Hearing a hundred times is not as good as practicing once.

For about 10 years, as I taught writing to office workers and university students, this was my motto. Writing is a skill; no matter how many tips students hear, they struggle to internalize them unless they practice. I explained that only through hands-on exercises could they improve, and I made sure to include practice writing sessions throughout. I used the example of football player Son Heung-min: “Even if you watch a shooting lecture from Son Heung-min with accompanying videos a hundred times, you cannot acquire his skills unless you try the movements yourself.”


Applying this motto to evaluation, it means: “The actual writing each student submits demonstrates their writing ability better than a hundred pieces of knowledge about writing.” That’s why I assessed participants’ work in writing courses based on their own written pieces.


Writing is a way to elevate and refine thinking. As we enter the age of artificial intelligence (AI), I basically agree with the National Education Council’s move to introduce essay-type and long-answer questions to the College Scholastic Ability Test. Based on my personal experience shared above, I also understand the approach that both fostering and measuring writing ability should involve actual writing tasks. However, I began to question the grading method. The Korea Institute for Curriculum and Evaluation is currently researching how to use AI models to grade essay and long-answer responses.


There is criticism that introducing essay-type questions and AI grading will drive an increase in private tutoring and raise concerns over fairness and accuracy in grading. Based on my years of experience devising writing questions and grading descriptive answers, these concerns have merit. First, let me explain the grading process. In class, I would provide an original passage and a revised alternative, explain the differences, and then ask students in exams to rewrite the original as their answer. The wrong answers were highly varied, some unlike anything I had seen before. I reviewed all responses before establishing grading criteria, and then applied those criteria to each answer—deducting points depending on how many and how much their answer deviated from the model.


When hundreds of thousands submit essay answers, there will inevitably be incorrect answers the test creator never anticipated. Even with pre-learned model answers and grading rubrics, it seems virtually impossible for an AI system to grade such answers without any room for controversy. In my own trial and error, I changed the format of the writing questions. However, in a high-stakes exam like the CSAT, the first mistake may be seen as a failure and could well lead to the abandonment of essay and long-answer formats altogether.


The three key writing guidelines I distilled are: first-sentence main idea, organization (using tables of contents and paragraphs), and MECE (mutually exclusive, collectively exhaustive). While originally I had students write descriptive answers, for the sake of grading efficiency and objectivity, I switched to multiple-choice questions. With clear answers, multiple-choice questions leave little room for dispute. I concluded that multiple-choice formats can still convey these core writing principles as effectively as writing exercises.


Before focusing on how grading is performed, education authorities must first create question types and formats that enable objective grading. If clear question types are developed and taught in schools, there will be no need for private tutoring.



Baek Woojin, Economics Columnist


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing