83.6% of Canadian teachers see their students' critical thinking declining with generative AI

By Telegraph-Journal · October 3, 2026.

A Fraser Institute survey of 756 teachers across Canada, cited in an opinion piece published by Telegraph-Journal, puts numbers to a suspicion that has been circulating in staff rooms for months: 95.6% of teachers believe that at least some of their students use generative artificial intelligence (AI) to do their homework, and 55.9% go further and think AI does the work for them outright, rather than supporting it. The figure that should worry most is a different one: 83.6% of respondents perceive a decline in their students' critical thinking compared with five years ago, and point to AI as one of the main suspects.

That last figure should be read with a caveat: it is teacher perception, not a controlled measurement of cognitive skills using standardized tests before and after. But the direction matches the first-hand account of Asyle Nahas, the Grade 11 high school student who signs the original article in Telegraph-Journal: nights facing a blank essay that once meant creative block and today are resolved with a two-sentence prompt the day before the deadline.

The piece also quotes Victor Lee, head of AI and education at Stanford University, with an idea that should be at the center of the debate and almost never is: "it ultimately depends on how it is used". According to Lee, basic skills and prior mastery of a subject determine whether AI acts as a tool that amplifies (an "augmentative" use, in his words) or as a substitute that thinks for the student. Put another way: someone who already knows how to write an essay uses AI as a proofreader or for brainstorming; someone who does not hands over authorship entirely. The problem is not the tool, it is at what point on the learning curve it is introduced.

New Brunswick, the Canadian province where this article is published, has already made a move with a curriculum framework that explicitly promotes "learning without AI", precisely to protect those foundational skills before students delegate to the machine. It is a more nuanced response than an outright ban or adoption without criteria: it does not bar the tool, it sets limits on when and how it is used.

Our reading is that this debate reproduces, in miniature, the same tension we already see in the labor market: AI does not spread its effects evenly, it spreads them according to who uses it and with what prior grounding. In the classroom, as in the office, the winner is whoever already had judgment of their own and uses AI to multiply it, and the loser is whoever uses it to avoid developing it. The scientific-calculator analogy the article itself raises is only half right: a calculator does not write the reasoning for you, and a language model can if you let it, which demands far more deliberate instructional design than simply "allowing" or "banning" the use of AI in a subject.

In the long run, a generation that learns to orchestrate AI instead of delegating its thinking to it is exactly the profile best prepared for an economy of material abundance: able to devote its own effort to judgment, creativity and discernment, the pieces no AI fully replaces. But that outcome is neither automatic nor free: it requires schools, families and curriculum officials —like the framework New Brunswick is already trialing— to make active decisions now, while the generation currently in high school finishes forming the thinking habits it will use for the rest of its life.