Unpacking the relationship between self-control and academic engagement: the mediating role of meaning in life and the moderating role of GenAI dependence

Abstract

Background

With the rapid integration of generative artificial intelligence (GenAI) into higher education, understanding how technology shapes students’ psychological motivation has become increasingly critical. Self-control is widely recognized as a core internal resource that sustains academic engagement, yet little is known about how this self-regulatory capacity operates when external technological reliance intervenes in learning. Addressing this gap, the present study introduces an innovative framework that integrates Self-Determination Theory (SDT) and Path Dependence Theory (PDT) to examine how meaning in life—defined as individuals’ perceived sense of purpose, coherence, and value in life—mediates the relationship between self-control and academic engagement, and how GenAI dependence moderates this pathway. This approach provides a novel perspective on the interaction between inner volitional strength and external technological dependence in shaping academic motivation.

Method

A cross-sectional survey was conducted among 1,139 university students in China. Validated self-report questionnaires were used to assess self-control, meaning in life, GenAI dependence, and academic engagement. Structural equation modeling (SEM) was performed using Mplus 8.3 to construct latent variables and test the hypothesized moderated mediation model.

Results

The findings indicated that self-control significantly predicted academic engagement. Meaning in life partially mediated this relationship. Furthermore, GenAI dependence negatively moderated the pathway from self-control to meaning in life—specifically, the positive effect of self-control on meaning in life was weaker among students with higher GenAI dependence. These results reveal a dynamic interaction between internal psychological strengths and external technological reliance in shaping academic engagement.

Conclusion

This study proposes and validates a novel moderated mediation mechanism linking self-control, meaning in life, and academic engagement within the emerging context of AI-augmented learning. By integrating motivational and technological perspectives, it contributes a new theoretical model of “motivational regulation under technological mediation.” Practically, the findings underscore the dual importance of cultivating students’ self-regulation and existential meaning-making capacity while promoting reflective and autonomous GenAI use. These insights offer evidence-based guidance for fostering sustainable academic motivation and psychological well-being in digitalized higher education environments.

Introduction

Academic engagement has long been a central topic in educational research and practice. It is widely regarded as a key indicator of students’ learning quality and academic performance, and an essential manifestation of learning competence. In the broader literature, academic engagement has been conceptualized from two distinct but complementary traditions. The North American perspective, largely established by Fredricks, Blumenfeld, and Paris, defines engagement as a multidimensional construct encompassing behavioral, emotional, and cognitive dimensions, emphasizing students’ observable participation, affective involvement, and strategic thinking processes[1]. In contrast, the European perspective—rooted in occupational psychology and represented by the Utrecht Work Engagement Scale—conceptualizes engagement as a sustained, positive, and fulfilling psychological state characterized by vigor, dedication, and absorption[2]. Given that the present study focuses on Chinese university students’ internal motivational processes and emotional-cognitive persistence rather than their observable classroom behavior, the European conceptualization provides a more theoretically coherent and contextually appropriate framework for capturing the depth and quality of students’ learning engagement.

Defined as a sustained, positive, and focused cognitive-emotional state during the learning process, academic engagement primarily comprises three dimensions: vigor, dedication, and absorption[3]. At the individual level, academic engagement reflects students’ learning adaptability and serves as a significant predictor of developmental outcomes such as academic achievement and mental health[4]. At the systemic level, it functions as a core psychological mechanism for realizing educational goals and improving educational quality[5]. Both conceptualizations emphasize that engagement is not a static trait but a dynamic process influenced by an interplay of psychological and contextual determinants.

A growing body of research has identified multiple psychological antecedents of engagement, including grit[6], resilience, academic self-efficacy, self-regulated learning strategies, and positive emotions, all of which enhance persistence and intrinsic motivation[7, 8]. Contextual determinants—such as supportive teacher–student relationships, parental autonomy support, and peer collaboration—also play crucial roles by shaping students’ perceived autonomy, competence, and relatedness[7, 9]. However, while these studies have provided valuable insights into the drivers of engagement, the literature remains fragmented in integrating internal self-regulatory capacities (e.g., self-control) with higher-order motivational constructs (e.g., meaning in life) within contemporary technology-embedded learning environments. In particular, limited attention has been paid to how students’ psychological resources interact with external technological dependencies—such as generative artificial intelligence (GenAI)—to influence engagement outcomes. This omission leaves a critical gap in understanding the evolving mechanisms of motivation activation and academic persistence in digitalized higher education contexts.

Among these psychological predictors, self-control—a core regulatory capacity that enables individuals to suppress impulses, delay gratification, and persist toward long-term goals—has been consistently linked to academic persistence, learning motivation, and achievement outcomes[10]. Meanwhile, meaning in life—a motivational foundation grounded in the perception of purpose, value, and direction in one’s existence—has emerged as a key construct that enriches learning with significance and coherence. Individuals with a high sense of meaning in life tend to demonstrate greater persistence and goal-directedness in learning contexts[11].

Importantly, the rapid development of artificial intelligence (AI), particularly GenAI tools such as ChatGPT, DeepSeek, and Kimi, has profoundly reshaped university students’ learning practices. From information retrieval and comprehension to writing assistance and task automation, GenAI tools are increasingly embedded into students’ academic routines[12]. While these tools can enhance learning efficiency and cognitive accessibility, they have also raised concerns about potential overreliance, which may compromise students’ autonomy, self-regulation, and intrinsic motivation[13]. Some studies have warned that habitual dependence on GenAI tools may undermine students’ initiative and creativity, ultimately disrupting their goal setting, learning agency, and achievement motivation[14]. Hence, GenAI should be considered a novel environmental factor that may exert a moderating effect on internal psychological processes, warranting further empirical investigation.

Therefore, to bridge the aforementioned theoretical gaps, the present study integrates both psychological and contextual perspectives to examine how internal self-regulatory mechanisms (self-control) and motivational meaning-making (meaning in life) jointly predict academic engagement, while considering the moderating influence of GenAI dependence as a salient environmental factor. By situating the study within China’s rapidly digitalizing higher education system, it also highlights the cultural and regional significance of understanding engagement not merely as behavioral participation, but as a psychologically rich and value-oriented process.

Specifically, three research questions are examined: (1) Does self-control significantly predict academic engagement? (2) Does meaning in life mediate the relationship between self-control and academic engagement? (3) Does GenAI dependence moderate the relationship between self-control and meaning in life, thereby indirectly influencing engagement levels? Through addressing these questions, this study not only contributes to a more integrative understanding of the motivational foundations of academic engagement but also provides context-sensitive implications for fostering adaptive learning in AI-augmented environments.

Literature review and hypotheses

Theoretical framework

This study is grounded in two core theoretical perspectives: Self-Determination Theory (SDT) and Path Dependence Theory (PDT). Based on these frameworks, a moderated mediation model was developed in which self-control serves as the antecedent variable, meaning in life as the mediating variable, GenAI dependence as the moderating variable, and academic engagement as the outcome variable. The model aims to explore how university students’ internal regulatory mechanisms influence their learning behaviors in technology-embedded environments. Drawing upon SDT, the model conceptualizes self-control as a self-regulatory capability that supports the fulfillment of basic psychological needs (autonomy, competence, and relatedness), whereas meaning in life represents a higher-order motivational outcome that integrates these needs into a coherent sense of purpose and value. PDT, in contrast, provides a contextual lens for understanding how habitual reliance on GenAI technologies can alter or constrain these motivational processes through mechanisms of cognitive inertia and behavioral lock-in.

According to SDT, academic engagement is primarily driven by intrinsic motivation, which is activated when individuals’ basic psychological needs—autonomy, competence, and relatedness—are satisfied[15]. When students perceive a sense of self-direction, efficacy, and meaningful connection with others in the learning process, they are more likely to engage in learning activities in a sustained and positive manner, thereby enhancing both learning outcomes and subjective well-being[16]. When these needs are fulfilled, learners experience a sense of volition, capability, and connectedness, which sustains engagement even in challenging tasks[17]. Academic engagement, therefore, reflects not only the activation of intrinsic motivation but also the internalization of learning goals as personally meaningful and self-endorsed[18].

Within this motivational system, self-control can be understood as a volitional process that facilitates autonomous regulation rather than externally imposed control. Although self-control is not a core construct explicitly defined within SDT, it has been conceptualized as a self-regulatory mechanism that enables individuals to translate autonomous motivation into consistent goal-directed behavior[19]. Students who effectively manage impulses and sustain effort toward long-term learning goals demonstrate higher perceived competence and autonomy satisfaction, thereby enhancing engagement quality. Empirical evidence supports this view, showing that self-control positively predicts need satisfaction, autonomous motivation, and engagement across diverse cultural contexts [20].

Meaning in life, in turn, embodies the process of motivational internalization emphasized by SDT. As individuals integrate learning experiences into their broader life goals and values, they develop a sense of coherence and purpose that sustains intrinsic motivation [21]. Recent studies have demonstrated that meaning in life mediates the relationship between basic psychological need satisfaction and engagement, functioning as a meta-motivational resource that provides existential significance to learning activities [22]. Hence, within the SDT framework, self-control may indirectly promote academic engagement through its facilitation of meaning-making processes that transform extrinsic demands into personally meaningful pursuits.

However, in contemporary learning contexts where GenAI technologies are increasingly embedded, self-regulated learning processes do not operate in isolation. PDT offers a complementary explanation for how technological dependence can disrupt or reshape motivational pathways. Originating from institutional and behavioral economics [23, 24], PDT posits that once a particular behavioral pattern or technology use trajectory becomes established, it generates increasing returns, cognitive inertia, and structural lock-in, leading individuals to persist in habitual behaviors even when more adaptive alternatives are available [25, 26]. Applied to digital learning, GenAI dependence can be conceptualized as a path-dependent behavior where repeated reliance on AI-generated outputs fosters cognitive offloading and undermines self-initiated regulation [27].

This theoretical integration suggests that while SDT explains how internal resources drive engagement, PDT elucidates how external technological structures can constrain or distort these motivational dynamics. Specifically, habitual GenAI use may reduce autonomy and competence satisfaction by outsourcing cognitive effort, thereby weakening the positive linkage between self-control and meaning in life. In this sense, PDT does not replace SDT but complements it, providing a structural account of how technology-mediated environments may moderate self-regulatory processes predicted by SDT.

In summary, this study integrates SDT’s motivational framework with PDT’s contextual logic to construct a moderated mediation model in which self-control enhances academic engagement through meaning in life, while GenAI dependence serves as a contextual moderator that potentially weakens this indirect pathway. This synthesis not only aligns the study’s conceptual model with established theoretical traditions but also extends SDT to account for motivational adaptation under conditions of technological dependence.

Self-Control and academic engagement

In exploring the internal mechanisms underlying academic engagement, self-control has been increasingly recognized as a cornerstone of motivational self-regulation and goal-directed behavior within the framework of SDT. Self-control refers to an individual’s ability to consciously regulate thoughts, emotions, and behaviors in the face of immediate temptations or impulses, thereby enabling the pursuit of more valuable long-term goals[28]. From an SDT perspective, self-control represents a volitional process that enables learners to align their behavior with personally endorsed goals rather than externally imposed demands [29]. This form of autonomous regulation promotes sustained engagement by satisfying the basic needs for autonomy and competence—key psychological drivers of intrinsic motivation [30].

According to the Resource Allocation Model of self-control, individuals are willing to mobilize limited cognitive resources to sustain regulatory efforts when they recognize the importance of a task and perceive self-control behaviors as instrumental to goal attainment—even under conditions of resource depletion [31]. This model aligns closely with SDT’s notion of need-based motivation: when learners internalize the value of academic tasks, they are more likely to invest cognitive and emotional effort despite temporary fatigue or frustration [32].

Self-control not only facilitates volitional persistence but also operates as a meta-regulatory capacity that integrates emotion regulation, attention control, and delay of gratification. Students with high levels of self-control are more likely to effectively manage their study schedules, organize tasks strategically, and proactively resist external distractions. They tend to exhibit higher levels of task focus and emotional stability, allowing for greater consistency and goal-directed behavior throughout the learning process [33]. In turn, these self-regulatory processes contribute to sustained engagement by reducing self-discrepancy, enhancing task enjoyment, and reinforcing perceived competence [34, 35].

Empirical evidence has repeatedly confirmed the positive relationship between self-control and academic engagement across cultural and disciplinary contexts. For example, Yang et al. found that self-control indirectly promotes learning engagement through resilience and positive emotions among Chinese college students[10]. Similarly, de la Fuente et al. demonstrated that university students with higher self-control exhibit stronger cognitive persistence and lower burnout, mediated by academic self-efficacy[36]. Other studies reported that self-control predicts better time management, higher goal orientation, and less academic procrastination—factors that are strongly associated with engagement and achievement [37, 38]. Meta-analytic findings further indicate that self-control contributes not only to behavioral persistence but also to emotional and cognitive engagement through enhanced need satisfaction and reduced ego-depletion[29].

From a contemporary perspective, self-control may also serve as a critical psychological buffer in AI-enhanced learning environments. Students with higher self-control are better equipped to critically evaluate the utility of GenAI tools, maintain focus on long-term learning goals, and resist the temptation to over rely on algorithmic outputs for convenience. In this sense, self-control strengthens learners’ autonomous regulation, enabling them to use technology as a tool for competence enhancement rather than as a substitute for effort or cognition [39].

Based on this theoretical and empirical foundation, the following hypothesis is proposed:

H1: Self-Control positively predicts academic engagement.

Meaning in life as a mediator

Academic engagement is not only influenced by cognitive and behavioral factors but is also deeply shaped by individuals’ understanding of life’s meaning, which serves as a key psychological mediator linking self-regulation and motivation. Meaning in life is commonly defined as a person’s perception that life is purposeful, coherent, and significant [21, 40]. It is generally conceptualized along two dimensions: the presence of meaning, which reflects the extent to which individuals perceive their lives as meaningful, and the search for meaning, which indicates a proactive motivation to explore and construct personal meaning[41].

According to SDT, meaning in life can be understood as the outcome of motivational internalization—the process through which extrinsic goals are transformed into personally valued and self-endorsed motives [42]. When individuals perceive their learning as aligned with deeply held values and life purposes, they experience greater autonomy and competence satisfaction, which in turn strengthens intrinsic motivation. Thus, meaning in life serves as a meta-motivational construct that organizes self-regulatory efforts into coherent, value-consistent action patterns [41].

Empirical studies provide strong evidence for this theoretical mechanism. For instance, Cai et al. found that the presence of meaning promotes academic engagement through enhanced hope and motivational persistence, forming a reciprocal, dynamic feedback loop [41]. Similarly, Yuen and Datu demonstrated that students with a high sense of meaning exhibit stronger academic self-efficacy, better time management, and higher emotional regulation, which in turn foster sustained engagement and problem-solving [43]. Other studies within existential and positive psychology frameworks have shown that meaning in life mediates the relationship between basic psychological need satisfaction and well-being or engagement [44, 45], supporting its integrative motivational role.

Within the SDT framework, meaning in life can also be seen as the experiential manifestation of fully internalized motivation. Individuals who perceive coherence and purpose in their studies are more likely to engage deeply, not merely for external rewards but because learning expresses who they are and who they aspire to become. This process reflects identified and integrated regulation, the most self-determined forms of extrinsic motivation [42]. Consequently, meaning in life bridges the transition from self-control (volitional regulation) to academic engagement (motivated action), serving as the psychological channel through which effort becomes intrinsically energizing and sustainable.

Self-control has also been empirically identified as a significant antecedent of meaning in life. Li et al. demonstrated that self-control enhances individuals’ perceived meaning both directly and indirectly through the reinforcement of self-efficacy and future orientation [46]. Similarly, Cheng et al. found that young adults with higher self-control report greater purpose clarity and existential fulfillment, suggesting that volitional self-regulation supports meaning construction by enabling individuals to align behavior with long-term goals [47]. Additional research indicates that self-controlled individuals engage more frequently in reflective processing, prosocial goal pursuit, and deliberate life planning—processes that strengthen existential coherence and perceived purpose [20, 34].

Taken together, these findings support a theoretically grounded motivational pathway in which self-control promotes meaning in life by facilitating the internalization of values and goals, and meaning in life, in turn, enhances academic engagement by activating autonomous and purpose-driven learning motivation. Accordingly, the following hypothesis is proposed:

• H2: Meaning in life mediates the relationship between self-control and academic engagement.

The moderating role of GenAI dependence

GenAI dependence can be broadly defined as a psychological, cognitive, and behavioral state in which individuals exhibit excessive trust in and reliance on the functionalities of GenAI tools, often accompanied by a loss of perceived autonomy and self-efficacy in their learning or decision-making processes [13]. When such tools become unavailable, individuals may experience withdrawal-like symptoms—such as anxiety, helplessness, or confusion—reflecting impaired cognitive control and motivational disruption [48]. This dependence thus goes beyond instrumental use, evolving into a habitual and emotionally reinforced reliance on AI-mediated cognition.

From the perspective of PDT, GenAI dependence exemplifies a process of technological lock-in shaped by cumulative behavioral reinforcement and cognitive inertia [24]. Once learners adopt GenAI tools as default problem-solving aids, the immediate convenience and positive feedback they provide generate increasing returns, reinforcing habitual reliance. Over time, this habitual usage pattern becomes path-dependent: learners persist in relying on GenAI even when independent strategies might be more adaptive, due to sunk cognitive costs, reduced self-efficacy, and motivational inertia [39]. In this sense, GenAI dependence represents a behavioral lock-in mechanism that can reshape students’ motivational structures by shifting effort from intrinsic to extrinsic control sources.

Building on Park’s typology, GenAI dependence can be divided into two interrelated forms: Functional Dependence (FD)—a pragmatic overreliance on AI for task execution—and Existential Dependence (ED)—a deeper psychological attachment where GenAI becomes integrated into one’s identity and sense of capability [49, 50]. Functional dependence reflects users’ cognitive substitution of self-regulation with AI guidance: they may uncritically accept algorithmic outputs, gradually eroding critical thinking and self-initiated decision-making. Existential dependence, by contrast, represents an emotional and motivational attachment to AI, wherein individuals experience diminished meaning, confidence, and agency when detached from the system. Together, these forms illustrate how technological tools, when deeply embedded in personal and educational contexts, can influence both volitional regulation and existential motivation.

Integrating PDT with SDT clarifies how GenAI dependence moderates the motivational pathway between self-control and meaning in life. SDT posits that autonomy, competence, and relatedness are essential for intrinsic motivation and meaning construction [42]. However, under high GenAI dependence, these needs may be frustrated rather than satisfied, as cognitive and emotional regulation are externally offloaded to AI systems. This externalization of control undermines perceived autonomy and competence [27], thereby weakening the self-determined regulation that links self-control to meaning-making. In contrast, students with low dependence maintain a stronger sense of volitional agency, allowing self-control to effectively foster purpose, coherence, and intrinsic engagement [51].

Empirical findings align with this theoretical proposition. Prior studies have shown that high-frequency GenAI users report lower self-control, weaker reflective judgment, and diminished future orientation [52]. Overreliance on GenAI also increases decision inertia, reduces metacognitive monitoring, and delays self-initiated actions [53]. Furthermore, research in AI ethics and psychology suggests that excessive technological mediation may erode authentic goal construction, displacing intrinsic value with algorithmic convenience and leading to what Nyholm and Rüther (2023) describe as “existential outsourcing”[54]. Such tendencies directly counteract SDT’s principle of autonomous motivation and undermine the internalization process through which meaning in life is developed.

In summary, PDT provides a structural explanation for how habitual GenAI use can entrench behavioral dependence, while SDT explains how such dependence impairs motivational autonomy and meaning-making. By linking these perspectives, the current study posits that GenAI dependence attenuates the positive association between self-control and meaning in life—that is, when dependence is high, self-control is less effective in fostering existential purpose and coherent motivation.

Accordingly, the following hypothesis is proposed:

• H3: GenAI dependence negatively moderates the relationship between self-control and meaning in life.

Method

Procedure and sample

This study employed a cross-sectional survey design, with participants comprising undergraduate and graduate students (including both master’s and doctoral levels) from 20 provinces across China. A total of 1,409 questionnaires were distributed, and after excluding incomplete, duplicate, or logically inconsistent responses, 1,139 valid questionnaires were retained, resulting in an effective response rate of 80.84%. Among the valid respondents, 463 were male (40.6%) and 676 were female (59.4%). The average age of participants was 20.51 ± 2.65 years. Demographic information included gender and age, while discipline (science/engineering vs. humanities/social sciences) was reported solely to describe the sample composition and was not included as an analytical variable. Students majoring in science, engineering, agriculture, or related fields accounted for 476 participants (41.8%), while those from the humanities and social sciences—including physical education and the arts—accounted for 663 participants (58.2%). In terms of academic level, 962 undergraduate students (84.5%) and 177 graduate students (15.5%) participated in the study. The final sample size (N= 1,139) meets and exceeds the recommended thresholds for structural equation modeling (SEM). According to Kline and Wolf et al. [55, 56], samples larger than 500 are considered highly robust for models involving multiple latent variables. Therefore, the current sample provides sufficient statistical power and parameter stability for the hypothesized model.

Data were collected using an online questionnaire administered through the “Wenjuanxing” platform, which generated both a QR code and a direct survey link.

Prior to the formal administration of the survey, a pilot test was conducted to ensure the clarity of item wording, accuracy of content, and rationality of the scoring logic. Using a convenience sampling method, the study recruited 20 university students (10 males and 10 females) to participate in the pilot test. The primary objectives were to examine the clarity, interpretative consistency, contextual appropriateness of the questionnaire items, and the logical soundness of reverse-scored items.

During the pilot test, participants were asked to complete the questionnaire and subsequently rephrase the meaning of selected items in their own words. This procedure was implemented to assess whether their interpretation aligned with the original intent of the item design. Items for which the participants’ paraphrasing showed substantial deviation from the intended meaning were flagged as potentially ambiguous or unclear. Additionally, response patterns for both positively and negatively worded items were reviewed to verify that participants could correctly discern the meaning of reverse-scored items, thereby minimizing confusion attributable to item phrasing.

Based on participant feedback and paraphrasing results, several items were revised and optimized. For instance, “arts and sports categories” were incorporated into the “humanities and social sciences category” to maintain classification consistency and conceptual coverage. Another example includes modifying the original item “I believe AI tools will prioritize user interests over their own” to “I believe AI tools will place user interests above their own,” thereby enhancing semantic naturalness and logical fluency.

Results from the pilot test indicated that 18 out of the 20 participants (90%) were able to accurately paraphrase the meaning of all items. Only two participants exhibited interpretative deviations on specific items, which were accordingly revised based on their feedback. The overall questionnaire completion process proceeded smoothly, with an average completion time of 7.4 min. Participants generally considered the number of items appropriate, the wording clear, and the expressions natural. After revisions, the questionnaire achieved a high standard in terms of linguistic expression, content logic, and interpretative consistency, thereby establishing a solid reliability foundation and feasibility assurance for the formal survey administration.

The questionnaires were distributed with the assistance of educational administrators at participating universities. Participation was anonymous and voluntary.

This study was reviewed and approved by the Ethics Committee of Capital University of Physical Education and Sports, and research permission was obtained from all participating institutions.

Measurement instruments

Self-Control Scale (SCS)

Self-Control was measured using the Self-Control Scale (SCS) developed by Tangney et al.[28]and revised by Tan et al.[57]. The scale consists of 19 items covering five dimensions: resisting temptation, healthy habits, limiting entertainment, impulse control, and task focus. Sample items include “I am good at resisting temptation” and “I have a hard time breaking bad habits.” Responses were rated on a 5-point Likert scale ranging from 1 (very much like me) to 5 (not at all like me), with higher scores indicating greater self-control. In the original Chinese version of the study, the scale demonstrated an overall Cronbach’s α coefficient of 0.86. In the present investigation, the scale yielded a Cronbach’s α coefficient of 0.87, which is closely aligned with previous findings. This consistency indicates that the scale maintains satisfactory internal consistency reliability and measurement stability within the current sample.

Meaning in Life Questionnaire (MLQ)

The Meaning in Life Questionnaire (MLQ), originally developed by Steger et al.[40]and revised by Wang et al. for use in Chinese populations[58], was employed in this study. The scale consists of 10 items covering two dimensions: presence of meaning and search for meaning. Example items include “I understand my life’s meaning” and “I am looking for something that makes my life feel meaningful.” Responses were recorded on a 7-point Likert scale ranging from 1 (not at all true) to 7 (absolutely true), with higher scores indicating a stronger sense of meaning in life. The original Chinese version of the scale demonstrated sound internal consistency and temporal stability, with Cronbach’s α coefficients of 0.85 and 0.82 for the “Presence of Meaning” and “Search for Meaning” subscales, respectively. In the present study, the scale achieved an overall Cronbach’s α coefficient of 0.87, with coefficients of 0.94 and 0.79 for the two dimensions. These results indicate that the scale maintains high internal consistency reliability within the current sample, confirming the stability and reliability of the measurement outcomes. The observed variations in dimension-level coefficients fall within acceptable methodological expectations and align with established psychometric standards.

Utrecht Work Engagement Scale (UWES)

The Utrecht Work Engagement Scale (UWES) developed by Schaufeli et al. [3]and revised by Fang et al.[59] was used to assess students’ engagement in learning. The scale consists of 17 items covering three dimensions: vigor, dedication, and absorption. Example items include “I feel bursting with energy when studying,” “I am enthusiastic about my studies,” and “I am immersed in my studies.” Responses were rated on a 7-point Likert scale ranging from 1 (never) to 7 (always), with higher scores indicating a greater level of academic engagement. Prior research has established that this scale possesses high reliability within Chinese university student populations, reporting an overall Cronbach’s α coefficient of 0.95. The reliability coefficients for its constituent dimensions were 0.86 for Vigor, 0.91 for Dedication, and 0.91 for Absorption, collectively indicating satisfactory internal consistency and construct validity. In the current investigation, the scale demonstrated an overall Cronbach’s α coefficient of 0.95, with dimension-level coefficients of 0.88, 0.90, and 0.91, respectively. These estimates are closely aligned with those reported in previous studies, thus substantiating the scale’s robust measurement stability within the present sample.

Large Language Models Dependence Scale (LDS)

The GenAI dependence was measured using the Large Language Models Dependence Scale (LDS) developed by Li et al.[48]. To enhance participants’ understanding and ensure smooth completion of the questionnaire, minor wording adjustments were made during administration. Specifically, the original term “Large Language Models (LLMs)” was replaced with “GenAI” to improve clarity, while preserving the construct’s original meaning. Given that LLMs represent a specific subset of GenAI, we adopted the broader term “GenAI” to better capture users’ general perceptions and dependencies on generative AI systems within educational contexts. The scale comprises 18 items covering two dimensions: functional dependency and existential dependency. Sample items include “I believe GenAI are honest” and “I trust that GenAI will prioritize the user’s interests over their own.” A 5-point Likert scale was used, ranging from 1 (strongly disagree) to 5 (strongly agree), with higher scores indicating greater dependency on large language models. In the original study by Li et al., the scale demonstrated overall Cronbach’s α coefficients of 0.87 and 0.89 across two independent samples, with reliability coefficients for both subdimensions (FD and ED) exceeding 0.85, indicating satisfactory internal consistency and measurement reliability. In the present study, the scale yielded an overall Cronbach’s α coefficient of 0.71, with subdimensional coefficients of 0.84 and 0.89, respectively. These results confirm that the scale maintains adequate reliability within the current sample and remains suitable for assessing university students’ dependency characteristics toward generative artificial intelligence.

Data analysis procedures

To ensure data quality and the validity of subsequent analyses, this study systematically conducted data preprocessing and hypothesis testing using SPSS 26.0, following these specific procedures: First, missing value analysis was performed, which indicated no missing data in the dataset. Second, univariate and multivariate outlier screening was conducted using the ± 3 standard deviations (SD) criterion and Mahalanobis distance, respectively, leading to the identification and removal of extreme values. After this procedure, the valid sample size was reduced from 1,409 to 1,139, resulting in a usable response rate of 80.84%. Third, normality was assessed by examining the skewness and kurtosis values of all observed variables. The results showed that the absolute values of both skewness and kurtosis for all variables were below 2, indicating that the data reasonably met the univariate normality assumption. Fourth, multicollinearity was diagnosed by calculating the Variance Inflation Factor (VIF) and Tolerance statistics. All VIF values were below 10 and all Tolerance values exceeded 0.10, confirming the absence of substantial multicollinearity.

Since the scales utilized both 5-point and 7-point Likert-type response formats, all variables were standardized (Z-scores) prior to analysis to mitigate scale disparity and ensure coefficient comparability in the mediation and moderation analyses. Following data screening, the hypothesized model was tested using Structural Equation Modeling (SEM) in Mplus 8.3 with the Maximum Likelihood Estimation (MLE) method. To examine the mediation and moderation effects, the bias-corrected bootstrap method was employed with 5,000 resamples and a 95% confidence interval. Effects were considered statistically significant if their confidence intervals did not contain zero. All reported path coefficients are standardized estimates, with the significance level set at α < 0.05.

Results

Common method bias

To control for and assess potential common method bias (CMB), several procedural remedies were implemented during questionnaire design, including randomization of item order, inclusion of reverse-coded items, and assurances of anonymity and confidentiality. After data collection, Harman’s single-factor test was conducted using exploratory factor analysis (EFA) without rotation to statistically evaluate the presence of CMB[60]. Results indicated that 10 factors with eigenvalues greater than 1 were extracted, with the first factor accounting for only 20.34% of the total variance—well below the critical threshold of 40%[60]. This suggests that no single factor dominated the variance structure, indicating a low likelihood of serious common method bias.

In addition, confirmatory factor analysis (CFA) was performed using a single-factor model including all self-reported items to further test for CMB. The model fit was found to be poor: χ²/df = 14.67, RMSEA = 0.11, SRMR = 0.14, CFI = 0.36, GFI = 0.38, AGFI = 0.34, NFI = 0.35, TLI = 0.34, RFI = 0.32. These results provide further evidence that common method bias is not a serious concern in this study.

Descriptive statistics and correlation analysis

After controlling for age and gender, Pearson correlation analysis was conducted among self-control, meaning in life, academic engagement, and GenAI dependence. The results are presented in Table 1. Self-control was positively correlated with meaning in life (r = 0.25, p < 0.001) and academic engagement (r = 0.35, p < 0.001), but negatively correlated with GenAI dependence (r = −0.19, p < 0.001). Meaning in life showed a significant positive correlation with academic engagement (r = 0.44, p < 0.001), but was not significantly associated with GenAI dependence (r = 0.02, p = 0.44). Additionally, GenAI dependence was negatively correlated with academic engagement (r = −0.08, p = 0.01). These preliminary findings provide foundational support for subsequent hypothesis testing.

Table 1.

Descriptive statistics and partial correlations among variables (N = 1139)

Variable	M	SD	1	2	3	4
1.Self-Control	58.83	9.94	-
2.Meaning in Life	47.43	8.46	0.25***	-
3.Academic Engagement	79.84	14.62	0.35***	0.44***	-
4.GenAI Dependence	53.84	5.34	−0.19***	0.02		-

*p<0.05

**p<0.01

***p<0.001

Measurement invariance test

To examine whether the measurement structure of the scale used in this study is consistent across different gender groups, tests for configural, metric, scalar, and strict invariance were conducted respectively. The model fit indices were as follows: configural invariance model (CFI = 0.98, RMSEA = 0.05), metric invariance model (CFI = 0.98, RMSEA = 0.05), scalar invariance model (CFI = 0.98, RMSEA = 0.05), and strict invariance model (CFI = 0.98, RMSEA = 0.05). Model comparisons revealed that the changes in CFI (ΔCFI) between adjacent models ranged from 0 to 0.003, all below the critical cutoff of 0.01, while the changes in RMSEA (ΔRMSEA) ranged from 0.001 to 0.002, all below the critical cutoff of 0.015. These results demonstrate that the measurement instrument used in this study has achieved full measurement invariance across gender groups.

Testing the mediating effect of meaning in life

Through structural equation modeling, the mediating effect of self-control as the independent variable, meaning in life as the mediator, and learning engagement as the dependent variable was examined. The model demonstrated a good fit: χ²/df = 7.61, RMSEA = 0.07, SRMR = 0.04, CFI = 0.98, TLI = 0.96. After controlling for gender and age, self-control significantly predicted meaning in life (β = 0.38, p < 0.001), and meaning in life significantly predicted academic engagement (β = 0.53, p < 0.001). The indirect effect of self-control on academic engagement through meaning in life was significant (β = 0.20, p < 0.001), accounting for 52.63% of the total effect. In addition, the direct effect of self-control on academic engagement remained significant even after including the mediator (β = 0.18, p < 0.001), explaining 47.37% of the total effect. These results indicate that meaning in life significantly mediated the relationship between self-control and academic engagement. The detailed results are presented in Table 2.

Table 2.

Testing the mediating effect of meaning in life

Path relationship	β	SE	95%CI	p	Ratio
SC→ML	0.38	0.04	[0.29,0.47]	< 0.001	-
ML→AE	0.53	0.06	[0.42,0.64]	< 0.001	-
SC→ML→AE	0.20	0.03	[0.13,0.26]	< 0.001	52.63%
SC→AE	0.18	0.04	[0.10,0.27]	< 0.001	47.37%
Total effect	0.38	0.03	[0.30,0.43]	< 0.001	100%

SC Self-Control, ML Meaning in Life, AE Academic Engagement

Testing the moderating effect of GenAI dependence on the mediating path

To examine whether GenAI dependence moderates the mediating pathway “self-control → meaning in life → learning engagement,” this study constructed a latent interaction model using Mplus and compared it with a baseline mediation model that did not include the interaction term. The results indicated that the baseline model demonstrated a good fit to the data (χ²/df = 4.14, RMSEA = 0.05, SRMR = 0.03, CFI = 0.98, TLI = 0.97). After incorporating the latent interaction term between self-control and GenAI dependence, the full model showed a decrease in the Akaike Information Criterion (AIC) by 6.43. Furthermore, a likelihood ratio test yielded a significant result (Δ−2LL = 8.43, Δdf = 1, p < 0.01), indicating that the model including the moderation term provided a better fit to the data and possessed superior explanatory power compared to the baseline model. These findings support the further investigation of the moderating effect of GenAI dependence.

At the structural path level (see Fig. 1), self-control was found to significantly and positively predict learning engagement (β = 0.19, p < 0.001) and meaning in life (β = 0.40, p < 0.001). Meaning in life, in turn, also demonstrated a significant positive prediction on learning engagement (β = 0.51, p < 0.001). Furthermore, GenAI dependence significantly and positively predicted meaning in life (β = 0.10, p = 0.01). More critically, the interaction term between self-control and GenAI dependence exhibited a significant negative predictive effect on meaning in life (β = −0.11, p < 0.01). This indicates that GenAI dependence exerted a significant negative moderating effect on the first stage of the mediating pathway (self-control → meaning in life). Specifically, as the level of GenAI dependence increases, the positive influence of self-control on meaning in life is significantly attenuated.

Fig. 1.

Path diagram of the moderated mediation model. Note: * p＜0.05, ** p＜0.01, *** p＜0.001

To further elucidate the specific pattern of the moderating effect, simple slopes for the effect of self-control on meaning in life were plotted at high (+ 1 SD) and low (−1 SD) levels of GenAI dependence (see Fig. 2). The results revealed that the positive predictive effect of self-control on meaning in life was significant and stronger in the low GenAI dependence group (−1 SD). In the high GenAI dependence group (+ 1 SD), the positive relationship between self-control and meaning in life remained significant, but the slope was noticeably attenuated. This pattern provides clear visual evidence that higher levels of GenAI dependence weaken the positive role of self-control in fostering an individual’s sense of meaning in life, thereby confirming a negative moderating effect.

Fig. 2.

The moderating effect of GenAI dependence on the relationship between self-control and meaning in life

To more rigorously test whether GenAI dependence influences the strength of the entire mediation pathway (i.e., whether moderated mediation exists), this study further compared the conditional indirect effects at different levels of GenAI dependence. The results indicated that under the condition of low GenAI dependence (−1 SD), the indirect effect of self-control on learning engagement via meaning in life was 0.25, with a 95% confidence interval of [0.18, 0.32], which did not include zero, indicating a significant mediation effect. Under the condition of high GenAI dependence (+ 1 SD), this indirect effect decreased to 0.14, with a 95% confidence interval of [0.08, 0.20], which also did not include zero, suggesting that the mediation effect remained significant but was attenuated in magnitude. More critically, the difference in the indirect effects between the two conditions was − 0.11, with a 95% confidence interval of [−0.18, −0.03], which again did not include zero. This demonstrates that a higher level of GenAI dependence significantly weakened the strength of the indirect pathway “self-control→meaning in life→learning engagement”. Taken together, these findings provide comprehensive support for the hypothesized moderated mediation model.

Discussion

This study constructed a moderated mediation model to explore the psychological mechanisms underlying academic engagement among university students. Specifically, self-control was proposed as the independent variable, meaning in life as the mediating variable, and GenAI dependence as the moderating variable. The results revealed that: (1) self-control positively predicted academic engagement; (2) meaning in life mediated the relationship between self-control and academic engagement; and (3) GenAI dependence negatively moderated the path from self-control to meaning in life. Together, these findings highlight how the dynamic interaction between internal psychological resources and external technological dependence shapes students’ academic engagement in technology-rich learning environments.

The influence of Self-Control on academic engagement

This study found that self-control significantly and positively predicted academic engagement, supporting Hypothesis 1. This result aligns with the Self-Determination Theory (SDT), which conceptualizes self-control not merely as behavioral restraint but as a volitional capacity enabling learners to act in accordance with self-endorsed goals. By regulating impulses in a way that sustains autonomy and competence satisfaction, self-control reinforces intrinsic motivation and energizes sustained engagement [29, 42]. The current finding also corroborates earlier empirical studies demonstrating that students with higher self-control tend to show stronger learning persistence, lower procrastination, and higher emotional stability across diverse educational contexts [10, 20, 33].

From a theoretical perspective, this study extends prior SDT-based research by demonstrating that self-control functions as a volitional bridge between basic psychological needs and sustained academic engagement. Whereas SDT emphasizes the role of autonomy and competence in promoting motivation, our findings suggest that self-control provides the regulatory strength necessary to translate these needs into consistent, goal-directed learning behavior. Thus, self-control represents the operational mechanism through which motivational energy is stabilized and directed toward long-term academic goals.

Two potential mechanisms help explain this predictive relationship. First, self-control facilitates the management of psychological conflicts between multiple goals, thereby sustaining continuous academic engagement. Consistent with SDT’s notion of integrated regulation, learners high in self-control can align short-term behaviors with long-term values, maintaining motivational coherence even under conflicting demands. Berrios et al. noted that when individuals face goal conflicts, they often experience emotional tension between long-term aspirations and immediate desires. These mixed emotions can activate self-regulatory efforts, helping individuals make decisions that favor long-term development [61]. Sharabi and Roth further showed that students who regulate emotions effectively after academic failure display higher adaptive persistence—supporting the idea that self-control enhances engagement through emotion-regulatory pathways consistent with SDT’s competence need [62].

Second, self-control enhances persistence in learning through the development of delayed gratification. According to Mischel’s Cognitive-Affective Personality System theory, an individual’s ability to activate the “cool system” to inhibit the “hot system” in tempting situations reflects the level of self-control [63]. Within the SDT framework, this ability corresponds to maintaining autonomy-driven regulation rather than impulsive, externally controlled reactions. Students who exercise delayed gratification are better able to sustain autonomy and competence satisfaction, both of which directly promote academic engagement. They are more likely to employ cognitive strategies—such as task visualization, self-monitoring, and goal planning—to redirect attention from instant gratification to academic tasks [64]. Over the course of long-term academic engagement, these strategies enable individuals to continually mobilize internal resources in response to complex and dynamic academic contexts, thereby sustaining motivation and efficacy and fostering persistent, goal-oriented academic behavior. This mechanism parallels findings by Liu et al. [38]and Tangney et al.[28], who reported that self-control enhances engagement by reinforcing academic self-efficacy and resilience—two constructs conceptually rooted in competence satisfaction within SDT.

Compared with previous studies that treated self-control as a static personality trait, the present research highlights its dynamic, process-oriented nature. Our results indicate that self-control interacts with motivational systems to promote sustained engagement, thereby refining SDT by identifying a volitional mechanism underlying motivational maintenance. This integration not only broadens the theoretical application of SDT to self-regulatory behavior but also complements recent models emphasizing the synergy between emotion regulation, need satisfaction, and goal persistence.

Practically, these results have actionable implications for educational policy and psychological support. Universities should implement interventions that cultivate both the cognitive and emotional components of self-control—such as goal-setting workshops, time management training, and emotional regulation programs—designed to strengthen students’ volitional persistence. At the same time, educational counselors can use self-control indicators to identify students at risk of disengagement, allowing early intervention before motivational decline occurs. In learning environments increasingly shaped by GenAI tools, institutions should also emphasize the responsible use of technology, encouraging students to view GenAI as an assistive resource rather than a substitute for effort and autonomy. Such approaches can foster balanced, autonomy-supportive learning environments that align with the psychological mechanisms revealed in this study.

The mediating role of meaning in life between Self-Control and academic engagement

This study found that meaning in life partially mediated the relationship between self-control and academic engagement, supporting Hypothesis 2. This result aligns with the core tenets of SDT, which posits that students are more likely to engage in learning activities in a persistent and positive manner when they perceive autonomy, competence, and relatedness in the learning process, ultimately enhancing both learning quality and well-being [16]. Meaning in life reflects the internalization and integration of these basic needs into a coherent sense of purpose and direction, transforming learning from an externally driven activity into a self-endorsed pursuit. By providing existential coherence and value alignment, meaning in life acts as a motivational bridge that translates self-regulatory strength into sustained academic engagement.

Theoretically, this finding extends SDT by identifying meaning in life as a higher-order construct that captures the outcome of full motivational internalization. When students’ efforts are guided by a clear sense of purpose and personal values, their regulatory behavior (supported by self-control) becomes self-concordant rather than externally enforced [21] Hence, meaning in life operationalizes the transition from controlled to autonomous regulation, representing a psychological mechanism through which volitional self-control transforms into intrinsic motivation and engagement.

Several mechanisms may explain this mediating process. First, self-control facilitates the construction of meaning in life by enabling future-oriented thinking, goal integration, and self-efficacy. Li et al. found that individuals with high self-control tend to focus their attention on the future, actively setting goals and envisioning successful outcomes[46]. This future-oriented mindset helps individuals integrate their understanding of life goals, direction, and values into their actions, thereby fostering a stronger sense of meaning in life[46]. In addition, Cheng et al., in a study of Chinese youth, reported that self-control enhances individuals’ action-oriented and coping self-efficacy, which in turn promotes goal persistence and self-identification in the face of challenges, leading to a deeper experience of meaning in life[47]. These findings suggest that self-control nurtures meaning construction by sustaining volitional regulation and enabling students to integrate effort with life direction, thereby reinforcing SDT’s principles of autonomy and competence satisfaction.

Second, meaning in life contributes to academic engagement through both indirect and direct pathways—by enhancing hope and by directly motivating learning behavior. Cai et al. highlighted hope as a key psychological resource linking meaning in life with academic engagement, emphasizing that students with a strong sense of meaning perceive clearer goals, higher attainability, and stronger agency in learning. This motivational clarity increases both willingness and effort investment, leading to higher behavioral and emotional engagement [43]. Beyond hope, meaning in life fosters a sense of coherence and control over one’s learning environment [55]. Students who find meaning in academic pursuits tend to interpret challenges as purposeful and self-defining experiences, reinforcing intrinsic motivation and resilience—key elements of sustained engagement [44].

The current findings not only align with but also expand prior empirical evidence. While previous research has linked self-control to engagement primarily through self-efficacy or resilience[10, 33], the present study demonstrates that meaning in life provides an additional existential–motivational pathway that complements these cognitive mechanisms. This integrative perspective bridges volitional regulation (self-control) and existential motivation (meaning), thereby refining SDT’s explanatory scope by connecting motivational energy with value-based purpose.

Practically, these findings carry important implications for higher education and mental health promotion. On one hand, educators should go beyond cultivating behavioral discipline to foster students’ reflective understanding of “why they learn”, helping them connect coursework to personal growth, career goals, and societal contribution. Programs in life design, reflective journaling, and purpose-oriented mentoring can strengthen students’ sense of meaning and internal motivation. On the other hand, psychological support services should address the absence or fragmentation of meaning, which may underlie academic disengagement, burnout, or procrastination. Interventions that integrate self-control training with meaning-centered education (e.g., purpose workshops, existential goal-setting) can enhance both motivational depth and engagement persistence, contributing to students’ long-term well-being and learning success.

The moderating effect of GenAI dependence on the mediating path

This study found that GenAI dependence significantly moderated the effect of self-control on meaning in life in a negative direction, supporting Hypothesis 3. This finding integrates the perspectives of SDT and PDT, revealing that technological dependence can alter motivational processes by constraining psychological autonomy and volitional regulation. Specifically, while SDT emphasizes that meaning-making emerges from the internalization of self-endorsed goals and satisfaction of autonomy and competence needs [42], PDT posits that once habitual behavioral trajectories are established, individuals tend to persist in them due to cognitive inertia and increasing returns[23]. In this study, habitual reliance on GenAI exemplifies such a path-dependent process, which disrupts the motivational sequence linking self-control to meaning in life.

From a theoretical standpoint, this moderating effect illustrates how external technological structures can constrain the internal motivational dynamics proposed by SDT. When students rely excessively on GenAI for cognitive processing, feedback, or decision support, the satisfaction of autonomy and competence becomes externally regulated rather than self-determined. Over time, this externalization of agency undermines the internal motivational mechanisms that typically connect self-control with purpose construction. Hence, high GenAI dependence represents an environmental constraint that weakens self-determined functioning and impedes the volitional transformation of effort into existential meaning.

Two primary mechanisms help explain this moderating pattern. First, GenAI dependence may undermine individuals’ metacognitive regulation abilities, thereby reducing their psychological resilience when facing setbacks and challenges. In contexts where GenAI provides highly convenient and immediate feedback, students may progressively offload complex cognitive and evaluative functions to technology—a process described as cognitive offloading [65]. While such offloading can temporarily enhance performance, persistent reliance can lead to “metacognitive inertia,” reducing self-monitoring, reflection, and goal regulation [66]. Studies have shown that excessive GenAI use diminishes knowledge transfer and long-term retention, and significantly decreases self-regulatory behaviors such as planning and self-assessment [67]. This attenuation of metacognitive activity constrains the functional impact of self-control on meaning-making, since students become less able to connect their academic effort with internalized goals and life values.

Second, GenAI dependence may erode psychological security and autonomous stability, fostering a structural reliance on external feedback systems. Students accustomed to instantaneous responses and standardized evaluations in GenAI-driven learning environments often experience reduced tolerance for uncertainty and ambiguity [66]. When confronted with complex, ill-structured, or delayed-feedback tasks, such individuals are prone to anxiety and helplessness, reflecting diminished autonomy and competence satisfaction[54, 65]. As a result, their self-control no longer serves as a stabilizing force for self-determined action but becomes contingent upon external guidance. This shift from internal to external regulation interrupts the motivational translation of self-control into meaning in life, as students increasingly depend on technological scaffolds rather than intrinsic goals to maintain direction and purpose.

Together, these mechanisms underscore that GenAI dependence disrupts the self-regulatory pathway between self-control and meaning in life by externalizing cognitive control, undermining autonomy, and weakening volitional resilience. Within the SDT framework, this pattern reflects a regression from autonomous to controlled motivation, illustrating how technological path dependence can reshape motivational architecture in higher education contexts.

The findings carry critical theoretical and practical implications. Theoretically, this study extends SDT by demonstrating that digital path dependence constitutes an external constraint on need satisfaction, clarifying how autonomy frustration emerges in technology-mediated environments. At the same time, it enriches PDT by applying it to psychological processes, showing that path-dependent reliance on GenAI can lead to motivational lock-in and reduced adaptive flexibility. Practically, these insights suggest that while GenAI can enhance learning efficiency, educators and institutions must cultivate digital self-regulation and reflective use habits to preserve students’ motivational autonomy.

Programs promoting “technology as support, not substitution” should be embedded in curricula—emphasizing metacognitive monitoring, critical evaluation of AI outputs, and autonomous problem-solving. Training modules on digital literacy and reflective technology use can mitigate overdependence and sustain the internal motivational processes necessary for meaning construction. In psychological counseling and academic advising, practitioners should assess not only students’ self-control levels but also their patterns of GenAI reliance, helping them re-establish volitional balance between technological aid and intrinsic motivation.

Theoretical and practical implications

Grounded in SDT and PDT, this study constructed a moderated mediation model linking self-control, meaning in life, and academic engagement, systematically revealing the dynamic interplay between internal motivational regulation and external technological dependence in GenAI-driven learning contexts. The results collectively indicate that while self-control and meaning in life function as core psychological resources promoting autonomous engagement, habitual GenAI reliance may act as a structural constraint that undermines these internal processes.

Theoretical implications

First, this study enriches SDT by empirically identifying “meaning in life” as a higher-order outcome of motivational internalization. Within SDT, motivation quality depends on the satisfaction of autonomy, competence, and relatedness [42]. Our findings extend this principle by showing that meaning in life operationalizes the full internalization of these needs—transforming self-control-driven regulation into intrinsically meaningful and value-consistent learning behavior [21]. This conceptual expansion clarifies how existential purpose emerges as a product of sustained volitional effort, thereby deepening SDT’s explanatory scope regarding long-term motivation.

Second, the study advances SDT’s understanding of volitional regulation by specifying self-control as a functional mechanism linking need satisfaction with persistence and engagement. Prior research has emphasized self-control as a trait-like predictor of achievement [33], but our model reframes it as a process-oriented volitional capacity that stabilizes motivation under fluctuating emotional and contextual demands. This refinement highlights self-control as an active enabler of autonomy-driven motivation rather than a purely inhibitory force, integrating it more coherently within SDT’s framework of self-determined action [17, 29].

Third, the inclusion of GenAI dependence as a moderating factor extends motivational theory into the digital era by bridging SDT and PDT. While SDT focuses on internal regulation, PDT explains how external technological systems create “behavioral lock-in” that constrains psychological autonomy [23, 26]. The observed negative moderation of GenAI dependence demonstrates that path-dependent reliance on technology can weaken the volitional and existential pathways essential for meaning-making. This integration expands the boundary of SDT, providing a theoretical foundation for understanding “motivational alienation under technological mediation”—a novel concept describing how digital convenience can erode self-determined motivation and existential coherence in learning contexts [27, 54].

Together, these contributions clarify how internal and external regulatory forces interact, offering a more comprehensive theoretical model for explaining student engagement in technology-intensive environments.

Practical implications

Practically, the results provide actionable strategies for enhancing academic engagement and psychological resilience among university students.

Educational institutions should implement structured interventions that develop students’ cognitive, emotional, and behavioral self-regulation skills, such as time-management workshops, mindfulness-based self-control training, and goal-monitoring systems. Programs emphasizing delayed gratification, emotion regulation, and reflective decision-making can strengthen autonomy and competence satisfaction—key mechanisms linking self-control to engagement [18, 38].

Learning support systems should include meaning-oriented curricula that encourage students to explore “why they learn.” Universities can embed purpose workshops, reflective writing, and career-life planning modules into general education programs, helping students align academic tasks with personal values and life goals. Such interventions have been shown to reduce burnout and increase sustained engagement through enhanced existential purpose [41, 43].

In an era of pervasive AI technologies, educators should adopt a dual strategy of technological empowerment and psychological protection. While GenAI tools can enhance learning efficiency, students must be guided to use them as cognitive aids rather than substitutes for effort. Higher education institutions should integrate digital self-regulation training into curricula—covering metacognitive monitoring, critical evaluation of AI outputs, and self-reflective learning design. This approach upholds the SDT principle of autonomy by ensuring that technology supports rather than replaces self-directed motivation [65, 66].

Academic advisors and mental health counselors should assess both students’ self-control and GenAI usage patterns, identifying those at risk of motivational decline. Interventions combining self-control enhancement with meaning-centered therapy can restore volitional balance and psychological well-being, mitigating risks of disengagement, dependence, or burnout.

In summary, this study offers a theoretically grounded and practically actionable framework for promoting student engagement in AI-embedded learning environments. By integrating volitional training, meaning construction, and responsible technology use, universities can foster autonomous, purposeful, and resilient learners capable of thriving amid rapid digital transformation.

Limitations and future research

Despite its contributions, this study has several limitations that should be addressed in future research.

First, the study employed a cross-sectional survey design, which can reveal associations among variables but cannot establish causal relationships or capture their dynamic evolution over time. In reality, the development of self-control, the construction of meaning in life, and changes in academic engagement often occur in stages and accumulate gradually, influenced by time-related factors. Therefore, future research should consider longitudinal designs or experimental methods to better track changes over time and uncover the causal mechanisms and developmental trajectories of the proposed model.

Second, the study relied primarily on self-reported questionnaires. Although validated scales were used and efforts were made to control for measurement error, the risk of common method bias and social desirability effects cannot be entirely ruled out. To improve the reliability and explanatory power of future findings, researchers may adopt a mixed-methods approach by incorporating behavioral data, interview materials, or physiological indicators.

Third, the sample consisted mainly of university students from mainland China, which may limit the generalizability of the findings. Future studies should expand the sample base and adopt a cross-cultural perspective to explore the moderating roles of factors such as academic discipline, digital literacy, and cultural values. This would help provide more context-sensitive and nuanced theoretical support for educational interventions in digital learning environments.

Furthermore, the study’s sample was drawn from 20 provinces, with participants exhibiting variability in educational level and disciplinary background. To account for this potential nested group structure, we attempted to conduct a clustered (multilevel) model analysis using Mplus to examine potential clustering effects attributable to province, educational level, and disciplinary category. However, during model specification, the overall fit of the multilevel models failed to achieve an acceptable standard, precluding reliable parameter estimation. Preliminary comparisons indicated limited inter-provincial differences, reducing the empirical necessity for multilevel modeling. Consequently, the primary analyses did not further incorporate clustering effects. This limitation implies that unmodeled provincial or educational background variations may affect the precision of parameter estimates, thereby presenting potential implications for the generalizability of the findings. Future research could re-examine clustering effects within sample structures more suitable for multilevel modeling to enhance the robustness of the conclusions.

Conclusion

Drawing on SDT and PDT, this study systematically examined the mechanisms through which self-control, meaning in life, and GenAI dependence influence university students’ academic engagement. The results demonstrated that self-control positively predicted academic engagement both directly and indirectly by enhancing individuals’ sense of meaning in life. This confirms the mediating role of meaning in life in the relationship between self-control and academic engagement. Moreover, GenAI dependence was found to significantly moderate the indirect pathway by weakening the positive effect of self-control on meaning in life, highlighting a disruptive mechanism of technological dependence in the formation of academic motivation. By integrating both psychological regulation and technological impact, the study offers a novel theoretical perspective for understanding how students adapt to learning in AI-embedded educational contexts.

Based on these findings, it is recommended that universities strengthen the systematic cultivation of students’ self-control abilities and guide them in developing a strong sense of meaning in life to stimulate intrinsic learning motivation. At the same time, educators should remain vigilant about the risks of path dependence associated with GenAI tools. Establishing appropriate usage norms and promoting students’ metacognitive skills and self-regulation capacity are essential for fostering rational and healthy learning habits in the age of intelligent technology.

Author contributions

XC: Investigation, data management, software, chart making, and writing. MC: Investigation and data management. PS, XX and KM: Investigation and resources. LY: Fund acquisition, methodology, investigation, resources, supervision, and writing.

Funding

This work was supported by the Science and Technology School Strengthen Project of CUPES (155225002/009); China-Montenegro Science and Technology Cooperation Committee “China-Montenegro National Report on Physical Literacy for Children and Adolescents in Schools” Project (4 − 2).

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Declarations

Ethics approval and consent to participate

The study adhered to the ethical principles outlined in the Declaration of Helsinki and received approval from the Ethics Committee of Capital University of Physical Education and Sports (Approval No.2029A092). Prior to participation, all participants were informed about the purpose and procedures of the study. Completion of the questionnaire was considered to indicate informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.