Contemporary technological advancements, particularly in artificial intelligence (AI), are fundamentally transforming multiple domains of human activity, including education, healthcare, and business (Kumar, 2019). Within education settings, AI-powered tools have attracted considerable scholarly interest due to their potential to optimize learning experiences and academic performance (Xu, 2024). However, the accelerated integration of AI in educational settings has also prompted concerns regarding its possible adverse effects on student well-being (Klimova & Pikhart, 2025). While AI applications have demonstrated efficacy in facilitating personalized learning, automating academic tasks, and enhancing student engagement, excessive or inappropriate usage may cause detrimental outcomes, including heightened stress, technological dependency, and cognitive overload (Nizamani et al., 2024; Sajja et al., 2024). This dichotomy has led researchers to characterize AI’s influence on student life as a “double-edged sword” (Chen & Lin, 2024, p. 147).

 

In educational contexts, well-being has been found to have a robust correlation with various factors, including academic achievement, mental and emotional well-being, social cohesion, and overall life satisfaction (Kaya & Erdem, 2021; Michalos, 1991). As AI tools become increasingly pervasive in students’ daily academic routines, understanding their impact on well-being is imperative for developing strategies that harness their benefits while minimizing potential harm. Prior research has underscored the importance of individual differences in shaping the relationship between technology use and well-being, with self-efficacy and ego depletion emerging as critical factors (Huang et al., 2024; Reinecke et al., 2014). Self-efficacy, defined as an individual’s confidence in their capacity to successfully execute tasks (Bandura, 1995), has been identified as a mediator between technology usage and well-being (Huang et al., 2024). Conversely, ego depletion, which denotes the diminishment of cognitive resources following sustained self-regulatory effort, may amplify the negative consequences of AI overuse, thereby impairing well-being (Luqman et al., 2021).

 

Despite burgeoning scholarship on AI in education, it remains unclear whether AI usage can predict student well-being through the mediating mechanisms of self-efficacy and ego depletion. While existing studies have predominantly emphasized the positive contributions of AI to learning outcomes, little attention has been paid to its potential drawbacks, especially concerning mental and emotional health. This study sought to bridge these gaps by investigating the double-edged sword effect of AI usage on student well-being, with a specific focus on the mediating roles of self-efficacy and ego depletion. By doing so, I aimed to advance a more nuanced understanding of AI tools for student well-being, ultimately informing evidence-based strategies for its responsible and beneficial integration into educational ecosystems.

This study employed conservation of resources (COR) theory (Hobfoll et al., 2003) to explain the dual pathways through which AI usage can predict student well-being. COR theory posits that resource loss primarily drives psychological distress, while resource gain enhances well-being and fosters resilience against future losses. Within modern educational ecosystems, AI tools function as both potential resource enhancers and drains, yielding complex effects on student well-being.

AI facilitates the creation of personalized learning paths that cater to the distinct learning style and pace of individual students (Deng et al., 2024). Adaptive learning platforms leverage algorithms to assess student performance and offer tailored recommendations for supplementary resources or practice activities. This personalized approach empowers students to exert greater control over their learning journey, thereby bolstering their confidence and self-efficacy (Savio-Ramos, 2015). Furthermore, AI provides students with instant feedback on their work and tracks their progress over time (Harry & Sayudin, 2023). This feedback enables students to pinpoint areas for improvement and adjust their learning strategies accordingly, thereby enhancing their self-efficacy (Duijnhouwer et al., 2010). Additionally, AI automates routine tasks, freeing up students’ time to engage in more meaningful learning activities (Haluza & Jungwirth, 2023). This reduction in workload boosts motivation and contributes to high self-efficacy (Bandura, 1995).

 

Empirical evidence has consistently demonstrated that self-efficacy positively predicts well-being (Yu & Luo, 2018). High self-efficacy enables individuals to effectively navigate challenges, whereas low self-efficacy exacerbates perceived task difficulty, amplifying stress and anxiety (Çakar, 2012; Pajares, 2003). Students with robust self-efficacy exhibit greater perceived control over their academic environment, which correlates with improved well-being (Santos et al., 2014). Thus, this study proposed the following hypothesis:

Hypothesis 1: Artificial intelligence usage will have a positive indirect relationship with well-being through the mediator of increased self-efficacy.

A critical adverse effect of AI usage is cognitive overload (Nizamani et al., 2024). The fragmented, multiple-task-oriented nature of AI interactions strains working memory capacity, as students must continuously process and integrate AI-generated information (Naseer et al., 2025). Similar to findings in social media research, this cognitive fragmentation depletes attentional resources, impairing academic performance (Zhang & Mao, 2023). Overreliance on AI may also erode self-regulatory capacity, as students tend toward AI-assisted task completion, thereby exacerbating ego depletion (Inzlicht & Schmeichel, 2012; Rodríguez-Ruiz et al., 2025). Furthermore, excessive dependence on AI compromises critical thinking and intellectual autonomy, further depleting cognitive resources (Shanmugasundaram & Tamilarasu, 2023).

 

Well-being is contingent upon self-control resources (Orkibi & Ronen, 2017). Students with adequate self-control are better equipped to manage their time, set goals, and overcome negative emotions, all of which contribute to enhanced well-being (King & Gaerlan, 2014). However, when these resources are depleted due to AI usage, students may encounter difficulties in maintaining their well-being (Jin et al., 2024). Thus, this study proposed the following hypothesis:

Hypothesis 2: Artificial intelligence usage will have a negative indirect relationship with well-being through the mediator of increased ego depletion.

 

The research model of this study is shown in Figure 1.

The study employed a cross-sectional research design, conducting data collection using electronic devices under the supervision of a research assistant. I selected participants from management degree programs based on convenience and accessibility. Initial contact was established with university professors overseeing eligible college groups to solicit their collaboration in data collection during scheduled class sessions. Following professor approval, students were briefed on the objectives of the study, assured of the anonymity and confidentiality of their responses, and informed that participation was voluntary, with the option to withdraw at any time. Those who consented to participate accessed the survey via a QR code during regular class hours. This study received approval from the Ethical Committee of Putian University (2025-009).

 

I distributed 600 questionnaires, yielding 487 valid responses for an effective recovery rate of 74.7%. The sample comprised 240 (49.28%) women and 247 (50.72%) men. Respondents were predominantly aged 18–23 years (M = 19.65, SD = 1.46). In terms of class distribution, there were 131 (26.90%) freshmen, 141 (28.95%) sophomores, 124 (25.46%) juniors, and 91 (18.69%) seniors.

This study assessed frequency of AI usage using one item  from Rodríguez-Ruiz et al. (2025): “How often have you used AI tools in the last 12 months?” Responses were measured using a 7-point Likert scale ranging from 1 = never to 7 = almost every day.

 

This study used a three-item scale adapted from Spreitzer (1995) to measure self-efficacy. A sample item is “I am self-assured about my capabilities to perform my academic tasks.” Responses were measured using a 7-point Likert scale ranging from 1 = strongly disagree to 7 = strongly agree.

 

This study adopted Twenge et al.’s (2004) widely used five-item scale (see Han et al., 2023) to measure ego depletion. A sample item is “I feel like my willpower is gone.” Responses were measured using a 7-point scale ranging from 1 = strongly disagree to 7 = strongly agree.

 

This study used the World Health Organization-Five Well-Being Index (see Fung et al., 2022) to assess well-being over the past 2 weeks. One item was excluded due to a low corrected item–total correlation (< .30), resulting in a final four-item scale (e.g., “I have felt calm and relaxed”). Responses were measured using a 7-point Likert scale ranging from 1 = strongly disagree to 7 = strongly agree.

I conducted a hierarchical regression analysis to examine the incremental effect of the predictive variables, and tested the mediation effects using the PROCESS macro for SPSS 25.0.

I conducted a confirmatory factor analysis for each scale using Amos 22.0. The results showed that the factor loadings were higher than .60. Moreover, composite reliability values were above .70 and average variance extracted values exceeded the threshold of .50, indicating good reliability and validity. The factor loadings and Cronbach’s alpha values for each variable are presented in Table 1.

The regression analysis results are presented in Table 2. Models 1 and 2 show the regression outcomes with self-efficacy and ego depletion as the dependent variables, respectively. The findings indicated that AI usage exerted a significant positive influence on both self-efficacy and ego depletion. AI usage, self-efficacy, and ego depletion as independent variables were incorporated separately in Models 3 to 5, with well-being serving as the dependent variable. Model 6 represents a comprehensive model that includes all three variables simultaneously. The regression coefficients for well-being with respect to AI usage and self-efficacy were found to be positive and significant, whereas the regression coefficient for well-being with respect to ego depletion was negative and significant. The 95% confidence interval (CI) for the indirect effect of AI usage on well-being through self-efficacy was positive and significant, .04, 95% CI [0.01, 0.07]. Conversely, the indirect effect of AI usage on well-being through ego depletion was negative and significant, −.03, 95% CI [−0.06, −0.02]. These findings indicated the concurrent existence of both pathways, thereby supporting Hypotheses 1 and 2.

This study has contributed to the expanding scholarly discourse regarding the paradoxical impacts of digital technology on human well-being, with a specific focus on the ramifications of AI usage on the well-being of students. I found that AI simultaneously enhanced well-being through increased self-efficacy, while diminishing it through ego depletion, presenting a nuanced perspective on the role of technology in educational contexts.

 

The positive pathway via self-efficacy corroborates and extends existing literature on the empowering potential of technology (Bandura, 1995; Savio-Ramos, 2015). AI-driven tools, such as adaptive learning platforms and intelligent tutoring systems, leverage machine-learning algorithms to create personalized learning trajectories that dynamically adjust to individual performance metrics and cognitive preferences (Deng et al., 2024). This granular personalization not only optimizes learning efficiency but also reinforces students’ perceived competency, a core dimension of self-efficacy (Bandura, 1995). These findings align with the resource gain tenet of COR theory (Hobfoll et al., 2003), positioning AI as a technological resource that augments psychological resources, such as self-efficacy. By enhancing students’ belief in their abilities to navigate academic challenges, AI indirectly cultivates resilience against stressors, thereby elevating overall well-being (Yu & Luo, 2018).

 

Conversely, the negative pathway through ego depletion reveals the understudied cognitive costs of AI integration. Rapid switching between AI-generated prompts, automated grading systems, and algorithmically curated content impose significant demands on working memory, precipitating cognitive overload (Nizamani et al., 2024; Zhang & Mao, 2023). Moreover, overreliance on AI-generated solutions erodes metacognitive skills, such as self-regulated learning and error detection (Shanmugasundaram & Tamilarasu, 2023). Over time, such dependency can create a “cognitive outsourcing” loop (Wang et al., 2025, p. 967), diminishing students’ capacity for deep, sustained intellectual engagement (Rodríguez-Ruiz et al., 2025). These findings substantiate the resource loss principle of COR theory, illustrating how AI, when overutilized, acts as a resource drain that depletes finite cognitive reserves essential for maintaining well-being (Hobfoll et al., 2003).

 

The coexistence of dual pathways underscores the complex and multifaceted relationship between AI usage and student well-being, highlighting how AI can predict well-being through distinct mediating mechanisms. This dynamic is anchored in COR theory, which posits that individuals proactively engage in the acquisition and preservation of resources to sustain their well-being (Hobfoll et al., 2003). Building on this framework, in the current study I demonstrated the dual role of AI as a resource-enhancing and resource-depleting factor within the educational ecosystem. Specifically, AI operates as a contextual resource that fosters self-efficacy through adaptive scaffolding, while simultaneously causing ego depletion via cognitive overextension and autonomy erosion.

 

Educators and institutions must acknowledge that the value of AI resides not in its uncritical adoption, but in its strategic integration. To optimize gains in self-efficacy, AI tools should be designed to prioritize adaptive feedback and personalized learning pathways while mitigating cognitive overload through intuitive design. For example, AI platforms could incorporate mindful breaks (Ayub et al., 2025) or hybrid interfaces that blend analytical and creative tasks, mirroring the benefits of balanced AI utilization (Zhang & Mao, 2023). Additionally, students should be instructed to utilize AI as a supplement rather than a substitute. Workshops on digital literacy could emphasize self-regulation strategies, such as time-blocking AI usage and reflections on its emotional ramifications, to reduce ego depletion.

A primary constraint of this study is its cross-sectional design, which limits the capacity to establish causal relationships. Longitudinal investigations are needed in the future to address causality. Second, the data-collection method of self-reported measures may have introduced potential response bias and common method variance. Future research would benefit from triangulated data-collection methods, integrating objective behavioral metrics (e.g., AI interaction logs). Third, AI usage was assessed using a single-item measure that focused solely on frequency. This approach confounds multiple aspects of AI usage, including the intensity, duration, and the types of AI tools being utilized. Therefore, future studies could employ more comprehensive measurement instruments that assess the diversity and complexity of AI usage.

Ayub, F., E-Sehar, G., Javed, M., & Iqbal, H. (2025). Mindfulness and focus in AI based learning platforms: A cognitive behavioral analysis. The Critical Review of Social Sciences Studies, 3(3), 778–796. https://doi.org/10.59075/zpy09450

 

Bandura, A. (1995). Self-efficacy in changing societies. Cambridge University Press. https://doi.org/10.1017/CBO9780511527692

 

Çakar, F. S. (2012). The relationship between the self-efficacy and life satisfaction of young adults. International Education Studies, 5(6), 123–130. https://doi.org/10.5539/ies.v5n6p123

 

Chen, J. J., & Lin, J. C. (2024). Artificial intelligence as a double-edged sword: Wielding the POWER principles to maximize its positive effects and minimize its negative effects. Contemporary Issues in Early Childhood, 25(1), 146–153. https://doi.org/10.1177/14639491231169813

 

Deng, W., Wang, L., & Deng, X. (2024). Strategies for optimizing personalized learning pathways with artificial intelligence assistance. International Journal of Advanced Computer Science and Applications, 15(6), 605–616. https://doi.org/10.14569/IJACSA.2024.0150662

 

Duijnhouwer, H., Prins, F. J., & Stokking, K. M. (2010). Progress feedback effects on students’ writing mastery goal, self-efficacy beliefs, and performance. Educational Research and Evaluation, 16(1), 53–74. https://doi.org/10.1080/13803611003711393

 

Fung, S.-F., Kong, C. Y. W., Liu, Y.-M., Huang, Q., Xiong, Z., Jiang, Z., Zhu, F., Chen, Z., Sun, K., Zhao, H., & Yu, P. (2022). Validity and psychometric evaluation of the Chinese version of the 5-item WHO Well-Being Index. Frontiers in Public Health, 10, Article 872436. https://doi.org/10.3389/fpubh.2022.872436

 

Haluza, D., & Jungwirth, D. (2023). Artificial intelligence and ten societal megatrends: An exploratory study using GPT-3. Systems, 11(3), Article 120. https://doi.org/10.3390/systems11030120

 

Han, C., Zhang, R., Liu, X., Wang, X., & Liu, X. (2023). The virus made me lose control: The impact of COVID-related work changes on employees’ mental health, aggression, and interpersonal conflict. Frontiers in Public Health, 11, Article 1119389. https://doi.org/10.3389/fpubh.2023.1119389

 

Harry, A., & Sayudin, S. (2023). Role of AI in education. Journal of Interdisciplinary Studies, 2(3), 260–268. https://doi.org/10.58631/injurity.v2i3.52

 

Hobfoll, S. E., Johnson, R. J., Ennis, N., & Jackson, A. P. (2003). Resource loss, resource gain, and emotional outcomes among inner city women. Journal of Personality and Social Psychology, 84(3), 632–643. https://doi.org/10.1037/0022-3514.84.3.632

 

Huang, F., Wang, Y., & Zhang, H. (2024). Modelling generative AI acceptance, perceived teachers’ enthusiasm and self-efficacy to English as a foreign language learners’ well-being in the digital era. European Journal of Education, 59(4), Article e12770. https://doi.org/10.1111/ejed.12770

 

Inzlicht, M., & Schmeichel, B. J. (2012). What is ego depletion? Toward a mechanistic revision of the resource model of self-control. Perspectives on Psychological Science, 7(5), 450–463. https://doi.org/10.1177/1745691612454134

 

Jin, G., Jiang, J., & Liao, H. (2024). The work affective well-being under the impact of AI. Scientific Reports, 14(1), Article 25483. https://doi.org/10.1038/s41598-024-75113-w

 

Kaya, M., & Erdem, C. (2021). Students’ well-being and academic achievement: A meta-analysis study. Child Indicators Research, 14(5), 1743–1767. https://doi.org/10.1007/s12187-021-09821-4

 

King, R. B., & Gaerlan, M. J. M. (2014). High self-control predicts more positive emotions, better engagement, and higher achievement in school. European Journal of Psychology of Education, 29(1), 81–100. https://doi.org/10.1007/s10212-013-0188-z

 

Klimova, B., & Pikhart, M. (2025). Exploring the effects of artificial intelligence on student and academic well-being in higher education: A mini-review. Frontiers in Psychology, 16, Article 1498132. https://doi.org/10.3389/fpsyg.2025.1498132

 

Kumar, P. (2019). Artificial intelligence: Reshaping life and business. BPB Publications.

 

Luqman, A., Masood, A., Shahzad, F., Shahbaz, M., & Feng, Y. (2021). Untangling the adverse effects of late-night usage of smartphone-based SNS among university students. Behaviour & Information Technology, 40(15), 1671–1687. https://doi.org/10.1080/0144929X.2020.1773538

 

Michalos, A. C. (1991). Global report of student well-being: Volume IV: Religion, Education, Recreation, and Health. Springer.

 

Naseer, A., Ahmad, N. R., & Chishti, M. A. (2025). Psychological impacts of AI dependence: Assessing the cognitive and emotional costs of intelligent systems in daily life. Review of Applied Management and Social Sciences, 8(1), 291–307. https://doi.org/10.47067/ramss.v8i1.458

 

Nizamani, M., Ramzan, F., Fatima, M., & Asif, M. (2024). Investigating how frequent interactions with AI technologies impact cognitive and emotional processes. Bulletin of Business and Economics, 13(3), 316–325.

 

Orkibi, H., & Ronen, T. (2017). Basic psychological needs satisfaction mediates the association between self-control skills and subjective well-being. Frontiers in Psychology, 8, Article 936. https://doi.org/10.3389/fpsyg.2017.00936

 

Pajares, F. (2003). Self-efficacy beliefs, motivation, and achievement in writing: A review of the literature. Reading & Writing Quarterly, 19(2), 139–158. https://doi.org/10.1080/10573560308222

 

Reinecke, L., Hartmann, T., & Eden, A. (2014). The guilty couch potato: The role of ego depletion in reducing recovery through media use. Journal of Communication, 64(4), 569–589. https://doi.org/10.1111/jcom.12107

 

Rodríguez-Ruiz, J., Marín-López, I., & Espejo-Siles, R. (2025). Is artificial intelligence use related to self-control, self-esteem and self-efficacy among university students? Education and Information Technologies, 30(2), 2507–2524. https://doi.org/10.1007/s10639-024-12906-6

 

Sajja, R., Sermet, Y., Cikmaz, M., Cwiertny, D., & Demir, I. (2024). Artificial intelligence-enabled intelligent assistant for personalized and adaptive learning in higher education. Information, 15(10), Article 15100596. https://doi.org/10.3390/info15100596

 

Santos, M. C. J., Magramo, C. S., Jr., Oguan, F., Jr., & Paat, J. J. (2014). Establishing the relationship between general self-efficacy and subjective well-being among college students. Asian Journal of Management Sciences & Education, 3(1), 1–12.

 

Savio-Ramos, C. A. (2015). A study of the self-efficacy of personalized learning as a remediation tool in Algebra [Doctoral dissertation]. Arizona State University. https://keep.lib.asu.edu/items/154035

 

Shanmugasundaram, M., & Tamilarasu, A. (2023). The impact of digital technology, social media, and artificial intelligence on cognitive functions: A review. Frontiers in Cognition, 2, Article 1203077. https://doi.org/10.3389/fcogn.2023.1203077

 

Spreitzer, G. M. (1995). Psychological empowerment in the workplace: Dimensions, measurement, and validation. Academy of Management Journal, 38(5), 1442–1465. https://doi.org/10.2307/256865

 

Twenge, J., Muraven, M., & Tice, D. M. (2004). Measuring state self-control: Reliability, validity, and correlations with physical and psychological stress. San Diego State University.

 

Wang, F., Tang, X., & Yu, S. (2025). Cognitive outsourcing based on generative artificial intelligence: An analysis of interactive behavioral patterns and cognitive structural features [In Chinese]. Acta Psychologica Sinica, 57(6), 967–986. https://doi.org/10.3724/SP.J.1041.2025.0967

 

Xu, Z. (2024). AI in education: Enhancing learning experiences and student outcomes. Applied and Computational Engineering, 51(1), 104–111. https://doi.org/10.54254/2755-2721/51/20241187

 

Yu, Y., & Luo, J. (2018). Dispositional optimism and well-being in college students: Self-efficacy as a mediator. Social Behavior and Personality: An international journal, 46(5), 783–792. https://doi.org/10.2224/sbp.6746

 

Zhang, G., & Mao, S. (2023). A double-edged sword: Social media use and creativity. Information Technology & People, 38(2), 660–685. https://doi.org/10.1108/ITP-09-2022-0682