Stabilize the difficulty level of the university entrance exam.

GD&TĐ - Following controversy over the difficulty of the university entrance exam, the South Korean Ministry of Education is considering using artificial intelligence to stabilize the difficulty level of the test.

However, many experts warn that technology cannot yet solve the root causes of the shortcomings in the education system.

In early March, the South Korean Ministry of Education announced plans to develop an AI system capable of generating English reading comprehension passages, typically excerpts from academic or journalistic materials. This system would not only assist in compilation but also help assess the relevance of the content. Testing could begin with mock exams in the near future.

This proposal emerged after the English section of the 2025 university entrance exam proved unusually difficult, exceeding the scope of textbook knowledge. The controversy surrounding the exam's difficulty became even more apparent when looking at the data. In the 2025 exam, only about 3.1% of candidates achieved the highest score in English, the lowest since the implementation of the perfect-scoring system in 2018. The significant fluctuations over the years, with some years seeing rates exceeding 12%, indicate a lack of stability in the current system.

The incident raises questions about the stability and fairness of the exam.

According to experts, one of the main reasons why exams become too difficult is last-minute revisions. Approximately 42% of questions are changed right before the exam day, leaving insufficient time for review to ensure accuracy. In this context, AI is expected to help standardize the exam creation process, from content generation to predicting the difficulty level of each question.

Furthermore, the current exam structure also has unintended consequences. The English exam is graded on a scale of absolute scores, while other subjects use a relative scoring system. This not only fails to reduce academic pressure but also causes spending on private education to shift towards younger age groups, as parents seek to prepare their children early.

Therefore, some experts view the use of AI as a highly promising proposal. Professor Lee Byung-min, who teaches at Seoul National University, commented: “There have been problems with paragraph selection, resulting in texts that are sometimes incoherent, too abstract, or too short to provide context. If AI can help evaluate and refine these aspects, it will certainly benefit test-takers.”

AI can act as a technical support tool, helping to improve content quality and minimize human error. Additionally, AI can analyze big data from previous exams to predict difficulty levels, thereby helping to maintain consistency over the years.

Conversely, Professor Yoon Hee-cheol, who teaches at Duksung Women's University, warns: “AI technology cannot solve the more fundamental problems. Applying AI can help adjust the difficulty of the exam instead of letting test-takers choose the passages themselves. But relying on AI-generated content can lead to passages lacking depth or not accurately reflecting the real-world context, thus affecting the ability to comprehensively assess language proficiency.”

Another worrying issue is the risk of increased inequality. As AI is integrated into exams, private educational institutions can quickly adapt and exploit this advantage.

Professor Yoon Hee-cheol, who teaches at Duksung Women's University in South Korea, warns: "The solution lies not only in technology but also in how to define 'appropriate difficulty levels.' We need clear guidelines or a broader consensus on the answer key system, aligned with societal expectations and educational policy directions."

According to Korea JoongAng Daily

Source: https://giaoducthoidai.vn/on-dinh-do-kho-ky-thi-dai-hoc-post771132.html