Understanding dynamic learning profiles in AI-supported environments in Vietnam: Evidence from a three-wave latent transition analysis
This longitudinal study of 725 Vietnamese undergraduates reveals that successful adaptation to AI-supported learning is characterized by high self-efficacy and metacognitive regulation with low cognitive offloading, a stable "Optimal" profile that consistently yields higher academic performance than maladaptive patterns of over-reliance on AI.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the modern classroom, a quiet revolution is reshaping how students learn. Artificial intelligence has moved from the periphery of education to the center, offering tools that can generate ideas, explain complex concepts, and solve problems. This shift changes the very nature of thinking. When a student uses an AI system, they are no longer working alone; they are working in a partnership with a machine. This new dynamic raises a fundamental question about the human mind: when we hand over the heavy lifting of thinking to a computer, what happens to our own ability to learn? To understand this, researchers look at three specific things. First is the student's confidence in using the technology. Second is the extent to which they let the machine do the actual work of generating ideas or solving problems, a process known as cognitive offloading. Third is the student's ability to keep a watchful eye on their own learning, checking if the machine is helping or if it is taking over too much. This balance between using a tool and staying in control is the key to understanding whether technology makes us smarter or just faster.
A team of researchers at Vietnam National University in Hanoi set out to map exactly how students navigate this new landscape. They followed 725 undergraduate students over the course of a single semester, tracking their habits and grades at three different points in time. The researchers were not just looking at how much AI was used, but at the specific patterns of behavior that emerged when confidence, delegation, and self-control mixed together. By observing these students as they moved through their coursework, the team discovered that learners did not fall into a single category of "good" or "bad" users. Instead, they clustered into three distinct groups, each with its own unique way of handling the technology.
The first group, which the researchers called the "Optimal" profile, consisted of students who were highly confident in their ability to use AI but kept a tight rein on their own thinking. These students used the technology as a powerful assistant but rarely let it take over the core work of understanding and reasoning. They maintained strong self-control, constantly checking the machine's output to ensure it made sense. The second group, the "Moderate" profile, showed a mix of traits. These students had average confidence and used the AI to a moderate degree, neither fully relying on it nor fully resisting it. They were in a transitional state, still figuring out the right balance. The third group, the "Maladaptive" profile, told a different story. These students were willing to use the technology, but they handed over the heavy cognitive work entirely. They let the AI generate ideas, construct explanations, and solve problems, often without enough oversight or critical evaluation.
The study revealed that these patterns were not just fleeting moments but stable ways of learning that persisted throughout the semester. The researchers found that the students in the Optimal group consistently achieved the highest academic grades. Their success was not simply because they used the technology, but because they used it while maintaining their own mental authority. In contrast, the students in the Maladaptive group, who relied heavily on the machine to do the thinking for them, ended up with the lowest grades. The data showed a strong link between the two: as students delegated more of their thinking to the AI, their ability to regulate their own learning dropped significantly. This suggests that the danger of AI is not the tool itself, but the loss of human oversight. When students stop monitoring the process, their learning becomes shallow.
Over the course of the semester, the researchers observed a gradual shift in the overall student body. On average, students became slightly more confident in using AI, and they began to rely on it a little less for the heavy lifting of thought. At the same time, their ability to monitor and control their own learning improved. This suggests that with time and experience, many students naturally learn to rebalance the relationship between human and machine, moving toward a more effective way of working. However, the study also noted that the patterns established early on were quite sticky. Students who started in the Maladaptive group tended to stay there, while those in the Optimal group remained there. The path to success was not about using more AI, but about using it with a steady hand.
The findings challenge the idea that simply having access to advanced technology guarantees better learning. The research indicates that the most successful students are those who treat AI as a partner rather than a replacement. They use the tool to expand their capabilities without surrendering the core processes of understanding. For educators and students alike, the lesson is clear: the value of artificial intelligence in education depends entirely on the human capacity to stay in charge. The technology can handle the data and the drafts, but the mind must remain the architect. Without that active, regulatory control, the efficiency of the machine comes at the cost of the depth of human thought.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.