[사설] 서·논술형 수능, 관건은 AI가 아니라 사회적 공감대다

The ongoing debate surrounding the reform of the university entrance examination system in South Korea has centered around the introduction of a written and argumentative evaluation format for the college scholastic ability test. Concerns such as fairness and reliability in grading, increased private education costs, social inequality, and the burden on teachers have been raised, making the formation of a social consensus a challenging task.

Given the uncertainty surrounding the trustworthiness of the artificial intelligence (AI) grading tool being considered by the national education committee, it is crucial to present concrete alternatives and foster a sense of shared public opinion during the public discourse phase. The national education committee is currently reviewing the utilization of AI as an auxiliary tool for grading the nationwide examination, which is taken by hundreds of thousands of students each year.

To overcome issues related to time, cost, and grading variance during the evaluation process, teachers would design the questions and grading criteria, while the AI would perform an initial grading based on exemplary and sample answers. Any discrepancies between the AI and teacher assessments would be reviewed by the teachers, resulting in a final score.

However, the AI grading tool has only partial applicability in the full-service written and argumentative examination format. A study conducted by the Korean Educational Development Institute in 2023 revealed that the AI grading tool tends to prioritize formatting elements over conceptual accuracy and logical reasoning, often favoring longer answers that include keywords from the prompt.

For instance, a teacher might assign a grade of C to an essay that strays from the given topic and includes irrelevant ideas, while the AI would award an A for an answer that is lengthy and incorporates many of the vocabulary words mentioned in the prompt. France, which has maintained a written and argumentative examination format for over two centuries, employs various strategies to ensure fairness in grading, including anonymous answer submissions, cross-grading of answer sheets among educational authorities, and the establishment of common grading guidelines.

However, they do not utilize AI grading tools. The fear of students submitting original and creative arguments that deviate from the exemplary answers in the AI tool is a concern, as the AI may misclassify such answers as incorrect. Even though France faces ongoing debates about the subjective nature of grading and the fairness of the examination, discussions advocating for a shift to objective multiple-choice tests are scarce.

Despite the potential to eliminate subjectivity in grading, the essence of education lies in fostering critical thinking and expression skills, which is best achieved through the written and argumentative evaluation format. A social consensus has formed around the reform's intention to move away from rote memorization and competition, with a significant number of citizens agreeing.

However, South Korea has long regarded the objective examination system as fair, as evidenced by the widespread belief that a small difference in exam scores can determine one's admission to a prestigious university. For the written and argumentative examination to be accepted by the public, it is imperative to provide evaluative and grading alternatives that minimize confusion in the field and facilitate the formation of a social consensus through public discourse.