The era of banning AI is coming to an end. Shouldn't universities next consider 'assessment design'?

Generative AI is rapidly permeating educational settings.

Generative AI, including ChatGPT, has become capable of handling tasks such as writing, summarizing, and even programming with high precision. Meanwhile, many universities are continuing to debate how much AI usage should be permitted in reports.

However, I have recently felt that this debate itself might be slightly off-track.

The era of determining 'whether AI was used' is coming to an end

Many universities are moving forward with creating rules such as:
・Prohibiting the use of AI
・Declaring if it is used
・Not submitting text created by AI
Of course, this is a necessary way of thinking to protect academic integrity.

However, in reality, the performance of generative AI is improving rapidly.

A few years ago, it was relatively easy to spot 'AI-like text.' But now, with just a little human intervention, it has become extremely difficult to distinguish it from text written by a human.

In a few years, it will likely be almost impossible for instructors to look only at the final product and judge,

'This is AI.'

In other words,

the game of trying to detect whether AI was used will eventually come to an end.

So, what should universities evaluate?

Here, we need to shift the axis of evaluation.

What is important is not

'whether AI was used.'

Isn't it rather,

'can you explain the submitted content in your own words?'

For example, suppose a student uses AI to write a report.

If you ask them about the content,

* Why did you reach this conclusion?

* Are there other ways of thinking about this?

* What did you read from this data?

* How would you respond to opposing views?
If they cannot explain it in their own words, it cannot be said that they truly understood it.
Conversely, if they use AI but interpret, think about, and explain it in their own way, it can be said that the student is using AI as a 'tool'.
What should be evaluated is not the use of AI itself, but the depth of that understanding.

However, this also has its challenges

A question arises here.

Serious students will likely follow the rules.

But what about students who want to get high grades?

It is entirely possible to have AI write a high-quality report and submit it after making minor revisions.

Furthermore, some students might even self-report that 'I did not use AI.'

Current university education assumes the honesty of students in many situations.

That is precisely why this problem cannot be solved by simple rules alone.

Toward assessment design that assumes AI usage

So, what should be done?

I think an era is coming where we will design assessments based on the premise of 'using AI'.

For example,

* Clearly state how AI was used if it was used
* Submit the prompts that were used
* Write about where you made revisions to the content output by AI
* Explain it orally after submission
* Confirm the level of understanding through discussions or presentations

With these assessment methods, even if AI is used, it is possible to confirm whether the student truly understands the material.

In other words,

evaluate the ability to 'explain' itself, rather than the final product.

It is a shift toward that kind of thinking.

Abilities that universities should cultivate in the AI era

Generative AI will continue to improve in performance.

The debate of just 'banning AI' will eventually reach its limit.

That is why universities need to change toward

education that does not prevent the use of AI,

but rather cultivates talent that can think with their own minds and explain in their own words, even when using AI.

What is being questioned in the AI era is not just 'writing ability'.

What is truly being questioned is the ability to understand, think, and explain.