Why loving AI is not the same as needing it: the hidden motivations behind student dependency

Generative artificial intelligence has woven itself into the daily routines of university students so quickly that researchers are still scrambling to understand what the technology is doing to the minds that use it. A new study published in Humanities and Social Sciences Communications offers one of the most psychologically detailed pictures yet of how students slide from enthusiastic, appropriate use of AI tools into something that looks far more like dependency. The research, led by Jianfeng Yin and Yueying Zhang of Qingdao University together with Runjie Jiang of Nanjing University, surveyed 405 university students and found that the road to AI dependency is not paved by attitude alone. Instead, the study argues, it is the underlying motivation behind each interaction with a chatbot or writing assistant that determines whether the technology becomes a helpful companion or a psychological crutch.

The theoretical engine driving the study is Self-Determination Theory, a decades-old framework in motivational psychology that distinguishes between reasons for acting that feel fully chosen and reasons that feel imposed. The researchers applied this framework to AI use by measuring several qualitatively different motivations. Intrinsic motivation describes using AI because the activity itself is enjoyable or intellectually stimulating. Identified regulation describes using AI because the student personally endorses the goal it serves, such as completing coursework more efficiently. External motivation describes using AI to satisfy outside pressures, such as grades, deadlines, or the expectations of instructors and peers. According to Self-Determination Theory, these motivations sit along a continuum of autonomy, and the position a student occupies on that continuum is thought to shape the psychological consequences of the behavior.

The survey results delivered a surprise that cuts against the intuitive assumption that students who love AI will inevitably come to depend on it. When the researchers ran their statistical models, they found that AI attitude did not directly predict AI dependency at all. A positive attitude toward generative AI, on its own, was psychologically inert with respect to dependency. Its influence, the study showed, was transmitted entirely through motivational pathways. This finding reframes the debate about AI in higher education: the question is not simply whether students view AI favorably, but why they reach for it in the first place.

Those motivational pathways pointed in opposite directions. Intrinsic motivation, the most autonomous form of drive in the Self-Determination Theory continuum, inhibited the development of AI dependency. Students who turned to AI out of genuine curiosity or enjoyment were, in effect, protected from the slide toward dependence, because their engagement with the technology remained an expression of their own interests rather than a substitute for them. Identified regulation, by contrast, facilitated dependency. When students used AI because they had internalized the goal of efficiency or academic performance, the tool became woven into their sense of what getting work done requires, and the boundary between assistance and reliance began to blur. The study thus identifies a motivational fork in the road: the same favorable attitude toward AI can lead to healthy engagement or creeping dependence depending on which type of regulation carries the behavior.

Beyond the mediation analysis, the researchers used a person-centered approach to identify distinct dependency profiles among the 405 students. Latent profile analysis, a statistical technique that groups individuals based on their response patterns rather than treating the sample as a uniform population, revealed three clusters. The researchers labeled them dependent, moderate, and conservative users. The stratification was clear and orderly: dependent users scored highest across every dimension of AI dependency measured in the study, moderate users occupied the middle ground, and conservative users recorded the lowest scores. The existence of these profiles suggests that AI dependency is not a single continuous variable along which all students are evenly spread, but a set of qualitatively distinct psychological states that different students inhabit.

The profile analysis also produced a striking connection between motivation and membership. Individuals with stronger motivations overall were more likely to be classified into the dependent user profile, and external motivation played a particularly salient role in that classification. This detail matters because external motivation is the least autonomous form of drive in the Self-Determination Theory framework. Students who use AI primarily to meet external demands, the study implies, are the group most at risk of crossing from appropriate use into dependency, perhaps because the technology becomes a coping mechanism for pressures they do not feel they have chosen. The finding echoes a broader principle in motivational psychology: behaviors regulated by external contingencies tend to be fragile, compulsive, and vulnerable to becoming habitual in ways that self-endorsed behaviors are not.

The psychological mechanism the study proposes is worth spelling out. Dependency, in this account, is not simply a matter of frequency of use. A student might consult an AI assistant many times a day out of curiosity and remain psychologically independent, while another student might use the tool less often yet feel unable to start an assignment without it. What distinguishes the two cases is the degree to which the behavior satisfies basic psychological needs for autonomy, competence, and relatedness. When AI use is autonomously motivated, it complements the student’s own agency. When it is controlled by external or internalized pressures, it can gradually displace the student’s sense of competence, leaving the individual feeling that the tool is necessary rather than optional. The study’s mediation results, showing that attitude effects flow through motivation rather than acting directly, fit this account precisely.

The practical implications reach several audiences at once. For university administrators, the findings suggest that policies aimed at AI in education should look beyond blanket permission or prohibition and instead attend to the motivational context of use. Encouraging students to engage with AI out of curiosity, exploration, and personal interest, rather than purely as a means to hit deadlines or maximize grades, may be the most effective way to keep the technology in its proper place. For tool developers, the study implies that design choices matter: interfaces that position AI as a thinking partner rather than an answer machine may foster the intrinsic motivation that protects against dependency. For educational researchers, the identification of three dependency profiles provides a template for future longitudinal work tracking how students move between clusters over time.

The study is not without boundaries that readers should keep in mind. The data come from a questionnaire administered at a single point in time, so the mediation pathways, however statistically well supported, remain correlational rather than causal. The sample of 405 students, while adequate for the analyses performed, was drawn from a specific institutional context, and dependency dynamics might differ across cultures, disciplines, and levels of study. Self-report measures also carry the usual limitations, since students may not accurately perceive or admit the degree to which they rely on AI tools. The authors obtained ethical approval from the Ethics Committee of Qingdao Medical College, Qingdao University, and conducted the survey in accordance with the Declaration of Helsinki, with informed consent from all participants.

Even with those caveats, the study lands at a moment when the question of AI dependency has become one of the most urgent in educational psychology. Generative AI tools are now embedded in essay writing, coding, research, and revision at nearly every university, and the pace of adoption has outstripped the pace of psychological understanding. By showing that attitude alone does not determine dependency, and that the motivational quality of use is the decisive variable, the research gives educators a concrete lever. The difference between a student who uses AI as an extension of their own thinking and a student who uses it as a replacement for that thinking may come down to something as subtle as why they opened the app. In the emerging science of human-AI interaction, motivation, it turns out, is the variable that matters most.

Subject of Research: AI dependency among university students

Article Title: From appropriate use to dependency: How AI attitude and use motivations shape university students’ AI dependency

Article References: Yin, J., Zhang, Y., & Jiang, R. (2026). From appropriate use to dependency: How AI attitude and use motivations shape university students’ AI dependency. Humanities and Social Sciences Communications. https://doi.org/10.1057/s41599-026-09477-x

Image Credits: AI Generated

DOI: 10.1057/s41599-026-09477-x

Keywords: artificial intelligence, AI dependency, university students, Self-Determination Theory, motivation, educational psychology, generative AI, higher education, psychology, appropriate, dependency, attitude