"The assignment is done, but I’m not sure what I learned" ... Studying in the AI era [AI Sapiens - Questioning intellectual sovereignty 3]

[Financial News]A university student in his 20s, identified as A, entered lecture materials into AI ahead of his midterm exams. After asking it to summarize the lengthy materials, he also received possible exam questions and answers to organize. The day before the test, he repeatedly read the AI-generated summary instead of the textbook.Similar questions appeared on the exam, but he found it difficult to answer when asked to explain the concepts in his own words. A said, "I finished the report and memorized the expected questions, but when I actually had to explain them, I wasn’t sure whether I was describing what I understood or what AI had organized for me."
A also turns to AI first when working on assignments. He uses it to create an outline that fits the topic, organize reference materials, and polish sentences. Before submitting, he compares the result with the original sources, but the time spent reading materials from the beginning and deciding on the key points has decreased. He said, "In the past, I would read the materials and look up anything I didn’t understand. Now I often just read what AI has organized and fix only the parts that are wrong."
Students are using AI for more than search and spell-checking. They use it to summarize lecture materials, generate likely exam questions, and decide the structure of reports. Even if they do not submit AI’s answers verbatim, having AI handle the early stages of learning can reduce the time students spend reading and thinking on their own.

AI is being used throughout the study process, from summarizing lecture materials and drafting likely questions to organizing presentation scripts. The method is to receive the key points before reading the full material, turn difficult concepts into simpler language, and even get possible exam questions in advance.
According to the '2025 User Panel Survey on the Intelligent Information Society' released by the Korea Media and Communications Commission (KMCC) and the Korea Information Society Development Institute, the share of respondents with experience using generative AI rose from 12.3% in 2023 to 24.0% in 2024 and 38.9% in 2025. The average daily usage time among AI users also increased from 41.0 minutes in 2023 to 45.9 minutes in 2024 and 49.6 minutes in 2025. The survey was not limited to students. It was based on smartphone users and examined their experience with AI and the time they spent using it.
Students say AI helps reduce study time because it lets them check the key points before reading the full lecture materials, rephrase difficult concepts in simpler terms, and receive possible exam questions. When reading foreign-language materials or drafting presentation scripts, they can also request translation and sentence editing at the same time.
But reading a summary does not mean understanding the original text. Even if students feel they know the content after seeing AI-generated answers, they may struggle when asked to explain the basis for those answers or apply them to different cases. The memory, comprehension, and application skills required in exams cannot be built simply by reading sentences produced by AI.


The changes that occur when AI replaces the early stages of learning also appear in research on writing assignments. If students receive AI-generated text before reading the materials and forming sentences themselves, assignments can be completed quickly. But whether the result remains in their own memory and understanding is a separate issue.
In a preprint released in 2025, researchers at the MIT Media Lab analyzed how writing assignments using AI affect cognitive activity in the paper titled 'Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task.' The researchers divided participants into a large language model (LLM) group, a search engine group, and a group that used no tools, then asked them to write essays. Fifty-four people took part in the first through third experiments, and 18 participated in the fourth.
The researchers used EEG measurements, text analysis, and evaluations by teachers and AI. As a result, the group that used no tools showed the broadest and strongest brain connectivity, while the search engine group showed a moderate level of cognitive engagement. By contrast, the LLM group showed the weakest brain connectivity and the lowest sense of ownership over the essays they wrote. LLM users also had difficulty accurately quoting their own writing again.
However, because the study is only a preprint, further verification is needed regarding the sample size and research design. It is difficult to generalize the findings to all students’ use of AI. Still, the study can serve as a useful reference for how receiving AI-generated text first and then revising it may affect students’ memory and ability to explain ideas in their own words.
A also experienced something similar while preparing for exams. He could easily find answers when solving AI-generated practice questions, but he struggled to explain concepts when faced with cases not covered in class. The difference lay between memorizing sentences organized by AI and applying what he learned to a different situation.

AI can help shorten long texts, explain difficult concepts in simpler terms, and generate new questions. If students compare AI’s answers with the textbook and correct mistakes themselves, it can be used as a review tool. That is why the way AI is used, rather than its functions alone, affects learning outcomes.
On the other hand, if students leave everything from topic selection to conclusion writing to AI, there is less for them to do on their own. Deciding what to wonder about, checking whether sources are reliable, and comparing different arguments are not merely secondary steps in an assignment. Through that process, students understand the content and rebuild it in their own language.
But the fact that an assignment has been completed is not the same as the student being able to explain what was learned. AI can help polish outlines and sentences to finish a submission, but what the student read and how the student judged the material must be checked separately.
A said he plans to read the lecture materials first before entering questions into AI for his next assignment. He will mark the parts he does not understand in the materials and compare his own answers with AI’s responses. He said, "I realized that instead of memorizing what AI organized, I need to know what I don’t understand first if I want to explain it during a presentation."


Experts say AI can help students complete assignments faster, but it should not replace the process of asking questions and making judgments on their own. Ki Hye-seon, head of the Literacy Education and Culture Research Institute, explained how students use generative AI and what that means for education during a forum held at the National Assembly in August last year titled 'The Direction and Tasks of Student Ethics Education in the AI Era.'
Ki said that in educational settings, students turn to AI first even when making ethical judgments, and she warned that they have become accustomed to "copy-and-paste" behavior. She added that rather than trying to generate information themselves, students remain "information consumers" who depend on machines. On AI literacy education, she said, "Students must be able to go beyond simply using AI as a tool and develop critical reflection and ethical judgment." She added, "Education should also include experiential learning that enables them to apply it practically in their own lives."
[email protected] Han Seung-gon Reporter