Essay-Style CSAT Graded by AI? We Took the Test Ourselves

[Anchor]

Proposals are under review to introduce essay and subjective-response questions to the College Scholastic Ability Test (CSAT) and have artificial intelligence grade them.

To see if any issues might arise if AI grades the answers, Reporter Jeong Banseok took an essay-style test himself.

[Reporter]

This high school in Gyeonggi Province is already utilizing AI to grade performance assessments.

With the school's permission, our reporting team sat in on a class and took the lesson alongside the students.

[Lee Yoon-gyu / Science Teacher, Pyeongtaek Cheongbuk High School: Let us learn about the specific conditions under which something becomes a fossil.]

After solving an essay question along with the first-year students here, we will have our answers graded by AI.

Once the completed answer sheet is scanned, the handwriting is automatically recognized, and grading is completed in less than a minute based on the grading criteria preset by the teacher.

Making final score adjustments is up to the teacher.

[Lee Yoon-gyu / Science Teacher, Pyeongtaek Cheongbuk High School: The first reason provided is incorrect. The AI judged it to be worth 4 points, but I think I would give it 3 points.]

Out of a possible 10 points, the reporting team received 7, while the students received 8.

This approach has the advantage of reducing the time spent deciphering the poor handwriting of some students.

[Lee Yoon-gyu / Science Teacher, Pyeongtaek Cheongbuk High School: It used to take a whole day just to grade. Now, grading a single class takes anywhere from a short 30 minutes to an hour. And now, it can even read graphs and tables entirely.]

Answer sheets graded by AI in schools under the jurisdiction of the Gyeonggi Provincial Office of Education have reached 3.37 million cases over the past year.

However, a survey of 105 teachers revealed a problem: scores fluctuated whenever the AI was made to grade the exact same answer sheet multiple times.

Because of this, teachers gave lower scores to consistency and reliability in grading compared to other categories.

There are also concerns that it could fuel private education.

[Kim Hee-jung / Spokesperson, Teachers' Union Federation (High School Korean Language Teacher): Answers involving context or background knowledge are definitely not at a level that AI can grade yet. Above all, there are a tremendous number of complaints in school settings regarding essay and subjective-response questions. People even bring in doctoral dissertations from cram schools and argue, "This is accepted, so why is that not?"]

The National Education Commission's stance is to review the introduction of essay and subjective-response questions to the CSAT as early as 2032, but only after going through a public consensus process.

Although AI technology will advance in the meantime, building trust among students and parents remains the top priority, given that hundreds of thousands of students take the CSAT simultaneously and admissions outcomes are decided by a single-point margin.

(Video Shooting: Joo Yong-jin, Video Editing: Kim Jin-won, Design: Kim Min-young)
※ Please note: This article was translated by AI and may contain errors.