Series Index

Part 1 | When AI Enters the Classroom, What Choice Will Korean Education Make?
Part 2 | AI Has Broken Korea’s School Assessments

 

By Jung Yeon-Tae,
President of the Korea Information Service of the Construction Industry and Essayist.

Many teachers say the same thing these days: “Performance assessments are the hardest part now.” The reason is simple. You can no longer judge a student’s learning by looking only at the final product.

Reports have become cleaner, sentences more polished, and structures more sophisticated. But that level of finish no longer guarantees genuine understanding or strong thinking. Since the rise of generative AI, performance tasks and written reports have already collapsed at a structural level. The bigger problem is that schools continue to act as if nothing has changed, trying to preserve an old assessment system in a new learning reality.

The Assumptions Behind Performance Assessment Have Fallen Apart

Traditional performance assessment rested on a set of assumptions. Students find sources on their own, think independently, write in their own words, and the quality of the final submission reflects the depth of learning. Before AI, these assumptions mostly held. Today, they do not.

AI now collects information, builds a structure, refines expressions, and organizes logic. In that environment, asking “Did you write this by yourself?” has become an empty question. Students did not write it alone, because the learning environment itself has changed.

If schools continue to grade only the output, assessment turns into a formal procedure that no longer reflects reality. Output-centered scoring also rewards the easiest path to high marks. It becomes the moment when assessment shifts from “verifying learning” to “mastering submission.”

Outcome-Based Assessment Is Powerless in Front of AI

Most performance assessments today remain outcome-centered. Page count, sentence quality, formatting accuracy, and tidy references are treated as core criteria. But these are precisely the criteria where AI has the strongest advantage. They evaluate the appearance of the product, not the student’s thinking process.

When outcome-based assessment continues, predictable classroom behaviors follow. Students can earn grades without thinking, they don’t need to ask real questions, and revision and verification disappear from the learning cycle. What students end up practicing is not scholarship, but automated submission skills.

As assessment points toward the final output, learning slips away from the process. The classroom becomes a place optimized for “completion,” not “understanding.” AI is just a tool, but the assessment system measures exactly what that tool does best.

If We Want to Keep Performance Assessment, We Must Change the Questions

The most important question is no longer “Did you use AI?” The real question is “How did you think?” What performance assessment must now examine is not the final product, but the trace of reasoning behind it.

To do that, assessment must be redesigned around different questions. How did you define the topic? What was the first question you asked? Which sources did you choose, and why? What did you accept from AI’s responses, and what did you reject? What are the limits of your final conclusion? These questions do not ask for a polished output—they ask for the flow of thought.

At that point, students no longer need to hide AI use. Instead, how they used AI becomes part of learning capability, because the quality of that use is directly tied to the quality of thinking.

Process-Based Assessment Is Not a New Burden

Many teachers feel process-based assessment is too demanding. They worry it will increase documentation and make criteria vague. In reality, it works the other way. It is far more stable to verify a structured process than to doubt the authenticity of a single final submission.

Process assessment is not a demand to record everything. It is a method of checking a few critical points. Even a small number of checkpoints can reveal a student’s reasoning level and depth of understanding.

For example, just five items can be enough: the initial problem-definition sentence, a short summary of prompts given to AI, a list of references and reasons for selecting them, one or two mid-course revisions, and the final claim with acknowledged limitations. With only these, the student’s thinking becomes visible.

Performance Tasks Can Be Rebuilt Without Starting From Scratch

There is no need to reinvent performance assessment from the ground up. Even a slight redesign of existing assignments can transform the entire structure. The most important factor is not length, but sequence—the structure that reveals the movement of thought.

A typical assignment looks like this: “Write a report on the topic.” A redesigned assignment can say: “Explain how you defined the topic, summarize your research process including AI use, and describe your final conclusion and its limitations.”

This becomes far more powerful than plagiarism detection. It evaluates not just what a student submitted, but how they arrived there. It shifts assessment away from surface-level output and toward the trajectory of reasoning.

What Changes When Process-Based Assessment Takes Root

Once process-based assessment becomes established, the atmosphere in the classroom changes. Students think more carefully about questions, value evidence over answers, stop accepting AI responses at face value, and treat revision and reflection as part of learning itself. The center of learning moves back from the product to the process.

In that setting, AI is no longer something that “ruins learning.” It becomes a mirror that reveals thinking. What the student questioned, what they selected, and what they doubted becomes clearer than ever. AI can provide an answer, but learning is revealed in how that answer is handled.

Teachers, too, shift roles. They are no longer enforcers policing results, but professionals trained to read and guide thinking. Teaching and assessing cognition returns to the center of education.

Rebuilding Starts Only When We Admit the Collapse

The collapse of performance assessment is not the student’s fault. It is not AI’s fault, either. The problem is that the era has changed while the assessment criteria have not. The change has already happened, and the old standards no longer explain the new reality.

The more schools cling to outcome-centered grading, the more teachers grow suspicious, students grow secretive, and education is reduced to formality. When assessment loses trust, learning loses direction. Shifting to process-based assessment is not optional. It is the minimum response required to catch up with what has already changed.

In the next article, the focus turns to the most sensitive issue in this transition: why so many teachers and professors remain silent even though they know AI is here, and why this should be understood not as an individual attitude problem, but as a structural problem built into the system.

[Part 3 Preview]
Teachers’ Dignity and the Institutional Trap 
— Why Do Teachers Fall Silent in Front of AI?

Author Bio
Jung Yeon-Tae, President of the Korea Information Service of the Construction Industry. He completed his master’s and doctoral coursework at Stevens Institute of Technology in the United States. He worked in science, technology, and information and communications industries in New York and New Jersey.

He later served in leadership roles in public enterprises and private research institutions in Korea. He writes extensively on artificial intelligence, education, and civilizational transformation.