The advent of generative artificial intelligence (AI) has revolutionized various sectors, with academia being a significant domain experiencing its transformative impact. A recent study published in BMC Psychology unravels the intricate relationship between academic stress and university students’ increasing reliance on generative AI technologies, employing a sophisticated multiple mediation model grounded in partial least squares structural equation modeling (PLS-SEM). This research not only sheds light on the psychological dynamics underpinning AI dependency among students but also offers a technical framework for understanding the mediating factors that play pivotal roles in this emerging phenomenon.
With the rise of tools like large language models (LLMs) and AI-powered writing assistants, university students today face unprecedented choices in how they approach their academic tasks. The pressure to perform, coupled with the convenience offered by AI, has led many to develop a dependency on these technologies. The study, conducted by Liu et al., bridges the gap between psychological stressors inherent in academic life and the behavioral adaptations that students exhibit in response to generative AI availability. The researchers tapped into PLS-SEM, a multifaceted statistical technique well-suited for analyzing complex relationships among observed and latent variables, providing a rigorous methodological backbone to the inquiry.
The investigation begins by contextualizing academic stress as a multifaceted construct that includes perceived workload, time pressures, performance anxiety, and social expectations. These stressors cumulatively impact students’ mental health and coping mechanisms. The authors posit that generative AI tools serve as both a coping strategy and potentially an avoidance mechanism, highlighting the dual-edged nature of AI’s integration into academic activities. They suggest that while AI can enhance productivity and learning, unchecked reliance might lead to dependency, thereby affecting students’ cognitive autonomy and critical thinking capabilities.
One of the critical contributions of the study is the deployment of a multiple mediation model to dissect how academic stress influences dependency on generative AI. Unlike simple direct effect models, multiple mediation allows the researchers to unravel indirect pathways through which stress impacts AI dependency. The team explored variables such as anxiety levels, self-efficacy in academic skills, and perceived usefulness of AI tools as mediators. These factors collectively elucidate the psychological processes through which stress translates into behavioral inclination towards AI usage.
PLS-SEM, the analytical tool of choice in the study, is a variance-based structural equation modeling approach that excels in handling small to medium sample sizes and complex model specifications. This method also accommodates measurement error and enables the simultaneous assessment of multiple relationships. Liu and colleagues meticulously validated their scales for constructs like academic stress, anxiety, self-efficacy, and AI dependency using confirmatory factor analysis within the PLS framework, ensuring the robustness of their findings. Their model fit and reliability indices affirmed the suitability of the hypothesized pathways, providing credible empirical support for the theoretical constructs proposed.
The data, collected from a diverse cohort of university students across multiple disciplines, painted a nuanced picture. Academic stress was positively associated with increased anxiety, which in turn diminished students’ confidence in their academic abilities—a phenomenon known as reduced self-efficacy. This diminished self-efficacy then correlated with higher perceived usefulness of generative AI tools, reflecting how students sought external scaffolding to compensate for their self-doubt. Ultimately, a higher perceived usefulness translated into greater dependency on generative AI, underscoring the mediatory role of psychological states in technology reliance.
Beyond these direct relationships, the study’s findings also illustrate a feedback loop where continued dependency on AI may exacerbate academic stress over time, potentially due to concerns over skill degradation or ethical dilemmas linked to AI-assisted work. This cyclical interaction poses important questions about the long-term implications of integrating these technologies into academic processes. The authors call for interventions that balance AI usage with skill development to prevent detrimental effects on students’ learning journeys.
The significance of this research extends into pedagogical and policy domains. As educational institutions increasingly incorporate AI into curricula and resource ecosystems, understanding the psychosocial consequences is paramount. This study serves as a clarion call for educators to foster environments where generative AI is deployed as an augmentative tool rather than a crutch. Strategies that enhance students’ self-efficacy and resilience to academic stress can mediate their reliance on AI, ultimately promoting healthier digital habits.
Furthermore, the ethical tensions addressed indirectly in the study resonate with ongoing debates surrounding academic integrity in the AI era. Dependency on generative AI raises questions about originality, plagiarism, and the essential skill sets that education aims to cultivate. The research by Liu et al. indirectly underscores the necessity for clear guidelines and transparent communication around AI use, ensuring that generative technologies are aligned with educational values and objectives.
The technical rigor of this study, especially its use of PLS-SEM, sets a precedent for future research exploring AI-human interaction within psychological frameworks. By leveraging this sophisticated modeling approach, researchers can unpack complex, multidimensional phenomena that traditional methods might obscure. The authors advocate for continued exploration of cognitive and emotional variables influencing AI engagement, promoting a holistic understanding of the student experience in digitally augmented educational contexts.
In conclusion, this pioneering study delivers critical insights into how academic stress propels university students towards dependency on generative AI technologies, mediated by anxiety, self-efficacy, and perceived tool usefulness. The findings illuminate the delicate interplay between psychological well-being and technological adoption, urging educators, policymakers, and technologists to collaboratively cultivate supportive academic ecosystems. As generative AI continues to evolve, understanding its psychological ramifications will be essential for leveraging its benefits while safeguarding student development and authenticity.
Subject of Research:
The influence of academic stress on university students’ dependency on generative artificial intelligence, analyzed through a multiple mediation model utilizing partial least squares structural equation modeling (PLS-SEM).
Article Title:
Academic stress and university students’ dependency on generative artificial intelligence: a multiple mediation model using PLS-SEM.
Article References:
Liu, X., Liu, Y., Dai, Y. et al. Academic stress and university students’ dependency on generative artificial intelligence: a multiple mediation model using PLS-SEM. BMC Psychol (2026). https://doi.org/10.1186/s40359-026-03986-9
Image Credits: AI Generated
