What Are We Measuring?

If our classrooms are still stopping at teaching bubble sort, I think that's a grave mistake. The classroom should go further: “Hey look, this is what an LLM is doing. Here is the code. What do you understand about it? What would you accept? What would you question?” What kind of data are we sorting? How large is it? Is it already nearly sorted? What does “improve” mean here? Now go ahead, use an LLM, and improve it. Then explain what you have changed, test whether it works, and justify why it is better for this situation. That, I think, is how it should go. Maybe.

In 1995, wolves were reintroduced to Yellowstone National Park after being absent for decades. Their return changed much more than the population of elk they hunted. Changes in elk populations and grazing patterns contributed to vegetation recovery in some areas, with effects extending to birds, beavers, and other wildlife. The wolves did not directly change everything. Their presence altered relationships within the ecosystem, setting off a chain of changes.

Now, look at our own education ecosystem. AI has entered the system. And if we are still thinking that it will change only one part of our education system, perhaps how we teach, how students write assignments, or how we conduct examinations, how wrong are we? The arrival of AI is not merely the addition of another tool. It changes the relationships between teachers, students, knowledge, learning, and assessment. Just as the return of wolves affected Yellowstone in ways that went far beyond the wolves themselves, AI has the potential to reshape the entire educational ecosystem.

How long are we going to limit our research to comparing students who used AI with those who did not? Such comparisons are important, but shouldn't we be moving beyond them? Shouldn't we be asking what kind of AI use improves learning, what creates dependency, and how our assessments need to change? Look, AI has already entered the system. It is part of the ecosystem now. There is no going back.

Rather than continuing to measure learning through the lens of "with AI" and "without AI," shouldn't we be questioning the way we assess learning itself? Why are we still relying on the same assessment methods when the ways students learn, think, and solve problems are changing? If students are learning, thinking, and solving problems differently, shouldn't our assessments evolve too? We cannot introduce something that changes the entire ecosystem and then expect the old ways of measuring learning to remain equally meaningful.

What if the problem is not with how students are using AI, but with how we are still assessing them? The question is no longer how to keep AI out of education, but how to make learning better with AI in it.

The deeper you study, the less you seek an award for it.