A recent study suggests that AI dependency among students is not simply a question of how often they use ChatGPT. The research points to a more layered pattern: academic stress, confidence in one’s own academic ability, and expectations about what AI can deliver may interact in ways that push some students toward heavier reliance on AI tools.
The study, conducted by researchers from Sungkyunkwan University in Seoul and Korea University and led by Jang Hyun Kim, surveyed 300 students who had used ChatGPT. It was published in the International Journal of Educational Technology in Higher Education.
What the researchers studied
The research team used the I-PACE (Interaction of Person-Affect-Cognition-Execution) model as its theoretical framework. Within that framework, the study looked at possible factors behind AI addiction among students, with a particular focus on ChatGPT use.
The researchers defined AI dependency as "excessive reliance on AI technologies and applications across various aspects of life, including academic studies, daily routines, and social interactions." They also emphasized that the issue is not only about frequency of use. According to the researchers, this dependency also involves "significant psychological dependence."
That distinction matters. A student may use ChatGPT as an aid without necessarily being dependent on it. The study, however, was concerned with cases where reliance becomes excessive and psychologically important to the student’s academic or everyday functioning.
Stress appears to be a key link
One of the study’s central findings concerns academic self-efficacy, or students’ belief in their own ability to handle academic work. The researchers initially expected a direct significant relationship between academic self-efficacy and AI dependence. They did not find one.
Instead, the relationship appeared to be indirect. Students with low academic self-efficacy were more prone to academic stress, and that stress was associated with greater reliance on AI tools.
In plain terms, the study suggests that the pathway may not be: a student lacks confidence, so the student becomes dependent on AI. The more precise pattern reported by the researchers is: a student with low academic self-efficacy may experience more academic stress, and that stress may contribute to heavier dependence on AI.
This makes the role of stress especially important. The source article does not support a causal conclusion, but it does show that stress sits between low academic self-efficacy and AI dependence in the researchers’ findings.
Expectations of AI also matter
The study also found that performance expectations of AI had a significant impact on dependency. Students who expected more from AI were more likely to overuse it.
This creates a more complex picture than a simple story about students choosing convenience. According to the study, students with low self-efficacy may experience more stress. That stress can lead to higher expectations of AI technology, and those higher expectations may ultimately be tied to greater AI dependence.
The result is a feedback-like pattern in which AI becomes more than a tool for support. For some students, it may become a coping mechanism for academic pressure, especially when they believe the technology can improve performance or reduce the burden of difficult tasks.
The study does not say that every use of ChatGPT is harmful. It does, however, raise a clear warning: when students expect AI to solve too much, they may become more vulnerable to overuse.
What students said could go wrong
The study also asked students about negative effects of AI dependency. The most commonly cited concern was an increase in laziness, with 113 mentions. Limited creativity followed closely, with 112 mentions.
Other reported effects included the spread of false information, with 67 mentions, and a decrease in critical and independent thinking, with 56 and 47 mentions respectively. Students also cited limited problem-solving skills and an increased risk of plagiarism and copyright issues.
Those concerns can be grouped into several broad areas:
- Work habits: students pointed to laziness and reduced independence.
- Thinking skills: students cited reduced creativity, critical thinking, independent thinking, and problem-solving skills.
- Information quality: students mentioned the spread of false information.
- Academic integrity: students raised plagiarism and copyright issues.
These are self-assessed consequences, not independently measured outcomes. Still, they show that students themselves recognize risks when AI use shifts from assistance to dependence.
What the study cannot prove
The researchers also noted important limitations. The data was collected at only one point in time, so the study cannot draw causal conclusions. It can identify relationships, but it cannot prove that one factor directly caused another.
The results also refer only to South Korean students. That means the findings should not automatically be treated as universal for all students in all education systems.
Another limitation is that the consequences of AI dependence were examined through students’ self-assessments. Those assessments are described as highly subjective, which means they may reflect students’ perceptions as much as measurable effects.
The study also does not distinguish between acceptable use of AI as an aid and unacceptable use for cheating purposes. That distinction is important because not all student use of ChatGPT has the same academic meaning or ethical weight.
The researchers plan to dig deeper in future studies. For now, the findings offer a careful but useful signal: AI dependency may be shaped by academic stress, students’ confidence in their own abilities, and the expectations they place on tools like ChatGPT.