Educators need to reassess the objective for students to learn in each class, and then design the participation to accomplish that in an AI world. For example, if the objective is to be able to write a essay, they should have them write by hand in class. This is the way things were done just a few decades ago, we should be flexible in reevaluating the best method in this new world
I agree. Anything assigned as take-home will just be offloaded to AI. Maybe we could take a page from some successful countries' book and do more work in school while leaving afterschool as a more open time for kids to explore their interests and hobbies (assuming they don't fill it with screentime)
I forgot that the Trump sledge hammer also dismantled the department of education. What a train wreck.
I am on a committee regarding AI use and cheating in our school. It looks quite different and has different solutions in different subjects.
Generally, the solution is live assessment, oracy, but it takes time. Pencil to paper works, socratic discussion works. Providing frames and heuristics and applying to real world examples works. The only drawback is coverage. Less coverage of standards because assessment without paper writing takes a long time.
AI isn't helping young people cheat. It's forcing adults to finally ask the real questions: what's actually worth learning, and how do we make that learning mean something.
School was outdated and broken long before AI showed up. So the knee-jerk reaction, that we just need to keep kids from "cheating," misses the entire point.
I've got at least 10,000 words on this. It's complicated. But start here:
Think about the last time you really learned something. What was the motivation? How did you do it? Seriously, pause and sit with it. My guess is it wasn't for a grade.
Yes, 100%. Grades were an imperfect convenience to assess competence and lesson plans were the failure of a teacher to adapt to an individual student. AI solves both problems.
The undetectable/detectable debate of AI is a red herring, the underlying issue is that we, for much longer than AI has been around, have few tools or methods to verify who wrote or did something. You could cognitively surrender work to your friend, mom, sister, or a paid proxy long before you could to a chat or.
While i agree with the some of commentary and understand the data set examples, it is a real stretch to blame AI's evolution for the defeciencies in math, science , reading and critical thinking and problem solving. (National Assessment of Education Progress , ( 2025). The US instructional designs used in public and private education has not helped either. Critical thinking and problem solving skills are just as residentally deficient across generations.
In this age of the new knowledge economy, we should resist the temptation of being too aggresive and provacative about the perils of AI, rather it may be more productive to work collectively to reform teaching preparation, adapt to the current audience of both student and adult learners by using more innovative learning models such as expanding the use of self-directed and experiential learning integration both in the classrooms and outside the classroom. In American education..... from K-12 to all forms of learning design, its time to reform and change the lens we are using which got its start in the one room school house in the 1800's. Let's spend more time on learning about this emerging technology ensuring AI is the tool it can become not the boggey man.
The first two graphs depict reading and math scores plummeting since 2020, long before AI. It’s much more plausible that remote learning after the Covid pandemic is the real culprit.
When I was at school, my Dad used to do my art homework for me - not to encourage me to cheat, but to encourage me to maximise my time on the subjects I really cared about.
There’s nothing wrong with outsourcing some tasks or some aspects of a task. But there are limits. Or there ought to be!
The funding data on this is striking too, since the best-funded category in AI-education VC by far is habit-forming practice apps in the Duolingo mold (streaks, instant answers, repetition). Meanwhile tools that actually verify whether real learning happened are comparatively starved of investment, described by one research firm as underfunded relative to its classroom importance. Feels like the market is structurally rewarding exactly the "give the answer, skip the struggle" pattern you're describing, mainly because engagement metrics are easier to sell to investors than proof of actual learning. Here is the RichStorm article to read more: https://www.richstorm.co/post/ai-in-education-where-things-actually-stand-in-2026
Technically, the Covid period was a time when cheating was even easier than today. Even if it were a live test we could have Google open right in front of us. I think AI is pretty much comparable to the Covid period.
Wow. Excellent piece. I think many schools pride themselves in being “tech forward” and “innovative,” so they throw AI/iPads/Chromebooks into the classroom, but this is literally worst-case scenario. All that does is digitize a broken model. We’re truly at an inflection point where we need to reimagine the model itself. And I 100% agree that with AI tutors, there is tremendous hope.
I think the idea of “alternative” schooling freaks people out, but everything about our world is changing, and school needs to be the place that prepares our kids for the real world. (Otherwise, what’s the point of school?) Thank you for writing.
You hit on a critical point about how commercial AI tools encourage cognitive offloading. However, the fact that a generic tool can so easily pass our school exams or write these essays doesn't just expose the flaws of tech—it exposes that our school systems have been rewarding the mere reproduction of knowledge for far too long.
Learning is inherently a messy, slow process that requires deep cognitive friction, which commercial shortcuts easily wipe out. But since AI is already here and blanket bans are rarely effective, the real work lies in how we respond. Beyond changing what we test, our best path forward might be moving away from these generic platforms and focusing on intentional 'Educational AI'—technology specifically designed to prompt deep, critical reasoning and sharpen human agency, rather than replacing it.
I think this is an extensive conversation that needs to take place in educational spaces. One way that I combat the downside of AI is eliminating the "boogey man" cloak it's usually draped in when brought up. I discuss the pros and cons of AI in my classroom, utilize it as a tool with my students, and train them to use it as a learning tool --- never a replacement of research, thinking, etc. This is the short answer, but I hope more people continue the conversation so we can use AI as the tool and not the "thinker".
The tool-removal studies are the part that should worry people most, and they also settle a question sitting underneath this whole debate. A skill that disappears when the tool is taken away was never transferred to the student. It was rented. That is the line between offloading and dependency, and your own evidence runs it: the gains vanish on removal, so this is dependency, not a quiet trade of detail-level skill for higher-level skill.
Which is why the same tool does opposite things depending on timing. Someone who built the foundation and now offloads detail to work at a higher level is multiplying a skill they own. A student who offloads before building it never forms the skill at all. Same behavior, opposite result. The teacher who said "multiplication tables before the calculator" had the whole thing.
The harder question is what mass education was actually built to produce. If the honest answer is a trustworthy signal of capability for employers, more than learning for its own sake, then AI is not only degrading learning. It is degrading the signal. A degree was valuable because it was expensive to fake. Once it is cheap to fake, it stops carrying information, and some of the panic about thinking is really alarm about a signal collapsing. That does not make the alarm wrong. It makes it deeper.
I ran almost this exact structure on professional sport recently, another system whose stated purpose and real function have quietly come apart. Different system, same question. What does it actually run on, and what is it spending to keep going?
"The harder question is what mass education was actually built to produce." Amen brother.
Right now it just produces debt. The college industrial complex is handicapping our economy. You are right that we have to have these first principles discussions because otherwise schools are just going to keep finding new boogey men.
This feels like the practical edge of the question: not whether students can produce answers faster, but whether school still gives them enough practice becoming people who can think without the tool.
Educators need to reassess the objective for students to learn in each class, and then design the participation to accomplish that in an AI world. For example, if the objective is to be able to write a essay, they should have them write by hand in class. This is the way things were done just a few decades ago, we should be flexible in reevaluating the best method in this new world
I agree. Anything assigned as take-home will just be offloaded to AI. Maybe we could take a page from some successful countries' book and do more work in school while leaving afterschool as a more open time for kids to explore their interests and hobbies (assuming they don't fill it with screentime)
I forgot that the Trump sledge hammer also dismantled the department of education. What a train wreck.
I am on a committee regarding AI use and cheating in our school. It looks quite different and has different solutions in different subjects.
Generally, the solution is live assessment, oracy, but it takes time. Pencil to paper works, socratic discussion works. Providing frames and heuristics and applying to real world examples works. The only drawback is coverage. Less coverage of standards because assessment without paper writing takes a long time.
AI isn't helping young people cheat. It's forcing adults to finally ask the real questions: what's actually worth learning, and how do we make that learning mean something.
School was outdated and broken long before AI showed up. So the knee-jerk reaction, that we just need to keep kids from "cheating," misses the entire point.
I've got at least 10,000 words on this. It's complicated. But start here:
Think about the last time you really learned something. What was the motivation? How did you do it? Seriously, pause and sit with it. My guess is it wasn't for a grade.
Yes, 100%. Grades were an imperfect convenience to assess competence and lesson plans were the failure of a teacher to adapt to an individual student. AI solves both problems.
The undetectable/detectable debate of AI is a red herring, the underlying issue is that we, for much longer than AI has been around, have few tools or methods to verify who wrote or did something. You could cognitively surrender work to your friend, mom, sister, or a paid proxy long before you could to a chat or.
Big tech and the 0.00I% class forced ai on society for a specific reason.
While i agree with the some of commentary and understand the data set examples, it is a real stretch to blame AI's evolution for the defeciencies in math, science , reading and critical thinking and problem solving. (National Assessment of Education Progress , ( 2025). The US instructional designs used in public and private education has not helped either. Critical thinking and problem solving skills are just as residentally deficient across generations.
In this age of the new knowledge economy, we should resist the temptation of being too aggresive and provacative about the perils of AI, rather it may be more productive to work collectively to reform teaching preparation, adapt to the current audience of both student and adult learners by using more innovative learning models such as expanding the use of self-directed and experiential learning integration both in the classrooms and outside the classroom. In American education..... from K-12 to all forms of learning design, its time to reform and change the lens we are using which got its start in the one room school house in the 1800's. Let's spend more time on learning about this emerging technology ensuring AI is the tool it can become not the boggey man.
The first two graphs depict reading and math scores plummeting since 2020, long before AI. It’s much more plausible that remote learning after the Covid pandemic is the real culprit.
When I was at school, my Dad used to do my art homework for me - not to encourage me to cheat, but to encourage me to maximise my time on the subjects I really cared about.
There’s nothing wrong with outsourcing some tasks or some aspects of a task. But there are limits. Or there ought to be!
The funding data on this is striking too, since the best-funded category in AI-education VC by far is habit-forming practice apps in the Duolingo mold (streaks, instant answers, repetition). Meanwhile tools that actually verify whether real learning happened are comparatively starved of investment, described by one research firm as underfunded relative to its classroom importance. Feels like the market is structurally rewarding exactly the "give the answer, skip the struggle" pattern you're describing, mainly because engagement metrics are easier to sell to investors than proof of actual learning. Here is the RichStorm article to read more: https://www.richstorm.co/post/ai-in-education-where-things-actually-stand-in-2026
Technically, the Covid period was a time when cheating was even easier than today. Even if it were a live test we could have Google open right in front of us. I think AI is pretty much comparable to the Covid period.
Wow. Excellent piece. I think many schools pride themselves in being “tech forward” and “innovative,” so they throw AI/iPads/Chromebooks into the classroom, but this is literally worst-case scenario. All that does is digitize a broken model. We’re truly at an inflection point where we need to reimagine the model itself. And I 100% agree that with AI tutors, there is tremendous hope.
I think the idea of “alternative” schooling freaks people out, but everything about our world is changing, and school needs to be the place that prepares our kids for the real world. (Otherwise, what’s the point of school?) Thank you for writing.
You hit on a critical point about how commercial AI tools encourage cognitive offloading. However, the fact that a generic tool can so easily pass our school exams or write these essays doesn't just expose the flaws of tech—it exposes that our school systems have been rewarding the mere reproduction of knowledge for far too long.
Learning is inherently a messy, slow process that requires deep cognitive friction, which commercial shortcuts easily wipe out. But since AI is already here and blanket bans are rarely effective, the real work lies in how we respond. Beyond changing what we test, our best path forward might be moving away from these generic platforms and focusing on intentional 'Educational AI'—technology specifically designed to prompt deep, critical reasoning and sharpen human agency, rather than replacing it.
Does this surprise anyone?
I think this is an extensive conversation that needs to take place in educational spaces. One way that I combat the downside of AI is eliminating the "boogey man" cloak it's usually draped in when brought up. I discuss the pros and cons of AI in my classroom, utilize it as a tool with my students, and train them to use it as a learning tool --- never a replacement of research, thinking, etc. This is the short answer, but I hope more people continue the conversation so we can use AI as the tool and not the "thinker".
The tool-removal studies are the part that should worry people most, and they also settle a question sitting underneath this whole debate. A skill that disappears when the tool is taken away was never transferred to the student. It was rented. That is the line between offloading and dependency, and your own evidence runs it: the gains vanish on removal, so this is dependency, not a quiet trade of detail-level skill for higher-level skill.
Which is why the same tool does opposite things depending on timing. Someone who built the foundation and now offloads detail to work at a higher level is multiplying a skill they own. A student who offloads before building it never forms the skill at all. Same behavior, opposite result. The teacher who said "multiplication tables before the calculator" had the whole thing.
The harder question is what mass education was actually built to produce. If the honest answer is a trustworthy signal of capability for employers, more than learning for its own sake, then AI is not only degrading learning. It is degrading the signal. A degree was valuable because it was expensive to fake. Once it is cheap to fake, it stops carrying information, and some of the panic about thinking is really alarm about a signal collapsing. That does not make the alarm wrong. It makes it deeper.
I ran almost this exact structure on professional sport recently, another system whose stated purpose and real function have quietly come apart. Different system, same question. What does it actually run on, and what is it spending to keep going?
"The harder question is what mass education was actually built to produce." Amen brother.
Right now it just produces debt. The college industrial complex is handicapping our economy. You are right that we have to have these first principles discussions because otherwise schools are just going to keep finding new boogey men.
https://thecompoundcomeback.substack.com/p/gen-ed-requirements-are-the-reason
This feels like the practical edge of the question: not whether students can produce answers faster, but whether school still gives them enough practice becoming people who can think without the tool.