Vietnamese High School Students in the AI Era: Examining the Control Threshold and Knowledge Paradox in Academic Integrity

Vietnamese High School Students in the AI Era: Examining the Control Threshold and Knowledge Paradox in Academic Integrity | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Vietnamese High School Students in the AI Era: Examining the Control Threshold and Knowledge Paradox in Academic Integrity Tinh Le, Doanh Ho, Hung Tran This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8458923/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract As artificial intelligence blurs the boundaries between authorship and assistance, traditional frameworks for understanding academic dishonesty are becoming obsolete. This study employs a data-driven approach to decode the complex ethical decision-making of adolescents in a developing digital economy. Analyzing data from 923 secondary students in Vietnam, we combine the Theory of Planned Behavior (TPB) with machine learning (XGBoost) and SHAP interpretability techniques. The XGBoost model achieved superior predictive accuracy ( R 2 = 0.634) compared to traditional linear regression ( R 2 = 0.604). The study offers two major contributions to the discourse on technology and ethics. First, it exposes a ‘knowledge paradox’ where increased awareness of academic integrity rules is significantly associated with higher plagiarism frequency ( B = 1.12, p < .001), indicating a disconnect between digital literacy and ethical practice. Second, XAI analysis reveals a nonlinear threshold in Perceived Behavioral Control: when the perceived difficulty of avoiding plagiarism exceeds a critical tipping point (score > 3.5 on a 5-point scale), the likelihood of offending escalates disproportionately, regardless of students' moral attitudes. These findings suggest that purely informational campaigns are insufficient in the AI era. Instead, promoting integrity requires systemic interventions that address the psychological thresholds of control and the unintended consequences of digital literacy education. Theory of Planned Behavior Academic integrity Tech-facilitated cheating Explainable AI (XAI) SHAP explainability Figures Figure 1 Figure 2 Figure 3 1. Introduction Turnitin’s 2024 report, based on the analysis of 200 million submissions worldwide, identified more than 22 million documents showing substantial AI involvement, underscoring how deeply AI has penetrated contemporary educational practices (Turnitin, 2024). This evidence points to a worrying rise in technology-supported academic dishonesty, challenging long-standing notions of authorship and merit in the digital era. As teaching and learning increasingly move into digital spaces, secondary school students are now routinely exposed to AI writing tools and extensive online information sources. Although these technologies can enhance access to knowledge and support more efficient writing processes (Banks et al., 2024 ), they also complicate students’ understanding of originality and academic integrity (Foltynek et al., 2023 ; Lim et al., 2023 ). In this sense, generative AI functions as a “double-edged sword”, making the boundary between acceptable support and unethical practice increasingly unclear and more difficult to monitor (Bin-Nashwan et al., 2023 ). Empirical evidence further illustrates the seriousness of this issue. Studies suggest that up to 43% of high school students admit to some form of cheating behavior (Birks & Clare, 2023a ; Mavrinac et al., 2010 ). Despite this, most existing research has focused primarily on higher education, as shown in recent bibliometric reviews (Watrianthos et al., 2023 ; Cotton et al., 2024 ). This imbalance leaves a notable gap in our understanding of academic integrity during adolescence, a critical developmental stage when learning habits, values, and moral reasoning begin to stabilize (Chan, 2025 ; Macdonald & Carroll, 2006 ). During this period, the use of AI as a form of “cognitive shortcut” may contribute to the normalization of plagiarism, particularly when students hold incomplete or inaccurate views of intellectual ownership in open digital environments (Birks & Clare, 2023b ; Nguyen & Goto, 2024 ; Eret & Ok, 2014 ). From a theoretical perspective, the Theory of Planned Behavior (TPB) (Ajzen, 1991a ) offers a useful framework for examining such behaviors, emphasizing the role of attitudes, perceived social expectations, and perceived behavioral control. Recent studies on technology use suggest that perceived benefits often outweigh ethical concerns when individuals decide whether to adopt new tools (Ivanov et al., 2024 ). While TPB has demonstrated strong predictive value across various academic integrity studies (Stone et al., 2009a ; Uzun & Kilis, 2020a ), applying the model directly to secondary education can overlook an important factor: students’ knowledge of plagiarism. Limited or incorrect understanding of academic rules may lead to accidental misconduct, or alternatively, to more deliberate attempts to bypass detection-dimensions that traditional intention-based models often fail to address adequately (Johansen et al., 2022a ; Maxwell et al., 2008 ). Beyond conceptual concerns, methodological issues also constrain existing research. Much of the TPB literature relies on self-reported data and linear statistical models. However, recent findings indicate that adolescents frequently misestimate their own technology use, calling into question the accuracy of self-report measures (Li et al., 2025 ). In addition, linear models may oversimplify students’ decision-making processes in technology-rich environments, where behaviors are shaped by complex interactions and threshold effects (Linardatos et al., 2020 ). In response to these challenges, machine learning techniques such as XGBoost have demonstrated strong potential for identifying subtle patterns and improving predictive performance (Feldman-Maggor et al., 2025 ). Nevertheless, the limited transparency of such models remains a concern in educational and social research, where understanding underlying mechanisms is as important as prediction accuracy (Swamy et al., 2023 ). Recent advances in Explainable AI (XAI), particularly the SHAP (SHapley Additive exPlanations) method, provide a promising way forward by clarifying how individual variables contribute to model outcomes (Zeng et al., 2023 ). When used alongside established theoretical frameworks like TPB, SHAP helps connect predictive strength with meaningful interpretation. Building on this approach, the present study proposes an integrated framework that extends TPB by incorporating knowledge-related variables, employs hierarchical regression to test theory-driven hypotheses, and applies XGBoost with SHAP to capture non-linear relationships. Rather than focusing solely on predicting plagiarism behavior, the study aims to uncover the cognitive and contextual factors that shape students’ decisions, thereby offering insights to inform more effective academic integrity policies in the age of AI. 2. Literature Review 2.1 Plagiarism behavior in K-12 education While academic dishonesty has been extensively explored in higher education, its prevalence and determinants in secondary schools are not fully understood and reported results are mixed. This gap is particularly notable because adolescence is a critical period in the development of academic ethics and behavioral norms (Macdonald & Carroll, 2006 ). Recent studies have shown that plagiarism is already evident in this period. A review by Désiron & Petko ( 2023 ) found that approximately 43 percent of secondary school students admitted to plagiarism and that most did not recognize that it was an ethical violation. Results from Stoesz & Yudintseva ( 2018 ) point to a similar conclusion, with more than a third of European students reporting copying or paraphrasing without citing. However, most existing studies are still descriptive or methodologically limited, relying heavily on self-reports and lacking a strong theoretical foundation (Nora & Zhang, 2010 ; Johansen et al., 2022b ). The number of studies that apply rigorous behavioral frameworks or computational methods to model the complex antecedents of cheating behavior is still very small. At the same time, as Çelik & Razı ( 2023 ) note, many students have difficulty distinguishing between academic collaboration, resource use, and plagiarism, while their ability to think ethically is not fully developed and their training in academic skills is limited. To move the field further, three interrelated aspects need to be considered. The first aspect relates to the diversity of misconduct. Plagiarism in the K–12 environment does not take a single form but ranges from verbatim copying and patchwork plagiarism to contract cheating and unauthorized forms of collusion (Pike et al., 2025 ). It is important to note that students’ motivations, whether opportunistic, unintentional, or driven by external pressure, often determine whether they perceive the behavior as a violation (Zhao et al., 2022a ). The second aspect relates to developmental factors. At this age, metacognitive maturity is limited, self-regulation is not yet established, and peer pressure can be strong. These factors make students more vulnerable to dishonest behavior than older age groups (Sisti, 2007 ; Stone et al., 2010 ). Results from the OECD PISA 2022 assessment also show an increase in the number of rule violations in online tests, reflecting that the digital environment is increasing both the opportunity and the temptation for misconduct. A third aspect has come to the fore in recent years. Technologies powered by artificial intelligence are blurring the line between authorship and technical assistance. Tools such as automated paraphrasing software, summarization bots, or large language models such as ChatGPT make it harder to determine the originality of learning products. Studies by Johnston et al. ( 2024 ) and Ortiz Bonnin & Blahopoulou (2025) show that some students view AI-generated content as a form of scaffolding rather than plagiarism, especially in contexts where teachers have not provided clear guidance on the use of these tools. Despite the growing complexity of these issues, the number of studies that combine these three dimensions into a unified behavioral model is still very small. Work that operationalizes them using advanced predictive techniques that go beyond linear regression is even rarer. The application of foundational theories such as the Theory of Planned Behavior remains fragmented and rarely calibrated to the ecological context of high school. In addition, existing research has not yet clearly explained how digital competence and perceived norms interact with personal beliefs to shape cheating intentions. To fill this gap, the current study extends the TPB by adding the variable knowledge about plagiarism, which has been shown to moderate the gap between intentions and behavior but has often been overlooked in previous studies (Juan et al., 2022 ). At the same time, we use the state-of-the-art XGBoost machine learning model combined with the SHAP interpretation algorithm to uncover non-linear relationships and variable interactions that are difficult to identify with traditional analysis methods. This approach is both explanatory and predictive, based on a behavioral framework adapted to the digital learning context of today's high school students. 2.2 The Theory of Planned Behavior and academic dishonesty The Theory of Planned Behavior (TPB) (Ajzen, 1991b ) has long provided a central framework for understanding academic misconduct, particularly cheating and plagiarism. At its core, TPB proposes that behavioral intentions, shaped by attitudes, subjective norms, and perceived behavioral control (PBC), are the most proximate predictors of actual behavior. While its predictive validity has been consistently demonstrated in higher education (Curtis et al., 2018b ; Zhang, 2024 ), the application of TPB in secondary education remains limited and, in many cases, under-theorized (Choi & Suh, 2022 ). This gap is notable given the socio-cognitive peculiarities of adolescents. High school students, unlike their college-age peers, are still in the process of consolidating their moral identities and tend to be more responsive to external influences (e.g., peer behavior, teacher expectations, and institutional cues) when forming moral judgments (Malti et al., 2021 ; Irani et al., 2024 ). Therefore, applying the TPB in a high school context requires careful attention to how school culture and local socialization processes shape subjective norms. For example, normative beliefs about “copying” can vary significantly across school districts or geographic areas (Zhao et al., 2022b), and institutional enforcement practices can disproportionately influence students’ perceptions of their ability to control their own behavior. The original TPB formulation also offers limited guidance for understanding behavior in AI-mediated learning environments, where technological capabilities and emerging ambiguities around authorship reshape decision-making conditions (Leaton Gray et al., 2025 ; Bin-Nashwan et al., 2023 ). In digital environments that provide instant access to paraphrasing tools, automated citation tools, or Large Language Models (LLMs), the line between legitimate assistance and plagiarism may become blurred. These tools may increase students’ perceived sense of control over their behavior (e.g., “it’s easier to avoid detection”), while widespread use of such technologies may subtly alter subjective norms, making unauthorized assistance seem more acceptable (Clarke et al., 2023 ). Without a theoretical update, TPB-based models risk overlooking forms of moral disengagement that arise from these technologies. Another conceptual concern relates to the TPB’s implicit assumption that individuals accurately determine the behavior under consideration. However, empirical research suggests that misunderstandings about citation practices, paraphrasing, and authorship conventions (collectively referred to as “plagiarism knowledge”) are key factors contributing to both unintentional and intentional plagiarism among adolescents (Prashar et al., 2024 ). Incorporating plagiarism knowledge into the TPB thus strengthens the model by acknowledging that students cannot form an intention to avoid a behavior they do not explicitly perceive as problematic. To apply the TPB to AI-integrated secondary education, scholars have proposed extending the model or integrating it with additional theoretical perspectives. For example, Bandura’s (1991) Moral Disengagement Theory provides a useful perspective to explain how students justify unethical behavior, especially when institutional enforcement is perceived as inconsistent or weak. Similarly, adding technology-related constructs (e.g., digital competence or the perceived legitimacy of AI support) could improve the model’s explainability in modern learning environments (see Table 1 ). Table 1 TPB components Dimension Traditional TPB Extended TPB (AI-Era, Adolescents) Norm source General social expectations School-specific, peer-group norms, regional culture Behavioural control General perceived ease/difficulty Technology-mediated control (e.g., AI tools, plagiarism bots) Attitude Personal evaluation of behaviour Shaped by exposure to digital content and ethical ambiguity Knowledge (added) Often excluded Formalised as cognitive antecedent to intention/behaviour Moral disengagement (optional) Not modelled Complements TPB when behaviour contradicts internalised norms Taken together, these considerations suggest that the original TPB framework, while conceptually robust, requires meaningful adaptation to reflect the developmental, contextual, and technological realities of secondary school students. Specifically, integrating plagiarism knowledge as a cognitive component addresses the formative nature of academic literacy in this population, while accounting for the influence of AI tools helps clarify how perceived control and social norms are reshaped in digital learning environments. These theoretical refinements inform the present study’s conceptual model, which extends TPB by (1) incorporating students’ knowledge about plagiarism, and (2) situating behavioural predictors within an AI-mediated context. 2.3 Machine learning approaches in academic integrity research Machine learning (ML) is increasingly permeating educational and psychological research, offering analytical capabilities that go beyond the limitations of traditional parametric modeling. While ML has been successfully deployed to predict student engagement, dropout risk, and overall academic performance (Kemper et al., 2020 ; Villar & de Andrade, 2024 ), its application in modeling academic misconduct, specifically, plagiarism in high school students, has not been fully explored. Existing literature focuses primarily on higher education or technical detection algorithms, leaving a gap in understanding behavioral predictors in the K-12 context (Alsabhan, 2023 ). The main advantage of ML lies in its ability to handle nonlinear interactions on multidimensional data. In the context of plagiarism, psychological constructs (e.g., attitudes, subjective norms), contextual factors (e.g., school culture), and technological capabilities (e.g., GenAI tools) often interact in complex and dynamic ways (Ghimire et al., 2024 ; Lu et al., 2023 ). Standard linear models often fail to capture these subtleties. However, a persistent criticism of ML is the “black box” phenomenon, in which predictive power comes at the expense of interpretability. To mitigate this, the field is turning to explainable AI (XAI), particularly SHAP (SHapley Additive Explanations) (Guleria & Sood, 2023 ). SHAP provides a mathematically consistent framework for quantifying the contribution of each predictor at both the global and local levels, making it highly compatible with psychological frameworks such as the Theory of Planned Behavior (TPB) (Johora et al., 2025 ). Unlike alternative explanatory approaches such as LIME, SHAP offers theoretical consistency and model-independent flexibility (Ahmed et al., 2025 ), allowing researchers to analyze the relative importance of variables such as Perceived Behavioral Control or Subjective Norms while still accounting for non-linear effects. Despite these methodological advances, the integration of SHAP-enhanced Machine Learning (ML) into the study of academic integrity in secondary schools remains in its infancy. Although Molnar et al. ( 2020 ) used Interpretable Machine Learning to examine AI-facilitated plagiarism, their study was limited to higher education. There is little research that uses these tools in theory-based approaches, such as TPB, to investigate plagiarism in adolescents. To address this gap, the current study integrates an extended TPB framework with XGBoost and SHAP analysis. This methodological combination aims to connect theoretical understanding with data-driven accuracy, providing deeper insights into the mechanisms of academic cheating and informing targeted educational interventions. 2.4. This study In summary, while TPB offers a strong theoretical foundation for understanding academic dishonesty, its application to plagiarism behavior in K-12 students remains underexplored. Moreover, although machine learning (ML) methods show potential for modeling complex behavioral patterns, they have not yet been widely applied or integrated with TPB frameworks in this context. This study seeks to address these gaps by combining an extended TPB model-which incorporates knowledge about plagiarism-with ML-based predictive modeling and interpretable ML techniques. In doing so, we aim to contribute both to theory and to practical efforts for promoting academic integrity in schools. The study is guided by the following research questions: RQ1 To what extent do attitudes, subjective norms, perceived behavioural control, and knowledge about plagiarism predict self-reported plagiarism behaviour among secondary school students? RQ2 Are there significant group differences in TPB components, knowledge, and plagiarism behavior across student gender, grade level, and school type? RQ3 To what extent can machine learning models (specifically XGBoost) enhance the prediction of plagiarism behaviour compared to traditional hierarchical regression models? RQ4 What insights do interpretable machine learning techniques (specifically SHAP analysis) provide regarding the relative importance of predictors, and potential interaction patterns, in explaining plagiarism behavior among secondary school students? Through this integrated approach, we aim to deepen the understanding of plagiarism behavior among K-12 students and to demonstrate how theory-driven and data-driven methods can complement each other in educational research. 3. Methods 3.1 Participants and procedure This study involved a nationwide sample of secondary school students from various regions of Vietnam. Participants were recruited through an online survey distributed via Google Forms. The survey link was shared through networks of current university interns and former university students who are now teachers at secondary and high schools across the country. This recruitment approach enabled access to a diverse sample from both urban and rural schools. Data collection was conducted online during the Spring semester of 2025. After applying standard data cleaning procedures to remove incomplete and inconsistent responses, the final sample included 900 students. Of these, 660 were male (73.3%) and 240 were female (26.7%). In terms of school type, 789 students (87.7%) attended public institutions, while 111 students (12.3%) were enrolled in private schools. With respect to grade level, 89 students (9.9%) were in lower secondary school (Grades 6–9), and 811 students (90.1%) were in upper secondary school (Grades 10–12). The observed gender imbalance-predominantly male-is consistent with national enrollment trends in STEM-focused and digitally enriched academic tracks from which many participants were drawn. Participation in the study was entirely voluntary. No personally identifying information was collected, and participants’ anonymity was maintained throughout the process. Informed consent was obtained from all respondents, with additional parental consent secured where required by school or institutional policy. 3.2 Measures The survey instrument was grounded in the Theory of Planned Behavior (TPB) framework (Beck & Ajzen, 1991 ) and adapted to the Vietnamese secondary education context. The instrument comprised five distinct sections: plagiarism knowledge, the three TPB constructs (Attitudes, Subjective Norms, Perceived Behavioral Control), and self-reported plagiarism behavior. Table 2 summarizes the instrument structure. Plagiarism Knowledge. Objective understanding of academic integrity was assessed using a 12-item multiple-choice test. Items were developed based on international standards (e.g., APA, COPE) and covered common plagiarism scenarios. Scoring was binary (1 = Correct, 0 = Incorrect/Don’t know), with higher aggregate scores reflecting greater conceptual knowledge. Content validity was established through review by two academic integrity experts and two secondary school instructors. TPB Constructs . We measured Attitudes (ATT), Subjective Norms (SN), and Perceived Behavioral Control (PBC) using items adapted from existing literature on academic dishonesty (Stone et al., 2009b ; Uzun & Kilis, 2020b ) (see Table 2 ). To ensure ecological validity, items were modified to reflect local cultural and school context. All TPB items utilized a 5-point Likert scale ranging from 1 ( Strongly disagree ) to 5 ( Strongly agree ). Higher scores denote more permissive attitudes, stronger perceived social approval, and lower behavioral control (i.e., higher difficulty), respectively. The initial item pool underwent expert review for clarity and cultural appropriateness prior to deployment. Self-Reported Plagiarism Behavior . A six-item scale quantified the frequency of specific plagiarism behaviors (e.g., unauthorized paraphrasing, reusing work, copying without attribution). Unlike standard frequency scales, responses were captured on a 7-point scale designed to distinguish between historical and current behavior: 1 ( Never ), 2 ( Unsure ), 3–4 ( Past behavior only ), and 5–7 ( Current frequency: Rarely to Always ). Higher scores indicate active or frequent engagement in plagiarism. A pilot study ( N = 30 ) confirmed item clarity and response variability. Demographics. We collected data on gender, school type (public vs. private), and grade level (6–12) to facilitate group comparison analyses. Psychometric Properties. Construct validity and reliability were rigorously tested. Exploratory Factor Analysis (EFA) using principal axis factoring (Promax rotation) confirmed the three-factor structure for the TPB constructs. The first three eigenvalues (5.66, 1.90, 1.72) explained 64.58% of the total variance. Sampling adequacy was satisfactory (KMO = .87; Bartlett’s test p < .001). Internal consistency (Cronbach’s α was robust across all constructs: ATT (.83), SN (.80), PBC (.73), and self-reported behavior (.91). The plagiarism knowledge test demonstrated strong internal consistency (KR-20 = .85), with item difficulties ranging from .42 to .83 and discrimination indices between .29 and .59. Table 2 Summary of study variables and measurement Scale Name Items (English Translation) Rating Scale Notes Plagiarism Knowledge 12 items (e.g., Plagiarism is using another person’s ideas, processes, results, or words without giving appropriate credit. ) Correct (1) / Incorrect (0) / Don’t know (0) Total score: 0–12 Attitude toward the Behavior (ATT) 18 items; (e.g., Plagiarizing is as bad as stealing an exam.) 1 (Strongly disagree) to 5 (Strongly agree) Higher = more permissive attitude Subjective Norms (SN) 6 items; (e.g., If my friend allows me to copy, it is acceptable.) 1 (Strongly disagree) to 5 (Strongly agree) Higher = stronger perceived social approval Perceived Behavioral Control (PBC) 4 items; (e.g., I could not write a scientific paper without plagiarizing.) 1 (Strongly disagree) to 5 (Strongly agree) Item 15 reverse coded; Item 23 cross-loaded in ATT & PBC Self-reported Plagiarism Behavior 6 items. (e.g., Have you ever copied and pasted a section of someone else’s work without attribution?) 1 (Never), 2 (Not sure / Don’t remember), 3 (Done a few times in the past, no longer do it), 4 (Done many times in the past, no longer do it), 5 (Rarely do it), 6 (Often do it), 7 (Always do it) Higher = more frequent plagiarism behavior. Note: Item 6 (“use plagiarism detection software”) may not reflect plagiarism behavior - consider analyzing separately. Demographics Gender, School type, Grade level Nominal / Categorical Used for group comparisons 3.3 Data handling and assumption testing The final dataset included in the analysis included 923 valid observations after removing missing data. The proportion of missing data on the variables was negligible (< 2.5%) and was processed using the multivariate imputation through linked equations (MICE) technique to maximize statistical power and minimize bias (White et al., 2011; Schafer & Graham, 2002). Prior to conducting the regression, the prerequisite statistical assumptions, including multicollinearity, normality, and homogeneity of variance, were tested. The diagnostic results showed that there was no multicollinearity, with the variance inflation factor (VIF) of all independent variables ranging from 1.14 to 2.77, which is within the safe threshold (< 5) recommended by Hair et al. (2019). Regarding the normal distribution assumption, the Shapiro-Wilk test shows that the residuals of the model deviate from the ideal normal distribution (W = 0.958; p < 0.001 $ ). However, this is a common phenomenon in studies with large sample sizes, when the Shapiro//-Wilk test becomes too sensitive even to insignificantly small deviations (Field, 2018). More importantly, according to the Central Limit Theorem, with large sample sizes (N > 30 $ ), the regression estimates still ensure robustness and are not seriously affected by violations of the normality assumption of the residuals (Lumley et al., 2002; Schmidt & Finan, 2018). The final analytic sample included 923 complete cases after removing missing data. Missingness across items was minimal (< 2.5%) and addressed using multiple imputation via chained equations (MICE), although regression analyses used only cases with full data. Prior to regression, assumptions of multicollinearity, normality, and homoscedasticity were tested. Variance inflation factors (VIFs) for all independent variables ranged from 1.14 to 2.77, well below the standard threshold of 5, indicating no multicollinearity concerns. However, the residuals from the regression model deviated from normality, as indicated by a Shapiro-Wilk test result of W = .958, p < .001. Despite this, the large sample size provides robustness to mild violations of normality, which was noted as a limitation. 3.4 Analytical strategy The study adopted a two-tiered analytical strategy, combining a theory-based statistical model with a machine learning model to simultaneously test hypotheses and optimize predictions. RQ1: Hierarchical multiple regression was used to test the predictive power of TPB components (Attitude, Subjective Norms, PBC) and plagiarism knowledge on self-reported behavior. Demographic variables were entered first, followed by TPB variables, and finally plagiarism knowledge. Assumptions of multicollinearity and residual distribution were checked before analysis. RQ2: Independent samples t-tests were conducted to determine differences by gender, school type, and grade level for TPB variables and plagiarism behavior. RQ3: An XGBoost model was built to evaluate the nonlinear predictive ability and compared directly with OLS regression on the same test set (80% training, 20% testing). The model parameters were optimized through 5-fold cross-validation. The model performance was evaluated through R² and RMSE. RQ4: To explain the predictive mechanism of the machine learning model within the TPB framework, SHAP analysis was applied to evaluate the importance of features and explore nonlinear relationships, including threshold effects. Statistical analyses were performed using SPSS v.26. The machine learning models and SHAP analysis were implemented on the Google Colab platform with Python v.3.11, using the xgboost v.2.0.3 and shap v.0.41.0 libraries. 4. Results 4.1 Theory-based prediction of plagiarism behavior (RQ1) A hierarchical regression analysis was conducted to examine the predictive power of attitudes toward plagiarism, subjective norms, perceived behavioral control (PBC), and plagiarism knowledge in relation to students’ self-reported plagiarism behavior. In Step 1, demographic variables-gender, educational level (lower vs. upper secondary), and school type (public vs. private)-explained 20.9% of the variance in plagiarism behavior (R² = .209, p < .001). In Step 2, the inclusion of the TPB components (attitudes, subjective norms, and PBC) significantly improved model performance, yielding an additional 36.3% of explained variance (ΔR² = .363, p < .001), bringing the cumulative explained variance to R² = .572. In Step 3, the inclusion of the knowledge variable contributed a further 3.2% of explained variance (ΔR² = .032, p < .001), resulting in a final model that accounted for 60.4% of the total variance in plagiarism behavior (R² = .604). All predictors in the final model were statistically significant (p < .01). Perceived behavioral control emerged as the strongest negative predictor (B = -0.862, p < .001), indicating that students who felt more capable of resisting plagiarism were less likely to engage in it. Subjective norms had a strong positive effect (B = 0.668, p < .001), suggesting that permissive peer influences increased the likelihood of plagiarism. Attitudes toward plagiarism also significantly predicted behavior in the expected direction (B = -0.703, p < .001). Notably, knowledge about plagiarism was positively associated with plagiarism behavior (B = 1.123, p < .001), a surprising result that suggests students who are more knowledgeable may still choose to plagiarize-possibly due to external pressures or increased sophistication in using tools. No multicollinearity concerns were present; all variance inflation factor (VIF) values were below 3 (VIF_Gender = 1.12; VIF_Grade = 1.24; VIF_School Type = 1.02; VIF_Attitude = 2.70; VIF_Social Norms = 1.68; VIF_PBC = 2.95; VIF_Knowledge = 1.23). The model’s prediction error was acceptable, with RMSE = 0.788 and MAE = 0.564. Full regression coefficients are presented in Table 2 . Table 2 Hierarchical multi-regression predicting plagiarism behavior Predictor B SE t p Step 1 (Demographics) Intercept 3.177 0.120 26.433 < .001 Gender (0 = Male, 1 = Female) 0.699 0.086 8.157 < .001 Educational Level (0 = Lower, 1 = Upper) 1.287 0.124 10.409 < .001 School Type (0 = Public, 1 = Private) -0.364 0.112 -3.237 .001 Step 2 (TPB components) Attitudes -0.703 0.089 -7.878 < .001 Subjective Norms (SN) 0.668 0.067 9.966 < .001 PBC -0.862 0.063 -13.670 < .001 Step 3 (Knowledge) Knowledge 1.123 0.132 8.532 < .001 Model summary Step 1 R² .209 Step 2 R² .572 Step 3 R² .604 Furthermore, to address concerns about the gender imbalance in the sample (73.3% male), we conducted a robustness check by running separate regression models for the male and female subsamples. The “knowledge paradox” in which higher knowledge of plagiarism predicts increased misconduct, remained statistically significant and positive in both groups ( B male = 0.93, p < .001; B female = 1.02, p = .001). This consistency suggests that the counterintuitive relationship between knowledge and behavior is a universal mechanism driven by the instrumentalization of normative knowledge, rather than a gender-specific trait related to risk-taking tendencies. 4.2 Group differences (RQ2) To examine group differences in plagiarism behavior, attitudes, and knowledge, a series of independent-samples t-tests were conducted across gender, school type, and educational level. Specifically, male students reported significantly higher plagiarism frequency than female students (M = 5.04) vs. (M = 4.15), with the difference reaching statistical significance (t(906) = 8.41, p < .001). This suggests that the level of academic dishonesty tends to be higher in the male group, despite a higher understanding of plagiarism (M = 0.82 vs. 0.72; t(906) = 5.75, p < .001). It is worth noting that attitudes toward plagiarism between the two genders do not completely correlate with actual behavior. Female students, although having more tolerant attitudes towards plagiarism (M = 3.02 compared to 2.80 for males), reported significantly lower levels of plagiarism, with the difference also being statistically significant (t(906) = -4.99, p < .001). This result suggests a non-linear motivational pattern, in which stricter attitudes do not necessarily lead to higher compliance. The discrepancy between knowledge, attitudes and behavior, especially among males, can be interpreted as a manifestation of the perception-practice gap in the context of academic ethics. Regarding school type, students from public schools engaged in more plagiarism than those in private schools (t (906) = 2.84, p = .005). However, no significant differences were found between school types in either attitudes or knowledge. Significant differences also emerged across educational levels. Students in upper secondary school (Grades 10–12) reported substantially higher levels of plagiarism behavior than those in lower secondary school (Grades 6–9) (t(906) = -10.29, p < .001). Conversely, lower secondary students displayed more lenient attitudes (t(906) = 4.51, p < .001), while upper secondary students demonstrated significantly higher knowledge scores (t(906) = -5.58, p < .001). Group means and statistical details are reported in Table 3 . Table 3 Group Differences in Plagiarism-Related Constructs by Educational Level, Gender, and School Type Variable Group 1 M (SD) Group 2 M (SD) t p Cohen’s d Educational Level (Lower vs. Upper) Knowledge 0.63 (0.29) 0.81 (0.20) -5.58 < .001 0.75 Attitude 3.23 (0.87) 2.81 (0.42) 4.51 < .001 -0.61 Social Norm 2.83 (0.73) 3.08 (0.67) -4.04 < .001 0.36 PBC 2.96 (0.84) 3.56 (0.71) -8.22 < .001 0.76 Plagiarism Behavior 3.41 (1.37) 4.95 (1.13) -10.29 < .001 1.24 Gender (Male vs. Female) Knowledge 0.82 (0.20) 0.72 (0.24) 5.75 < .001 0.45 Attitude 2.80 (0.43) 3.02 (0.63) -4.99 < .001 -0.42 Social Norm 2.95 (0.68) 2.92 (0.72) 0.52 .601 0.04 PBC 3.33 (0.78) 3.30 (0.84) 0.43 .667 0.03 Plagiarism Behavior 5.04 (1.04) 4.15 (1.50) 8.41 < .001 0.73 School Type (Public vs. Private) Knowledge 0.79 (0.22) 0.80 (0.22) -0.70 .483 0.05 Attitude 2.83 (0.47) 2.95 (0.63) -1.97 .051 -0.22 Social Norm 2.93 (0.71) 2.93 (0.69) 0.00 .999 0.00 PBC 3.32 (0.80) 3.33 (0.81) -0.14 .890 -0.01 Plagiarism Behavior 4.86 (1.22) 4.50 (1.26) 2.84 .005 0.29 4.3 Machine learning-based prediction (RQ3 ) To assess the added value of machine learning in predicting plagiarism behavior, a Gradient Boosting regression model-analogous in structure to XGBoost-was trained using TPB constructs, plagiarism knowledge, and demographic predictors. The model was evaluated on a held-out test set comprising 20% of the data. It achieved an R² of .634, exceeding the hierarchical regression model’s explanatory power (R² = .604). The model’s prediction errors were within acceptable limits, with a root mean squared error (RMSE) of 0.788 and a mean absolute error (MAE) of 0.564. To evaluate generalizability, a 5-fold cross-validation procedure was conducted, yielding a mean R² of .615 (SD = .016), suggesting that the model's predictive performance was stable across folds. These results demonstrate the potential of data-driven approaches to enhance behavior prediction in educational contexts, particularly when theory-based variables are integrated with real-world demographic complexity. 4.4 Interpretability of the machine learning model (RQ4) SHAP analysis was conducted to interpret the XGBoost model and assess the relative contribution of each predictor (see Table 4 ). Perceived behavioral control (PBC) emerged as the most influential variable, with higher levels (red points) strongly associated with reduced predicted plagiarism behavior-aligning with theoretical expectations from the TPB framework (see Fig. 1 ). This was followed by subjective norms, plagiarism knowledge, and attitudes, all contributing meaningfully but with more variable effects across students. In contrast, demographic features such as gender and school type had minimal impact. SHAP dependence plots further revealed potential non-linear patterns, particularly for PBC and subjective norms, suggesting threshold effects at extreme values. Taken together, these findings demonstrate the added value of interpretable machine learning in uncovering nuanced, theory-aligned behavioral patterns not easily captured by traditional regression. Table 4 SHAP analysis: Feature importance in predicting plagiarism behavior (XGBoost model) Feature SHAP importance PBC 0.589 Social Norm 0.123 Knowledge 0.099 Attitude 0.076 Grade 0.060 Gender 0.039 School type 0.014 Analysis of the SHAP Dependence plot shows that the XGBoost model can reproduce nonlinear behavioral mechanisms that linear regression cannot capture. For the PBC variable, the results show a clear nonlinear pattern of impact. Specifically, between 1.0 and 3.5, the impact of PBC on plagiarism is almost constant, reflecting a plateau where perceived control does not yet play a decisive role. However, a tipping point is observed at 3.5: when perceived lack of control over the behavior exceeds this threshold, the SHAP value increases rapidly in the positive direction (see Fig. 2 ). This indicates that the loss of control can act as a strong trigger, significantly increasing the probability of plagiarism. In the case of Subjective Norms, the data show a more complex nonlinear dynamic. In the early stages (1.0 to 3.5), social pressure exhibits a cumulative effect: the higher the level of environmental acceptance, the higher the risk of plagiarism. However, at a threshold of approximately 3.5, this effect reaches a saturation point and begins to decrease slightly (see Fig. 3 ). Through the lens of machine learning models such as XGBoost, the results suggest that the influence of social norms is not only linear, but also has a threshold effect structure, implying the existence of a social limit in regulating misconduct. 5. Discussion This study set out to decode the complex behavioral drivers of plagiarism among adolescents in a developing digital economy. By integrating the Theory of Planned Behavior (TPB) with Explainable AI (XGBoost and SHAP), the findings offer empirical evidence that challenges prevailing assumptions about academic integrity. We argue that technology is not a neutral tool but a force that actively reshapes social norms and reveals the systemic vulnerabilities of traditional education. 5.1. The non-neutrality of technology and the knowledge paradox Our findings indicate that digital technologies are not neutral tools but actively reshape the relationship between knowledge and ethical behavior. Although the core components of the Theory of Planned Behavior remain meaningful, the emergence of what we describe as a knowledge paradox (where greater awareness of plagiarism rules is associated with a higher likelihood of misconduct) signals a troubling shift in students’ engagement with academic norms. This pattern is consistent with recent evidence reported by Huang et al. ( 2025 ) and Campo et al. ( 2025 ), which shows that students who possess a clearer understanding of academic regulations are, paradoxically, often more inclined to violate them. In the context of AI-mediated learning, knowledge appears to be used less as a guide for ethical compliance and more as a resource for strategic rule navigation. This observation echoes Akbulut et al.’s ( 2008 ) concept of “Internet-enabled academic dishonesty,” whereby technologically skilled students leverage their competence to engage in increasingly sophisticated forms of misconduct. Similarly, Bin-Nashwan et al. ( 2023 ) characterize AI as a “double-edged sword,” noting that students with high levels of digital literacy are better positioned to operate within the grey zones of AI assistance while avoiding detection by existing monitoring systems. In addition, the strong predictive role of Subjective Norms lends support to the notion of pluralistic ignorance (McCabe et al., 2001 ; Comas-Forgas & Sureda-Negre, 2010), whereby students overestimate their peers’ acceptance of cheating and, in doing so, gradually normalize such behavior. Within a digital learning environment where AI use is widespread yet weakly regulated, this misperception contributes to a self-reinforcing cycle. Practices driven by technological convenience become socially accepted, and completing academic work without AI support may increasingly be viewed as inefficient or even disadvantageous. 5.2 Contextual and Institutional Influences on Academic Misconduct Linear regression analysis results show that Perceived Behavioral Control (PBC) and negative attitudes toward plagiarism act as important protective factors. This finding is consistent with previous studies applying Theory of Planned Behavior (TPB) in the context of generative AI (GenAI). For example, Ivanov et al. ( 2024 ) asserted that PBC and attitudes are key antecedents in predicting the intention to use technology in education. Similarly, Bin-Nashwan et al. ( 2023 ) showed that academic integrity is often inversely related to the use of ChatGPT, thus acting as an ethical barrier to cheating. However, XGBoost–SHAP analysis provides a deeper insight that linear models struggle to fully reflect: student behavior does not change in a uniform manner but exhibits a distinct threshold effect. When PBC falls below a certain level, the probability of plagiarism increases sharply, indicating a rapid shift from a state of self-regulation to a feeling of loss of control. This result is consistent with studies on the motivations for using AI, where stress and time pressure are considered the main driving factors that lead users to seek out automated tools. Notably, Bin-Nashwan et al. ( 2023 ) also pointed out an important paradox: while academic integrity often limits the use of AI, under high pressure or the need to save time, even individuals with high levels of integrity tend to increase their use of these tools. Therefore, academic dishonesty should not be viewed solely as an intentional ethical violation. The results of this study, along with existing evidence, suggest that plagiarism in the context of GenAI may function as a coping strategy when students are overwhelmed. When academic demands exceed coping capabilities, AI tools offer an immediate solution, helping students regain a sense of control and maintain performance. This raises a crucial implication for education and school administration: instead of focusing solely on prohibitions or technical surveillance, more attention should be paid to the pressures of the learning environment and supporting mental health, because plagiarism in the AI ​​age may reflect systemic overload rather than individual moral decay. Besides individual factors, research findings also show that dishonest academic behavior is significantly influenced by the social and institutional context in which students participate. Specifically, significant differences by gender and educational level suggest that the degree of risk avoidance and sensitivity to school expectations are not uniform across demographic groups. This suggests that academic ethical norms are not received and internalized in the same way across all students but are influenced by different experiences and value orientations during the learning process. Notably, the observed differences between public and private schools highlight the role of organizational culture in shaping learner behavior. This finding is consistent with the argument of Ellis and Murdoch ( 2024 ), who argue that the institutional climate and how schools convey academic values ​​can directly influence how students interpret and practice ethical norms. In this context, the “Knowledge Paradox” is not merely a matter of individual perception or competence, but also a consequence of the learning environment where institutional signals about performance, competition, and compliance may inadvertently encourage the instrumental use of AI. In other words, the organizational climate can act as a moderator, either undermining or exacerbating the paradoxical relationship between awareness of plagiarism and dishonest behavior in the AI ​​age. 5.3 Policy implications Overall, the findings point to the need for a clear shift in educational policy: from restrictive methods to more adaptive and systematic forms of governance. First, the research results suggest that reforming assessment should be prioritized over expanding supervision. The existence of the analytical paradoxes above indicates that awareness campaigns and rule clarification alone are insufficient to prevent misconduct. Instead of increasing supervision or detection technology, educational institutions should reconsider how they design assessments. As Banks et al. ( 2024 ) and Dubljević ( 2024 ) argued, there should be a greater focus on learning processes, reflection, and repetitive work, rather than narrowly evaluating final outcomes. Such reforms could reduce the incentive to misuse AI while making assessment activities more aligned with genuine learning objectives. Secondly, instead of simply enforcing norms, secondary schools need to pay more attention to developing digital competencies. Beyond simply communicating rules, schools need to actively cultivate students' digital competencies. Jeon et al. ( 2025 ) emphasize that ethical behavior must be learned through practice, not instruction. In this regard, policies need to clearly state acceptable forms of human-AI collaboration and reflect these expectations in daily teaching and assessment. In this way, technical skills that could otherwise be used to circumvent regulations can be redirected toward the responsible and transparent use of AI, forming a more meaningful model of ethical digital competence. Finally, policymakers must confront the growing tension between efficiency and equity in education systems increasingly shaped by AI (Lim & Lee, 2025). If left unaddressed, integrating AI into the learning process could inadvertently reward students who prioritize speed and efficiency, while putting those who strictly adhere to academic rules at a disadvantage. Therefore, effective governance frameworks need to ensure that ethical compliance is not punished and that the benefits of AI-assisted learning are distributed in a way that remains socially equitable and legitimate. 5.4 Limitations and directions for future research Although the study provides important insights into the predictors of plagiarism among secondary and high school students in Vietnam, some limitations should be considered in interpreting the results. First, the scope of the model mainly focused on the components of the Theory of Planned Behavior (TPB) and the level of knowledge about plagiarism. Future studies should expand the analytical framework by adding other important variables such as moral disengagement, peer group norms, academic pressure, digital competence, or institutional climate. This would help to build a more comprehensive explanatory model of dishonest behavior in the context of digital learning. Second, the study was conducted in the Vietnamese educational context, where cultural factors such as hierarchical teacher-student relationships, strong exam orientation, and collectivist values ​​have a significant influence on students’ behavioral motivation and moral reasoning. These factors may make the results difficult to compare directly to Western educational systems or other cultural contexts. Therefore, cross-cultural comparison studies are needed to test the generalizability of the model. Third, the cross-sectional design of the study limits the ability to draw causal relationships. Future longitudinal studies are needed to track how attitudes, subjective norms, PBC, and knowledge change over time, and how they relate to actual plagiarism. Finally, self-reported data may be subject to social desirability bias or misunderstanding of plagiarism-related terminology. Although the questionnaire was clearly designed, future studies should adopt a multi-method approach, including behavioral data (e.g., plagiarism detection results), teacher or peer ratings, to enhance reliability and construct validity. Thus, future research directions should expand theoretical scope, implement cross-cultural comparisons, apply longitudinal methods, and integrate multiple data sources to more fully capture the complexity of academic dishonesty in diverse educational contexts. 6. Conclusion This study helps to better understand why high school students may engage in plagiarism in a digital learning environment. The results show that behavioral control, peer pressure, and personal attitudes remain important factors. However, a notable finding is that understanding plagiarism rules does not necessarily lead to better compliance; in some cases, it even leads to higher rates of violation. This suggests that in a digital context, knowledge is sometimes used to circumvent rules rather than comply with them. Furthermore, data-driven analyses indicate that students' behavior does not change gradually. When their sense of control drops below a certain level, the likelihood of violation increases rapidly. This suggests that technology-related plagiarism stems not only from personal choice but also from a reaction to academic pressure and overload. These findings suggest that simply reminding students of regulations or increasing control is insufficient. Education needs to focus more on helping students learn to self-regulate, learn responsibly when using AI, and understand social norms in the digital environment. Furthermore, the differences between student groups and school types indicate that solutions need to be flexible and tailored to each specific context. Declarations Funding delaration This research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number NCUD.05-2022.21. Ethic Delaration : This study was conducted in accordance with the principles of the Declaration of Helsinki. The research protocol was approved by the Committee of the National Science and Technology Development Fund (Approval Number: NCUD. 05-2022.21). Author Contribution Tinh T.T. Le: Conceptualization, Methodology, Data curation, Formal analysis, Writing – original draft.Doanh V.Q. Ho: Data collection, Writing – review & editing.Hung V. Tran: Methodology, Formal analysis, Writing – review & editing. Acknowledgement This research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number NCUD.05-2022.21 Data Availability Data will be made available on request. References Aamir Raoof Memon (2020) Knowledge, Attitudes, and Practices of Plagiarism as Reported by Participants Completing the AuthourAID MOOC on Research Writing , Science and Engineeing Ethics. 42 Ahmed S, Kaiser MS, Hossain S, M., Andersson K (2025) A Comparative Analysis of LIME and SHAP Interpreters With Explainable ML-Based Diabetes Predictions. IEEE Access 13:37370–37388. https://doi.org/10.1109/ACCESS.2024.3422319 Ajzen I (1991a) The theory of planned behavior. Organ Behav Hum Decis Process 50(2):179–211. https://doi.org/10.1016/0749-5978(91)90020-T Ajzen I (1991b) The theory of planned behavior. Organ Behav Hum Decis Process 50(2):179–211. https://doi.org/10.1016/0749-5978(91)90020-T Akbulut Y et al (2008) Exploring the types and reasons of Internet-triggered academic dishonesty among Turkish undergraduate students: Development of Internet-Triggered Academic Dishonesty Scale (ITADS). Comput Educ 51(1):463–473. https://doi.org/10.1016/j.compedu.2007.06.003 Ali O, Murray PA, Momin M, Al-Anzi FS (2023) The knowledge and innovation challenges of ChatGPT: A scoping review. Technol Soc 75:102402 Alsabhan W (2023) Student Cheating Detection in Higher Education by Implementing Machine Learning and LSTM Techniques. Sensors 23(8):4149. https://doi.org/10.3390/s23084149 Ari Santoso & Fitra Roman Cahaya (2018) Factors influencing plagiarism by accounting lectures. Acc Educ. https://doi.org/10.1080/09639284.2018.1523736 Banks D, Bosone C, Carpenter B, Shah T, Shi C (2024) Large Language Models: Trust and Regulation. Harv Data Sci Rev, 6(3) Barry D, Smith (1998) Psychology: Science & Understanding. McGraw-Hill Beck L, Ajzen I (1991) Predicting dishonest actions using the theory of planned behavior. J Res Pers 25(3):285–301. https://doi.org/10.1016/0092-6566(91)90021-H Bin-Nashwan SA, Sadallah M, Bouteraa M (2023) Use of ChatGPT in academia: Academic integrity hangs in the balance. Technol Soc 75:102370 Bin-Nashwan SA, Sadallah M, Bouteraa M (2023) Use of ChatGPT in academia: Academic integrity hangs in the balance. Technol Soc 75:102370. https://doi.org/10.1016/j.techsoc.2023.102370 Birks D, Clare J (2023a) Linking artificial intelligence facilitated academic misconduct to existing prevention frameworks. Int J Educational Integr 19(1):20. https://doi.org/10.1007/s40979-023-00142-3 Birks D, Clare J (2023b) Linking artificial intelligence facilitated academic misconduct to existing prevention frameworks. Int J Educational Integr 19(1):20. https://doi.org/10.1007/s40979-023-00142-3 Campo L, Delgado N, Etxabe-Urbieta JM, Kanso H, Gueye ML (2025) Relationship Between the Use of ChatGPT and Plagiarism in Higher Education: The Influence of Gender, Age and Previous Academic Results. Innovative Higher Education . https://doi.org/10.1007/s10755-025-09830-z Çelik Ö, Razı S (2023) Facilitators and barriers to creating a culture of academic integrity at secondary schools: an exploratory case study. Int J Educational Integr 19(1):4. https://doi.org/10.1007/s40979-023-00125-4 Chan CKY (2025) Students’ perceptions of ‘AI-giarism’: investigating changes in understandings of academic misconduct. Educ Inform Technol 30(6):8087–8108. https://doi.org/10.1007/s10639-024-13151-7 Choi Y, Suh KH (2022) Verifying the usefulness of the theory of planned behavior model for predicting illegal use of online content: the role of outcome expectancies and social loafing. BMC Psychol 10(1):263. https://doi.org/10.1186/s40359-022-00978-3 Clarke O, Chan WYD, Bukuru S, Logan J, Wong R (2023) Assessing knowledge of and attitudes towards plagiarism and ability to recognize plagiaristic writing among university students in Rwanda. High Educ 85(2):247–263. https://doi.org/10.1007/s10734-022-00830-y Cotton DRE, Cotton PA, Shipway JR (2024) Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations Educ Teach Int 61(2):228–239. https://doi.org/10.1080/14703297.2023.2190148 Curtis GJ, Vardanega L (2016) Is plagiarism changing over time? A 10-year time-lag study with three points of measurement. High Educ Res Dev 35(6):1167–1179. https://doi.org/10.1080/07294360.2016.1161602 Curtis GJ, Cowcher E, Greene BR, Rundle K, Paull M, Davis MC (2018a) Self-Control, Injunctive Norms, and Descriptive Norms Predict Engagement in Plagiarism in a Theory of Planned Behavior Model. J Acad Ethics 16(3):225–239. https://doi.org/10.1007/s10805-018-9309-2 Curtis GJ, Cowcher E, Greene BR, Rundle K, Paull M, Davis MC (2018b) Self-Control, Injunctive Norms, and Descriptive Norms Predict Engagement in Plagiarism in a Theory of Planned Behavior Model. J Acad Ethics 16(3):225–239. https://doi.org/10.1007/s10805-018-9309-2 Deepti M (2020) Factors Influencing Higher Similarity and Plagiarism amongst Research Scholars: A Comparative Case Study of two Indian Universities - Centers of Higher Education , Library Philosophy and Practice (e-journal), 4287. https://digitalcommons.unl.edu/libphilprac/4287 Désiron JC, Petko D (2023) Academic dishonesty when doing homework: How digital technologies are put to bad use in secondary schools. Educ Inform Technol 28(2):1251–1271. https://doi.org/10.1007/s10639-022-11225-y Drew W (1996) Psychology: Mind, Brain & Culture. John Wiley & Sons, Inc. Dubljević V (2024) Colleges and universities are important stakeholders for regulating large language models and other emerging AI. Technol Soc 77:102480 Phyo EM, Lwin T, Tun HP Zaw Zaw Oo, Kyaw Swa Mya & Henry Silverman (2023), Knowledge, attitudes, and practices regarding plagiarism of postgraduate students in Myanmar . Account Res, 43 https://doi.org/10.1080/08989621.2022.2077643 Ellis C, Murdoch K (2024) The educational integrity enforcement pyramid: a new framework for challenging and responding to student cheating. Assess Evaluation High Educ 49(7):924–934. https://doi.org/10.1080/02602938.2024.2329167 Eret E, Ok A (2014) Internet plagiarism in higher education: tendencies, triggering factors and reasons among teacher candidates. Assess Evaluation High Educ 39(8):1002–1016. https://doi.org/10.1080/02602938.2014.880776 Feldman-Maggor Y, Cukurova M, Kent C, Alexandron G (2025) The Impact of Explainable AI on Teachers’ Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations. Int J Artif Intell Educ. https://doi.org/10.1007/s40593-025-00486-6 Foltynek T, Bjelobaba S, Glendinning I, Khan ZR, Santos R, Pavletic P, Kravjar J (2023) ENAI Recommendations on the ethical use of Artificial Intelligence in Education. Int J Educational Integr 19(1):12. https://doi.org/10.1007/s40979-023-00133-4 Ghimire S, Abdulla S, Joseph LP, Prasad S, Murphy A, Devi A, Barua PD, Deo RC, Acharya R, Yaseen ZM (2024) Explainable artificial intelligence-machine learning models to estimate overall scores in tertiary preparatory general science course. Computers Education: Artif Intell 7:100331. https://doi.org/10.1016/j.caeai.2024.100331 Guleria P, Sood M (2023) Explainable AI and machine learning: performance evaluation and explainability of classifiers on educational data mining inspired career counseling. Educ Inform Technol 28(1):1081–1116. https://doi.org/10.1007/s10639-022-11221-2 Hou S, Zhang X, Yi B, Tang Y (2022) Public attitudes on open source communities in China: A text mining analysis. Technol Soc 71:102112 Huang CL, Chen Y, Zhang S, Yang SC (2025) Is copy and paste part of academic misconduct? The roles of attitude, experience and self-efficacy in judgment. Contemp Educ Psychol 82:102402. https://doi.org/10.1016/j.cedpsych.2025.102402 Icek Ajzen (1991) The Theory of Planned Behavior. Organ Behav Hum Decis Process, 50(2) Irani F, Muotka J, Lyyra P, Parviainen T, Monto S (2024) Social influence in adolescence: Behavioral and neural responses to peer and expert opinion. Soc Neurosci 19(1):25–36. https://doi.org/10.1080/17470919.2024.2323745 Ismail K (2018) Perceptions of Plagiarism Among Medical and Nursing Students in Erbill, Iraq. Sultan Qaboos Univ Med J. https://doi.org/10.18295/squmj.2018.18.02.012 Ivanov S, Soliman M, Tuomi A, Alkathiri NA, Al-Alawi AN (2024) Drivers of generative AI adoption in higher education through the lens of the Theory of Planned Behaviour. Technol Soc 77:102521 Jeon J, Kim L, Park J (2025) The ethics of generative AI in social science research: A qualitative approach for institutionally grounded AI research ethics. Technol Soc 102836. https://doi.org/10.1016/j.techsoc.2025.102836 Johansen MW, Goddiksen MP, Centa M, Clavien C, Gefenas E, Globokar R, Hogan L, Merit MT, Nielsen SS, Olsson IAS, Poškutė M, Quinn U, Santos JB, Santos R, Schöpfer C, Strahovnik V, Wall PJ, Sandøe P, Lund TB (2022a) Lack of ethics or lack of knowledge? European upper secondary students’ doubts and misconceptions about integrity issues. Int J Educational Integr 18(1):20. https://doi.org/10.1007/s40979-022-00113-0 Johansen MW, Goddiksen MP, Centa M, Clavien C, Gefenas E, Globokar R, Hogan L, Merit MT, Nielsen SS, Olsson IAS, Poškutė M, Quinn U, Santos JB, Santos R, Schöpfer C, Strahovnik V, Wall PJ, Sandøe P, Lund TB (2022b) Lack of ethics or lack of knowledge? European upper secondary students’ doubts and misconceptions about integrity issues. Int J Educational Integr 18(1):20. https://doi.org/10.1007/s40979-022-00113-0 Johnston H, Wells RF, Shanks EM, Boey T, Parsons BN (2024) Student perspectives on the use of generative artificial intelligence technologies in higher education. Int J Educational Integr 20(1):2. https://doi.org/10.1007/s40979-024-00149-4 Johora FT, Hasan MN, Rajbongshi A, Ashrafuzzaman M, Akter F (2025) An explainable AI-based approach for predicting undergraduate students academic performance. Array 26:100384. https://doi.org/10.1016/j.array.2025.100384 Juan LX, Tao WY, Veloo PK, Supramaniam M (2022) Using Extended TPB Models to Predict Dishonest Academic Behaviors of Undergraduates in a Chinese Public University. Sage Open 12(4). https://doi.org/10.1177/21582440221140391 Kam CCS, Hue MT, Cheung HY (2018) Academic dishonesty among Hong Kong secondary school students: application of theory of planned behaviour. Educational Psychol 38(7):945–963. https://doi.org/10.1080/01443410.2018.1454588 Kemper L, Vorhoff G, Wigger BU (2020) Predicting student dropout: A machine learning approach. Eur J High Educ 10(1):28–47. https://doi.org/10.1080/21568235.2020.1718520 Leaton Gray S, Edsall D, Parapadakis D (2025) AI-Based Digital Cheating At University, and the Case for New Ethical Pedagogies. J Acad Ethics 23(4):2069–2086. https://doi.org/10.1007/s10805-025-09642-y Li M, Patel J, Katapally TR (2025) Towards ethical surveillance of smartphone use among youth: exploratory digital citizen science approaches shaping the understanding of ubiquitous technology use. Technol Soc 83:103012 Liddell J (2003) A Comprehensive Definition of Plagiarism, Community and Junior College Lim WM, Gunasekara A, Pallant JL, Pallant JI, Pechenkina E (2023) Generative AI and the future of education: Ragnarök or reformation? A paradoxical perspective from management educators. Int J Manage Educ 21(2):100790. https://doi.org/10.1016/j.ijme.2023.100790 Linardatos P, Papastefanopoulos V, Kotsiantis S (2020) Explainable AI: A Review of Machine Learning Interpretability Methods. Entropy 23(1):18. https://doi.org/10.3390/e23010018 Lu S-C, Swisher CL, Chung C, Jaffray D, Sidey-Gibbons C (2023) On the importance of interpretable machine learning predictions to inform clinical decision making in oncology. Frontiers in Oncology , 13 . https://doi.org/10.3389/fonc.2023.1129380 Macdonald R, Carroll J (2006) Plagiarism-a complex issue requiring a holistic institutional approach. Assess Evaluation High Educ 31(2):233–245. https://doi.org/10.1080/02602930500262536 Malti T, Galarneau E, Peplak J (2021) Moral Development in Adolescence. J Res Adolescence 31(4):1097–1113. https://doi.org/10.1111/jora.12639 Mavrinac M, Brumini G, Bilić-Zulle L, Petrovečki M (2010) Construction and Validation of Attitudes Toward Plagiarism Questionnaire. Croatian Med J 51(3):195–201. https://doi.org/10.3325/cmj.2010.51.195 Maxwell A, Curtis GJ, Vardanega L (2008) Does culture influence understanding and perceived seriousness of plagiarism? Int J Educational Integr 4(2). https://doi.org/10.21913/IJEI.v4i2.412 McCabe DL, Trevino LK (1997) Individual and Contextual Influences on Academic Dishonesty: A Multicampus Investigation. Res High Educt 38(3):379–396. https://doi.org/10.1023/A:1024954224675 McCabe DL, Trevino LK, Butterfield KD (2001) Cheating in Academic Institutions: A Decade of Research. Ethics Behav 11(3):219–232. https://doi.org/10.1207/S15327019EB1103_2 McCabe DL, Treviño LK, Butterfield KD (2002) Honor Codes and Other Contextual Influences on Academic Integrity: A Replication and Extension to Modified Honor Code Settings. Res High Educt 43(3):357–378. https://doi.org/10.1023/A:1014893102151 Molnar C, Casalicchio G, Bischl B (2020) Interpretable Machine Learning - A Brief History, State-of-the-Art and Challenges (pp. 417–431). https://doi.org/10.1007/978-3-030-65965-3_28 Nazia S (2023) Factors Influencing among Faculty members and Research scholars in The Higher Education System. DESIDOC Journal of Library & Information Technology. https://digitalcommons.unl.edu/libphilprac/4287 Nguyen HM, Goto D (2024) Unmasking academic cheating behavior in the artificial intelligence era: Evidence from Vietnamese undergraduates. Educ Inform Technol. https://doi.org/10.1007/s10639-024-12495-4 Njuguna GN, Qingfei M (2026) How would prompt completion editing impact user experience scores in academic research with large language models? Technol Soc 84:103080 Nora WLY, Zhang KC (2010) Motives of cheating among secondary students: the role of self-efficacy and peer influence. Asia Pac Educ Rev 11(4):573–584. https://doi.org/10.1007/s12564-010-9104-2 Ortiz-Bonnin S, Blahopoulou J (2025) Chat or cheat? Academic dishonesty, risk perceptions, and ChatGPT usage in higher education students. Soc Psychol Educ 28(1):113. https://doi.org/10.1007/s11218-025-10080-2 Perry AH, Rettinger DA, Stephens JM, Anderman EM, McTernan ML, Tatum H, McNally D, Cullen C, Gallant B, T (2025) From institutional climate to moral attitudes: examining theoretical models of academic misconduct. Ethics Behav 1–18. https://doi.org/10.1080/10508422.2025.2514577 Phyo EM, Lwin T, Pyae Tun H, Zaw Zaw O, Swa Mya K, Silverman H (2023) Knowledge, Attitudes, and Practices Regarding Plagiarism of Postgraduate Students in Myanmar , Accountability in Research. https://doi.org/10.1080/08989621.2022.2077643 , Tr. 672–691 Pike RK, Buck LA, Tsoukkas E, Bell E (2025) From collaboration to contract cheating: exploring staff and student perceptions of the grey areas of academic outsourcing. Assess Evaluation High Educ 1–21. https://doi.org/10.1080/02602938.2025.2539291 Polona Sprajc (2017) Reasons for Plagiarism in Higher Education, De Gruyter , Tr. 33–45 Prashar A, Gupta P, Dwivedi YK (2024) Plagiarism awareness efforts, students’ ethical judgment and behaviors: a longitudinal experiment study on ethical nuances of plagiarism in higher education. Stud High Educ 49(6):929–955. https://doi.org/10.1080/03075079.2023.2253835 Rettinger DA, Kramer Y (2009) Situational and Personal Causes of Student Cheating. Res High Educt 50(3):293–313. https://doi.org/10.1007/s11162-008-9116-5 Ronald W, Belter (2009) A Strategy to reduce Plagiarism in an Undergraduate Course. Taylor & Francis Group, Tr, pp 257–261 Sisti DA (2007) How Do High School Students Justify Internet Plagiarism? Ethics Behav 17(3):215–231. https://doi.org/10.1080/10508420701519163 Stoesz BM, Yudintseva A (2018) Effectiveness of tutorials for promoting educational integrity: a synthesis paper. Int J Educational Integr 14(1):6. https://doi.org/10.1007/s40979-018-0030-0 Stone TH, Jawahar IM, Kisamore JL (2009a) Using the theory of planned behavior and cheating justifications to predict academic misconduct. Career Dev Int 14(3):221–241. https://doi.org/10.1108/13620430910966415 Stone TH, Jawahar IM, Kisamore JL (2009b) Using the theory of planned behavior and cheating justifications to predict academic misconduct. Career Dev Int 14(3):221–241. https://doi.org/10.1108/13620430910966415 Stone TH, Jawahar IM, Kisamore JL (2009c) Using the theory of planned behavior and cheating justifications to predict academic misconduct. Career Dev Int 14(3):221–241. https://doi.org/10.1108/13620430910966415 Stone TH, Jawahar IM, Kisamore JL (2010) Predicting Academic Misconduct Intentions and Behavior Using the Theory of Planned Behavior and Personality. Basic Appl Soc Psychol 32(1):35–45. https://doi.org/10.1080/01973530903539895 Swamy V, Du S, Marras M, Kaser T (2023) Trusting the Explainers: Teacher Validation of Explainable Artificial Intelligence for Course Design. LAK23: 13th International Learning Analytics and Knowledge Conference , 345–356. https://doi.org/10.1145/3576050.3576147 Tindall IK, Fu KW, Tremayne K, Curtis GJ (2021) Can negative emotions increase students’ plagiarism and cheating? Int J Educational Integr 17(1):25. https://doi.org/10.1007/s40979-021-00093-7 Towler G, Shepherd R (1991) Modification of Fishbein and Ajzen’s theory of reasoned action to predict chip consumption. Food Qual Prefer 3(1):37–45. https://doi.org/10.1016/0950-3293(91)90021-6 Uzun AM, Kilis S (2020a) Investigating antecedents of plagiarism using extended theory of planned behavior. Comput Educ 144:103700. https://doi.org/10.1016/j.compedu.2019.103700 Uzun AM, Kilis S (2020b) Investigating antecedents of plagiarism using extended theory of planned behavior. Comput Educ 144:103700. https://doi.org/10.1016/j.compedu.2019.103700 Villar A, de Andrade CRV (2024) Supervised machine learning algorithms for predicting student dropout and academic success: a comparative study. Discover Artif Intell 4(1):2. https://doi.org/10.1007/s44163-023-00079-z Wang M, Yang Y, Zhong P (2025) How does AI affect the self-actualization of content creators in dynamic environments? A knowledge management perspective. Technol Soc 102855. https://doi.org/10.1016/j.techsoc.2025.102855 Watrianthos R, Ahmad ST, Muskhir M (2023) Charting the growth and structure of early ChatGPT-education research: A bibliometric study. J Inform Technol Education: Innovations Pract 22:235–253 Zeng K, Wang Z, Lu T, Chen J, Han Z (2023) Implicit space pose consistent transfer network for deep face verification. Pattern Recognit Lett 176:1–6. https://doi.org/10.1016/j.patrec.2023.10.017 Zhang Y (2024) Academic cheating as planned behavior: the effects of perceived behavioral control and individualism-collectivism orientations. High Educ 87(3):567–590. https://doi.org/10.1007/s10734-023-01024-w Zhao L, Mao H, Compton BJ, Peng J, Fu G, Fang F, Heyman GD, Lee K (2022a) Academic dishonesty and its relations to peer cheating and culture: A meta-analysis of the perceived peer cheating effect. Educational Res Rev 36:100455. https://doi.org/10.1016/j.edurev.2022.100455 Additional Declarations No competing interests reported. Supplementary Files AppendixA.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 14 May, 2026 Reviews received at journal 23 Apr, 2026 Reviewers agreed at journal 28 Mar, 2026 Reviewers agreed at journal 04 Mar, 2026 Reviewers invited by journal 27 Feb, 2026 Editor assigned by journal 09 Feb, 2026 Submission checks completed at journal 19 Jan, 2026 First submitted to journal 18 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8458923","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":600204051,"identity":"dacca2d9-307e-44e4-a7c0-69bf36fa28a3","order_by":0,"name":"Tinh Le","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuElEQVRIiWNgGAWjYBACPhCRwHBADkgZMDCwEaGFjY2ZsQGoxZhELQwMBxIbiNci33/8wcMdd9I3HG/ewMxTdofB4HYDEbYknnmWu+HMsQJmnnPPGAzuHCBGS9vh3A03cgyYedsOM0jOSCBOS7rB/TckakkwuMED0cIvQVBLsuEMoBbDmWfSCg7OOXeYh6AWfuaDDz7+bDssz3f88MYHb8oOy7ER0oICDgAxDwnqR8EoGAWjYBTgAgA+oELUvUZN/wAAAABJRU5ErkJggg==","orcid":"","institution":"The University of Da Nang, University of Science and Education","correspondingAuthor":true,"prefix":"","firstName":"Tinh","middleName":"","lastName":"Le","suffix":""},{"id":600204052,"identity":"d0a88735-4f49-4de1-a263-8adf3f9d39c7","order_by":1,"name":"Doanh Ho","email":"","orcid":"","institution":"The University of Da Nang, University of Science and Education","correspondingAuthor":false,"prefix":"","firstName":"Doanh","middleName":"","lastName":"Ho","suffix":""},{"id":600204053,"identity":"25693ba7-a110-4b62-827a-88e561065ecf","order_by":2,"name":"Hung Tran","email":"","orcid":"","institution":"The University of Da Nang, University of Science and Education","correspondingAuthor":false,"prefix":"","firstName":"Hung","middleName":"","lastName":"Tran","suffix":""}],"badges":[],"createdAt":"2025-12-27 06:08:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8458923/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8458923/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104401770,"identity":"db35f90c-f794-4df8-8d67-cf85f5ee20f3","added_by":"auto","created_at":"2026-03-11 12:13:30","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":49597,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP sumary flot (factors impact on plagiarism behavior)\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8458923/v1/c956f83cfe4ff554da02d199.png"},{"id":103939287,"identity":"b31fc606-3c93-4b8f-8b4d-22ba6e303fe5","added_by":"auto","created_at":"2026-03-04 18:42:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":85605,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP dependence plot: PBC\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8458923/v1/5d52082d305bcaa84451258f.png"},{"id":103939289,"identity":"43b6abdd-71ee-48f6-abff-6f0d3c132a94","added_by":"auto","created_at":"2026-03-04 18:42:18","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":84016,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP dependence plot: Subjective norms\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8458923/v1/fd3c1ac9b528bfab1db44609.png"},{"id":104408353,"identity":"00debfc0-1825-4691-832b-697825eeeb28","added_by":"auto","created_at":"2026-03-11 12:42:16","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1400397,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8458923/v1/5eeef5c5-7a3f-4edd-971a-1f8b26114882.pdf"},{"id":104401880,"identity":"6c9120b8-b58d-4a1c-a46c-d5b8a2543c65","added_by":"auto","created_at":"2026-03-11 12:13:48","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":16615,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixA.docx","url":"https://assets-eu.researchsquare.com/files/rs-8458923/v1/8e9bc087628fec9bce6683ad.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Vietnamese High School Students in the AI Era: Examining the Control Threshold and Knowledge Paradox in Academic Integrity","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eTurnitin\u0026rsquo;s 2024 report, based on the analysis of 200\u0026nbsp;million submissions worldwide, identified more than 22\u0026nbsp;million documents showing substantial AI involvement, underscoring how deeply AI has penetrated contemporary educational practices (Turnitin, 2024). This evidence points to a worrying rise in technology-supported academic dishonesty, challenging long-standing notions of authorship and merit in the digital era. As teaching and learning increasingly move into digital spaces, secondary school students are now routinely exposed to AI writing tools and extensive online information sources. Although these technologies can enhance access to knowledge and support more efficient writing processes (Banks et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), they also complicate students\u0026rsquo; understanding of originality and academic integrity (Foltynek et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Lim et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). In this sense, generative AI functions as a \u0026ldquo;double-edged sword\u0026rdquo;, making the boundary between acceptable support and unethical practice increasingly unclear and more difficult to monitor (Bin-Nashwan et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eEmpirical evidence further illustrates the seriousness of this issue. Studies suggest that up to 43% of high school students admit to some form of cheating behavior (Birks \u0026amp; Clare, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2023a\u003c/span\u003e; Mavrinac et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Despite this, most existing research has focused primarily on higher education, as shown in recent bibliometric reviews (Watrianthos et al., \u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Cotton et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This imbalance leaves a notable gap in our understanding of academic integrity during adolescence, a critical developmental stage when learning habits, values, and moral reasoning begin to stabilize (Chan, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Macdonald \u0026amp; Carroll, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). During this period, the use of AI as a form of \u0026ldquo;cognitive shortcut\u0026rdquo; may contribute to the normalization of plagiarism, particularly when students hold incomplete or inaccurate views of intellectual ownership in open digital environments (Birks \u0026amp; Clare, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023b\u003c/span\u003e; Nguyen \u0026amp; Goto, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Eret \u0026amp; Ok, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFrom a theoretical perspective, the Theory of Planned Behavior (TPB) (Ajzen, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e1991a\u003c/span\u003e) offers a useful framework for examining such behaviors, emphasizing the role of attitudes, perceived social expectations, and perceived behavioral control. Recent studies on technology use suggest that perceived benefits often outweigh ethical concerns when individuals decide whether to adopt new tools (Ivanov et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). While TPB has demonstrated strong predictive value across various academic integrity studies (Stone et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2009a\u003c/span\u003e; Uzun \u0026amp; Kilis, \u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e2020a\u003c/span\u003e), applying the model directly to secondary education can overlook an important factor: students\u0026rsquo; knowledge of plagiarism. Limited or incorrect understanding of academic rules may lead to accidental misconduct, or alternatively, to more deliberate attempts to bypass detection-dimensions that traditional intention-based models often fail to address adequately (Johansen et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e; Maxwell et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBeyond conceptual concerns, methodological issues also constrain existing research. Much of the TPB literature relies on self-reported data and linear statistical models. However, recent findings indicate that adolescents frequently misestimate their own technology use, calling into question the accuracy of self-report measures (Li et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). In addition, linear models may oversimplify students\u0026rsquo; decision-making processes in technology-rich environments, where behaviors are shaped by complex interactions and threshold effects (Linardatos et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). In response to these challenges, machine learning techniques such as XGBoost have demonstrated strong potential for identifying subtle patterns and improving predictive performance (Feldman-Maggor et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Nevertheless, the limited transparency of such models remains a concern in educational and social research, where understanding underlying mechanisms is as important as prediction accuracy (Swamy et al., \u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRecent advances in Explainable AI (XAI), particularly the SHAP (SHapley Additive exPlanations) method, provide a promising way forward by clarifying how individual variables contribute to model outcomes (Zeng et al., \u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). When used alongside established theoretical frameworks like TPB, SHAP helps connect predictive strength with meaningful interpretation. Building on this approach, the present study proposes an integrated framework that extends TPB by incorporating knowledge-related variables, employs hierarchical regression to test theory-driven hypotheses, and applies XGBoost with SHAP to capture non-linear relationships. Rather than focusing solely on predicting plagiarism behavior, the study aims to uncover the cognitive and contextual factors that shape students\u0026rsquo; decisions, thereby offering insights to inform more effective academic integrity policies in the age of AI.\u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Plagiarism behavior in K-12 education\u003c/h2\u003e \u003cp\u003eWhile academic dishonesty has been extensively explored in higher education, its prevalence and determinants in secondary schools are not fully understood and reported results are mixed. This gap is particularly notable because adolescence is a critical period in the development of academic ethics and behavioral norms (Macdonald \u0026amp; Carroll, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). Recent studies have shown that plagiarism is already evident in this period. A review by D\u0026eacute;siron \u0026amp; Petko (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) found that approximately 43 percent of secondary school students admitted to plagiarism and that most did not recognize that it was an ethical violation. Results from Stoesz \u0026amp; Yudintseva (\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) point to a similar conclusion, with more than a third of European students reporting copying or paraphrasing without citing. However, most existing studies are still descriptive or methodologically limited, relying heavily on self-reports and lacking a strong theoretical foundation (Nora \u0026amp; Zhang, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Johansen et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2022b\u003c/span\u003e). The number of studies that apply rigorous behavioral frameworks or computational methods to model the complex antecedents of cheating behavior is still very small. At the same time, as \u0026Ccedil;elik \u0026amp; Razı (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) note, many students have difficulty distinguishing between academic collaboration, resource use, and plagiarism, while their ability to think ethically is not fully developed and their training in academic skills is limited.\u003c/p\u003e \u003cp\u003eTo move the field further, three interrelated aspects need to be considered. The first aspect relates to the diversity of misconduct. Plagiarism in the K\u0026ndash;12 environment does not take a single form but ranges from verbatim copying and patchwork plagiarism to contract cheating and unauthorized forms of collusion (Pike et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). It is important to note that students\u0026rsquo; motivations, whether opportunistic, unintentional, or driven by external pressure, often determine whether they perceive the behavior as a violation (Zhao et al., \u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe second aspect relates to developmental factors. At this age, metacognitive maturity is limited, self-regulation is not yet established, and peer pressure can be strong. These factors make students more vulnerable to dishonest behavior than older age groups (Sisti, \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Stone et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Results from the OECD PISA 2022 assessment also show an increase in the number of rule violations in online tests, reflecting that the digital environment is increasing both the opportunity and the temptation for misconduct.\u003c/p\u003e \u003cp\u003eA third aspect has come to the fore in recent years. Technologies powered by artificial intelligence are blurring the line between authorship and technical assistance. Tools such as automated paraphrasing software, summarization bots, or large language models such as ChatGPT make it harder to determine the originality of learning products. Studies by Johnston et al. (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and Ortiz Bonnin \u0026amp; Blahopoulou (2025) show that some students view AI-generated content as a form of scaffolding rather than plagiarism, especially in contexts where teachers have not provided clear guidance on the use of these tools.\u003c/p\u003e \u003cp\u003eDespite the growing complexity of these issues, the number of studies that combine these three dimensions into a unified behavioral model is still very small. Work that operationalizes them using advanced predictive techniques that go beyond linear regression is even rarer. The application of foundational theories such as the Theory of Planned Behavior remains fragmented and rarely calibrated to the ecological context of high school. In addition, existing research has not yet clearly explained how digital competence and perceived norms interact with personal beliefs to shape cheating intentions. To fill this gap, the current study extends the TPB by adding the variable knowledge about plagiarism, which has been shown to moderate the gap between intentions and behavior but has often been overlooked in previous studies (Juan et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). At the same time, we use the state-of-the-art XGBoost machine learning model combined with the SHAP interpretation algorithm to uncover non-linear relationships and variable interactions that are difficult to identify with traditional analysis methods. This approach is both explanatory and predictive, based on a behavioral framework adapted to the digital learning context of today's high school students.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 The Theory of Planned Behavior and academic dishonesty\u003c/h2\u003e \u003cp\u003eThe Theory of Planned Behavior (TPB) (Ajzen, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1991b\u003c/span\u003e) has long provided a central framework for understanding academic misconduct, particularly cheating and plagiarism. At its core, TPB proposes that behavioral intentions, shaped by attitudes, subjective norms, and perceived behavioral control (PBC), are the most proximate predictors of actual behavior. While its predictive validity has been consistently demonstrated in higher education (Curtis et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018b\u003c/span\u003e; Zhang, \u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), the application of TPB in secondary education remains limited and, in many cases, under-theorized (Choi \u0026amp; Suh, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis gap is notable given the socio-cognitive peculiarities of adolescents. High school students, unlike their college-age peers, are still in the process of consolidating their moral identities and tend to be more responsive to external influences (e.g., peer behavior, teacher expectations, and institutional cues) when forming moral judgments (Malti et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Irani et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Therefore, applying the TPB in a high school context requires careful attention to how school culture and local socialization processes shape subjective norms. For example, normative beliefs about \u0026ldquo;copying\u0026rdquo; can vary significantly across school districts or geographic areas (Zhao et al., 2022b), and institutional enforcement practices can disproportionately influence students\u0026rsquo; perceptions of their ability to control their own behavior.\u003c/p\u003e \u003cp\u003eThe original TPB formulation also offers limited guidance for understanding behavior in AI-mediated learning environments, where technological capabilities and emerging ambiguities around authorship reshape decision-making conditions (Leaton Gray et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Bin-Nashwan et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). In digital environments that provide instant access to paraphrasing tools, automated citation tools, or Large Language Models (LLMs), the line between legitimate assistance and plagiarism may become blurred. These tools may increase students\u0026rsquo; perceived sense of control over their behavior (e.g., \u0026ldquo;it\u0026rsquo;s easier to avoid detection\u0026rdquo;), while widespread use of such technologies may subtly alter subjective norms, making unauthorized assistance seem more acceptable (Clarke et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Without a theoretical update, TPB-based models risk overlooking forms of moral disengagement that arise from these technologies.\u003c/p\u003e \u003cp\u003eAnother conceptual concern relates to the TPB\u0026rsquo;s implicit assumption that individuals accurately determine the behavior under consideration. However, empirical research suggests that misunderstandings about citation practices, paraphrasing, and authorship conventions (collectively referred to as \u0026ldquo;plagiarism knowledge\u0026rdquo;) are key factors contributing to both unintentional and intentional plagiarism among adolescents (Prashar et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Incorporating plagiarism knowledge into the TPB thus strengthens the model by acknowledging that students cannot form an intention to avoid a behavior they do not explicitly perceive as problematic.\u003c/p\u003e \u003cp\u003eTo apply the TPB to AI-integrated secondary education, scholars have proposed extending the model or integrating it with additional theoretical perspectives. For example, Bandura\u0026rsquo;s (1991) Moral Disengagement Theory provides a useful perspective to explain how students justify unethical behavior, especially when institutional enforcement is perceived as inconsistent or weak. Similarly, adding technology-related constructs (e.g., digital competence or the perceived legitimacy of AI support) could improve the model\u0026rsquo;s explainability in modern learning environments (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eTPB components\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDimension\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraditional TPB\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eExtended TPB (AI-Era, Adolescents)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNorm source\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGeneral social expectations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSchool-specific, peer-group norms, regional culture\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBehavioural control\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGeneral perceived ease/difficulty\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTechnology-mediated control (e.g., AI tools, plagiarism bots)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAttitude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePersonal evaluation of behaviour\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eShaped by exposure to digital content and ethical ambiguity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnowledge (added)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOften excluded\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFormalised as cognitive antecedent to intention/behaviour\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMoral disengagement (optional)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNot modelled\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eComplements TPB when behaviour contradicts internalised norms\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTaken together, these considerations suggest that the original TPB framework, while conceptually robust, requires meaningful adaptation to reflect the developmental, contextual, and technological realities of secondary school students. Specifically, integrating plagiarism knowledge as a cognitive component addresses the formative nature of academic literacy in this population, while accounting for the influence of AI tools helps clarify how perceived control and social norms are reshaped in digital learning environments. These theoretical refinements inform the present study\u0026rsquo;s conceptual model, which extends TPB by (1) incorporating students\u0026rsquo; knowledge about plagiarism, and (2) situating behavioural predictors within an AI-mediated context.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Machine learning approaches in academic integrity research\u003c/h2\u003e \u003cp\u003eMachine learning (ML) is increasingly permeating educational and psychological research, offering analytical capabilities that go beyond the limitations of traditional parametric modeling. While ML has been successfully deployed to predict student engagement, dropout risk, and overall academic performance (Kemper et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Villar \u0026amp; de Andrade, \u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), its application in modeling academic misconduct, specifically, plagiarism in high school students, has not been fully explored. Existing literature focuses primarily on higher education or technical detection algorithms, leaving a gap in understanding behavioral predictors in the K-12 context (Alsabhan, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe main advantage of ML lies in its ability to handle nonlinear interactions on multidimensional data. In the context of plagiarism, psychological constructs (e.g., attitudes, subjective norms), contextual factors (e.g., school culture), and technological capabilities (e.g., GenAI tools) often interact in complex and dynamic ways (Ghimire et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lu et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Standard linear models often fail to capture these subtleties. However, a persistent criticism of ML is the \u0026ldquo;black box\u0026rdquo; phenomenon, in which predictive power comes at the expense of interpretability. To mitigate this, the field is turning to explainable AI (XAI), particularly SHAP (SHapley Additive Explanations) (Guleria \u0026amp; Sood, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). SHAP provides a mathematically consistent framework for quantifying the contribution of each predictor at both the global and local levels, making it highly compatible with psychological frameworks such as the Theory of Planned Behavior (TPB) (Johora et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Unlike alternative explanatory approaches such as LIME, SHAP offers theoretical consistency and model-independent flexibility (Ahmed et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), allowing researchers to analyze the relative importance of variables such as Perceived Behavioral Control or Subjective Norms while still accounting for non-linear effects.\u003c/p\u003e \u003cp\u003eDespite these methodological advances, the integration of SHAP-enhanced Machine Learning (ML) into the study of academic integrity in secondary schools remains in its infancy. Although Molnar et al. (\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) used Interpretable Machine Learning to examine AI-facilitated plagiarism, their study was limited to higher education. There is little research that uses these tools in theory-based approaches, such as TPB, to investigate plagiarism in adolescents. To address this gap, the current study integrates an extended TPB framework with XGBoost and SHAP analysis. This methodological combination aims to connect theoretical understanding with data-driven accuracy, providing deeper insights into the mechanisms of academic cheating and informing targeted educational interventions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. This study\u003c/h2\u003e \u003cp\u003eIn summary, while TPB offers a strong theoretical foundation for understanding academic dishonesty, its application to plagiarism behavior in K-12 students remains underexplored. Moreover, although machine learning (ML) methods show potential for modeling complex behavioral patterns, they have not yet been widely applied or integrated with TPB frameworks in this context. This study seeks to address these gaps by combining an extended TPB model-which incorporates knowledge about plagiarism-with ML-based predictive modeling and interpretable ML techniques. In doing so, we aim to contribute both to theory and to practical efforts for promoting academic integrity in schools. The study is guided by the following research questions:\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ1\u003c/strong\u003e \u003cp\u003eTo what extent do attitudes, subjective norms, perceived behavioural control, and knowledge about plagiarism predict self-reported plagiarism behaviour among secondary school students?\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ2\u003c/strong\u003e \u003cp\u003eAre there significant group differences in TPB components, knowledge, and plagiarism behavior across student gender, grade level, and school type?\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ3\u003c/strong\u003e \u003cp\u003eTo what extent can machine learning models (specifically XGBoost) enhance the prediction of plagiarism behaviour compared to traditional hierarchical regression models?\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eRQ4\u003c/strong\u003e \u003cp\u003eWhat insights do interpretable machine learning techniques (specifically SHAP analysis) provide regarding the relative importance of predictors, and potential interaction patterns, in explaining plagiarism behavior among secondary school students?\u003c/p\u003e \u003c/p\u003e \u003cp\u003eThrough this integrated approach, we aim to deepen the understanding of plagiarism behavior among K-12 students and to demonstrate how theory-driven and data-driven methods can complement each other in educational research.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Participants and procedure\u003c/h2\u003e \u003cp\u003eThis study involved a nationwide sample of secondary school students from various regions of Vietnam. Participants were recruited through an online survey distributed via Google Forms. The survey link was shared through networks of current university interns and former university students who are now teachers at secondary and high schools across the country. This recruitment approach enabled access to a diverse sample from both urban and rural schools. Data collection was conducted online during the Spring semester of 2025.\u003c/p\u003e \u003cp\u003eAfter applying standard data cleaning procedures to remove incomplete and inconsistent responses, the final sample included 900 students. Of these, 660 were male (73.3%) and 240 were female (26.7%). In terms of school type, 789 students (87.7%) attended public institutions, while 111 students (12.3%) were enrolled in private schools. With respect to grade level, 89 students (9.9%) were in lower secondary school (Grades 6\u0026ndash;9), and 811 students (90.1%) were in upper secondary school (Grades 10\u0026ndash;12).\u003c/p\u003e \u003cp\u003eThe observed gender imbalance-predominantly male-is consistent with national enrollment trends in STEM-focused and digitally enriched academic tracks from which many participants were drawn. Participation in the study was entirely voluntary. No personally identifying information was collected, and participants\u0026rsquo; anonymity was maintained throughout the process. Informed consent was obtained from all respondents, with additional parental consent secured where required by school or institutional policy.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Measures\u003c/h2\u003e \u003cp\u003eThe survey instrument was grounded in the Theory of Planned Behavior (TPB) framework (Beck \u0026amp; Ajzen, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) and adapted to the Vietnamese secondary education context. The instrument comprised five distinct sections: plagiarism knowledge, the three TPB constructs (Attitudes, Subjective Norms, Perceived Behavioral Control), and self-reported plagiarism behavior. Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes the instrument structure.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePlagiarism Knowledge.\u003c/b\u003e Objective understanding of academic integrity was assessed using a 12-item multiple-choice test. Items were developed based on international standards (e.g., APA, COPE) and covered common plagiarism scenarios. Scoring was binary (1\u0026thinsp;=\u0026thinsp;Correct, 0\u0026thinsp;=\u0026thinsp;Incorrect/Don\u0026rsquo;t know), with higher aggregate scores reflecting greater conceptual knowledge. Content validity was established through review by two academic integrity experts and two secondary school instructors.\u003c/p\u003e \u003cp\u003e \u003cb\u003eTPB Constructs\u003c/b\u003e. We measured \u003cem\u003eAttitudes\u003c/em\u003e (ATT), \u003cem\u003eSubjective Norms\u003c/em\u003e (SN), and \u003cem\u003ePerceived Behavioral Control\u003c/em\u003e (PBC) using items adapted from existing literature on academic dishonesty (Stone et al., \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e2009b\u003c/span\u003e; Uzun \u0026amp; Kilis, \u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e2020b\u003c/span\u003e) (see Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e). To ensure ecological validity, items were modified to reflect local cultural and school context. All TPB items utilized a 5-point Likert scale ranging from 1 (\u003cem\u003eStrongly disagree\u003c/em\u003e) to 5 (\u003cem\u003eStrongly agree\u003c/em\u003e). Higher scores denote more permissive attitudes, stronger perceived social approval, and lower behavioral control (i.e., higher difficulty), respectively. The initial item pool underwent expert review for clarity and cultural appropriateness prior to deployment.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSelf-Reported Plagiarism Behavior\u003c/b\u003e. A six-item scale quantified the frequency of specific plagiarism behaviors (e.g., unauthorized paraphrasing, reusing work, copying without attribution). Unlike standard frequency scales, responses were captured on a 7-point scale designed to distinguish between historical and current behavior: 1 (\u003cem\u003eNever\u003c/em\u003e), 2 (\u003cem\u003eUnsure\u003c/em\u003e), 3\u0026ndash;4 (\u003cem\u003ePast behavior only\u003c/em\u003e), and 5\u0026ndash;7 (\u003cem\u003eCurrent frequency: Rarely to Always\u003c/em\u003e). Higher scores indicate active or frequent engagement in plagiarism. A pilot study (\u003cem\u003eN\u0026thinsp;=\u0026thinsp;30\u003c/em\u003e) confirmed item clarity and response variability.\u003c/p\u003e \u003cp\u003e \u003cb\u003eDemographics.\u003c/b\u003e We collected data on gender, school type (public vs. private), and grade level (6\u0026ndash;12) to facilitate group comparison analyses.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePsychometric Properties.\u003c/b\u003e Construct validity and reliability were rigorously tested. Exploratory Factor Analysis (EFA) using principal axis factoring (Promax rotation) confirmed the three-factor structure for the TPB constructs. The first three eigenvalues (5.66, 1.90, 1.72) explained 64.58% of the total variance. Sampling adequacy was satisfactory (KMO\u0026thinsp;=\u0026thinsp;.87; Bartlett\u0026rsquo;s test \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001).\u003c/p\u003e \u003cp\u003eInternal consistency (Cronbach\u0026rsquo;s α was robust across all constructs: ATT (.83), SN (.80), PBC (.73), and self-reported behavior (.91). The plagiarism knowledge test demonstrated strong internal consistency (KR-20\u0026thinsp;=\u0026thinsp;.85), with item difficulties ranging from .42 to .83 and discrimination indices between .29 and .59.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of study variables and measurement\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eScale Name\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eItems (English Translation)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRating Scale\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNotes\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePlagiarism Knowledge\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12 items (e.g., \u003cem\u003ePlagiarism is using another person\u0026rsquo;s ideas, processes, results, or words without giving appropriate credit.\u003c/em\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCorrect (1) / Incorrect (0) / Don\u0026rsquo;t know (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTotal score: 0\u0026ndash;12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAttitude toward the Behavior (ATT)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18 items; (e.g., \u003cem\u003ePlagiarizing is as bad as stealing an exam.)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (Strongly disagree) to 5 (Strongly agree)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigher\u0026thinsp;=\u0026thinsp;more permissive attitude\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSubjective Norms (SN)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 items; (e.g., \u003cem\u003eIf my friend allows me to copy, it is acceptable.)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (Strongly disagree) to 5 (Strongly agree)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigher\u0026thinsp;=\u0026thinsp;stronger perceived social approval\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePerceived Behavioral Control (PBC)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 items; (e.g., \u003cem\u003eI could not write a scientific paper without plagiarizing.)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (Strongly disagree) to 5 (Strongly agree)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eItem 15 reverse coded; Item 23 cross-loaded in ATT \u0026amp; PBC\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSelf-reported Plagiarism Behavior\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 items. (e.g., \u003cem\u003eHave you ever copied and pasted a section of someone else\u0026rsquo;s work without attribution?)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (Never), 2 (Not sure / Don\u0026rsquo;t remember), 3 (Done a few times in the past, no longer do it), 4 (Done many times in the past, no longer do it), 5 (Rarely do it), 6 (Often do it), 7 (Always do it)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigher\u0026thinsp;=\u0026thinsp;more frequent plagiarism behavior. Note: Item 6 (\u0026ldquo;use plagiarism detection software\u0026rdquo;) may not reflect plagiarism behavior - consider analyzing separately.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDemographics\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGender, School type, Grade level\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNominal / Categorical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eUsed for group comparisons\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Data handling and assumption testing\u003c/h2\u003e \u003cp\u003eThe final dataset included in the analysis included 923 valid observations after removing missing data. The proportion of missing data on the variables was negligible (\u0026lt;\u0026thinsp;2.5%) and was processed using the multivariate imputation through linked equations (MICE) technique to maximize statistical power and minimize bias (White et al., 2011; Schafer \u0026amp; Graham, 2002). Prior to conducting the regression, the prerequisite statistical assumptions, including multicollinearity, normality, and homogeneity of variance, were tested. The diagnostic results showed that there was no multicollinearity, with the variance inflation factor (VIF) of all independent variables ranging from 1.14 to 2.77, which is within the safe threshold (\u0026lt;\u0026thinsp;5) recommended by Hair et al. (2019). Regarding the normal distribution assumption, the Shapiro-Wilk test shows that the residuals of the model deviate from the ideal normal distribution (W\u0026thinsp;=\u0026thinsp;0.958; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001\u003cspan\u003e$\u003c/span\u003e). However, this is a common phenomenon in studies with large sample sizes, when the Shapiro//-Wilk test becomes too sensitive even to insignificantly small deviations (Field, 2018). More importantly, according to the Central Limit Theorem, with large sample sizes (N\u0026thinsp;\u0026gt;\u0026thinsp;30\u003cspan\u003e$\u003c/span\u003e), the regression estimates still ensure robustness and are not seriously affected by violations of the normality assumption of the residuals (Lumley et al., 2002; Schmidt \u0026amp; Finan, 2018).\u003c/p\u003e \u003cp\u003eThe final analytic sample included 923 complete cases after removing missing data. Missingness across items was minimal (\u0026lt;\u0026thinsp;2.5%) and addressed using multiple imputation via chained equations (MICE), although regression analyses used only cases with full data. Prior to regression, assumptions of multicollinearity, normality, and homoscedasticity were tested. Variance inflation factors (VIFs) for all independent variables ranged from 1.14 to 2.77, well below the standard threshold of 5, indicating no multicollinearity concerns. However, the residuals from the regression model deviated from normality, as indicated by a Shapiro-Wilk test result of W\u0026thinsp;=\u0026thinsp;.958, p\u0026thinsp;\u0026lt;\u0026thinsp;.001. Despite this, the large sample size provides robustness to mild violations of normality, which was noted as a limitation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Analytical strategy\u003c/h2\u003e \u003cp\u003eThe study adopted a two-tiered analytical strategy, combining a theory-based statistical model with a machine learning model to simultaneously test hypotheses and optimize predictions.\u003c/p\u003e \u003cp\u003eRQ1: Hierarchical multiple regression was used to test the predictive power of TPB components (Attitude, Subjective Norms, PBC) and plagiarism knowledge on self-reported behavior. Demographic variables were entered first, followed by TPB variables, and finally plagiarism knowledge. Assumptions of multicollinearity and residual distribution were checked before analysis.\u003c/p\u003e \u003cp\u003eRQ2: Independent samples t-tests were conducted to determine differences by gender, school type, and grade level for TPB variables and plagiarism behavior.\u003c/p\u003e \u003cp\u003eRQ3: An XGBoost model was built to evaluate the nonlinear predictive ability and compared directly with OLS regression on the same test set (80% training, 20% testing). The model parameters were optimized through 5-fold cross-validation. The model performance was evaluated through R\u0026sup2; and RMSE.\u003c/p\u003e \u003cp\u003eRQ4: To explain the predictive mechanism of the machine learning model within the TPB framework, SHAP analysis was applied to evaluate the importance of features and explore nonlinear relationships, including threshold effects.\u003c/p\u003e \u003cp\u003eStatistical analyses were performed using SPSS v.26. The machine learning models and SHAP analysis were implemented on the Google Colab platform with Python v.3.11, using the xgboost v.2.0.3 and shap v.0.41.0 libraries.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Theory-based prediction of plagiarism behavior (RQ1)\u003c/h2\u003e \u003cp\u003eA hierarchical regression analysis was conducted to examine the predictive power of attitudes toward plagiarism, subjective norms, perceived behavioral control (PBC), and plagiarism knowledge in relation to students\u0026rsquo; self-reported plagiarism behavior.\u003c/p\u003e \u003cp\u003eIn Step 1, demographic variables-gender, educational level (lower vs. upper secondary), and school type (public vs. private)-explained 20.9% of the variance in plagiarism behavior (R\u0026sup2; = .209, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). In Step 2, the inclusion of the TPB components (attitudes, subjective norms, and PBC) significantly improved model performance, yielding an additional 36.3% of explained variance (ΔR\u0026sup2; = .363, p\u0026thinsp;\u0026lt;\u0026thinsp;.001), bringing the cumulative explained variance to R\u0026sup2; = .572. In Step 3, the inclusion of the knowledge variable contributed a further 3.2% of explained variance (ΔR\u0026sup2; = .032, p\u0026thinsp;\u0026lt;\u0026thinsp;.001), resulting in a final model that accounted for 60.4% of the total variance in plagiarism behavior (R\u0026sup2; = .604).\u003c/p\u003e \u003cp\u003eAll predictors in the final model were statistically significant (p\u0026thinsp;\u0026lt;\u0026thinsp;.01). Perceived behavioral control emerged as the strongest negative predictor (B = -0.862, p\u0026thinsp;\u0026lt;\u0026thinsp;.001), indicating that students who felt more capable of resisting plagiarism were less likely to engage in it. Subjective norms had a strong positive effect (B\u0026thinsp;=\u0026thinsp;0.668, p\u0026thinsp;\u0026lt;\u0026thinsp;.001), suggesting that permissive peer influences increased the likelihood of plagiarism. Attitudes toward plagiarism also significantly predicted behavior in the expected direction (B = -0.703, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). Notably, knowledge about plagiarism was positively associated with plagiarism behavior (B\u0026thinsp;=\u0026thinsp;1.123, p\u0026thinsp;\u0026lt;\u0026thinsp;.001), a surprising result that suggests students who are more knowledgeable may still choose to plagiarize-possibly due to external pressures or increased sophistication in using tools.\u003c/p\u003e \u003cp\u003eNo multicollinearity concerns were present; all variance inflation factor (VIF) values were below 3 (VIF_Gender\u0026thinsp;=\u0026thinsp;1.12; VIF_Grade\u0026thinsp;=\u0026thinsp;1.24; VIF_School Type\u0026thinsp;=\u0026thinsp;1.02; VIF_Attitude\u0026thinsp;=\u0026thinsp;2.70; VIF_Social Norms\u0026thinsp;=\u0026thinsp;1.68; VIF_PBC\u0026thinsp;=\u0026thinsp;2.95; VIF_Knowledge\u0026thinsp;=\u0026thinsp;1.23). The model\u0026rsquo;s prediction error was acceptable, with RMSE\u0026thinsp;=\u0026thinsp;0.788 and MAE\u0026thinsp;=\u0026thinsp;0.564. Full regression coefficients are presented in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eHierarchical multi-regression predicting plagiarism behavior\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePredictor\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003et\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eStep 1\u003c/b\u003e (Demographics)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntercept\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.177\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.120\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e26.433\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender (0\u0026thinsp;=\u0026thinsp;Male, 1\u0026thinsp;=\u0026thinsp;Female)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.699\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.086\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.157\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEducational Level (0\u0026thinsp;=\u0026thinsp;Lower, 1\u0026thinsp;=\u0026thinsp;Upper)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.287\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.124\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e10.409\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSchool Type (0\u0026thinsp;=\u0026thinsp;Public, 1\u0026thinsp;=\u0026thinsp;Private)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.364\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.112\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-3.237\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eStep 2\u003c/b\u003e (TPB components)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAttitudes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.703\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.089\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-7.878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubjective Norms (SN)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.668\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.067\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.862\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.063\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-13.670\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eStep 3\u003c/b\u003e (Knowledge)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.123\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.132\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.532\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eModel summary\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStep 1 R\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.209\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStep 2 R\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.572\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStep 3 R\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.604\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFurthermore, to address concerns about the gender imbalance in the sample (73.3% male), we conducted a robustness check by running separate regression models for the male and female subsamples. The \u0026ldquo;knowledge paradox\u0026rdquo; in which higher knowledge of plagiarism predicts increased misconduct, remained statistically significant and positive in both groups (\u003cem\u003eB\u003c/em\u003e\u003csub\u003e\u003cem\u003emale\u003c/em\u003e\u003c/sub\u003e = 0.93, p\u0026thinsp;\u0026lt;\u0026thinsp;.001; \u003cem\u003eB\u003c/em\u003e\u003csub\u003e\u003cem\u003efemale\u003c/em\u003e\u003c/sub\u003e = 1.02, p\u0026thinsp;=\u0026thinsp;.001). This consistency suggests that the counterintuitive relationship between knowledge and behavior is a universal mechanism driven by the instrumentalization of normative knowledge, rather than a gender-specific trait related to risk-taking tendencies.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Group differences (RQ2)\u003c/h2\u003e \u003cp\u003eTo examine group differences in plagiarism behavior, attitudes, and knowledge, a series of independent-samples t-tests were conducted across gender, school type, and educational level.\u003c/p\u003e \u003cp\u003eSpecifically, male students reported significantly higher plagiarism frequency than female students (M\u0026thinsp;=\u0026thinsp;5.04) vs. (M\u0026thinsp;=\u0026thinsp;4.15), with the difference reaching statistical significance (t(906)\u0026thinsp;=\u0026thinsp;8.41, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). This suggests that the level of academic dishonesty tends to be higher in the male group, despite a higher understanding of plagiarism (M\u0026thinsp;=\u0026thinsp;0.82 vs. 0.72; t(906)\u0026thinsp;=\u0026thinsp;5.75, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). It is worth noting that attitudes toward plagiarism between the two genders do not completely correlate with actual behavior. Female students, although having more tolerant attitudes towards plagiarism (M\u0026thinsp;=\u0026thinsp;3.02 compared to 2.80 for males), reported significantly lower levels of plagiarism, with the difference also being statistically significant (t(906) = -4.99, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). This result suggests a non-linear motivational pattern, in which stricter attitudes do not necessarily lead to higher compliance. The discrepancy between knowledge, attitudes and behavior, especially among males, can be interpreted as a manifestation of the perception-practice gap in the context of academic ethics.\u003c/p\u003e \u003cp\u003eRegarding school type, students from public schools engaged in more plagiarism than those in private schools (t (906)\u0026thinsp;=\u0026thinsp;2.84, p\u0026thinsp;=\u0026thinsp;.005). However, no significant differences were found between school types in either attitudes or knowledge.\u003c/p\u003e \u003cp\u003eSignificant differences also emerged across educational levels. Students in upper secondary school (Grades 10\u0026ndash;12) reported substantially higher levels of plagiarism behavior than those in lower secondary school (Grades 6\u0026ndash;9) (t(906) = -10.29, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). Conversely, lower secondary students displayed more lenient attitudes (t(906)\u0026thinsp;=\u0026thinsp;4.51, p\u0026thinsp;\u0026lt;\u0026thinsp;.001), while upper secondary students demonstrated significantly higher knowledge scores (t(906) = -5.58, p\u0026thinsp;\u0026lt;\u0026thinsp;.001). Group means and statistical details are reported in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGroup Differences in Plagiarism-Related Constructs by Educational Level, Gender, and School Type\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGroup 1\u003c/p\u003e \u003cp\u003eM (SD)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGroup 2\u003c/p\u003e \u003cp\u003eM (SD)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003et\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eCohen\u0026rsquo;s d\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"1\" nameend=\"c7\" namest=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEducational Level\u003c/b\u003e \u003cem\u003e(Lower vs. Upper)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"1\" nameend=\"c7\" namest=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.63 (0.29)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.81 (0.20)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-5.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAttitude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.23 (0.87)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.81 (0.42)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e-0.61\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSocial Norm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.83 (0.73)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.08 (0.67)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-4.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.36\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.96 (0.84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.56 (0.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-8.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlagiarism Behavior\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.41 (1.37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.95 (1.13)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-10.29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e1.24\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e \u003cem\u003e(Male vs. Female)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.82 (0.20)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.72 (0.24)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.45\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAttitude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.80 (0.43)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.02 (0.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-4.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e-0.42\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSocial Norm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.95 (0.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.92 (0.72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.601\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.33 (0.78)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.30 (0.84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.667\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlagiarism Behavior\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.04 (1.04)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.15 (1.50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.73\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSchool Type\u003c/b\u003e \u003cem\u003e(Public vs. Private)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.79 (0.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.80 (0.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.483\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAttitude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.83 (0.47)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.95 (0.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.051\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e-0.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSocial Norm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.93 (0.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.93 (0.69)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.999\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.00\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.32 (0.80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.33 (0.81)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.890\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e-0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlagiarism Behavior\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.86 (1.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.50 (1.26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.29\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e\u003cb\u003e4.3 Machine learning-based prediction (RQ3\u003c/b\u003e)\u003c/h2\u003e \u003cp\u003eTo assess the added value of machine learning in predicting plagiarism behavior, a Gradient Boosting regression model-analogous in structure to XGBoost-was trained using TPB constructs, plagiarism knowledge, and demographic predictors. The model was evaluated on a held-out test set comprising 20% of the data. It achieved an R\u0026sup2; of .634, exceeding the hierarchical regression model\u0026rsquo;s explanatory power (R\u0026sup2; = .604). The model\u0026rsquo;s prediction errors were within acceptable limits, with a root mean squared error (RMSE) of 0.788 and a mean absolute error (MAE) of 0.564.\u003c/p\u003e \u003cp\u003eTo evaluate generalizability, a 5-fold cross-validation procedure was conducted, yielding a mean R\u0026sup2; of .615 (SD\u0026thinsp;=\u0026thinsp;.016), suggesting that the model's predictive performance was stable across folds. These results demonstrate the potential of data-driven approaches to enhance behavior prediction in educational contexts, particularly when theory-based variables are integrated with real-world demographic complexity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Interpretability of the machine learning model (RQ4)\u003c/h2\u003e \u003cp\u003eSHAP analysis was conducted to interpret the XGBoost model and assess the relative contribution of each predictor (see Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Perceived behavioral control (PBC) emerged as the most influential variable, with higher levels (red points) strongly associated with reduced predicted plagiarism behavior-aligning with theoretical expectations from the TPB framework (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This was followed by subjective norms, plagiarism knowledge, and attitudes, all contributing meaningfully but with more variable effects across students. In contrast, demographic features such as gender and school type had minimal impact. SHAP dependence plots further revealed potential non-linear patterns, particularly for PBC and subjective norms, suggesting threshold effects at extreme values. Taken together, these findings demonstrate the added value of interpretable machine learning in uncovering nuanced, theory-aligned behavioral patterns not easily captured by traditional regression.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSHAP analysis: Feature importance in predicting plagiarism behavior (XGBoost model)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFeature\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSHAP importance\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.589\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSocial Norm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.123\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.099\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAttitude\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.076\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.060\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSchool type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.014\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAnalysis of the SHAP Dependence plot shows that the XGBoost model can reproduce nonlinear behavioral mechanisms that linear regression cannot capture. For the PBC variable, the results show a clear nonlinear pattern of impact. Specifically, between 1.0 and 3.5, the impact of PBC on plagiarism is almost constant, reflecting a plateau where perceived control does not yet play a decisive role. However, a tipping point is observed at 3.5: when perceived lack of control over the behavior exceeds this threshold, the SHAP value increases rapidly in the positive direction (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This indicates that the loss of control can act as a strong trigger, significantly increasing the probability of plagiarism. In the case of Subjective Norms, the data show a more complex nonlinear dynamic. In the early stages (1.0 to 3.5), social pressure exhibits a cumulative effect: the higher the level of environmental acceptance, the higher the risk of plagiarism. However, at a threshold of approximately 3.5, this effect reaches a saturation point and begins to decrease slightly (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Through the lens of machine learning models such as XGBoost, the results suggest that the influence of social norms is not only linear, but also has a threshold effect structure, implying the existence of a social limit in regulating misconduct.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThis study set out to decode the complex behavioral drivers of plagiarism among adolescents in a developing digital economy. By integrating the Theory of Planned Behavior (TPB) with Explainable AI (XGBoost and SHAP), the findings offer empirical evidence that challenges prevailing assumptions about academic integrity. We argue that technology is not a neutral tool but a force that actively reshapes social norms and reveals the systemic vulnerabilities of traditional education.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e5.1. The non-neutrality of technology and the knowledge paradox\u003c/h2\u003e \u003cp\u003eOur findings indicate that digital technologies are not neutral tools but actively reshape the relationship between knowledge and ethical behavior. Although the core components of the Theory of Planned Behavior remain meaningful, the emergence of what we describe as a \u003cem\u003eknowledge paradox\u003c/em\u003e (where greater awareness of plagiarism rules is associated with a higher likelihood of misconduct) signals a troubling shift in students\u0026rsquo; engagement with academic norms. This pattern is consistent with recent evidence reported by Huang et al. (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) and Campo et al. (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), which shows that students who possess a clearer understanding of academic regulations are, paradoxically, often more inclined to violate them.\u003c/p\u003e \u003cp\u003eIn the context of AI-mediated learning, knowledge appears to be used less as a guide for ethical compliance and more as a resource for strategic rule navigation. This observation echoes Akbulut et al.\u0026rsquo;s (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2008\u003c/span\u003e) concept of \u0026ldquo;Internet-enabled academic dishonesty,\u0026rdquo; whereby technologically skilled students leverage their competence to engage in increasingly sophisticated forms of misconduct. Similarly, Bin-Nashwan et al. (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) characterize AI as a \u0026ldquo;double-edged sword,\u0026rdquo; noting that students with high levels of digital literacy are better positioned to operate within the grey zones of AI assistance while avoiding detection by existing monitoring systems. In addition, the strong predictive role of Subjective Norms lends support to the notion of \u003cem\u003epluralistic ignorance\u003c/em\u003e (McCabe et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Comas-Forgas \u0026amp; Sureda-Negre, 2010), whereby students overestimate their peers\u0026rsquo; acceptance of cheating and, in doing so, gradually normalize such behavior. Within a digital learning environment where AI use is widespread yet weakly regulated, this misperception contributes to a self-reinforcing cycle. Practices driven by technological convenience become socially accepted, and completing academic work without AI support may increasingly be viewed as inefficient or even disadvantageous.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Contextual and Institutional Influences on Academic Misconduct\u003c/h2\u003e \u003cp\u003eLinear regression analysis results show that Perceived Behavioral Control (PBC) and negative attitudes toward plagiarism act as important protective factors. This finding is consistent with previous studies applying Theory of Planned Behavior (TPB) in the context of generative AI (GenAI). For example, Ivanov et al. (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) asserted that PBC and attitudes are key antecedents in predicting the intention to use technology in education. Similarly, Bin-Nashwan et al. (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) showed that academic integrity is often inversely related to the use of ChatGPT, thus acting as an ethical barrier to cheating.\u003c/p\u003e \u003cp\u003eHowever, XGBoost\u0026ndash;SHAP analysis provides a deeper insight that linear models struggle to fully reflect: student behavior does not change in a uniform manner but exhibits a distinct threshold effect. When PBC falls below a certain level, the probability of plagiarism increases sharply, indicating a rapid shift from a state of self-regulation to a feeling of loss of control. This result is consistent with studies on the motivations for using AI, where stress and time pressure are considered the main driving factors that lead users to seek out automated tools. Notably, Bin-Nashwan et al. (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) also pointed out an important paradox: while academic integrity often limits the use of AI, under high pressure or the need to save time, even individuals with high levels of integrity tend to increase their use of these tools.\u003c/p\u003e \u003cp\u003eTherefore, academic dishonesty should not be viewed solely as an intentional ethical violation. The results of this study, along with existing evidence, suggest that plagiarism in the context of GenAI may function as a coping strategy when students are overwhelmed. When academic demands exceed coping capabilities, AI tools offer an immediate solution, helping students regain a sense of control and maintain performance. This raises a crucial implication for education and school administration: instead of focusing solely on prohibitions or technical surveillance, more attention should be paid to the pressures of the learning environment and supporting mental health, because plagiarism in the AI ​​age may reflect systemic overload rather than individual moral decay.\u003c/p\u003e \u003cp\u003eBesides individual factors, research findings also show that dishonest academic behavior is significantly influenced by the social and institutional context in which students participate. Specifically, significant differences by gender and educational level suggest that the degree of risk avoidance and sensitivity to school expectations are not uniform across demographic groups. This suggests that academic ethical norms are not received and internalized in the same way across all students but are influenced by different experiences and value orientations during the learning process. Notably, the observed differences between public and private schools highlight the role of organizational culture in shaping learner behavior. This finding is consistent with the argument of Ellis and Murdoch (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), who argue that the institutional climate and how schools convey academic values ​​can directly influence how students interpret and practice ethical norms. In this context, the \u0026ldquo;Knowledge Paradox\u0026rdquo; is not merely a matter of individual perception or competence, but also a consequence of the learning environment where institutional signals about performance, competition, and compliance may inadvertently encourage the instrumental use of AI. In other words, the organizational climate can act as a moderator, either undermining or exacerbating the paradoxical relationship between awareness of plagiarism and dishonest behavior in the AI ​​age.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Policy implications\u003c/h2\u003e \u003cp\u003eOverall, the findings point to the need for a clear shift in educational policy: from restrictive methods to more adaptive and systematic forms of governance. First, the research results suggest that reforming assessment should be prioritized over expanding supervision. The existence of the analytical paradoxes above indicates that awareness campaigns and rule clarification alone are insufficient to prevent misconduct. Instead of increasing supervision or detection technology, educational institutions should reconsider how they design assessments. As Banks et al. (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and Dubljević (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) argued, there should be a greater focus on learning processes, reflection, and repetitive work, rather than narrowly evaluating final outcomes. Such reforms could reduce the incentive to misuse AI while making assessment activities more aligned with genuine learning objectives.\u003c/p\u003e \u003cp\u003eSecondly, instead of simply enforcing norms, secondary schools need to pay more attention to developing digital competencies. Beyond simply communicating rules, schools need to actively cultivate students' digital competencies. Jeon et al. (\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) emphasize that ethical behavior must be learned through practice, not instruction. In this regard, policies need to clearly state acceptable forms of human-AI collaboration and reflect these expectations in daily teaching and assessment. In this way, technical skills that could otherwise be used to circumvent regulations can be redirected toward the responsible and transparent use of AI, forming a more meaningful model of ethical digital competence.\u003c/p\u003e \u003cp\u003eFinally, policymakers must confront the growing tension between efficiency and equity in education systems increasingly shaped by AI (Lim \u0026amp; Lee, 2025). If left unaddressed, integrating AI into the learning process could inadvertently reward students who prioritize speed and efficiency, while putting those who strictly adhere to academic rules at a disadvantage. Therefore, effective governance frameworks need to ensure that ethical compliance is not punished and that the benefits of AI-assisted learning are distributed in a way that remains socially equitable and legitimate.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Limitations and directions for future research\u003c/h2\u003e \u003cp\u003eAlthough the study provides important insights into the predictors of plagiarism among secondary and high school students in Vietnam, some limitations should be considered in interpreting the results. First, the scope of the model mainly focused on the components of the Theory of Planned Behavior (TPB) and the level of knowledge about plagiarism. Future studies should expand the analytical framework by adding other important variables such as moral disengagement, peer group norms, academic pressure, digital competence, or institutional climate. This would help to build a more comprehensive explanatory model of dishonest behavior in the context of digital learning. Second, the study was conducted in the Vietnamese educational context, where cultural factors such as hierarchical teacher-student relationships, strong exam orientation, and collectivist values ​​have a significant influence on students\u0026rsquo; behavioral motivation and moral reasoning. These factors may make the results difficult to compare directly to Western educational systems or other cultural contexts. Therefore, cross-cultural comparison studies are needed to test the generalizability of the model. Third, the cross-sectional design of the study limits the ability to draw causal relationships. Future longitudinal studies are needed to track how attitudes, subjective norms, PBC, and knowledge change over time, and how they relate to actual plagiarism. Finally, self-reported data may be subject to social desirability bias or misunderstanding of plagiarism-related terminology. Although the questionnaire was clearly designed, future studies should adopt a multi-method approach, including behavioral data (e.g., plagiarism detection results), teacher or peer ratings, to enhance reliability and construct validity. Thus, future research directions should expand theoretical scope, implement cross-cultural comparisons, apply longitudinal methods, and integrate multiple data sources to more fully capture the complexity of academic dishonesty in diverse educational contexts.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study helps to better understand why high school students may engage in plagiarism in a digital learning environment. The results show that behavioral control, peer pressure, and personal attitudes remain important factors. However, a notable finding is that understanding plagiarism rules does not necessarily lead to better compliance; in some cases, it even leads to higher rates of violation. This suggests that in a digital context, knowledge is sometimes used to circumvent rules rather than comply with them. Furthermore, data-driven analyses indicate that students' behavior does not change gradually. When their sense of control drops below a certain level, the likelihood of violation increases rapidly. This suggests that technology-related plagiarism stems not only from personal choice but also from a reaction to academic pressure and overload. These findings suggest that simply reminding students of regulations or increasing control is insufficient. Education needs to focus more on helping students learn to self-regulate, learn responsibly when using AI, and understand social norms in the digital environment. Furthermore, the differences between student groups and school types indicate that solutions need to be flexible and tailored to each specific context.\u003c/p\u003e"},{"header":"Declarations","content":" \u003cp\u003e \u003cstrong\u003eFunding delaration\u003c/strong\u003e \u003cp\u003e \u003cem\u003eThis research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number NCUD.05-2022.21.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eEthic Delaration\u003c/b\u003e: \u003cem\u003eThis study was conducted in accordance with the principles of the Declaration of Helsinki. The research protocol was approved by the Committee of the National Science and Technology Development Fund (Approval Number: NCUD. 05-2022.21).\u003c/em\u003e\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eTinh T.T. Le: Conceptualization, Methodology, Data curation, Formal analysis, Writing \u0026ndash; original draft.Doanh V.Q. Ho: Data collection, Writing \u0026ndash; review \u0026amp; editing.Hung V. Tran: Methodology, Formal analysis, Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThis research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number NCUD.05-2022.21\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData will be made available on request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAamir Raoof Memon (2020) \u003cem\u003eKnowledge, Attitudes, and Practices of Plagiarism as Reported by Participants Completing the AuthourAID MOOC on Research Writing\u003c/em\u003e, Science and Engineeing Ethics. 42\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmed S, Kaiser MS, Hossain S, M., Andersson K (2025) A Comparative Analysis of LIME and SHAP Interpreters With Explainable ML-Based Diabetes Predictions. IEEE Access 13:37370\u0026ndash;37388. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ACCESS.2024.3422319\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2024.3422319\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAjzen I (1991a) The theory of planned behavior. Organ Behav Hum Decis Process 50(2):179\u0026ndash;211. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0749-5978(91)90020-T\u003c/span\u003e\u003cspan address=\"10.1016/0749-5978(91)90020-T\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAjzen I (1991b) The theory of planned behavior. Organ Behav Hum Decis Process 50(2):179\u0026ndash;211. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0749-5978(91)90020-T\u003c/span\u003e\u003cspan address=\"10.1016/0749-5978(91)90020-T\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAkbulut Y et al (2008) Exploring the types and reasons of Internet-triggered academic dishonesty among Turkish undergraduate students: Development of Internet-Triggered Academic Dishonesty Scale (ITADS). Comput Educ 51(1):463\u0026ndash;473. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compedu.2007.06.003\u003c/span\u003e\u003cspan address=\"10.1016/j.compedu.2007.06.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli O, Murray PA, Momin M, Al-Anzi FS (2023) The knowledge and innovation challenges of ChatGPT: A scoping review. Technol Soc 75:102402\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlsabhan W (2023) Student Cheating Detection in Higher Education by Implementing Machine Learning and LSTM Techniques. Sensors 23(8):4149. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/s23084149\u003c/span\u003e\u003cspan address=\"10.3390/s23084149\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAri Santoso \u0026amp; Fitra Roman Cahaya (2018) Factors influencing plagiarism by accounting lectures. Acc Educ. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/09639284.2018.1523736\u003c/span\u003e\u003cspan address=\"10.1080/09639284.2018.1523736\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBanks D, Bosone C, Carpenter B, Shah T, Shi C (2024) Large Language Models: Trust and Regulation. Harv Data Sci Rev, 6(3)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarry D, Smith (1998) Psychology: Science \u0026amp; Understanding. McGraw-Hill\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeck L, Ajzen I (1991) Predicting dishonest actions using the theory of planned behavior. J Res Pers 25(3):285\u0026ndash;301. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0092-6566(91)90021-H\u003c/span\u003e\u003cspan address=\"10.1016/0092-6566(91)90021-H\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBin-Nashwan SA, Sadallah M, Bouteraa M (2023) Use of ChatGPT in academia: Academic integrity hangs in the balance. Technol Soc 75:102370\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBin-Nashwan SA, Sadallah M, Bouteraa M (2023) Use of ChatGPT in academia: Academic integrity hangs in the balance. Technol Soc 75:102370. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.techsoc.2023.102370\u003c/span\u003e\u003cspan address=\"10.1016/j.techsoc.2023.102370\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBirks D, Clare J (2023a) Linking artificial intelligence facilitated academic misconduct to existing prevention frameworks. Int J Educational Integr 19(1):20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-023-00142-3\u003c/span\u003e\u003cspan address=\"10.1007/s40979-023-00142-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBirks D, Clare J (2023b) Linking artificial intelligence facilitated academic misconduct to existing prevention frameworks. Int J Educational Integr 19(1):20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-023-00142-3\u003c/span\u003e\u003cspan address=\"10.1007/s40979-023-00142-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampo L, Delgado N, Etxabe-Urbieta JM, Kanso H, Gueye ML (2025) Relationship Between the Use of ChatGPT and Plagiarism in Higher Education: The Influence of Gender, Age and Previous Academic Results. \u003cem\u003eInnovative Higher Education\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10755-025-09830-z\u003c/span\u003e\u003cspan address=\"10.1007/s10755-025-09830-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u0026Ccedil;elik \u0026Ouml;, Razı S (2023) Facilitators and barriers to creating a culture of academic integrity at secondary schools: an exploratory case study. Int J Educational Integr 19(1):4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-023-00125-4\u003c/span\u003e\u003cspan address=\"10.1007/s40979-023-00125-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChan CKY (2025) Students\u0026rsquo; perceptions of \u0026lsquo;AI-giarism\u0026rsquo;: investigating changes in understandings of academic misconduct. Educ Inform Technol 30(6):8087\u0026ndash;8108. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-13151-7\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-13151-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChoi Y, Suh KH (2022) Verifying the usefulness of the theory of planned behavior model for predicting illegal use of online content: the role of outcome expectancies and social loafing. BMC Psychol 10(1):263. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40359-022-00978-3\u003c/span\u003e\u003cspan address=\"10.1186/s40359-022-00978-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarke O, Chan WYD, Bukuru S, Logan J, Wong R (2023) Assessing knowledge of and attitudes towards plagiarism and ability to recognize plagiaristic writing among university students in Rwanda. High Educ 85(2):247\u0026ndash;263. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10734-022-00830-y\u003c/span\u003e\u003cspan address=\"10.1007/s10734-022-00830-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCotton DRE, Cotton PA, Shipway JR (2024) Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations Educ Teach Int 61(2):228\u0026ndash;239. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14703297.2023.2190148\u003c/span\u003e\u003cspan address=\"10.1080/14703297.2023.2190148\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurtis GJ, Vardanega L (2016) Is plagiarism changing over time? A 10-year time-lag study with three points of measurement. High Educ Res Dev 35(6):1167\u0026ndash;1179. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/07294360.2016.1161602\u003c/span\u003e\u003cspan address=\"10.1080/07294360.2016.1161602\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurtis GJ, Cowcher E, Greene BR, Rundle K, Paull M, Davis MC (2018a) Self-Control, Injunctive Norms, and Descriptive Norms Predict Engagement in Plagiarism in a Theory of Planned Behavior Model. J Acad Ethics 16(3):225\u0026ndash;239. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10805-018-9309-2\u003c/span\u003e\u003cspan address=\"10.1007/s10805-018-9309-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurtis GJ, Cowcher E, Greene BR, Rundle K, Paull M, Davis MC (2018b) Self-Control, Injunctive Norms, and Descriptive Norms Predict Engagement in Plagiarism in a Theory of Planned Behavior Model. J Acad Ethics 16(3):225\u0026ndash;239. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10805-018-9309-2\u003c/span\u003e\u003cspan address=\"10.1007/s10805-018-9309-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeepti M (2020) \u003cem\u003eFactors Influencing Higher Similarity and Plagiarism amongst Research Scholars: A Comparative Case Study of two Indian Universities - Centers of Higher Education\u003c/em\u003e, Library Philosophy and Practice (e-journal), 4287. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://digitalcommons.unl.edu/libphilprac/4287\u003c/span\u003e\u003cspan address=\"https://digitalcommons.unl.edu/libphilprac/4287\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD\u0026eacute;siron JC, Petko D (2023) Academic dishonesty when doing homework: How digital technologies are put to bad use in secondary schools. Educ Inform Technol 28(2):1251\u0026ndash;1271. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-022-11225-y\u003c/span\u003e\u003cspan address=\"10.1007/s10639-022-11225-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDrew W (1996) Psychology: Mind, Brain \u0026amp; Culture. John Wiley \u0026amp; Sons, Inc.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDubljević V (2024) Colleges and universities are important stakeholders for regulating large language models and other emerging AI. Technol Soc 77:102480\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePhyo EM, Lwin T, Tun HP Zaw Zaw Oo, Kyaw Swa Mya \u0026amp; Henry Silverman (2023), \u003cem\u003eKnowledge, attitudes, and practices regarding plagiarism of postgraduate students in Myanmar\u003c/em\u003e. Account Res, 43 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/08989621.2022.2077643\u003c/span\u003e\u003cspan address=\"10.1080/08989621.2022.2077643\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEllis C, Murdoch K (2024) The educational integrity enforcement pyramid: a new framework for challenging and responding to student cheating. Assess Evaluation High Educ 49(7):924\u0026ndash;934. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02602938.2024.2329167\u003c/span\u003e\u003cspan address=\"10.1080/02602938.2024.2329167\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEret E, Ok A (2014) Internet plagiarism in higher education: tendencies, triggering factors and reasons among teacher candidates. Assess Evaluation High Educ 39(8):1002\u0026ndash;1016. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02602938.2014.880776\u003c/span\u003e\u003cspan address=\"10.1080/02602938.2014.880776\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeldman-Maggor Y, Cukurova M, Kent C, Alexandron G (2025) The Impact of Explainable AI on Teachers\u0026rsquo; Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations. Int J Artif Intell Educ. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40593-025-00486-6\u003c/span\u003e\u003cspan address=\"10.1007/s40593-025-00486-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFoltynek T, Bjelobaba S, Glendinning I, Khan ZR, Santos R, Pavletic P, Kravjar J (2023) ENAI Recommendations on the ethical use of Artificial Intelligence in Education. Int J Educational Integr 19(1):12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-023-00133-4\u003c/span\u003e\u003cspan address=\"10.1007/s40979-023-00133-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGhimire S, Abdulla S, Joseph LP, Prasad S, Murphy A, Devi A, Barua PD, Deo RC, Acharya R, Yaseen ZM (2024) Explainable artificial intelligence-machine learning models to estimate overall scores in tertiary preparatory general science course. Computers Education: Artif Intell 7:100331. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.caeai.2024.100331\u003c/span\u003e\u003cspan address=\"10.1016/j.caeai.2024.100331\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuleria P, Sood M (2023) Explainable AI and machine learning: performance evaluation and explainability of classifiers on educational data mining inspired career counseling. Educ Inform Technol 28(1):1081\u0026ndash;1116. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-022-11221-2\u003c/span\u003e\u003cspan address=\"10.1007/s10639-022-11221-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHou S, Zhang X, Yi B, Tang Y (2022) Public attitudes on open source communities in China: A text mining analysis. Technol Soc 71:102112\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang CL, Chen Y, Zhang S, Yang SC (2025) Is copy and paste part of academic misconduct? The roles of attitude, experience and self-efficacy in judgment. Contemp Educ Psychol 82:102402. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cedpsych.2025.102402\u003c/span\u003e\u003cspan address=\"10.1016/j.cedpsych.2025.102402\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIcek Ajzen (1991) The Theory of Planned Behavior. Organ Behav Hum Decis Process, 50(2)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIrani F, Muotka J, Lyyra P, Parviainen T, Monto S (2024) Social influence in adolescence: Behavioral and neural responses to peer and expert opinion. Soc Neurosci 19(1):25\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/17470919.2024.2323745\u003c/span\u003e\u003cspan address=\"10.1080/17470919.2024.2323745\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIsmail K (2018) Perceptions of Plagiarism Among Medical and Nursing Students in Erbill, Iraq. Sultan Qaboos Univ Med J. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18295/squmj.2018.18.02.012\u003c/span\u003e\u003cspan address=\"10.18295/squmj.2018.18.02.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvanov S, Soliman M, Tuomi A, Alkathiri NA, Al-Alawi AN (2024) Drivers of generative AI adoption in higher education through the lens of the Theory of Planned Behaviour. Technol Soc 77:102521\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeon J, Kim L, Park J (2025) The ethics of generative AI in social science research: A qualitative approach for institutionally grounded AI research ethics. Technol Soc 102836. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.techsoc.2025.102836\u003c/span\u003e\u003cspan address=\"10.1016/j.techsoc.2025.102836\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohansen MW, Goddiksen MP, Centa M, Clavien C, Gefenas E, Globokar R, Hogan L, Merit MT, Nielsen SS, Olsson IAS, Poškutė M, Quinn U, Santos JB, Santos R, Sch\u0026ouml;pfer C, Strahovnik V, Wall PJ, Sand\u0026oslash;e P, Lund TB (2022a) Lack of ethics or lack of knowledge? European upper secondary students\u0026rsquo; doubts and misconceptions about integrity issues. Int J Educational Integr 18(1):20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-022-00113-0\u003c/span\u003e\u003cspan address=\"10.1007/s40979-022-00113-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohansen MW, Goddiksen MP, Centa M, Clavien C, Gefenas E, Globokar R, Hogan L, Merit MT, Nielsen SS, Olsson IAS, Poškutė M, Quinn U, Santos JB, Santos R, Sch\u0026ouml;pfer C, Strahovnik V, Wall PJ, Sand\u0026oslash;e P, Lund TB (2022b) Lack of ethics or lack of knowledge? European upper secondary students\u0026rsquo; doubts and misconceptions about integrity issues. Int J Educational Integr 18(1):20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-022-00113-0\u003c/span\u003e\u003cspan address=\"10.1007/s40979-022-00113-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnston H, Wells RF, Shanks EM, Boey T, Parsons BN (2024) Student perspectives on the use of generative artificial intelligence technologies in higher education. Int J Educational Integr 20(1):2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-024-00149-4\u003c/span\u003e\u003cspan address=\"10.1007/s40979-024-00149-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohora FT, Hasan MN, Rajbongshi A, Ashrafuzzaman M, Akter F (2025) An explainable AI-based approach for predicting undergraduate students academic performance. Array 26:100384. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.array.2025.100384\u003c/span\u003e\u003cspan address=\"10.1016/j.array.2025.100384\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJuan LX, Tao WY, Veloo PK, Supramaniam M (2022) Using Extended TPB Models to Predict Dishonest Academic Behaviors of Undergraduates in a Chinese Public University. Sage Open 12(4). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/21582440221140391\u003c/span\u003e\u003cspan address=\"10.1177/21582440221140391\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKam CCS, Hue MT, Cheung HY (2018) Academic dishonesty among Hong Kong secondary school students: application of theory of planned behaviour. Educational Psychol 38(7):945\u0026ndash;963. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01443410.2018.1454588\u003c/span\u003e\u003cspan address=\"10.1080/01443410.2018.1454588\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKemper L, Vorhoff G, Wigger BU (2020) Predicting student dropout: A machine learning approach. Eur J High Educ 10(1):28\u0026ndash;47. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/21568235.2020.1718520\u003c/span\u003e\u003cspan address=\"10.1080/21568235.2020.1718520\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeaton Gray S, Edsall D, Parapadakis D (2025) AI-Based Digital Cheating At University, and the Case for New Ethical Pedagogies. J Acad Ethics 23(4):2069\u0026ndash;2086. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10805-025-09642-y\u003c/span\u003e\u003cspan address=\"10.1007/s10805-025-09642-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi M, Patel J, Katapally TR (2025) Towards ethical surveillance of smartphone use among youth: exploratory digital citizen science approaches shaping the understanding of ubiquitous technology use. Technol Soc 83:103012\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiddell J (2003) A Comprehensive Definition of Plagiarism, \u003cem\u003eCommunity and Junior College\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLim WM, Gunasekara A, Pallant JL, Pallant JI, Pechenkina E (2023) Generative AI and the future of education: Ragnar\u0026ouml;k or reformation? A paradoxical perspective from management educators. Int J Manage Educ 21(2):100790. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijme.2023.100790\u003c/span\u003e\u003cspan address=\"10.1016/j.ijme.2023.100790\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLinardatos P, Papastefanopoulos V, Kotsiantis S (2020) Explainable AI: A Review of Machine Learning Interpretability Methods. Entropy 23(1):18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/e23010018\u003c/span\u003e\u003cspan address=\"10.3390/e23010018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu S-C, Swisher CL, Chung C, Jaffray D, Sidey-Gibbons C (2023) On the importance of interpretable machine learning predictions to inform clinical decision making in oncology. \u003cem\u003eFrontiers in Oncology\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fonc.2023.1129380\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2023.1129380\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacdonald R, Carroll J (2006) Plagiarism-a complex issue requiring a holistic institutional approach. Assess Evaluation High Educ 31(2):233\u0026ndash;245. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02602930500262536\u003c/span\u003e\u003cspan address=\"10.1080/02602930500262536\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalti T, Galarneau E, Peplak J (2021) Moral Development in Adolescence. J Res Adolescence 31(4):1097\u0026ndash;1113. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/jora.12639\u003c/span\u003e\u003cspan address=\"10.1111/jora.12639\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMavrinac M, Brumini G, Bilić-Zulle L, Petrovečki M (2010) Construction and Validation of Attitudes Toward Plagiarism Questionnaire. Croatian Med J 51(3):195\u0026ndash;201. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3325/cmj.2010.51.195\u003c/span\u003e\u003cspan address=\"10.3325/cmj.2010.51.195\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaxwell A, Curtis GJ, Vardanega L (2008) Does culture influence understanding and perceived seriousness of plagiarism? Int J Educational Integr 4(2). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21913/IJEI.v4i2.412\u003c/span\u003e\u003cspan address=\"10.21913/IJEI.v4i2.412\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCabe DL, Trevino LK (1997) Individual and Contextual Influences on Academic Dishonesty: A Multicampus Investigation. Res High Educt 38(3):379\u0026ndash;396. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1023/A:1024954224675\u003c/span\u003e\u003cspan address=\"10.1023/A:1024954224675\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCabe DL, Trevino LK, Butterfield KD (2001) Cheating in Academic Institutions: A Decade of Research. Ethics Behav 11(3):219\u0026ndash;232. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1207/S15327019EB1103_2\u003c/span\u003e\u003cspan address=\"10.1207/S15327019EB1103_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCabe DL, Trevi\u0026ntilde;o LK, Butterfield KD (2002) Honor Codes and Other Contextual Influences on Academic Integrity: A Replication and Extension to Modified Honor Code Settings. Res High Educt 43(3):357\u0026ndash;378. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1023/A:1014893102151\u003c/span\u003e\u003cspan address=\"10.1023/A:1014893102151\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMolnar C, Casalicchio G, Bischl B (2020) \u003cem\u003eInterpretable Machine Learning - A Brief History, State-of-the-Art and Challenges\u003c/em\u003e (pp. 417\u0026ndash;431). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-65965-3_28\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-65965-3_28\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNazia S (2023) Factors Influencing among Faculty members and Research scholars in The Higher Education System. DESIDOC Journal of Library \u0026amp; Information Technology. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://digitalcommons.unl.edu/libphilprac/4287\u003c/span\u003e\u003cspan address=\"https://digitalcommons.unl.edu/libphilprac/4287\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNguyen HM, Goto D (2024) Unmasking academic cheating behavior in the artificial intelligence era: Evidence from Vietnamese undergraduates. Educ Inform Technol. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-12495-4\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-12495-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNjuguna GN, Qingfei M (2026) How would prompt completion editing impact user experience scores in academic research with large language models? Technol Soc 84:103080\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNora WLY, Zhang KC (2010) Motives of cheating among secondary students: the role of self-efficacy and peer influence. Asia Pac Educ Rev 11(4):573\u0026ndash;584. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12564-010-9104-2\u003c/span\u003e\u003cspan address=\"10.1007/s12564-010-9104-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrtiz-Bonnin S, Blahopoulou J (2025) Chat or cheat? Academic dishonesty, risk perceptions, and ChatGPT usage in higher education students. Soc Psychol Educ 28(1):113. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11218-025-10080-2\u003c/span\u003e\u003cspan address=\"10.1007/s11218-025-10080-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerry AH, Rettinger DA, Stephens JM, Anderman EM, McTernan ML, Tatum H, McNally D, Cullen C, Gallant B, T (2025) From institutional climate to moral attitudes: examining theoretical models of academic misconduct. Ethics Behav 1\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10508422.2025.2514577\u003c/span\u003e\u003cspan address=\"10.1080/10508422.2025.2514577\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePhyo EM, Lwin T, Pyae Tun H, Zaw Zaw O, Swa Mya K, Silverman H (2023) \u003cem\u003eKnowledge, Attitudes, and Practices Regarding Plagiarism of Postgraduate Students in Myanmar\u003c/em\u003e, Accountability in Research. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/08989621.2022.2077643\u003c/span\u003e\u003cspan address=\"10.1080/08989621.2022.2077643\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, Tr. 672\u0026ndash;691\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePike RK, Buck LA, Tsoukkas E, Bell E (2025) From collaboration to contract cheating: exploring staff and student perceptions of the grey areas of academic outsourcing. Assess Evaluation High Educ 1\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02602938.2025.2539291\u003c/span\u003e\u003cspan address=\"10.1080/02602938.2025.2539291\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePolona Sprajc (2017) \u003cem\u003eReasons for Plagiarism in Higher Education, De Gruyter\u003c/em\u003e, Tr. 33\u0026ndash;45\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrashar A, Gupta P, Dwivedi YK (2024) Plagiarism awareness efforts, students\u0026rsquo; ethical judgment and behaviors: a longitudinal experiment study on ethical nuances of plagiarism in higher education. Stud High Educ 49(6):929\u0026ndash;955. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/03075079.2023.2253835\u003c/span\u003e\u003cspan address=\"10.1080/03075079.2023.2253835\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRettinger DA, Kramer Y (2009) Situational and Personal Causes of Student Cheating. Res High Educt 50(3):293\u0026ndash;313. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11162-008-9116-5\u003c/span\u003e\u003cspan address=\"10.1007/s11162-008-9116-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRonald W, Belter (2009) A Strategy to reduce Plagiarism in an Undergraduate Course. Taylor \u0026amp; Francis Group, Tr, pp 257\u0026ndash;261\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSisti DA (2007) How Do High School Students Justify Internet Plagiarism? Ethics Behav 17(3):215\u0026ndash;231. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10508420701519163\u003c/span\u003e\u003cspan address=\"10.1080/10508420701519163\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStoesz BM, Yudintseva A (2018) Effectiveness of tutorials for promoting educational integrity: a synthesis paper. Int J Educational Integr 14(1):6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-018-0030-0\u003c/span\u003e\u003cspan address=\"10.1007/s40979-018-0030-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStone TH, Jawahar IM, Kisamore JL (2009a) Using the theory of planned behavior and cheating justifications to predict academic misconduct. Career Dev Int 14(3):221\u0026ndash;241. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/13620430910966415\u003c/span\u003e\u003cspan address=\"10.1108/13620430910966415\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStone TH, Jawahar IM, Kisamore JL (2009b) Using the theory of planned behavior and cheating justifications to predict academic misconduct. Career Dev Int 14(3):221\u0026ndash;241. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/13620430910966415\u003c/span\u003e\u003cspan address=\"10.1108/13620430910966415\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStone TH, Jawahar IM, Kisamore JL (2009c) Using the theory of planned behavior and cheating justifications to predict academic misconduct. Career Dev Int 14(3):221\u0026ndash;241. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/13620430910966415\u003c/span\u003e\u003cspan address=\"10.1108/13620430910966415\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStone TH, Jawahar IM, Kisamore JL (2010) Predicting Academic Misconduct Intentions and Behavior Using the Theory of Planned Behavior and Personality. Basic Appl Soc Psychol 32(1):35\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01973530903539895\u003c/span\u003e\u003cspan address=\"10.1080/01973530903539895\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSwamy V, Du S, Marras M, Kaser T (2023) Trusting the Explainers: Teacher Validation of Explainable Artificial Intelligence for Course Design. \u003cem\u003eLAK23: 13th International Learning Analytics and Knowledge Conference\u003c/em\u003e, 345\u0026ndash;356. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3576050.3576147\u003c/span\u003e\u003cspan address=\"10.1145/3576050.3576147\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTindall IK, Fu KW, Tremayne K, Curtis GJ (2021) Can negative emotions increase students\u0026rsquo; plagiarism and cheating? Int J Educational Integr 17(1):25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40979-021-00093-7\u003c/span\u003e\u003cspan address=\"10.1007/s40979-021-00093-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTowler G, Shepherd R (1991) Modification of Fishbein and Ajzen\u0026rsquo;s theory of reasoned action to predict chip consumption. Food Qual Prefer 3(1):37\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0950-3293(91)90021-6\u003c/span\u003e\u003cspan address=\"10.1016/0950-3293(91)90021-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUzun AM, Kilis S (2020a) Investigating antecedents of plagiarism using extended theory of planned behavior. Comput Educ 144:103700. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compedu.2019.103700\u003c/span\u003e\u003cspan address=\"10.1016/j.compedu.2019.103700\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUzun AM, Kilis S (2020b) Investigating antecedents of plagiarism using extended theory of planned behavior. Comput Educ 144:103700. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compedu.2019.103700\u003c/span\u003e\u003cspan address=\"10.1016/j.compedu.2019.103700\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVillar A, de Andrade CRV (2024) Supervised machine learning algorithms for predicting student dropout and academic success: a comparative study. Discover Artif Intell 4(1):2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s44163-023-00079-z\u003c/span\u003e\u003cspan address=\"10.1007/s44163-023-00079-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang M, Yang Y, Zhong P (2025) How does AI affect the self-actualization of content creators in dynamic environments? A knowledge management perspective. Technol Soc 102855. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.techsoc.2025.102855\u003c/span\u003e\u003cspan address=\"10.1016/j.techsoc.2025.102855\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWatrianthos R, Ahmad ST, Muskhir M (2023) Charting the growth and structure of early ChatGPT-education research: A bibliometric study. J Inform Technol Education: Innovations Pract 22:235\u0026ndash;253\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZeng K, Wang Z, Lu T, Chen J, Han Z (2023) Implicit space pose consistent transfer network for deep face verification. Pattern Recognit Lett 176:1\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.patrec.2023.10.017\u003c/span\u003e\u003cspan address=\"10.1016/j.patrec.2023.10.017\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Y (2024) Academic cheating as planned behavior: the effects of perceived behavioral control and individualism-collectivism orientations. High Educ 87(3):567\u0026ndash;590. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10734-023-01024-w\u003c/span\u003e\u003cspan address=\"10.1007/s10734-023-01024-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao L, Mao H, Compton BJ, Peng J, Fu G, Fang F, Heyman GD, Lee K (2022a) Academic dishonesty and its relations to peer cheating and culture: A meta-analysis of the perceived peer cheating effect. Educational Res Rev 36:100455. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.edurev.2022.100455\u003c/span\u003e\u003cspan address=\"10.1016/j.edurev.2022.100455\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"international-journal-for-educational-integrity","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijei","sideBox":"Learn more about [International Journal for Educational Integrity](https://edintegrity.biomedcentral.com/)","snPcode":"40979","submissionUrl":"https://submission.springernature.com/new-submission/40979/3","title":"International Journal for Educational Integrity","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Theory of Planned Behavior, Academic integrity, Tech-facilitated cheating, Explainable AI (XAI), SHAP explainability","lastPublishedDoi":"10.21203/rs.3.rs-8458923/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8458923/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAs artificial intelligence blurs the boundaries between authorship and assistance, traditional frameworks for understanding academic dishonesty are becoming obsolete. This study employs a data-driven approach to decode the complex ethical decision-making of adolescents in a developing digital economy. Analyzing data from 923 secondary students in Vietnam, we combine the Theory of Planned Behavior (TPB) with machine learning (XGBoost) and SHAP interpretability techniques. The XGBoost model achieved superior predictive accuracy (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.634) compared to traditional linear regression (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.604). The study offers two major contributions to the discourse on technology and ethics. First, it exposes a \u0026lsquo;knowledge paradox\u0026rsquo; where increased awareness of academic integrity rules is significantly associated with higher plagiarism frequency (\u003cem\u003eB\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.12, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), indicating a disconnect between digital literacy and ethical practice. Second, XAI analysis reveals a nonlinear threshold in Perceived Behavioral Control: when the perceived difficulty of avoiding plagiarism exceeds a critical tipping point (score\u0026thinsp;\u0026gt;\u0026thinsp;3.5 on a 5-point scale), the likelihood of offending escalates disproportionately, regardless of students' moral attitudes. These findings suggest that purely informational campaigns are insufficient in the AI era. Instead, promoting integrity requires systemic interventions that address the psychological thresholds of control and the unintended consequences of digital literacy education.\u003c/p\u003e","manuscriptTitle":"Vietnamese High School Students in the AI Era: Examining the Control Threshold and Knowledge Paradox in Academic Integrity","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-04 18:42:13","doi":"10.21203/rs.3.rs-8458923/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-14T07:21:17+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-23T18:18:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"150475721619381522712633642155181416477","date":"2026-03-28T14:55:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"212638307456812694989098249133280743182","date":"2026-03-04T07:53:47+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-27T13:08:40+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-09T23:26:12+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-19T05:01:40+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal for Educational Integrity","date":"2026-01-18T23:41:55+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-for-educational-integrity","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijei","sideBox":"Learn more about [International Journal for Educational Integrity](https://edintegrity.biomedcentral.com/)","snPcode":"40979","submissionUrl":"https://submission.springernature.com/new-submission/40979/3","title":"International Journal for Educational Integrity","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e2f21de2-00a6-4ad6-b986-45f7a8646468","owner":[],"postedDate":"March 4th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-14T07:21:17+00:00","index":24,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-04T18:42:13+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-04 18:42:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8458923","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8458923","identity":"rs-8458923","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}