Assessment in the Age of AI: What to Redesign, and How

In an era when a plausible report takes an hour at home, assessment that looks only at "the finished product" starts to wobble. It is hard to be sure who wrote it, and the score does not reflect the student's actual ability. The easier AI makes it to produce a product, the more assessment has to shift its center of gravity from the result to the process. Let's look at how to redesign it.

Assessment under strain, and its purpose revisited

The real substance of the assessment crisis is not policing cheating but asking again what assessment was trying to measure. These need to be told apart.

• What has become hard to measure: The polish of a finished piece of writing, the volume of information — the things AI can supply on a student's behalf.
• What still matters: The process of thinking, the reasoning behind choices, the traces of revision — the things only a person can show.
• What we now have to look at: The ability to use AI appropriately and critically.

What a student thought and chose on the way to a result is a more honest object of assessment than what they turned in.

So moving assessment's focus from the product to the learning journey is the most fundamental response.

A process for designing process-based assessment

Turning an abstract principle into an actual lesson takes a procedure. Here is the order I would recommend.

1. Require a record of the process: Have students submit drafts, revisions, and notes together so the flow of thinking is preserved.
2. Disclose AI use: Have them state where they used AI and what they did themselves.
3. Check orally: Have students explain the key sections in their own words to gauge understanding.
4. Make it performance-based: Increase tasks that require live demonstration — discussion, experiments, presentations.

When one middle school's Korean language department allocated half of the report score to the draft and revision process, students started spending their time revising instead of copying a result. Change the assessment criteria and learning behavior follows. Assessment is a signal you send to students.

Process-based assessment has its own trap, though. Demand a large number of drafts and revisions for their own sake, and students can manufacture a fake process that satisfies the form. So the focus of assessment has to be the traces of change, meaning what was revised and why rather than the volume. It also matters to avoid penalizing use itself and instead assess how appropriately AI was used, so that a student who honestly discloses it is not the one who loses out. That is what gets students to reveal their use and learn together rather than hide it.

Key takeaways

Assessment in the age of AI has to shift its center of gravity from the finished product to the process of thinking. With a record of the process, disclosed AI use, oral checks, and performance tasks, you can design assessment that looks into the learning journey. Remember that the assessment criteria are what set the direction of student learning. Please start with one small change on your next performance assessment: collect the draft and the revision history along with the product. Take the criteria you validated on one task and widen them to another subject next semester, and assessment will slowly turn from a place where students are scored into a place that leads them to learn more deeply.