From Banning AI to Redesigning Assessment: What Has Changed Since I Wrote About Generative AI in Education

In October 2023, almost one year after generative AI entered the mainstream conversation, I wrote an article titled “Leveraging Generative AI in Education: A Balanced Perspective.” At that time, the debate around tools such as ChatGPT was still relatively new. The central question was simple: should educators allow generative AI in classrooms, or should they prohibit it?

My position then was that the answer could not be a simple yes or no.

I argued that generative AI was a double-edged sword. On one side, it could help students improve writing, language, research, and learning support. It could also help teachers reduce some repetitive tasks and provide more personalized guidance. On the other side, it raised legitimate concerns around cheating, overdependence, and the weakening of authentic student skills.

My recommendation at the time was not to ban AI blindly, but to govern its use carefully. I suggested four practical principles: clear communication, scope limitation, ongoing assessment, and teacher oversight.

Two and a half years later, I still believe those principles are valid. But I also believe the conversation has moved to a much deeper level.

The real question is no longer: Should students be allowed to use AI?

The real question is: What should education be testing in a world where AI is always available?

This question became even more urgent after Princeton University voted in May 2026 to require proctoring for all in-person exams, ending a 133-year tradition of unproctored examinations under its honor code. The new policy will take effect on July 1, 2026, and the decision was driven largely by concerns around AI-assisted cheating. (The Princetonian)

At first glance, this may look like a normal academic integrity response: students found a new way to cheat, and the university responded with stricter supervision. But I think this interpretation is too narrow.

AI did not simply create a cheating problem. It exposed a deeper weakness in the way we assess learning.

For more than a century, the traditional exam was built around memory, isolation, and controlled access to information. Students were placed in a room, separated from external resources, and asked to reproduce what they knew. That model made sense in a world where knowledge was scarce, difficult to access, and largely dependent on books, lectures, and personal memorization.

But we no longer live in that world.

Knowledge is now instantly accessible. AI tools can summarize, explain, translate, draft, code, calculate, and analyze within seconds. This does not mean students no longer need to learn. On the contrary, it means they need to learn differently.

The skill that matters today is not only whether a student can remember information under pressure. It is whether the student can ask the right question, evaluate the quality of an answer, identify errors, apply knowledge in context, and make sound judgments using the tools available.

This is why simply adding proctors to exam rooms feels more like a defensive reaction than a future-ready solution. Princeton is not alone. Stanford has also moved toward allowing broader proctoring of in-person exams following concerns around academic integrity and AI. (Stanford News)

These decisions are understandable. Universities must protect fairness, credibility, and trust. A degree must continue to mean something. But supervision alone does not solve the real problem. It protects the old assessment model without fully questioning whether that model still reflects the world students are entering.

This is where my 2023 view needs to evolve.

At that time, I focused on managing AI use in the classroom. Today, I believe we must go further. We need to redesign learning and assessment around the reality that AI is now part of the educational environment.

That does not mean allowing students to use AI without limits. It does not mean accepting AI-generated work as genuine student achievement. It also does not mean lowering standards.

It means changing what we measure.

Instead of asking students only to produce an answer, we should ask them to explain how they reached it. Instead of testing whether they can write without support, we should test whether they can think with support. Instead of banning tools that will exist in every workplace, we should teach students how to use them responsibly, critically, and ethically.

A better assessment model would test judgment, reasoning, originality, and application. It would include open-book and AI-assisted assignments where students must disclose their use of tools. It would include oral defense, scenario-based work, project-based assessment, peer review, and reflection. It would also continue to include controlled assessments where necessary, especially for foundational knowledge. But those controlled assessments should be part of a broader model, not the whole model.

This distinction is important.

Foundational knowledge still matters. A student cannot evaluate an AI-generated answer if they do not understand the subject. A doctor, engineer, teacher, lawyer, or data scientist cannot rely blindly on a tool. They need enough expertise to know when the tool is right, when it is wrong, and when the problem itself has been poorly framed.

But memorization alone is no longer enough. In many professional environments, the real value is not in recalling every detail. It is in knowing what exists, when to use it, how to validate it, and how to make decisions from it.

The same shift is happening in the workplace. Gartner reported in May 2026 that many organizations deploying AI or autonomous technologies had reduced their workforce, but workforce reduction did not correlate with higher return on investment. The stronger correlation was with what Gartner described as “people amplification”: using AI to make employees more productive by redesigning skills, roles, and operating models. (Gartner)

Education should take this lesson seriously.

The future is not about replacing the learner with AI. It is about amplifying the learner through AI.

The same applies to teachers. AI should not replace educators. It should give them better ways to personalize learning, identify gaps, support students, design richer assessments, and spend more time on the human parts of education: mentorship, motivation, ethical guidance, and critical dialogue.

Looking back at what I wrote in 2023, I would still keep the same foundation: generative AI in education requires balance, clear rules, limited and purposeful use, continuous evaluation, and teacher oversight.

But I would now add one more point: assessment redesign.

Without redesigning assessment, institutions will remain trapped in a cycle of reaction. Students adopt new tools. Universities restrict them. Students find new ways around the restrictions. Universities add more controls. This cycle may reduce some misconduct, but it does not prepare students for the AI-enabled world they will graduate into.

The stronger path is to build an assessment model where AI use is visible, guided, and meaningful. A model where students are not rewarded for pretending AI does not exist, but are assessed on how well they can use it with integrity, understanding, and judgment.

In 2023, the debate was about whether AI should enter the classroom.

In 2026, that debate is already outdated.

AI is already in the classroom. It is already in the workplace. It is already part of how knowledge is produced, accessed, and applied.

The challenge now is not to keep AI outside the exam room. The challenge is to build an education system where the presence of AI does not weaken learning, but forces us to define learning more seriously.

Because the purpose of education was never simply to produce answers.

The purpose of education is to develop people who can think, judge, question, create, and act responsibly.

AI did not break that mission.

It reminded us how important it is.

---

Originally published in the VerbionPulse newsletter on LinkedIn: From Banning AI to Redesigning Assessment: What Has Changed Since I Wrote About Generative AI in Education.