Add BrightSurf on Google Email

Uncovering multidimensional effects of generative ai on learning… "high dependency lowers actual test scores"

07.21.26 | Sungkyunkwan University External Affairs Division (PR team)
Apple iPhone 17 Pro

Apple iPhone 17 Pro delivers top performance and advanced cameras for field documentation, data collection, and secure research communications.

Sungkyunkwan University(SKKU) announced that a research team consisting of Researcher Somi Joo, Professor Changjun Lee, and Professor Daeho Lee from the Department of Artificial Intelligence Convergence has published the results of a study that multidimensionally analyzes the opportunities and risks that arise when using Generative Artificial Intelligence (GAI) in educational settings. This study is drawing significant attention as it scientifically verifies how generative AI, which students recently use as an

The research team conducted an experiment targeting 88 university students, utilizing evaluation questions equivalent to the domestic high school academic achievement level. The learning process was divided into three stages: 'Concept Understanding' to read and comprehend new content, 'Problem Solving' to solve problems based on what was learned, and 'Result Review' to check one's own answers. Subsequently, students were randomly assigned to use generative AI (GPT-4o) at different stages to track its effects.

The experimental results showed that the group that directly solved problems utilizing generative AI recorded higher scores on a subsequent test than the group that used AI simply during the concept understanding stage. This is because the hints and core summaries provided by the AI reduced excessive brain stress and unnecessary processes that occur during problem-solving, thereby drastically increasing learning efficiency.

However, an interesting twist emerged as the 'dependency' on generative AI increased. Students who usually relied heavily on AI tended to show lower actual test scores within the problem-solving group. This is because they accepted the answers provided by the AI without criticism, skipping the essential process of deeply contemplating and making the knowledge their own.

On the other hand, even though these highly dependent students experienced a drop in actual scores, their 'perceived learning'—the feeling that they studied well—and their 'learning satisfaction' were much higher. This is interpreted as a result of the AI organizing complex content in an easy-to-read manner and providing kind answers, which induces a sort of 'learning illusion' making students feel as if they have perfectly understood everything on their own.

Professor Changjun Lee of SKKU emphasized, "Generative AI is a powerful tool that increases learning efficiency, but relying on it uncritically can lead to the paradoxical result of lowering objective academic achievement." He added, "An educational guideline is urgently needed to foster students' independent questioning skills and critical verification abilities so that generative AI can function properly as a learning aid rather than a substitute."

This study is evaluated to have presented a specific direction on 'how to use it properly,' rather than the unconditional regulation or blind adoption of generative AI in the education sector. The study was conducted with support from the Ministry of Education and the National Research Foundation of Korea.

Interactive Learning Environments

10.1080/10494820.2026.2676669

Keywords

Article Information

Contact Information

Goeun KIM
Sungkyunkwan University External Affairs Division (PR team)
katekim0107@skku.edu

Source

This article is based on a news release from Sungkyunkwan University External Affairs Division (PR team). BrightSurf curates and republishes science news from research institutions worldwide; the original release is linked below.

How to Cite This Article

APA:
Sungkyunkwan University External Affairs Division (PR team). (2026, July 21). Uncovering multidimensional effects of generative ai on learning… "high dependency lowers actual test scores". Brightsurf News. https://www.brightsurf.com/news/8J4E03RL/uncovering-multidimensional-effects-of-generative-ai-on-learning-high-dependency-lowers-actual-test-scores.html
MLA:
"Uncovering multidimensional effects of generative ai on learning… "high dependency lowers actual test scores"." Brightsurf News, Jul. 21 2026, https://www.brightsurf.com/news/8J4E03RL/uncovering-multidimensional-effects-of-generative-ai-on-learning-high-dependency-lowers-actual-test-scores.html.