2021). Consequently, there is a crucial research gap regarding undergraduate students' voluntary adoption and utilization of Douyin for educational purposes. Addressing this gap could provide valuable insights into optimizing the educational potential of short-video platforms, aligning educational strategies more closely with contemporary students' digital lifestyles.
At the theoretical level, single-framework accounts are partial: acceptance-oriented perspectives illuminate platform-facing beliefs, while person-centered perspectives illuminate learners’ own determinants; few studies articulate both within one coherent specification tailored to education-oriented, rather than general, use. At the measurement level, prior work rarely models platform acceptance as a reflective second-order construct (Adoption indicated by perceived usefulness, perceived ease of use, and social influence) alongside a distinct first-order Personal Factor; clarifying these parallel routes helps avoid conflating entertainment use with education-oriented behaviors. Practically, a validated account can inform how courses scaffold platform tasks, how student support nudges purposeful use, and how platform policies align with higher education goals.
Given this context, this study aims to empirically investigate the determinants of undergraduates’ voluntary educational usage of Douyin by integrating constructs from the Technology Acceptance Model (TAM) and Social Cognitive Theory (SCT). Specifically, the study seeks to examine how undergraduates’ adoption of Douyin—reflected by perceived usefulness, perceived ease of use, and social influence—and their personal factors influence their voluntary educational engagement with the platform. In this study, “educational usage” denotes self-initiated, goal-directed engagement oriented to academic tasks or skills (rather than instructor-mandated activity). This investigation not only addresses the identified research gap but also provides educators and policymakers with practical guidance for harnessing the educational potential of short-video platforms in higher education.
Lıterature Revıew
TikTok and Douyin: A Rapidly Rising Digital Phenomenon
TikTok, launched in 2016 by the Chinese company ByteDance, has rapidly grown into a globally embraced platform known for short-form video content sharing (Liu, 2023). Characterized by innovative features that enable video creation with music and diverse audiovisual enhancements, TikTok has democratized content production, allowing users across different backgrounds to creatively express themselves and achieve widespread visibility (Yu, 2019; Fiallos et al., 2021). As of 2023, TikTok had amassed over 1.7 billion global users, with approximately 1 billion monthly active users, with particular dominance among ages 18 to 24 (Geyser, 2023). Its intuitive interface, incorporating functionalities like content browsing, social interaction, video creation, messaging, and live streaming, further facilitates broad user engagement and content diversity (Al-Khasawneh et al., 2022; Jin et al., 2022).
In China, TikTok’s domestic counterpart is Douyin. Both platforms share core functionalities but cater to distinct markets, and are distinct sociocultural contexts and regulatory environments (Kaye et al., 2021). Douyin's algorithmic content personalization strategy has significantly contributed to its massive user base, with approximately 997 million active users by 2022. College-aged users, particularly ages 18 to 23, represent nearly half of Douyin’s users, actively utilizing the platform not only for entertainment but increasingly for educational purposes (Oceanengine, 2022).
Educational Potential of Douyin: Engagement, Retention, and Digital Literacy
The educational affordances of TikTok and Douyin are grounded in their ability to condense complex topics into concise, engaging content suitable for digital-native learners (Fuad et al., 2023). Douyin’s growing popularity in education stems primarily from the platform's capacity for enhanced student engagement through interactive content creation, hashtags, challenges, and remixing capabilities (Rajan & Ismail, 2022; Wang et al., 2024). Such interactive features significantly increase the potential for self-directed learning, allowing undergraduates to explore topics beyond formal educational boundaries (Assad, 2024; Xu & He, 2023).
Moreover, the short-form video format of Douyin is particularly conducive to knowledge retention. According to Kolber (2024), concise, repeated exposure to targeted educational content aligns effectively with cognitive learning theories, enabling easier assimilation and reinforcement of knowledge. The platform’s personalized recommendation algorithms further enhance this retention by continually presenting users with related educational material tailored to their interests (Low et al., 2023).
In addition, Douyin fosters digital literacy, crucial for contemporary undergraduates. Active interaction with algorithms and multimodal content creation promote critical digital skills, including algorithmic literacy and media competency, increasingly vital for academic and professional success (Xu & He, 2023; Yu & Ding, 2022). Nonetheless, concerns persist regarding Douyin's capability to foster deep critical thinking due to inherent time constraints and an emphasis on entertainment. Yet, active student participation in content creation can partially address these limitations, encouraging critical synthesis and analysis (Vermeire et al., 2024). While these affordances are shared with entertainment use, outcomes diverge by orientation: in this study, education-oriented engagement is defined as intentional links to course or skill objectives rather than leisure consumption.
Empirical Insights into Educational Usage of Douyin
The educational usage of TikTok and Douyin has been empirically supported by multiple studies highlighting user-driven educational engagement. The LearnOnTikTok initiative, launched during the COVID-19 pandemic, exemplifies the platform's educational potential, accumulating over 72 billion views in its first year alone (Fiallos et al., 2021; Hutchinson, 2020). Research by Hayes et al. (2020) and Rajan and Ismail (2022) has demonstrated the platform’s capacity to contextualize and simplify complex educational topics, promoting active participation and improved comprehension across disciplines.
However, researchers have raised concerns about the brevity and entertainment-driven nature of Douyin's content, potentially limiting comprehensive knowledge transfer (Nguyen & Diederich, 2023). Algorithmic biases prioritizing engagement over educational quality further complicate its pedagogical use (Wang et al., 2023). Thus, leveraging Douyin's educational potential effectively requires balancing user engagement with academic rigor, especially within informal learning contexts driven by students’ voluntary use. Recent acceptance-oriented evidence in higher education also indicates that inhibitors such as perceived cyber risk can depress usefulness beliefs and adoption intentions, suggesting boundary conditions for education-oriented use on entertainment-first platforms (Al-Adwan et al., 2023).
Technology Acceptance Model in Social Media and Douyin Research
Originating from Davis’s (1989) work, the Technology Acceptance Model (TAM) remains a foundational framework for investigating technology adoption, emphasizing perceived usefulness (PU) and perceived ease of use (PEOU) as determinants of user acceptance. TAM has consistently demonstrated robust predictive capabilities across diverse educational technologies, including social media platforms such as Facebook, Twitter, and TikTok (Al-Khasawneh et al., 2022; Chipps et al., 2015; Farahat, 2012).
Specific research applying TAM to Douyin confirms the model’s validity. Studies by Xiang (2020) and Liu (2019) highlight perceived usefulness, perceived ease of use, and notably, social influence as critical predictors of Douyin adoption among university students. Social influence, conceptualized as the extent to which individuals perceive that significant others encourage platform adoption, has emerged as particularly influential in shaping user intentions and behaviors within social media contexts (Goli & Khan, 2022; Jia et al., 2023).
Complementing this line of work, recent extensions of TAM in higher education show that perceived usefulness tends to dominate the direct pathway to adoption-related outcomes, whereas perceived ease of use often operates indirectly via perceived usefulness or related appraisals; at the same time, perceived cyber risk functions as a salient inhibitor that undermines usefulness beliefs and intentions (Al-Adwan et al., 2023). This pattern strengthens the platform-facing route specified in the present study, where Adoption is operationalized as a reflective second-order construct indicated by PU, PEOU, and social influence.
Social Cognitive Theory in Social Media and Douyin Research
Social Cognitive Theory (SCT), introduced by Bandura (1988), offers a complementary perspective by emphasizing personal, environmental, and behavioral determinants interacting reciprocally to shape user behavior. SCT suggests that individuals actively engage with media, acquiring knowledge and behaviors through observational learning, self-efficacy, and outcome expectations (Bandura, 2001; Tseng, 2020).
Applied to social media research, SCT has elucidated user-generated content dynamics, persistent platform engagement, and online learning behaviors (Jin et al., 2022; Zhou et al., 2020). Within the specific context of Douyin, SCT research underscores personal factors—particularly self-efficacy, outcome expectations, and attitudes—as significant determinants influencing educational usage. Studies reveal that students with higher self-efficacy and positive attitudes toward technology demonstrate greater voluntary educational engagement with platforms such as Douyin (Deng & Zhang, 2023; Song, 2022).
Converging evidence from recent higher education work further suggests that acceptance-related factors and person-centered determinants jointly shape technology-supported learning: self-determination mechanisms can mediate between usefulness/ease of use and acceptance, and both acceptance and agentic dispositions contribute to engagement and downstream outcomes (Ma et al., 2025). In the present study, Personal Factor is modeled as a first-order construct measured by PF items, capturing students’ agentic dispositions relevant to education-oriented use without subdividing it into separate components.
Integration of TAM and SCT
Despite extensive research confirming the educational potential of Douyin and applying theoretical frameworks such as TAM and SCT to understand technology adoption, empirical studies explicitly integrating both theories remain limited. Prior research has individually emphasized perceived usefulness, perceived ease of use, and social influence (Al-Khasawneh et al., 2022; Liu, 2019; Xiang, 2020), and personal factor (Deng & Zhang, 2023; Zhou et al., 2020). However, few studies have combined these factors within a unified structural model to explore undergraduate students’ voluntary educational use of Douyin. Addressing this gap, this study integrates TAM and SCT constructs into a comprehensive research model to empirically investigate critical determinants shaping students’ educational engagement on Douyin.
Addressing this gap, we integrate TAM’s platform-facing beliefs—summarized by the reflective second-order Adoption— with SCT’s person-centered determinants —captured by the first-order Personal Factor— to examine two concurrent routes to Educational Usage in a single specification. This framing aligns with recent higher education evidence that usefulness-centric acceptance mechanisms and learner agency jointly underpin technology-supported learning behaviors, while also acknowledging potential inhibitors identified in acceptance extensions (Al-Adwan et al., 2023; Ma et al., 2025). The integrated lens thereby provides a clearer theoretical basis for the hypotheses developed in the next section.
Research Model and Hypotheses
This study integrates the Technology Acceptance Model (TAM) and Social Cognitive Theory (SCT) to examine determinants of Chinese undergraduates' voluntary educational usage of Douyin. As depicted in Figure 1, the research model specifies two latent predictors of Educational Usage: Adoption, modeled as a reflective second-order construct indicated by Perceived Usefulness (PU), Perceived Ease of Use (PEOU), and Social Influence (SI); and Personal Factor, modeled as a first-order construct grounded in SCT and measured by PF items. Both Adoption and Personal Factor are theorized to directly predict Educational Usage.
Research model.
Adoption (Reflective Second-order Construct)
In technology acceptance research, adoption generally refers to users’ acceptance and incorporation of a technological system into their routine activities (Guner & Acarturk, 2020). TAM and its extensions highlight perceived usefulness, perceived ease of use, and social influence as critical determinants of users’ intentions to adopt technologies such as social media platforms (Davis, 1989; Venkatesh et al., 2003).
Perceived Usefulness (PU) reflects users' beliefs that adopting a particular technology will enhance their performance or help achieve specific goals (Davis, 1989). Empirical studies have consistently identified perceived usefulness as a critical dimension reflecting users' adoption of social media platforms such as Douyin (Al-Khasawneh et al., 2022; Elkaseh et al., 2016). Recent higher-education evidence further indicates that usefulness tends to dominate the direct pathway to adoption-related outcomes on emerging learning platforms (Al-Adwan et al., 2023), which aligns with our platform-facing route.
Perceived Ease of Use (PEOU) denotes users' perceptions regarding the effortlessness associated with using a given technology (Davis, 1989). Douyin's intuitive design and user-friendly interface are critical in encouraging sustained voluntary educational use (Al-Azawei, 2018; Rauniar et al., 2014). At the same time, recent extensions of TAM suggest that ease of use frequently exerts its impact indirectly via usefulness or related appraisals in higher education settings (Al-Adwan et al., 2023). Treating PEOU as a first-order indicator of Adoption is therefore consistent with both classic TAM and recent findings.
Social Influence (SI) refers to users’ perceptions that important others (peers, friends, influencers) affect their adoption behavior (Venkatesh & Brown, 2001). Undergraduate students typically have robust social connections, and their behaviors are likely influenced by peers advocating Douyin as an educational tool (Goli & Khan, 2022; Jia et al., 2023). Therefore, social influence is a critical reflective dimension contributing to the adoption of Douyin.
In summary, Adoption is conceptualized in this study as a reflective second-order construct, manifested by perceived usefulness, perceived ease of use, and social influence.
Relationship between Adoption and Educational Usage
Previous research strongly supports the direct relationship between users’ adoption of technological platforms and their actual usage behavior (Arteaga Sánchez et al., 2014; Mazman & Usluel, 2010). According to TAM, users’ adoption significantly predicts actual platform usage behavior. Given Douyin's growing prominence as an informal educational tool, students’ overall adoption is expected to positively influence their voluntary educational usage of Douyin. Hence, the following hypothesis is proposed:
Personal Factor and Educational Usage
Social Cognitive Theory suggests personal factors—such as self-efficacy, outcome expectations, and attitudes—are crucial determinants of individual behaviors. In the context of Douyin usage, personal factors encompass students' self-efficacy (confidence in using Douyin for learning), outcome expectations (positive beliefs regarding learning outcomes), and attitudes (positive evaluations of Douyin as a learning platform) (Zhou et al., 2020; Deng & Zhang, 2023). Converging work in higher education shows that acceptance-related factors and person-centered determinants jointly shape technology-supported learning behaviors and engagement (Ma et al., 2025). These motivational and cognitive factors significantly foster active student engagement, content creation, and sustained educational participation on Douyin. Therefore, the following hypothesis is proposed:
Methodology
Instruments
A structured questionnaire was employed to collect empirical data from undergraduate students at Pingdingshan University, located in Henan Province, China. Adapted from previously validated instruments relevant to the research context, the questionnaire comprised four logically structured sections (see Appendix A). Section A gathered demographic information about the respondents. Section B included items measuring Adoption, operationalized through three dimensions: Perceived Usefulness, Perceived Ease of Use, and Social Influence. Section C focused explicitly on Personal Factor, and Section D contained items measuring respondents’ Educational Usage of Douyin.
Respondents were instructed to indicate their agreement on a seven-point Likert scale, ranging from “1 = strongly disagree" to "7 = strongly agree," facilitating intuitive and precise responses. Clear introductory instructions outlined the study's objectives, invited voluntary participation, and assured respondents of confidentiality. To ensure logical structure, coherence, and user-friendliness, the questionnaire underwent a rigorous review process involving three academic experts. A subsequent pilot test was conducted, resulting in minor revisions for improved clarity and consistency.
Sampling, Procedure, and Sample Size Rationale
The target population comprised full-time undergraduates enrolled at Pingdingshan University. A single-site cluster approach was adopted at the institutional level, followed by voluntary-response recruitment within the site across majors and year levels via official channels (course announcements and student mailing lists). Data were collected online between June and August 2024. Eligibility required respondents to confirm current full-time undergraduate status prior to accessing the questionnaire. The study received approval from the Human Research Ethics Committee of Universiti Pendidikan Sultan Idris; informed consent was obtained electronically, participation was voluntary, and responses were anonymous.
A total of 568 questionnaires were submitted. After listwise deletion of 21 cases with incomplete data, 547 complete responses remained. Multivariate outliers were then identified using Mahalanobis distance computed across all measurement indicators; cases exceeding the chi-square critical value at p < 0.001 (61 cases) were removed. The final analytic sample contained 486 respondents. This screening protocol limits undue leverage on parameter estimates and improves the robustness of model fit and structural paths.
The retained sample size (N = 486) satisfies widely cited adequacy benchmarks for confirmatory factor analysis and structural equation modeling, including the “N ≥≥ 200” guideline and observations-to-parameters ratios of at least 10 to 15:1 (Kline, 2023).
Participants
The demographic characteristics of the respondents, including gender, academic year, Douyin usage experience with Douyin, frequency of use, and daily time spent on Douyin, are summarized in Table 1. Of the 486 valid respondents, female students (53.1%) slightly outnumbered male students (46.9%). Regarding academic year distribution, freshmen constituted the largest group, accounting for 28.2% (n = 137) of participants, followed closely by sophomores (27.0%, n = 131). Juniors represented 24.5% (n = 119), and seniors and above accounted for the remaining 20.4% (n = 99).
Demographic Characteristics of Respondents.
Participants’ experience with Douyin varied considerably. A majority (33.7%, n = 164) reported using Douyin for 3–4 years, while 25.7% (n = 125) had been users for 1 to 2 years. Another substantial group (23.9%, n = 116) indicated using Douyin for more than 5 years, and a smaller portion (16.7%, n = 81) had less than 1 year of experience.
In terms of usage frequency, most respondents (65.4%, n = 318) used Douyin multiple times daily. Another group (13.4%, n = 65) accessed Douyin once daily. Participants who used Douyin 4 to 6 times weekly accounted for 8.6% (n = 42), while those using it 1 to 3 times weekly represented 4.5% (n = 22). Additionally, 8.0% (n = 39) of respondents reported accessing Douyin less than once per week.
Regarding daily time spent on Douyin, a majority of participants (55.3%, n = 269) indicated spending 1 to 3 hr daily on the platform, followed by those spending less than 1 hr per day (24.7%, n = 120). A smaller portion of respondents (17.5%, n = 85) reported spending 4–6 hr daily. A few participants spent longer durations per day on Douyin, with 1.6% (n = 8) spending 7 to 9 hr and 0.8% (n = 4) exceeding 10 hr per day.
Data Analysıs And Results
Structural Equation Modeling (SEM) has emerged as a pivotal and highly effective method for analyzing complex relationships in social science research (Anderson & Gerbing, 1988; Hair et al., 2010). Combining factor analysis and path analysis, SEM enables researchers to examine the structural relationships among observed variables and latent constructs, making it a widely adopted technique across disciplines (Kline, 2023). Its growing popularity, particularly in fields like information science (Hooper et al., 2008), stems from its capacity to estimate multiple and interrelated dependencies within a single analysis.
Confirmatory Factor Analysis
Following Anderson and Gerbing’s (1988) two-step procedure, we first estimated the measurement models in AMOS 24, allowing latent factors to intercorrelate freely. Model adequacy was evaluated with indices from the three families summarized by Byrne (2016)—absolute fit (GFI, RMSEA), incremental fit (CFI, TLI, AGFI), and parsimony fit (X2/df). Consistent with Hair et al. (2010), we adopted the following benchmarks: GFI/AGFI ≥ 0.90, CFI/TLI ≥ 0.90, RMSEA < 0.08, and X2/df < 3. Standardized factor loadings ≥0.60 were acceptable and ≥0.70 ideal (Hair et al., 2010); SMC > 0.50 was regarded as ideal for indicator reliability (Fornell & Larcker, 1981). Items failing these primary rules were candidates for removal.
Adoption (second-order). Adoption was specified as a reflective second-order factor with three first-order dimensions—Perceived Usefulness (PU), Perceived Ease of Use (PEOU), and Social Influence (SI)—each initially measured by five indicators. Preliminary estimates identified SI1 with loading = 0.68 (acceptable) but SMC = 0.46 (below the ideal threshold >0.50), and it was removed on that basis. Global fit remained suboptimal (CMIN/DF = 3.538, GFI = 0.919, AGFI = 0.889, RMSEA = 0.072). Two further items, PU5 and PEOU4, did not satisfy the primary rules (loading <.60 and/or SMC <0.50) and were removed. The resulting model exhibited excellent fit (CMIN/DF = 2.039, GFI = 0.966, AGFI = 0.948, TLI = 0.988, CFI = 0.990, RMSEA = 0.046).
A first-order CFA was conducted for the Personal Factor construct measured by seven observed items. Although all initial factor loadings were satisfactory, initial fit indices indicated a poor model fit (CMIN/DF = 8.541, AGFI = 0.868, RMSEA = 0.125). After removing item PF6 due to large residual correlations, the revised CFA yielded substantial improvement, presenting acceptable fit indices (CMIN/DF = 2.760, GFI = 0.984, AGFI = 0.962, TLI = 0.990, CFI = 0.994, RMSEA = 0.060).
Similarly, the Educational Usage construct was initially measured by seven observed variables. Initial CFA results indicated suboptimal fit indices (CMIN/DF = 9.030, AGFI = 0.851, RMSEA = 0.129). Items EU1 and EU6 showed large residuals and were subsequently removed. The refined CFA model demonstrated substantial improvement, achieving excellent fit indices (CMIN/DF = 2.747, GFI = 0.988, AGFI = 0.965, TLI = 0.993, CFI = 0.997, RMSEA = 0.060).
Measurement Model
Following the completion of confirmatory factor analyses for each construct, all latent variables were integrated into a unified measurement model (see Figure 2). The model exhibited a strong overall fit, with key indices meeting established benchmarks: CMIN/DF = 1.894, GFI = 0.931, AGFI = 0.915, TLI = 0.980, CFI = 0.983, and RMSEA = 0.043.
Measurement model for all constructs.
All standardized factor loadings were above the recommended threshold of 0.70, indicating satisfactory convergent validity across constructs. The second-order construct Adoption was well represented by its three dimensions: Perceived Usefulness (loading = 0.86), Perceived Ease of Use (loading = 0.95), and Social Influence (loading = 0.85). For the Personal Factor construct, loadings ranged from 0.80 to 0.91, while Educational Usage showed similarly strong loadings between 0.80 and 0.93. These findings confirmed the reliability and validity of the comprehensive measurement model, supporting its suitability for subsequent structural model analysis.
In addition to overall model fit, the reliability and validity of all latent constructs were assessed using standardized factor loadings, composite reliability (CR), average variance extracted (AVE), and inter-construct correlations. The results, presented in Table 2, indicate that all standardized factor loadings exceeded the recommended threshold of 0.70, and that all CR values were above 0.70, thereby demonstrating good internal consistency (Hair et al., 2010). Furthermore, the AVE values for all constructs were greater than 0.50, indicating adequate convergent validity.
Construct Validity and Reliability Metrics for Measurement Items.
*p < 0.05. **p < 0.01. ***p < 0.001.
To evaluate discriminant validity, the Fornell-Larcker criterion was applied. As shown in Table 3, the square roots of AVE for each construct (displayed on the diagonal) exceeded the correlations with other constructs, confirming satisfactory discriminant validity among the latent variables (Fornell & Larcker, 1981). Specifically, the square root of AVE for Adoption (0.900), Personal Factor (0.887), and Educational Usage (0.867) each exceeded their corresponding inter-construct correlation coefficients.
Discriminant Validity Analysis.
Diagonal shows the square root of AVE.
Prior to estimating structural paths, potential collinearity among the exogenous predictors of Educational Usage was assessed using ordinary least squares regression with factor scores to compute tolerance and variance inflation factor (VIF). In line with common guidelines, tolerance values above 0.10 and VIF values below 5 indicate no problematic multicollinearity (Hair et al., 2010). As reported in Table 4, Adoption displayed VIF = 2.352 (tolerance = 0.425) and Personal Factor displayed VIF = 1.721 (tolerance = 0.581), indicating acceptable collinearity levels and supporting the stability of subsequent path estimates.
Multicollinearity Assessment.
Dependent variable: educational usage.
Structural Model
Upon confirming the validity and reliability of the measurement model, the structural model was evaluated to test the hypothesized relationships among the latent constructs (see Figure 3). As shown in Table 5, the model demonstrated satisfactory goodness-of-fit, with all fit indices within the recommended thresholds: CMIN/DF = 1.894, GFI = 0.931, AGFI = 0.915, TLI = 0.980, CFI = 0.983, and RMSEA = 0.043. These results indicate that the proposed theoretical framework fits the empirical data well and is appropriate for further hypothesis testing.
Overall Structural Equation Model.
Fitness Indices for Overall Structural Equation Model.
The path analysis results are summarized in Table 6. Both hypothesized relationships were statistically significant and in the expected direction. The path from Adoption to Educational Usage yielded a standardized coefficient of β = .536 (S.E. = 0.051, C.R. = 11.988, p < 0.001), providing strong support for H1. This finding suggests that students who perceive Douyin as useful and easy to use, and who perceive social influence from important others, are more likely to engage with the platform for educational purposes. The strength of this path highlights the critical role of platform-related perceptions, as conceptualized in the Technology Acceptance Model (TAM), in shaping students' informal learning behavior.
Hypothesis Testing.
Similarly, the path from Personal Factor to Educational Usage was significant, with a standardized coefficient of β = .405 (S.E. = 0.049, C.R. = 9.971, p < .001), supporting H2. This result confirms that students’ internal beliefs—such as self-efficacy, outcome expectations, and positive attitudes—exert a meaningful influence on their voluntary engagement with Douyin for learning. This is consistent with the predictions of Social Cognitive Theory (SCT), which emphasizes the centrality of personal agency and motivation in behavioral outcomes.
These findings provide empirical validation for the integrated theoretical model, confirming that both technology-related perceptions (from TAM) and individual motivational factors (from SCT) play pivotal roles in shaping undergraduates’ voluntary educational use of Douyin.
Discussion, Implications, Limitations, And Conclusion
Discussion
This study examined the determinants of Chinese undergraduates’ voluntary educational usage of Douyin by integrating the Technology Acceptance Model (TAM) with Social Cognitive Theory (SCT). Using SEM on the screened analytic sample, the results indicate that Adoption—modeled as a reflective second-order construct indicated by perceived usefulness, perceived ease of use, and social influence—and the first-order Personal Factor both show positive, significant paths to Educational Usage. The effect of Adoption is larger, suggesting a predominant platform-facing route, with Personal Factor forming a complementary person-centered route toward education-oriented engagement.
These findings are consistent with TAM evidence underscoring the centrality of usefulness and the enabling role of ease of use in educational technology contexts (Davis, 1989; Chipps et al., 2015; Farahat, 2012). The salience of social influence aligns with research on peer effects in collectivist settings (Goli & Khan, 2022; Jia et al., 2023) and with observations about the social nature of Douyin among university students (Liu, 2019; Xiang, 2020). In parallel, the significance of Personal Factor accords with SCT’s emphasis on learner agency in shaping behavior (Bandura, 2001) and with empirical studies linking favorable person-centered determinants to purposeful engagement on social media (Deng & Zhang, 2023; Zhou et al., 2020). Together, the results address the study objective of explaining when and why students move beyond casual consumption to education-oriented use. While Douyin’s short-form, visually dynamic delivery and algorithmic personalization are compatible with micro-learning and repeated exposure (Fuad et al., 2023; Kolber, 2024; Low et al., 2023), repurposing the platform for learning appears to depend jointly on platform-facing appraisals summarized by Adoption and person-centered dispositions summarized by Personal Factor (Assad, 2024; Wang et al., 2024).
Taken together, the pattern we observe is likely context-sensitive. In institutions where platform governance surfaces education-oriented verticals and where course tasks align closely with short-video formats, the platform-facing route (Adoption) should strengthen via perceived usefulness; by contrast, in settings that cultivate self-regulated learning under higher academic demands, the person-centered route (Personal Factor) is expected to loom larger. In campuses characterized by dense peer networks and stronger collectivist norms, social influence may contribute more to Adoption.
Implications
Theoretical implications. The findings articulate an integrated two-route explanation for education-oriented usage on short-video platforms: a platform-facing route (Adoption as a reflective second-order construct indicated by usefulness, ease of use, and social influence) and a person-centered route (Personal Factor as a first-order construct). Modeling Adoption as second-order clarifies construct hierarchy and helps avoid conflating general entertainment usage with education-oriented behaviors while preserving TAM’s internal logic (e.g., the supportive role of ease of use for usefulness). Estimating both routes concurrently within a single SEM demonstrates complementarity, responding to calls to bridge acceptance-focused and learner-focused perspectives in informal digital learning.
Practical implications. For educators, course designs can leverage Adoption-relevant affordances while supporting the person-centered route—for example, curating short, concept-focused playlists aligned with assessments; scaffolding initial tasks to reduce effort and underscore usefulness; and activating peer mechanisms (e.g., class hashtags, peer endorsement) to harness social influence constructively. For platform and instructional designers, clearer labeling and retrieval of credible educational content, low-friction tools for assembling micro-modules, and prompts that nudge purposeful viewing sequences may strengthen the platform-facing route while aligning with learners’ agentic dispositions. For institutions, recognizing purposeful use (e.g., analytics-informed feedback, acknowledgement of high-quality academic channels) can consolidate normative support without compromising autonomy.
Limitations and Future Research
Three limitations bound inference. First, although data were collected at a single public university in Henan, the analytic sample spans multiple years and majors, and the measurement model met established reliability/validity benchmarks; future multi-site studies across regions and institution types can assess measurement/path invariance and further strengthen external validity. Second, the cross-sectional design does not establish temporal ordering; longitudinal panels (e.g., cross-lagged models) and course-embedded field studies can examine how Adoption and Personal Factor co-evolve with Educational Usage. Third, self-reports may introduce common-method bias; anonymity and clear instructions were used to mitigate this, and subsequent work could triangulate surveys with de-identified platform traces (e.g., watch/replay events) and brief qualitative follow-ups to deepen construct validity.
Generalizability is therefore bounded by institutional and cultural conditions. Future work should adopt multi-site sampling and conduct multi-group SEM across region and institution type to test path invariance, incorporate institution-level covariates in hierarchical models, and pursue cross-cultural replications (Douyin vs. TikTok) to probe whether the social-influence channel varies with normative environments.
Conclusion
Undergraduates’ educational usage of Douyin reflects the interplay of platform-facing adoption beliefs and person-centered dispositions. By specifying and testing these two routes within a unified model, the study clarifies how an entertainment-first platform can be repurposed for learning. As higher education continues to engage with short-video ecosystems, recognizing and designing for both routes will be essential to support voluntary, education-oriented engagement.
Questionnaire (survey).
Footnotes
Acknowledgements
The author would like to sincerely thank Prof. Dr. Ahmad Zamzuri bin Mohamad Ali from the Faculty of Art, Sustainability and Creative Industry, Universiti Pendidikan Sultan Idris (UPSI), Tanjung Malim, Perak, Malaysia, for his invaluable supervision, intellectual guidance, and generous support throughout the development of this research.
Ethical Considerations
This study was approved by the Human Research Ethics Committee of Sultan Idris Education University (Approval No. 2024-0419-01).
Consent to Participate
All participants were fully informed about the nature and purpose of the study and provided written informed consent prior to data collection.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.